a dissertation submitted to the department of …wr860nt2714/katherine roelofs... · thermodynamic...
TRANSCRIPT
INTERFACE ENGINEERING IN INORGANIC-ABSORBER
NANOSTRUCTURED SOLAR CELLS
A DISSERTATION
SUBMITTED TO THE DEPARTMENT OF MATERIALS SCIENCE &
ENGINEERING
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
KATHERINE ELIZABETH ROELOFS
August 2015
http://creativecommons.org/licenses/by-nc/3.0/us/
This dissertation is online at: http://purl.stanford.edu/wr860nt2714
© 2015 by Katherine Elizabeth Roelofs. 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.
Stacey Bent, 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.
Michael McGehee
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.
Friedrich Prinz
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
Katherine Roelofs Ph.D. Dissertation
iv
Katherine Roelofs Ph.D. Dissertation
v
Abstract
The focus of this work is variants on the dye-sensitized solar cell (DSSC) that
employ inorganic materials as the light absorber, replacing the organic dye molecules
used in DSSCs. Such DSSC-inspired devices are emerging technologies in the broader
class of thin film solar cells, and include quantum-dot sensitized solar cells (QDSSCs)
and perovskites solar cells (PSCs). Quantum-dot sensitized solar cells employ
semiconductor nanocrystals, or quantum dots, as the light absorber. The band gap of
quantum dots varies with size, allowing for a tunable absorption onset in these devices,
among other benefits. PSCs, in which the absorber is CH3NH3PbI3, or variants thereof,
with the perovskites crystal structure, first attracted attention in 2012 and have shown an
unprecedented rise in efficiency to current record values of 20.1%. QDSSCs and PSCs
can be fabricated completely from solution processed materials that can be low-purity,
contrasting favorably with the industrial standard, silicon solar cells, which require
expensively-processed, high-purity silicon. This tolerance to defects is partially due to
the nanostructured design of some PSCs and all QDSSCs, in which a nanostructured bulk
heterojunction is formed between the electron-transport material, the absorber, and the
hole-transport material. However, the high interfacial areas involved in such designs
leads to high rates of interfacial recombination, causing losses in photocurrent, and
limiting device efficiency.
In this work, I will present methods to reduce interfacial recombination in these
inorganic-absorber nanostructured solar cells though surface modifications. In QDSSCs,
these include growing ultra-thin insulating metal oxide films by atomic layer deposition
(ALD) at the interface and controlling of the nucleation and growth of the inorganic
absorber. These studies provide insight into the working mechanisms of QDSSCs,
through a combination of the highly-controlled nature of ALD, where films can be grown
a single atomic layer at a time and an interface can be atomically engineered, X-ray
absorption measurements of interfacial geometric and electronic structure, and detailed
studies of the resulting solar cell performance. I will also detail the use of ALD to grow
entire material layers in perovskites solar cells, both ALD TiO2 as the electron-transport
material, and ALD NiOx as the hole-transport material. Despite their high efficiencies,
Katherine Roelofs Ph.D. Dissertation
vi
PSCs are unstable and rapidly degrade when exposed to moisture or excessive heat. The
use of ultra-conformal inorganic layers grown by ALD to cap the perovskites absorber,
instead of the currently-employed organic layers, has the potential to improve the
stability, and thus efficiency, of perovskites solar cells.
Acknowledgements
As the culmination of my formal education, completing my doctorate could not
have been completed without the help of many people along the way – advisors, mentors,
colleagues, family, and friends. I am greatly indebted to everyone who has taught,
trained and supported me, and I hope I have already begun to pass along this mentoring
to the next generation as part of a lifetime goal of mine. As they say, it takes a village,
and I think I have had at least three villages worth of help!
First, I would like to thank my family – my aunts, uncles and cousins, my
siblings, and particularly my parents. I would not be here without my parents’ help. As
chemists, my parents encouraged me from an early age in the sciences, and particularly
when I entered college it was with their encouragement that I pursued engineering. I
think from my father I learned a lot about experimental design, such as the “on or off”
experimental method – before you do any close testing, double a variable and see if it
makes any difference. From my mother, I undoubtedly learned multi-tasking,
organization, and of particular use has been the advice to always have at least two
projects running at once. Having my sister Becca close by in Berkeley for the last few
years of my time here has been a great moral support and source of fun weekends in the
city. My older brother Kevin continues to provide a guide in navigating the paths of
upper education, and my younger brother Sean has excelled at being my favorite younger
brother. My aunt Libby and uncle Mark have served as a second family during my time
at Stanford, with Mark organizing expeditions and weekend morning Frisbee and Libby
teaching me to surf (and making me coffee that knocks my teeth off!); a great benefit of
going to Stanford has been getting to spend more time with them. And of course, my
cousins Kimmy, Emily, and Chelsea have brought a lot of fun and joy into my life out
here. My aunt Margie and uncle Bruce have encouraged me throughout my degree, with
Katherine Roelofs Ph.D. Dissertation
vii
Margie’s great advice on life and science (as an early example of a materials scientist!),
and Bruce’s reminders to take some time to play music and have fun. I hope to be seeing
more of my favorite Florida cousins Jen, Cynthia, and Annette now that I’m moving back
east. Finally, I’d like to thank my grandparents, Dawn and Gene, who have connected
and grounded my life from an early age. Dawn has always nurtured my right brain, while
Gene has given me an interest in history and heritage.
The second village that continued my scientific education was my undergraduate
experience in Chemical Engineering at Princeton University. A number of professors
were influential in giving me the tools and groundwork to tackle any research project in
the sciences. Early on, Robert Prudhomme guided my interests in the Bioengineering
certificate branch off of Chemical Engineering, as well as guiding my choice of summer
internships. It was in Athanassios Panagiotopoulos’s class that I first fell in love with
thermodynamics, and it is in his script from the handwritten class notes that I still see the
thermodynamic relations in my mind. My interest in materials science began in Richard
Register’s polymer class, and continued and expanded to inorganic materials in George
Scherer’s class on ceramics. Throughout my undergrad career, it was a pleasure to
interact with Lynn Loo, who has since shown her continued availability for help and
advice. My heartfelt thanks go always to Nan Yao, the advisor of my senior research
project, who opened my eyes to the field of microscopy and gave me a lasting love for
nanoscale imaging. I was extremely lucky that my senior project also brought me into
contact with Jerry Poirier formerly of the Imaging Analysis Center (IAC) at Princeton, a
scientist and environmentalist who drove me an hour each way from my hometown in
Delaware to Princeton over the summer for my senior research project.
At Stanford, my education was continued with a third group of mentors guiding
my graduate career. It is with deep gratitude that I thank my advisor Stacey Bent. I have
learned a lot from her these past five years. I am incredibly grateful for her organized,
responsive, and thoughtful management of the group and of my projects here. I am also
grateful that Stacey has always supported my long-term goals, both personal and
professional. Through the many courses I have taken, Stacey stands out as an incredible
teacher and lecture, in her organization and clarity of explanation. My knowledge of
Katherine Roelofs Ph.D. Dissertation
viii
surface science and materials characterization has really been formed through these past
five years of research with Stacey. In group meetings, I still do not know how she
manages to maintain razor focus on our experimental details and the implications of
various results of each of her students through four hours of meetings on Wednesdays,
with only a cup of tea for caffeine. I also feel I have learned a great deal from Stacey
about how to write scientific proposals; sometimes it seems as if Stacey possesses a
‘magic wand’ which transforms long-winded flat paragraphs into compelling statements.
Finally, I have learned about group management, lab safety, and a general approach to
making progress in scientific research. Stacey has even welcomed the group into her
home each year for a summer get-together, which has been a great pleasure. I could not
have asked for a better advisor.
I am lucky to have such an outstanding defense committee. Mike McGehee has
welcomed me into his lab and his help and that of his group has been instrumental in the
success of my projects. Moreover, I was able to take Mike’s solar cells course early on in
my graduate career, which gave me a strong grounding in the fundamentals and a broader
picture of the industrial aspects of solar cells. I still remember my first meeting with
Mike, where I was so nervous I read from pre-written questions from my notebook! I
had the fortune of seeking Chris Chidsey’s input on my research projects early on.
Chris’s questions cut down to the core of your project and get to what you are really
trying to learn and accomplish. I have benefited from his advice to inter-relate my
various hypotheses to create a full assessment of the system, and in mathematical terms if
possible, so that contradictions or issues of scale will be clear. I have only had the
pleasure of interacting with Fritz Prinz in the last two years of my time here, though I
have worked with his group closely throughout. It has been very fun to see another ALD
lab in action, and to learn tips and techniques on everything from ALD to TEM. I am
grateful for the interest he has taken in my projects, in providing career advice, and
serving as a member of my reading committee. Alberto Salleo has always brightened up
the materials science department, and I had an early interest in his research. Moreover,
his thermodynamics course was one of my favorite (and hardest!) core courses, and I
think I particularly benefitted both from his theoretical rigor and practical, everyday
examples. Moreover, I have had learned a lot from members of his group, on
Katherine Roelofs Ph.D. Dissertation
ix
synchrotron methods and characterizing the electrical properties of organic materials. It
was a true pleasure that Alberto agreed to be a member of my defense committee, and I
would like to thank him.
I would like to thank Tom Brennan, who mentored me closely in the Bent group
for the first three years of my PhD and completed several projects with me. Tom is both
a strategic researcher and a deep thinker; I remember our brainstorming sessions very
fondly. As a side-note, I have yet to come across a better file organization system than
Tom’s. I developed my writing style for scientific papers from reading Tom’s previous
work and seeing his editing process in our co-written papers. On a more technical side,
Tom modified the transient photovoltage and photocurrent systems, as well as passed on
Matlab and Python scripts which were a great help in getting me started on mass-
analyzing my data.
Tom helped begin an incredibly fruitful collaboration with Orlando Trejo and
John Xu from Fritz Prinz’s research group, which has produced great results over the past
years. I got to spend a lot of time with Orlando on campus and down and the synchrotron
and I greatly respect his thorough approach to studies. For example, in our last project
analyzing XANES data, from the raw spectra at least five different levels of analysis
were required, each with a different software to correctly calibrate, normalize, analyze,
and pair with simulation results. Through this Orlando jumped in to learn each new
software, and in the end we were able to really dig into the physical meaning behind the
results.
I am grateful to all the previous students in the Bent group; I would like to name
them here with a few details that come to mind. From Han-Bo-Ram Lee I learned to
make beautiful posters, presentations, and figures, and how to design robust experiments.
From Jon Bakke, I learned how to write personal statements and represent myself well
scientifically. From Marja Mullings I learned many tips about professional work
environment and presentation, and even ergonomics. Han Zhou was an inspiration for
making dedicated and consistent progress in experiments. Speaking of persistence, I
don’t think I could have seen a better example of working through experimental
equipment troubleshooting than Nid Methaapanon’s year-long struggle with the in-situ
Katherine Roelofs Ph.D. Dissertation
x
FTIR set up. From Keith Wong and Bonggeun Shong, I have learned a lot about the
simulation side of surface science. I have had the benefit of a series of particularly
insightful and helpful post-docs coming through the Bent group, namely Carl Hagglund,
Han-Bo-Ram Lee, Scott Geyer, Jukka Tanskanen, Chaiya Prasittichai, Adam Hultqvist,
Pralay Santra, Woohee Kim, and Adrie Mackus. All have helped sharpen my approach
to putting together a scientific papers and presentations, in addition to expanding my
knowledge of different experimental techniques.
From the students in the group close to my year and after, I have had a lot of fun
with in the lab and out of the lab. From the first day I joined the group, Katie Pickrahn
(“Katie P.” to my “Katie R.”) has been a great friend and colleague who is remarkable for
her willingness to lend her time and help to anyone in need, and it was a great joy to see
her become Katie Nardi last summer. Steve Herron has been my go-to for chemistry
knowledge/safety, as well as mental/physical fitness! The early duo of Art
Wangperawong’s expertise in electrical engineering and Steve’s in chemistry was very
fun to witness. I have greatly enjoyed having Yesheng Yee move through the program
with me; Yesheng seems perennially upbeat and engrossed in the science at hand,
offering great suggestions of experiments and simulations that could complement my
projects. I am deeply impressed by Nuoya Yang building her catalysis project from
scratch, both identifying an approachable system of interest and collecting a network of
equipment to fulfill her experimental needs. Fatemeh Hashemi has shown that it is
possible to make rapid experimental progress and achieve significant results through hard
work in the lab; she seems well on her way to finish a whole doctorate in just four years!
Axel Palmstrom has proved particularly helpful in thinking through deep concepts on the
operating principals of quantum dot solar cells, and incredibly capable in working as a
counter-point on several of my projects these last few months while pursuing his own as
well. Just when I thought I had finalized my presentation style, David Bergsman’s
examples in our group meetings showed there is always another level up in explaining
complicated concepts in beautiful clean crisp slides. The incoming students have made
the group even more social and enjoyable work environment. Tania Sandoval, Callisto
MacIsaac, and Takero Sone have brought a lot of joy to the lab, perhaps to even out
Richard Closser’s level-headed cynicism. Tania has been a great running partner, a true
Katherine Roelofs Ph.D. Dissertation
xi
force to be reckoned with on the trail or off! Joseph Singh has been a great desk-mate,
and it has been fun working on the NiOx ALD process with Jon Baker. It was a great
pleasure to work with Dara Bobb-Semple last fall, and have her join our group in the
spring; I can already see that she will be a great mentor in turn.
The Stanford undergraduate students and bay area K-12 teachers who I have
mentored over the course of my graduate career have enriched my experience of science
and my appreciation for all the state-of-the-art tools Stanford offers. These include Juan
Domingo, Janny Cahatol, Susan Cooper, Troy Yang, and Pan Chuen. In addition their
subprojects have greatly advanced and grounded my graduate research, often
systematically exploring an area that I would never get to do on my own.
I am also grateful to all the people in the McGehee group, who have helped me to
excel at solar cell fabrication and analysis. Colin Bailie has been a great source of help,
and I am forever impressed by his organizational methods to ensure clean, well-stocked,
well-placed lab materials and tools. I can’t count the number of times I have phoned him
at odd hours, asking where this or that is, and what exact temperature should be used. He
has been a helpful friend throughout, and a great sounding board for ideas. I am grateful
to I-Kang Ding, who first trained Colin and me on solar cell fabrication, and who
possessed legendarily steady hands. William Nguyen has also been a key contact, and a
great help to the whole team with his chemistry background. In terms of electrical
analysis, I am indebted to George Margulis, George Burkhard, Eric Hoke, and Tommaso
Giovenzana for help with the more complicated techniques, equipment, and Labview
programming. I have had the pleasure to count Toby Sachs-Quintana as a friend inside
and outside the lab. The other members of my cohort in the McGehee lab, Billy Mateker
and Jon Bartelt have been a welcoming presence. More recently for the perovskite
project, I have gotten to work with Rachel Beal, Becky Belisle, Andrea Bowring, Kevin
Bush, and Dan Slotcavage, who have gone above and beyond in bringing me up to speed
on perovskite solar cell fabrication. Sitting in on a few of the perovskite team meetings
has been quite special, with the fast flow of new concepts and ideas I have learned more
in one hour than I would from any amount of literature reading. Finally, Roong
Katherine Roelofs Ph.D. Dissertation
xii
Cheacharoen has been a great friend this past year, and it has been great fun to record a
duet together just this summer.
Finally, I would like to mention my friends at Stanford, who have been a constant
source of support throughout my time here. There are the early and lasting friendships I
made in Rains Housing: my roommate Leigh Harris, the bio dinner circle of Sarah
Carden, Emily Garbinsky, and Sharon Briggs, and the engineering
(wecandowhaterverwewant ©) dinner circle of Yashodhan Gogte, Mike Sprague, Shruti
Gupta, Julia Ling, Joe Funke, Xuefeng Chen, Alex Neckar, and Megan Tsai. My
materials science cohort, forged from the fire of our core classes, has been a great group
of talented, interesting people who have been fun to interact with inside and outside the
classroom. My quals study buddies of Vrinda Thareja and Krysta Biniek have been my
close friends throughout, and I am lucky to have found them.
Katherine Roelofs Ph.D. Dissertation
xiii
Table of Contents
Chapter 1. Introduction ....................................................................................................... 1
1.1. Main Goals and Organization of Dissertation ...................................................... 1
1.2. A Case for Solar Photovoltaics ............................................................................ 3
1.3. Importance of Nanostructured Solar Cells ........................................................... 5
1.4. Solid-State Devices .............................................................................................. 8
1.5. Solar Cell Operating Principles ............................................................................ 9
1.6. ALD for Interface Engineering .......................................................................... 11
1.7. Financial Support, Collaborations, and Copyrights ........................................... 14
1.8. References .......................................................................................................... 14
Chapter 2. Interface Engineering Strategies ..................................................................... 17
2.1. QDSSCs, ETASCs, and CQDSCs...................................................................... 17
2.2. Device Architectures and Efficiencies ............................................................... 19
2.3. Interfacial Charge Transfer Processes ................................................................ 23
2.4. Band Alignment by Molecular Dipole Layer ..................................................... 27
2.5. Improved CQD Film Mobility by Ligand Exchange ......................................... 31
2.6. Reduced Recombination by Interfacial Inorganic Layers .................................. 36
2.7. Outlook ............................................................................................................... 41
2.8. Financial Support, Collaborations, and Copyrights ........................................... 44
2.9. References .......................................................................................................... 45
Chapter 3. Experimental Methods .................................................................................... 50
3.1. Solar Cell Fabrication......................................................................................... 50
3.1.1. Substrate Preparation .................................................................................. 50
3.1.2. QD Deposition ............................................................................................ 55
3.1.3. Perovskite Deposition ................................................................................. 57
3.1.4. Hole-Transport Material and Metal Contact Deposition ............................ 58
3.1.5. Atomic Layer Deposition of TiO2 and NiOx............................................... 59
3.2. Materials Characterization ................................................................................. 61
3.2.1. Spectroscopy, Diffraction, and Microscopy ............................................... 62
3.2.2. Synchrotron Methods .................................................................................. 67
Katherine Roelofs Ph.D. Dissertation
xiv
3.3. Electrical Characterization Techniques .............................................................. 70
3.3.1. Current-Voltage Curves and Quantum Efficiency ...................................... 70
3.4. Financial Support, Collaborations, and Copyrights ........................................... 76
3.5. References .......................................................................................................... 77
Chapter 4. Al2O3 Recombination Barrier Layers by ALD in Solid-State CdS QDSSCs . 80
4.1. Introduction ........................................................................................................ 80
4.2. Experimental Section ......................................................................................... 84
4.3. Results and Discussion ....................................................................................... 87
4.3.1. TiO2/Al2O3/QD Device Performance ......................................................... 87
4.3.2. TiO2/QD/Al2O3 Device Performance ......................................................... 89
4.3.3. Comparison of Layer Placement and Role of Al2O3 Layer ........................ 91
4.3.4. Effect of Al2O3 on Electron Lifetimes ........................................................ 97
4.4. Conclusions ...................................................................................................... 100
4.5. Financial Support, Collaborations, and Copyrights ......................................... 101
4.6. References ........................................................................................................ 101
Chapter 5. PbS QD Synthesis by ALD ........................................................................... 104
5.1. Introduction ...................................................................................................... 104
5.2. Experimental Details ........................................................................................ 107
5.3. Results and Discussion ..................................................................................... 110
5.3.1. PbS QD growth by ALD and SILAR ....................................................... 110
5.3.2. SILAR studies to increase PbS nucleation................................................ 115
5.3.3. Device Performance .................................................................................. 119
5.4. Quantifying Geometric Strain at the ALD PbS QD/ TiO2 Interface ................ 127
5.5. Conclusions ...................................................................................................... 135
5.6. Financial Support, Collaborations, and Copyrights ......................................... 137
5.7. References ........................................................................................................ 138
Chapter 6. Increased QD Loading by pH Control Reduces Interfacial Recombination in
QDSSCs .......................................................................................................................... 141
6.1. Introduction ...................................................................................................... 141
6.2. Experimental Methods ..................................................................................... 146
6.3. Results and Discussion ..................................................................................... 149
Katherine Roelofs Ph.D. Dissertation
xv
6.3.1. QD Characterization ................................................................................. 149
6.3.2. Device Performance .................................................................................. 158
6.3.3. Interfacial Recombination ......................................................................... 164
6.4. Conclusions ...................................................................................................... 173
6.5. Derivation: Modeling Recombination Lifetime ............................................... 175
6.6. Financial Support, Collaborations, and Copyrights ......................................... 177
6.7. References ........................................................................................................ 178
Chapter 7. ALD Transport Layers in Perovskite Solar Cells.......................................... 183
7.1. Experimental Methods ..................................................................................... 184
7.2. Conformal TiO2 Blocking Layers by ALD ...................................................... 187
7.3. Results and Discussion - TiO2 Blocking Layers .............................................. 190
7.4. Wide Band Gap Perovskite Solar Cells............................................................ 200
7.5. ALD Encapsulation Layers .............................................................................. 206
7.6. Conclusions ...................................................................................................... 209
7.7. Financial Support, Collaborations, and Copyrights ......................................... 210
7.8. References ........................................................................................................ 211
Chapter 8. Conclusion and Outlook ................................................................................ 215
8.1. Summary of Work and Future Studies ............................................................. 215
8.2. Industrial Applications ..................................................................................... 218
8.3. Future Directions and Outlook ......................................................................... 219
8.4. References ........................................................................................................ 221
Appendix A. Quantifying Geometric Strain at the ALD PbS QD/ TiO2 Interface ..... 223
Appendix B. Details of NiOx ALD as a hole-transport layer in PSCs ....................... 236
Katherine Roelofs Ph.D. Dissertation
xvi
List of Tables
Table 2-1. List of open-circuit voltage (VOC), short-circuit current density (JSC), fill factor
(FF), and efficiency (η) for top devices of various architectures as of December, 2012.§ 21
Table 3-1. Electrical and material properties of common transparent conductive oxides.1
........................................................................................................................................... 51
Table 4-1. Device Parameters for TiO2/Al2O3/QD and TiO2/QD/Al2O3 Configurations.†
........................................................................................................................................... 91
Table 5-1. Average atomic concentrations measured using TEM-EDS. The atomic shells
used for each element are denoted inside parentheses. ................................................... 112
Table 5-2. Average device parameters for the five configurations tested. A substantial
improvement in efficiency was achieved by performing a single ALD cycle of Al2O3
prior to PbS growth whereas no improvement was observed with a single Al2O3 ALD
cycle after PbS growth. The efficiency improvement is driven by a large increase in JSC
and a smaller increase in VOC. Values were obtained by taking the average of the top two
of four devices (~10 mm2
in area) on a given substrate and then averaging over the four
substrates tested with each configuration. ...................................................................... 122
Table 6-1. Characterization of PbS QDs deposited under varying SILAR deposition
conditions. Listed are the QD band gap extracted from Tauc analysis of the UV-Vis
spectra, and the QD diameter expected from the Tauc band gaps, using a literature
calibration of PbS QD band gap and size.81
Also listed are the average QD diameter and
percent of the TiO2 surface area covered by QDs obtained from TEM, and the valence
band position expected from the TEM diameter, using a literature calibration of PbS QD
size and valence band position.82
Six TEM were analyzed for each sample,
corresponding to roughly 0.03 μm2 of TiO2 surface area and 300 QDs; error bars show
the standard deviation. .................................................................................................... 153
Table 6-2. Values of average ss-QDSSC metrics as plotted in Figure 6-8, for varying QD
deposition conditions. Device metrics were averaged from 10 devices (the two best cells
from five different substrates), and error bars represent standard deviations. ................ 160
Table 6-3. JSC values determined by integrating the EQE spectra from champion-
efficiency ss-QDSSC devices, plotted in Figure 6, along with JSC values of the same
devices as determined from the J-V curves..................................................................... 163
Table 7-1. Series resistance of the different compact TiO2 layers, measured by impedance
spectroscopy of full perovskite solar cells under dark conditions. These values are the
average series resistance across the different voltage setpoints. ..................................... 198
Katherine Roelofs Ph.D. Dissertation
xvii
List of Figures
Figure 1-1. Map showing a three-year average global solar irradiance in W/m3, including
nights and cloud coverage. Sunlight hitting the dark circles could, cumulatively, produce
18 TW of power if solar cells with a conversion efficiency of only 8% were installed.
Reprinted with permission from Matthias Loster.6 ............................................................. 4
Figure 1-2. Projected convergence of the cost of electricity produced by photovoltaics
(bounded by a high-cost curve, blue, and a low-cost curve, purple), and the rising cost of
electricity produced by conventional methods, which determine grid prices in the United
States. Reprinted with permission from Wolden, C. A., et al., Journal of Vacuum Science
& Technology A 2011, 29, 030801.7 ........................................................................... 5
Figure 1-3. Overview of wafer-based photovoltaic and thin film photovoltaic
technologies. Wafer-based technologies include indirect band gap absorbers, such as Si
and Ge, that require thick films to absorb all the incident light. Thin film technologies,
which exclusively deal with direct band gap absorbers, are designated as commercial or
emerging. The emerging thin films technologies of quantum dot solar cells and
perovskite solar cells are the focus of this work. Reproduced from Ref. 8 with
permission of The Royal Society of Chemistry. ................................................................. 7
Figure 1-4. Operation of a quantum-dot-sensitized solar cell (QDSSC). A photon
incident on the device is absorbed by the QD, exciting an electron and creating an
exciton. The electron is transferred to the electron-transport material, TiO2, and the hole
is transferred to the hole-transport material (HTM). The charges then travel to their
respective contacts. ........................................................................................................... 11
Figure 1-5. Schematic of atomic layer deposition (ALD). A substrate is placed under
vacuum, and a single ALD cycle is as follows. (1) The metalorganic reactant is pulsed
into the chamber, followed by (2) purging of the chamber with N2 purge gas; (3) pulsing
in the counter-reactant, such as H2O, O3, or H2S; and (4) purging with N2 to remove the
leaving ligands. (5) This process is repeated for several ALD cycles to build up a thin
film. Reproduced from Ref. 35 with permission from The Royal Society of Chemistry. 12
Figure 1-6. ALD can produce films or nanoparticles depending on deposition conditions
and specific film material and substrate material combination. (a) Conformal metal oxide
film growth on metal oxide substrate, for ALD In2O3 on TiO2 naoparticles. Adapted with
permission from Brennan, T. P., et al., The Journal of Physical Chemistry C 2013, 117
(46), 24138-24149. Copyright 2013 American Chemical Society.41
(b) Metal
chalcogenide nanocrystal growth on metal oxide substrate, for ALD PbS on SiO2
nanowires. Adapted with permission from Dasgupta, N. P., et al., Nano Letters 2011, 11
(3), 934-940. Copyright 2011 American Chemical Society.42
(c) ALD TiO2 film
Katherine Roelofs Ph.D. Dissertation
xviii
morphology and crystallinity depends on deposition conditions. Reproduced from Ref.
43 with permission of The Royal Society of Chemistry. .................................................. 13
Figure 2-1. Stylized schematic of device architectures of (a) quantum dot-sensitized solar
cells (QDSSCs), (b) extremely thin absorber solar cells (ETASCs), and (c) colloidal
quantum dot solar cells (CQDSCs). For each, the electron-transporting anode is shown in
dark blue, the absorber material in yellow, and the hole-transport material (HTM) in light
blue. While a sintered network of dark blue spherical nanoparticles is shown as the
anode in the QDSSCs and ETASCs above, other nanostructured photonanodes such as
nanorods or nano-pyramids have also been used in these devices. .................................. 19
Figure 2-2. Schematic of band energy levels in a QDSSC, ETASCs, and some CQDSCs.
Charge transfer steps involved in photocurrent collection are depicted by solid black
arrows for electron transfer and a hollow black arrow for hole transfer; recombination
processes are depicted by red arrows. Energetic overpotentials for electron injection and
hole transfer (regeneration of the neutral absorber species) are also shown. ................... 23
Figure 2-3. Schematic of band shifting in a quantum dot (QD) due to functionalization
with a surface dipole layer. Surface dipoles have been investigated explicitly in QDSSCs
via molecular dipole layers of varying strengths, although it is also possible for inorganic
layers to create a surface dipole layer. The band energy levels of the QD prior to the
addition of the surface dipole layer are shown in black: the conduction band energy level,
E°CB, and valence band energy level E°VB. The position of the bands after the application
of the surface dipole layer are indicated by EʹCB and EʹVB. ............................................... 28
Figure 2-4. Control of injection in QDSSCs by molecular dipole layers. (a) Incident
photon-to-current efficiency (IPCE) measurements of QDSSCs with benzenethiol
derivatives of varying dipole moment strengths. The ordered series of aromatic
functional groups (NO2, F, CH3, OCH3) were calculated to have increasingly negative
dipole moment values (D). (b) Photovoltage spectroscopy (PVS) was employed to
measure the electron injection from the CdSe QD to TiO2. (c) The shift in PV onset to
lower energetic values (higher wavelengths) was found to correspond with increasingly
negative dipole moments. This shift in PV onset to lower energies could be caused by
increased electron injection from QDs that, prior to molecular dipole treatment, had
previously had too low a CB for injection to occur efficiently into the TiO2. Adapted
with permission from Ref. 62. Copyright 2010 American Chemical Society. ............... 30
Figure 2-5. Schematic of colloidal quantum dot (CQD) film fabrication. During
colloidal synthesis of the quantum dots, long alkyl ligands are used to suspend the QDs in
the organic solvent. After the synthesis of the CQDs, the longer ligands are then
exchanged for shorter, often conjugated ligands, for better mobility properties in the final
CQD film. Due to the multi-layer processes used to grow CQD films of the desired film
thickness, ligand exchange is often conducted after CQD deposition on a substrate.
Ligand exchange at this stage also provides the change in CQD solubility required to
Katherine Roelofs Ph.D. Dissertation
xix
prevent dissolution of the CQDs back into the original solution. The exchange of longer
ligands for shorter ones results in a densification of the CQD film.................................. 32
Figure 2-6. Charge mobilities in PbSe CQD films were found to exponentially decay
with increased ligand length. The ligands studied were 1,2-ethanedithiol (EDT); 1,3-
propanedithiol (PDT, >98%); 1,4-butanedithiol (BuDT); 1,5-pentanedithiol (PenDT);
1,6-hexanedithiol (HDT). The electron and hole mobilities were measured in ambipolar
field-effect transistors (6.1 nm NCs). Reprinted with permission from Ref. 67.
Copyright 2010 American Chemical Society. .................................................................. 33
Figure 2-7. CQDSCs with ligand exchange down to a single-atom ligand, studying
ethanedithiol (EDT), 1,6-hexane dithiol (HDT), and mercaptocarboxylic acid (MPA), and
the single-atom ligand of Br. (a) Time-resolved infrared (TRIR) spectroscopy
measurements of CQD films; peaks in TRIR spectra correspond to trap-to-band
transitions. In the Brˉ-capped CQD films, the TRIR peak occurs at a lower energy level,
indicating that Brˉ-capped films have shallower trap states. (b) Decay times extracted
from the absorption decay curves of the TRIR peak, tracking the rate at which trap states
are depopulated. (c) Higher CQD film mobilities were found in CQD films with lower
trap-to-band transition energies. Reprinted with permission from Macmillan Publishers
Ltd: Nature Materials (Ref. 71), copyright 2011. ........................................................... 34
Figure 2-8. Interfacial band structure depicting the potential impacts of inorganic surface
treatments in QDSSCs. Here, a nanocrystal of the nanoporous anode (i.e., a TiO2
nanocrystal) is depicted in navy, the inorganic coating layer in magenta, and the light-
absorbing quantum dot (QD) in yellow. The hole-transport material (HTM) is shown in
light blue. (a) The interfacial band structure prior to deposition of the metal oxide layer.
Only recombination pathways from the photoanode to the QD are shown, from (α) the
anode conduction band and (β) the anode density of states (DOS), although
recombination can also occur with the HTM. (b) Recombination barrier layers will
reduce both (α) and (β) recombination processes due to the tunneling effect. (c)
Recombination barriers may also alter the DOS of the anode, by passivating trap states,
leading to further reductions in the (β) recombination path. (d) The recombination barrier
layer may also act as a surface dipole, shifting the anode conduction band upward. (e)
Due to the difficulties of conformally growing Ångstrom-thick inorganic layers on a
nanoporous anode, it is possible that multiple geometric configurations of the
anode/inorganic coating/QD exist within the same device. .............................................. 37
Figure 2-9. The role of TiCl4 treatment in QDSSCs (n-TiO2/CdS QDs/electrolyte). (a) J-
V curves as a function of TiO2 CBD time; inset shows dark J-V curves. The highest
efficiency is achieved at 30 min. (b) Open circuit voltage decay curves under AM 1.5
illumination, show that the highest electron lifetimes (slowest rate of VOC decay) are
achieved for the 30 min samples. (c) Schematic of the authors’ proposed mechanism.
With increasing TiO2 CBD times, the resulting TiO2 coating is believed to decrease
Katherine Roelofs Ph.D. Dissertation
xx
interfacial recombination, an effect which dominates at the 30 min CBD time point. For
thicker TiO2 coatings, though, it is believed that the detrimental effects of transport
resistance lead ultimately to the lower observed electron lifetimes at 60 min. Reprinted
from Ref. 81, Copyright 2012, with permission from Elsevier. ...................................... 38
Figure 2-10. (a) Inorganic layers of Al2O3, BaTiO3, and MgO deposited at the interface
in ETASCs. (b) All three metal oxide layers served to improve the VOC, a good
indication of decreased interfacial recombination, though only in the MgO case did JSC
values improve as well. (c) A double layer treatment of BaTiO3 followed by MgO,
provided even better improvements in performance than either treatment alone. The
device parameters of the J-V curves are included, and have the following units: JSC
(mA/cm2), VOC (V), FF (unitless), and η (%). Reprinted from reference 75. .................. 40
Figure 2-11. Inorganic surface modification of CdS QDSSCs by the deposition of ultra-
thin Al2O3 layers by atomic layer deposition (ALD). The Al2O3 layer thickness was
varied, with 0, 1, and 3 ALD cycles of Al2O3. The Al2O3 layers were deposited both after
QD deposition and before QD deposition, resulting in two different configurations: n-
TiO2/QD/Al2O3 (filled markers) and n-TiO2/Al2O3/QD (open markers). (a) Electron
lifetimes were extracted from transient photovoltage measurements, showing longer
carrier lifetimes in the case of the n-TiO2/Al2O3/QD devices. Standard deviations
represent the spread of lifetimes measured across three different batches of devices (each
data point represents ~ 6 devices in total). (b) One possible explanation for the results is
that the n-TiO2/Al2O3/QD devices could block two interfacial charge recombination
processes, whereas the n-TiO2/QD/Al2O3 devices could block only one. Reprinted from
reference 49. ...................................................................................................................... 43
Figure 3-1. (a) Large FTO substrate capable of producing 10 individual solar cell
substrates. (b) Masking with Scotch tape. (c) Zn and HCl etch removes the FTO from
un-masked area in the center. (d) Substrate is cleaned with soapy water and then
solvents. (e) After ozone-cleaning, the entire substrate is coated with a compact TiO2
layer, deposited by spray-pyrolysis or ALD. (f) For nanoporous architectures, a
nanoporous TiO2 layer is deposited. (g) The substrate is finished, and can be divided into
individual cells. (h) Complete device for the case of a quantum-dot-sensitized solar cell,
with a hole-transport material (HTM) and top metal contact, showing the active area set
by the original FTO etch and the metal contact on top. .................................................... 52
Figure 3-2. Spray pyrolysis set-up. (a) TiCl4 solution is a 1:9 volume ratio of titanium
diisopropoxide bis(acetylacetonate) to ethanol. The purple circles show the tip of the
solution-delivery bottle, and where it attaches to the spray-pyrolysis system. (b) The
spray-gun, with a yellow arrow indicating the N2 incoming line, a yellow circle showing
where the line attaches to the spray gun, and a red circle indicating the nob that turns the
N2 flow on or off. (c) Spray-pyrolysis onto a hot plate, using a reflective sleeve to protect
the hand and arm. Images courtesy of Colin Bailie. ........................................................ 53
Katherine Roelofs Ph.D. Dissertation
xxi
Figure 3-3. Transmission electron microscopy (TEM) images of a nanoporous TiO2 film
completed by doctor blading on a Si wafer. (a) Cross-section of the film showing details
of the film structure, with the scale bar indicating 1 μm. (b) Close view of an anatase
TiO2 nanoparticle, with scale bar indicating 5 nm. The nanoparticle lattice spacing is
3.58 Å, which corresponds to the lattice constant a in anatase TiO2. ............................... 55
Figure 3-4. Schematic of successive ion layer adsorption and reaction (SILAR), for the
deposition of CdS as an example. (a) The dip and rinse steps involved in one SILAR
cycle, and (b) the adsorption of the cadmium cation and then sulfur anion during this
process............................................................................................................................... 56
Figure 3-5. CdS QDs and PbS QDs grown with varying number of SILAR cycles on
nanoporous TiO2 films on microscopy slides. The CdS QD picture was taken with back-
lighting to emphasize the change in film color with SILAR cycle, whereas an opaque
background was sufficient to observe the change in PbS QD film coloration with SILAR
cycle. ................................................................................................................................. 56
Figure 3-6. (a) PbI2 films formed by spin-coating the substrate with a 1.3 M PbI2 in DMF
solution. The effect of increasing DMF concentration in the glovebox atmosphere is
shown, with the PbI2 film becoming noticeably rougher, resulting in a more opaque
yellow color, and ultimately brownish edges from scattered light. (b) A CH3NH3PbI3
perovskite film after dipping in MAI/IPA solution, and (c) a similar CH3NH3PbI3
perovskite film held up to a light source. .......................................................................... 58
Figure 3-7. (a) Scanning electron microscope (SEM) cross-sectional image of a
completed QDSSC device, taken in a region where the FTO has been etched away. The
spiro-OMeTAD layer is too thick in this device. (b) SEM image of a device region with
FTO, shown for comparison. Adapted with permission, courtesy of I-Kang Ding.2 ...... 59
Figure 3-8. Details of the compact TiO2 layer growth by TiCl4 and H2O. Film
thicknesses measured by ellipsometry on ozone-cleaned Si substrates. (a) Film thickness
as a function of ALD cycles, with varying stage temperatures. In one study, the samples
deposited at 300 °C were annealed in air at 450 °C for 2 hrs on a high-temperature hot
plate. (b) Corresponding growth rates as a function of ALD cycle, calculated by dividing
the overall film thickness by the number of ALD cycles. ................................................ 60
Figure 3-9. Schematic of the photoelectron effect, showing an incident photon with
energy hv striking an electron in atom, causing the electron to leave the atom with a
kinetic energy, KE. ............................................................................................................ 63
Figure 3-10. Schematic of the Auger effect. (a) An incident electron collides with an
electron in an inner shell, the 1s shell as shown here, ejecting a 1s electron. (b) The hole
in the inner shell is then filled by a higher energy shell electron, from the 2s shell as
shown here. The energy released by the 2s 1s transition is absorbed by an electron in
an outer shell, the 2p shell as shown here, causing that electron to leave. This third
Katherine Roelofs Ph.D. Dissertation
xxii
exiting electron is called the Auger electron. (c) A view of the transitions involved in the
process from an energy level diagram. ............................................................................. 64
Figure 3-11. Schematic of powder X-ray diffraction (XRD). (a) For a sample with
parallel planes of atoms, with a spacing distance dhkl, constructive interference applies
only under the conditions of Bragg’s law (Equation 3). (b) Experimental set-up of
powder XRD, where an X-ray incident on the sample at an angle θ is diffracted. The
scattering vector, q, is also shown. ................................................................................... 66
Figure 3-12. Schematic of X-ray absorption spectroscopy (XAS), showing (a) the
sections of an XAS spectra. (b) In XAS, a X-ray is absorbed by exciting an electron.
The electron can be excited to a higher-energy vacant state, or excited above the vacuum
energy level, with some additional kinetic energy. ........................................................... 68
Figure 3-13. Current-voltage (J-V) curve of a solar cell under illumination (black) and
dark (grey dashed) conditions. The short-circuit current, JSC, is the current density with
no applied potential, the open-circuit voltage, VOC, is the voltage with zero net current,
the max power point, Pmax, is given by the current multiplied by the voltage at the
maximum power point. The fill factor, FF, is the ratio of Pmax divided by the power
value given by JSC multiplied by VOC, or the ratio of the purple square to the blue square.
The series resistance, Rseries, in the inverse of the slope at a forward bias, and the shut
resistance, Rshunt, is the inverse of the slope at reverse bias. ............................................. 70
Figure 3-14. Light-soaking effect in solid-state CdS QDSSCs with spiro-OMeTAD as
the solid-state HTM. In the control cells (black), the light-soaking effect occurs on a
timescale of 100 min, roughly 10 times longer than that in solid-state DSSCs made with
spiro-OMeTAD. (a) Devices made with ALD Al2O3 barrier layers deposited after the
QDs, and (b) devices made with ALD Al2O3 barrier layers deposited before the QDs.
Curves are intended to guide the eye. ............................................................................... 72
Figure 3-15. Schematic of external quantum efficiency (EQE) measurement. A
Tungsten lamp lightsource is monochromated by an SP-150 Monocrhomator. A chopper
wheel is used to chop the monochromated light, after which the light is split. One beam
goes to the device, while the other goes to a Si reference photodiode, tracked by a
current-voltage sourcemeter (Keithley 6517A). Prior to the measurement, a calibration Si
photodiode is placed in the device position, and counts the total number of photons for a
light pulse at a given photon energy. During the EQE measurement, a current-voltage
sourcemeter (Keithley 236) counts the amount of current produced by the device for a
given light pulse. The EQE is the ratio of the current produced by the device over the
number of photons counted by the Si photodiode, as adjusted by the reference diode to
account for any slight variations in the light source intensity during the measurement. If
a bias light is used during the measurement, then the sourcemeter from the device, and
the sourcemeter from the reference photodiode are connected to a lock-in amplifier such
Katherine Roelofs Ph.D. Dissertation
xxiii
that current is only collected from the device when it is seen on the reference photodiode
as well. .............................................................................................................................. 73
Figure 3-16. Schematic of the transient photovoltage and photocurrent setup. A 1 sun
bias light is applied to the device, while a function generator (Agilent 3300) produces
pulsed signal sent to an LED held at roughly 1/20th
the intensity of the bias light. The
function generator signals an oscilloscope (Tektronix) to record voltage or current data at
the end of the pulsed light, which is measured by a current-voltage sourcemeter (Keithley
2400). In addition, the intensity of the bias LED array can be tuned by a programmable
power supply (Circuit Specialists CSI3644A). A LabVIEW software program
coordinates the interaction between these components and receives the data. ................. 75
Figure 3-17. Example of a transient photovoltage measurement. (a) The solar cell is
illuminated at 1 sun bias. A pulsed light provides an additional 0.05 sun illumination,
causing the J-V curve to shift outward. The current is held constant, and when the pulsed
light turns off, the decay in voltage from V2 back to V1 is measured. (b) Voltage-decay
curves at a given current setpoint, where the light pulse is turned off at t = 0 s. The
voltage decay rate changes with experimental parameters, such as in this case the
application of ALD-grown Al2O3 barrier layers of varying thickness. ............................ 75
Figure 4-1. Schematic of barrier layer configurations (not to scale) available in quantum
dot-sensitized solar cells: (a) TiO2/Al2O3/QD and (b) TiO2/QD/Al2O3, resulting
respectively from deposition of the Al2O3 layer before and after the CdS QDs. Spiro-
OMeTAD is employed as the hole-transport material (HTM). Arrows indicate
undesirable recombination pathways; pathways that may be blocked by the Al2O3 barrier
layer are shown by dashed arrows. ................................................................................... 82
Figure 4-2. Current density-voltage curves of representative devices in the
TiO2/Al2O3/QD configuration with increasing ALD cycles of Al2O3 (a) under 1 sun of
illumination and (b) in the dark. Al2O3 barrier layers are effective at suppressing the onset
of dark current, leading to increases in VOC. However, when more than 1 ALD cycle is
performed, the Al2O3 layer hinders electron injection thereby reducing JSC. ................... 88
Figure 4-3. UV-Vis absorption spectra of CdS QDs (which begin absorbing at 510 nm)
deposited on mesoporous TiO2 coated with varying ALD cycles of Al2O3. For the
TiO2/Al2O3/QD configuration (a), the presence of Al2O3 reduces CdS growth, but the
reduction is not sufficient to fully explain the drops in JSC observed. For the
TiO2/QD/Al2O3 configuration (b), no reduction in QD absorption is observed. .............. 89
Figure 4-4. Current density-voltage curves of representative devices in the
TiO2/QD/Al2O3 configuration increasing ALD cycles of Al2O3 (a) under 1 sun of
illumination and (b) in the dark. As with the TiO2/Al2O3/QD configuration, the Al2O3
layer effectively suppresses dark current but reduces JSC if too thick. ............................. 90
Katherine Roelofs Ph.D. Dissertation
xxiv
Figure 4-5. High resolution X-ray photoelectron spectroscopy scans of the Al 2p peak of
Al2O3 deposited by ALD on thick CdS films (grown by SILAR) or on thick TiO2 films
(grown by ALD) as a function of the number of Al2O3 ALD cycles. The data indicate that
Al2O3 readily grows on both TiO2 and CdS surfaces. ...................................................... 91
Figure 4-6. Percent change of device parameters for the TiO2/QD/Al2O3 configuration
averaged over three batches of devices, relative to the average value at 0 ALD cycles of a
given batch. Specifically, within a batch of n devices, the percent change of a parameter
for a given device (pn) was calculated using the value for that device (vn) and the average
value at 0 ALD cycles for devices in that batch:
. The values
of pn were then binned for all three batches and the average and standard deviation (error
bars) of this total set are plotted above. ............................................................................ 92
Figure 4-7. Comparison of device parameters for TiO2/Al2O3/QD and TiO2/QD/Al2O3
configurations under 1 sun of illumination with varying ALD cycles of Al2O3. Parameters
are the average of the top 50% of devices (ranked by efficiency); the set of the top 50%
range in number from 6-8 devices, and error bars indicate standard deviations. The
corresponding values of device parameters are given in Table 4-1. In both configurations,
device efficiency improves after 1 ALD cycle of Al2O3 but drops thereafter due to
decreases in JSC. ................................................................................................................ 94
Figure 4-8. Dark current onset for devices with TiO2/Al2O3/QD and TiO2/QD/Al2O3
configurations, with varying ALD cycles of Al2O3. The quantified dark current onset is
the voltage at which the current of a dark J-V curve reaches a threshold of 0.2 mA/cm2.
Error bars indicate standard deviation. The two configurations show similar suppression
of the dark current at thicker Al2O3 barrier layers. ........................................................... 96
Figure 4-9. Electron lifetimes for the TiO2/Al2O3/QD (open symbols) and
TiO2/QD/Al2O3 (closed symbols) as determined via transient photovoltage measurements
for varying ALD cycles of Al2O3. The larger lifetime improvements achieved with Al2O3
barrier layers in the TiO2/Al2O3/QD configuration are attributed to the fact that in that
configuration, recombination to both spiro-OMeTAD and oxidized quantum dots is
suppressed. ........................................................................................................................ 98
Figure 5-1. Transmission electron micrographs of PbS QDs on nanoporous TiO2 (a) with
(a) and (b,c) without an intermediate Al2O3 cycle. As seen in (a) and (b), the coverage of
the QDs is not affected by Al2O3. (c) Lattice diffraction of the rock salt PbS QDs is
captured with a measured lattice constant of 5.9 Å. ....................................................... 111
Figure 5-2. TEM images of TiO2 nanoparticles coated with PbS QDs grown by 6 SILAR
cycles, taken from the same film. (a), and (b) show PbS QD coverages typical of higher-
coverage regions observed, while (c) shows an example from a low QD-coverage region.
......................................................................................................................................... 113
Katherine Roelofs Ph.D. Dissertation
xxv
Figure 5-3. Auger electron spectroscopy cross-sectional line scans with SEM images of
nanoporous TiO2 films sensitized with PbS QDs. On the linescan position axis, 0 marks
the boundary between the substrate and TiO2 film, and ~2.5 μm marks the surface of the
TiO2 film. The variation in the background coloration (dark on the left and brighter on
the right) is due to variations in the signal collected from the film for the SEM, as one
TiO2 film was deposited on FTO and one on Si. (a) 6 SILAR cycles of PbS on
nanoporous TiO2 on F:SnO2 coated glass. (b) 10 ALD cycles of PbS on nanoporous TiO2
on a Si wafer. These deposition cycles of SILAR and ALD were chosen to give similar
sized QDs. (c) A comparison of the Pb:Ti and S:Ti ratios for the ALD-grown QDs and
SILAR-grown QDs. ........................................................................................................ 114
Figure 5-4. (a) UV-visible spectra of PbS QDs grown with 0.02 M Pb(NO3)2 aqueous
solution and 0.02 M Na2S aqueous solution for 8 SILAR cycles. (a) Dip times in the
Pb(NO3)2 and Na2S solutions, as well as the rinse times in the DI H2O solution were
varied. (b) Tauc Analysis of the UV-vis spectra, with an upper and lower tangent line to
represent the range of band gaps (QD diameters) in each sample. ................................. 116
Figure 5-5. (a) Schematic of approach of capping QDs with a TiO2 layer, with subsequent
SILAR deposition to nucleate new QDs. (b) J-V curves in light of three device
configurations: 3 SILAR cycles of PbS, 3 SILAR cycles of PbS capped with 2 ALD
cycles of TiO2 followed by an additional 3 SILAR cycles of PbS, and 6 SILAR cycles of
PbS. (c) Corresponding J-V dark curves of the devices in (b). ...................................... 118
Figure 5-6. (a) Current-voltage curve of solid-state QDSSCs with SILAR-grown or ALD-
grown PbS QDs. Shown are devices with the number of deposition cycles giving the
highest efficiency for each process. Device metrics are shown in the inset table. (b)
External quantum efficiency (EQE) of same samples, showing a peak EQE of 10%. ... 119
Figure 5-7. Schematics of the configurations studied. (a) Control devices with PbS
quantum dots grown on TiO2 nanoparticles, surrounded by the spiro-OMeTAD hole
transport material (HTM). (b) TiO2 deposited prior to QD growth. (c) Al2O3 deposited
prior to QD growth. In this configuration the TiO2 or Al2O3 barrier layer is expected to
slow recombination involving electrons in TiO2 and holes in the HTM. (d) TiO2
deposited after growth of QDs. (e) Al2O3 deposited after growth of QDs. In this
configuration the TiO2 or Al2O3 barrier layer is expected to slow recombination involving
electrons in TiO2 and holes in both the HTM and oxidized quantum dots. Configuration
(c) yielded the highest device efficiencies. ..................................................................... 120
Figure 5-8. UV-Vis spectra of full devices. No significant changes in light absorption
were observed when a single ALD cycle of TiO2 or Al2O3 preceded PbS deposition,
indicating that neither enhanced PbS nucleation. All spectra exhibit interference fringes
common in thin films. The UV-Vis spectra represent data averaged from multiple
devices............................................................................................................................. 121
Katherine Roelofs Ph.D. Dissertation
xxvi
Figure 5-9. Average device parameters for the five configurations tested, listed in Table
5-2, showing the standard deviations calculated from the best two devices of four
substrates tested, eight devices in total, for each configuration. ..................................... 123
Figure 5-10. (a) Dark current-voltage (J-V) curves for typical devices of the five
configurations studied. Whether deposited before or after the PbS quantum dots, a single
cycle of Al2O3 suppressed the onset of dark current to a similar degree. The value in
parentheses corresponds to a current-threshold voltage, that is, the voltage at which the
dark current in the devices exceeded an arbitrary value (here, 0.4 mA cm-2
). These
current threshold values are averaged over all the devices for that particular
configuration. (b) Illuminated current-voltage curves for typical devices of the five
configurations studied. (c) External quantum efficiency (EQE) spectra of devices in each
of the device configurations collected at zero bias, data is averaged from multiple
devices. When Al2O3 precedes PbS QD growth, a substantial improvement in both EQE
and JSC is observed. ......................................................................................................... 124
Figure 5-11. Electron lifetimes for the five different device configurations as determined
by fitting the photovoltage decay obtained from transient photovoltage experiments. At
higher photovoltages the recombination suppression from depositing Al2O3 prior to PbS
growth is readily apparent and is superior to Al2O3 deposited after PbS, although
recombination is also reduced in that configuration, particularly at lower photovoltages.
Error bars indicate standard deviations of the several devices tested. ............................ 127
Figure 5-12. Schematics of sample architectures. The sample architectures for the AN
(a) and AM (b) sample sets are shown. For both sets, 10, 20, and 40 ALD cycles (10x,
20x, and 40x) of PbS were deposited on the nanoporous TiO2 substrates. The resulting
PbS QDs were capped with 60 cycles of ALD TiO2, which is amorphous. ................... 129
Figure 5-13. Diffraction and TEM characterization of PbS QDs on TiO2 nanoparticles. (a)
XRD measurements show the presence of rocksalt PbS (111) and (200) peaks in the AM
set while there is no evidence of crystallinity other than the substrate for the AN set.
SAED measurements were done on 10x PbS and 40x PbS for the AN (b,c) and AM (d,e)
sets. BF-TEM images of (f) 10x PbS AN and (g) 10x PbS AM show SAED regions with
the embedded PbS QDs, respectively. ............................................................................ 131
Figure 5-14. X-ray absorption characterization of S K and O K edges. The sulfur and
oxygen XANES K-edges are shown for the reference samples (a, c) and the PbS QD
sample sets (b, d). (a) The reference samples for the S K-edge depict four features that are
due to electronic transitions into unoccupied orbitals (peaks A1 and A2) and features
resulting from multiple-scattering of photoelectrons (peaks A3 and A4). (b) Similarly, the
O K-edge can be analyzed from orbital contributions (peaks B1-B4) and multiple-
scattering effects (peak B5). The broadening of the features in amorphous TiO2 are due to
perturbations to Ti and O bonding interaction. The S-K XANES for the AM set look
similar to the references suggesting a rocksalt atomic environment (the features become
Katherine Roelofs Ph.D. Dissertation
xxvii
more pronounced from 10x to 40x PbS because as the QDs grow the number of sulfur
atoms found in the octahedral environment increases), while there are clear distinctions in
the AN set, specifically in the A4 peak (starred). (d) The O-K XANES for the AM and
AN sets have features that confirm that ALD TiO2 on anatase TiO2 and PbS QDs,
respectively, is amorphous. ............................................................................................. 134
Figure 6-1. Example of light-soaking behavior, shown for the control QDSSCs, with 2,
4, 6, and 8 SILAR deposition cycles of PbS QDs. The cell is exposed continuously to
light, and a J-V curve is measured at 10 min intervals, with each measurement taking
roughly 1 minute. Plotted are the resulting parameters (a) short-circuit current, (b) fill
factor, (c) open-circuit voltage, and (d) efficiency. ........................................................ 148
Figure 6-2. UV-Vis absorption of PbS QD-coated nanoporous TiO2 substrates. QDs
were grown for 2, 4, 6, and 8 SILAR cycles, with base-assisted QD growth employing
NaOH and ED. TEA-assisted QD growth is also plotted for comparison. .................... 150
Figure 6-3. UV-vis spectroscopy of nanoporous TiO2 films sensitized by PbS QDs.
Shown are (a) blank TiO2 films and 2 SILAR cycles of QDs grown by the standard
process (solid line) and in the presence of NaOH, ED, and TEA (dashed lines). Also
shown (b) is the trend with 2, 4, 6, and 8 SILAR cycles comparing the standard process
and TEA-assisted QD growth, with the corresponding Tauc analysis curves shown in the
inset. ................................................................................................................................ 151
Figure 6-4. Representative TEM images of PbS QDs grown on TiO2 nanocrystals, at the
same magnification. Shown are 2 SILAR cycles comparing (a) the standard process with
that with (b) NaOH, (c) ED, and (d) TEA. Also shown are higher SILAR cycles by (e),
(g), & (i) the standard process, and (f), (h), & (j) that with TEA. .................................. 154
Figure 6-5. HRTEM images showing rocksalt PbS QDs grown by the standard SILAR
process, on anatase TiO2 nanocrystals. In (a) the lattice spacing along the red line is 3.53
Å, fitting the PbS rocksalt d111=3.44 Å, and the lattice spacing along the yellow line is
3.02 Å, matching PbS rocksalt d200=2.98 Å. (b) A higher-resolution image, with a lattice
spacing of 2.91 Å along the red line, matching PbS rocksalt d200=2.98 Å. .................... 155
Figure 6-6. PbS QD size and loading calculated from TEM images, showing the average
QD diameter and percent of the TiO2 surface area covered by QDs. QDs were deposited
for varying SILAR cycles, and by the standard process and with NaOH, ED, and TEA-
assisted growth. At each SILAR cycle the points are staggered for visibility. Six TEM
images were analyzed for each sample, corresponding to roughly 0.03 μm2 of TiO2
surface area and 300 QDs; error bars show the standard deviation. ............................... 156
Figure 6-7. Auger electron spectroscopy line scan of sample in cross-section for
nanoporous TiO2 films on FTO, sensitized with 6 SILAR cycles of PbS QDs grown by
(a) the standard process, and (b) with TEA. Shown are the atomic concentrations with
the raw intensities at the bottom. .................................................................................... 157
Katherine Roelofs Ph.D. Dissertation
xxviii
Figure 6-8. J-V curves of the highest-efficiency solid state QDSSCs at each condition,
plotting curves collected under 1 sun illumination and in the dark. Shown are devices
grown with (a) 2, (b) 4, (c) 6, and (d) 8 SILAR cycles of PbS QDs with NaOH and ED-
assisted growth. ............................................................................................................... 159
Figure 6-9. Selected J-V curves from Figure 6-8 showing only the control versus TEA-
assisted growth at higher SILAR cycles. ........................................................................ 159
Figure 6-10. Average device parameters—power conversion efficiency (PCE), short-
circuit current (JSC), open circuit voltage (VOC) and fill factor (FF)—for varying QD
SILAR deposition cycles, by the standard process (control), as well as NaOH, ED, and
TEA-assisted growth. At each SILAR cycle, the points are staggered for visibility.
Device parameters were averaged from 10 devices (the two best cells from five different
substrates), and error bars represent standard deviations. Corresponding values are listed
in Table S2. ..................................................................................................................... 161
Figure 6-11. External quantum efficiency (EQE) spectra measured at short circuit of the
highest-efficiency devices with increasing SILAR cycles of PbS QDs compared for the
standard process (solid lines), and that with TEA (dashed lines). .................................. 162
Figure 6-12. Representative transient photovoltage decay curves, shown at the highest
light intensity collected ~1.1 sun, along with the lowest light intensity reported ~0.09 sun.
Plotted are the decay curves for 2 SILAR cycles of the control QDSSCs at (a) ~1.1 sun
and (b) ~0.09 sun, and 2 SILAR cycles with TEA-assisted QD growth at (c) ~1.1 sun and
(d) ~0.09 sun. .................................................................................................................. 165
Figure 6-13. Recombination lifetimes comparing (a) 2 SILAR cycles of QDs grown by
the standard process and in the presence of NaOH, ED, and TEA. (b) Recombination
lifetimes of 2, 4, 6, and 8 SILAR cycles comparing the standard process and TEA-
assisted QD growth. To guide the eye, the purple band highlights the standard process
(control), and the light blue band highlights the TEA data points. ................................. 166
Figure 6-14. Predicted dependence of recombination lifetimes from the valence band
model, equation (4). Experimental recombination lifetimes at VOC of 0.37 V are plotted
against the calculated QD valence band level. Included samples are for 4, 6, and 8
SILAR deposition cycles with base-assisted growth (NaOH, ED, and TEA). These
samples had similar QD coverages of the TiO2 surface (coverage ~ 10%), but varied in
QD size, and thus valence band position. ....................................................................... 171
Figure 6-15. Recombination lifetimes of QDs grown for 6 SILAR cycles by the standard
process (6 cyc), by TEA-assisted QD growth (6 cyc TEA), as compared to QDs grown by
the standard process with a post-QD-growth TEA treatment (6 cyc post-growth TEA).173
Figure 6-16. (a) J-V curves under 1 sun (AM 1.5) illumination and in the dark, and (b)
the EQE spectra for QDs grown for 6 SILAR cycles by the standard process (6 cyc), by
Katherine Roelofs Ph.D. Dissertation
xxix
TEA-assisted QD growth (6 cyc TEA), as compared to QDs grown by the standard
process with a post-growth TEA treatment (6cyc post-growth TEA). ........................... 173
Figure 7-1. (a) GIXRD data of 1700 ALD cycles of TiO2 (roughly 50 nm), initially
deposited at 300 °C and subsequently annealed at 450 °C for 1 hr. Shown are ALD TiO2
films deposited on a Si substrate, producing the anatase phase, and on an FTO substrate,
producing the rutile phase. Also shown are the GIXRD of a Si blank and FTO blank.
Anatase TiO2 peaks are labeled with circles, rutile TiO2 with red squares, and rutile SnO2
with diamonds. (b) GIXRD area pattern of 1700 ALD cycles of TiO2 on FTO, initially
deposited at 300 °C and subsequently annealed at 450 °C for 1 hr scale bar shown
ranging from high (red) to low (blue) intensity. ............................................................. 192
Figure 7-2. In-situ GIXRD taken during annealing of 1700 ALD cycle TiO2 film grown
on FTO at 100 °C. From the initially-amorphous TiO2 film, the (101) rutile TiO2 peak at
Q = 2.5 Å-1
is observed to form during annealing at 500 °C for 2 hrs. Significant heating
occurred during the temperature ramp-down, furthering the crystallization. ................. 194
Figure 7-3. Scanning electron microscopy (SEM) images of TiO2 films at the same
magnification, with scale bar of 200 nm shown. ALD TiO2 grown on Si: (a) 100 cyc at
100 °C, (b) 100 cyc at 300 °C, and (c) 1700 cyc at 300 °C. Also shown are TiO2 films
grown on FTO: (d) 50 nm spray-pyrolysis TiO2 at 450 °C as well as (e) 100 cyc ALD
TiO2 at 300 °C and (f) 1700 cyc ALD TiO2 at 300 °C. Films (e) and (f) were annealed at
450 °C to match device fabrication conditions. 1700 ALD cycles of TiO2 give a ~50 nm
thick film. ........................................................................................................................ 195
Figure 7-4. Performance of nanoporous perovskite solar cells with 50 nm thick compact
TiO2 layers deposited by spray-pyrolysis, compared to those made with varied
thicknesses of ALD TiO2 from 100 ALD cycles (~3 nm thick) to 1700 ALD cycles (~50
nm thick). Shown are the J-V performance under (a) 1 sun illumination and (b) dark
conditions, with an inset enlarging the low-current region. ............................................ 197
Figure 7-5. Series resistance of compact TiO2 layers, measured by impedance
spectroscopy of perovskite solar cells under dark conditions at different voltage setpoints.
......................................................................................................................................... 199
Figure 7-6. Impedance spectroscopy measurements of full perovskite solar cells with
varied compact TiO2 layers taken at a voltage setpoint of 0.7 V in dark conditions. Shown
are (a) an enlarged region near the origin, and (b) the full results. Traces for multiple
devices are shown for each condition, to give an idea of the variability of the
measurement. .................................................................................................................. 199
Figure 7-7. CH3NH3PbI3 perovskite films (left) along with wide band gap
CH3NH3PbIxBr(3-x) perovskite films (right). ................................................................... 200
Katherine Roelofs Ph.D. Dissertation
xxx
Figure 7-8. (a) Conventional n-i-p perovskite solar cell stack, along with (b) the inverted
p-i-n perovskite solar cell stack used in this work. ......................................................... 201
Figure 7-9. Conventional n-i-p perovskite solar cells in the planar architecture, showing
(a) the J-V curve for the best solar cell without and (b) with TiCl4 treatment. ............... 202
Figure 7-10. Statistics of devices fabricated without TiCl4 treatment as compared to
those fabricated with TiCl4, showing parameters determined from the J-V curves under 1
sun illumination, including (a) the VOC, (b) JSC, (c) RShunt, and (d) RSeries. ..................... 202
Figure 7-11. Optimization of standard n-i-p structure with wide band gap perovskite solar
cells fabricated by two-step deposition process. Best efficiency solar cells, with each of
the following modifications (a) two times the concentration of spiro-OMeTAD, with also
(b) two times the concentration of PbI2, and the further addition of (c) a mesoporous TiO2
layer................................................................................................................................. 203
Figure 7-12. Scanning electron microscopy (SEM) images of perovskite (PSK) films
fabricated by conversion of PbI2 layers deposited by different spin speeds. The film
morphology of (a) spin-coating the PbI2 at 2000 rpm followed by a 6000 rpm deposition
of PbI2, while (b) shows spin-coating of 6000 rpm PbI2 followed by a 2000 rpm
deposition where the lower layer was found to be PbI2-rich. ......................................... 204
Figure 7-13. Schematic of proposed mechanism for the effect of spin-coating speed of
PbI2 layers, showing (a) 2000 rpm followed by 6000 rpm producing a fully converted
perovskite film, while (b) 6000 rpm followed by 2000 rpm only converts the top layer of
the PbI2 to perovskite. (c) J-V curves of the completed devices. ................................... 205
Figure 7-14. Initial results showing ALD TiO2 capping layers in the inverted perovskite
solar cell. (a) Structure of the inverted device, (b) schematic of the energy levels in the
device, not to scale, (c) J-V curves collected under illumination collected at 0.5 s delay
between each 0.1 V step, and (d) performance metrics of the devices. .......................... 207
Figure 7-15. Optimization of the inverted wide band gap perovskite solar cells. (a) J-V
performance of devices with 10 nm thick NiOx layers, 3 nm thick TiO2 capping layers,
and varied PCBM:PS weight ratios of polystyrene. (b) J-V performance of devices with
5 nm thick NiOx layers, the optimized PCBM:PS ratio of 1.7 wt% polystyrene, and varied
TiO2 ALD capping layer thicknesses. ............................................................................. 208
Figure 7-16. Schematic of tandem design where an inverted wide band gap perovskite
solar cell can be paired with a Si solar cell to perform photocatalysis. .......................... 209
Katherine Roelofs Ph.D. Dissertation
1
Chapter 1. Introduction
In this chapter, the broad motivation and background concepts for this work are
presented. The importance of emerging thin film photovoltaic technologies is described,
specifically for the inorganic-absorber nanostructured solar cells studied in this work:
quantum-dot-sensitized solar cells (QDSSCs) and perovskite solar cells (PSCs).
Research on QDSSCs with solid-state hole-transport materials is motivated; in PSCs the
standard architectures already employ solid-state hole-transport materials. General solar
cell operating principles are discussed in the context of QDSSCs and PSCs. Further,
atomic layer deposition is introduced as a method to modify interfaces in these devices,
deposit semiconductor quantum dots as the absorber in QDSSCs, and grow thin films to
replace entire layers in perovskite solar cells.
1.1. Main Goals and Organization of Dissertation
The central goal of this work is to develop emerging thin film photovoltaic
technologies and further our understanding of interface effects in these devices. The
specific focus is variants on the dye-sensitized solar cell (DSSC) that employ inorganic
materials as the light absorber, replacing the organic dye molecules used in DSSCs. Such
DSSC-inspired devices are emerging technologies in the broader class of thin film solar
cells, and include quantum-dot sensitized solar cells (QDSSCs) and perovskite solar cells
(PSCs). In QDSSCs, semiconductor nanocrystals (quantum dots) are used as the light
absorber; the onset of absorption in quantum dots is tunable with the nanocrystal size.
The absorber in PSCs is CH3NH3PbI3, or a close variant, which has a perovskite crystal
structure. PSCs first attracted attention in 2012 and have shown an unprecedented rise in
efficiency to current record values of 20.1%. QDSSCs and PSCs can be fabricated
completely from solution processed materials that may be low-purity, contrasting
favorably with the industrial standard, silicon solar cells, which require expensively-
processed, high-purity silicon. This tolerance to defects is partially due to the
nanostructured design of some PSCs and all QDSSCs, in which a nanostructured bulk
heterojunction is formed between the electron-transport material, the absorber, and the
Katherine Roelofs Ph.D. Dissertation
2
hole-transport material (HTM). This tolerance for more defective materials is due to the
splitting of the electron and hole into layers in which the electron is transferred to an n-
type layer and the hole to a p-type layer. However, the high interfacial area involved in
such designs leads to high rates of interfacial recombination, causing losses in
photocurrent, and limiting device efficiency.
This thesis presents methods to reduce interfacial recombination in these
inorganic-absorber nanostructured solar cells though surface modifications. In QDSSCs,
such methods include the growth of ultra-thin insulating metal oxide films by atomic
layer deposition (ALD) at the interface and the control of the QD nucleation and growth
to increase loading and narrow the size distribution. These studies provide insight into
the working mechanisms of QDSSCs, through a combination of material deposition by
ALD (where films can be grown a single atomic layer at a time and an interface can be
atomically engineered), X-ray absorption measurements of interfacial geometric and
electronic structure, and detailed studies of the resulting solar cell performance. Also
described is the use of ALD to grow entire material layers in perovskite solar cells, by
using ALD TiO2 as the electron-transport layer, and/or ALD NiOx as the hole-transport
layer. Despite their high efficiencies, PSCs are unstable and degrade rapidly when
exposed to moisture. The use of ultra-conformal inorganic layers grown by ALD to cap
the perovskite absorber, instead of the currently-employed organic layers, has the
potential to improve the stability, and thus efficiency over time of perovskite solar cells.
This dissertation is organized as follows. Chapter 1 provides the motivation and
background concepts. Chapter 2 gives a detailed review of the field of inorganic-
absorber nanostructured solar cells, paying particular attention to common approaches to
engineer the interfaces within these devices. Chapter 3 proceeds with a description of the
experimental details, including the solar cell fabrication, material characterization
techniques, and electrical characterization techniques. Chapter 4 presents the use of
ultra-thin insulating Al2O3 barrier layers, deposited by ALD, at the TiO2/HTM interface
in QDSSC as a means to reduce interfacial recombination. Chapter 5 compiles several
studies on the nucleation and growth of QDs by ALD, as compared to the solution-based
method of successive ion layer adsorption and reaction (SILAR). Included are studies on
Katherine Roelofs Ph.D. Dissertation
3
the effect of crystalline versus amorphous TiO2 substrates on the growth of PbS QDs,
with the full results of quantifying the geometric strain at the ALD PbS QD/TiO2
interface included in Appendix A. Chapter 6 describes work on increased QD deposition
by base-assisted growth, and studies of how the number of QDs and QD size affect
recombination at the interface between the electron-transport material (TiO2), the QD,
and the hole-transport material. Chapter 7 presents the use of ALD TiO2 as the compact
TiO2 layer in perovskite solar cells, and the impact of crystallinity and conductivity of the
compact TiO2 layer on device performance. Chapter 7 also describes the use of ALD
layers to cap perovskite solar cells, in order to improve device stability. This includes the
use of ALD TiO2 on top of the perovskite as a passivating and/or encapsulating layer to
push wide band gap perovskite solar cells to even higher open-circuit voltage values in
the inverted p-i-n design. For this inverted structure, ALD NiOx was developed as an
inorganic hole-transport material underlying the perovskite layer. ALD NiOx was also
explored as a combined capping layer/hole-transport layer atop the perovskite film; full
details of this study are included in Appendix B. Chapter 8 gives the overall conclusions
and outlook.
1.2. A Case for Solar Photovoltaics
Solar photovoltaics are a leading candidate for terawatt (TW) scale renewable
energy production,1 and the global PV manufacturing capacity is the most rapidly
growing of any renewable energy source.2, 3
Of other renewable energy sources, the total
power available to be harnessed is limited: 2 TW tidal power, 4 TW wind power, 12 TW
geothermal power, 15 TW nuclear power (assuming 1 nuclear plant built per day for 40
years).4 These values pale in comparison to the total solar power availability of 120,000
TW incident on the earth’s surface.4 Even if only a tiny fraction of solar energy is
harvested, it can provide more than enough to meet the projected 2050 global power
demand of 30 TW.4 The drawback of solar power, as compared to traditional power
plants or nuclear power plants, is that solar power is diffuse and intermittent. To
overcome this drawback and harness solar energy, effective storage and transportation are
critical. For a typical latitude in the United States, a solar cell farm with 10% efficient
solar cells would only require an area of 1.6% of the total U.S. land area to serve as the
Katherine Roelofs Ph.D. Dissertation
4
country’s sole energy source;4 for comparison, this is comparable to the total land area
used by the U.S. highway system.5 Globally, land areas shown in Figure 1-1 would be
sufficient to meet the global power demand.6 Compared in another way, a total solar
panel area of ~1012
m2, for 10% net conversion efficiency, is required to supply 30 TW of
energy. If a 30 year panel lifetime can be reached, this implies the annual production of
solar panels needs to reach roughly 5 x 1010
m2/yr, which is on the same order of
magnitude as the global production of all flat glass (6 x 109 m
2/yr).
7
Figure 1-1. Map showing a three-year average global solar irradiance in W/m3,
including nights and cloud coverage. Sunlight hitting the dark circles could,
cumulatively, produce 18 TW of power if solar cells with a conversion efficiency of only
8% were installed. Reprinted with permission from Matthias Loster.6
The photovoltaic (PV) industry has undergone great changes in the past decade.
We have seen an emergence of many new PV technologies, as well as continued
improvements on the existing technologies. Broadly, PV technologies are grouped into
indirect band gap absorber materials, such as Si and GaAs, and direct band gap materials,
such as CdTe, CuInxGa(1-x)S2 (CIGS), and Cu2ZnSnS4 (CZTS).7 Indirect band gap
materials have lower absorption, and thus require thick films (on the order of 100 μm) in
order to capture all of the incident light. Direct band gap materials absorb strongly, and
only need a few microns of material to absorber all the light, and are thus called thin film
Katherine Roelofs Ph.D. Dissertation
5
technologies. This proliferation of PV technologies matches the broad range of
applications for solar cells, from large solar farms competing with traditional power
plants, to residential roofs to offset individual power use, to mobile solar applications.7
As shown in Figure 1-2, the projected continued decreasing cost in PV energy at a utility
scale is projected to cross the rising cost of grid electricity in the U.S. market, shown at
the convergence, the time at which solar cells might reach ‘grid parity’.7
As electricity
from solar cells is likely to reach grid parity within the next 10 to 15 years, this is a
critical and exciting time to further the development of PV technologies.
Figure 1-2. Projected convergence of the cost of electricity produced by photovoltaics
(bounded by a high-cost curve, blue, and a low-cost curve, purple), and the rising cost of
electricity produced by conventional methods, which determine grid prices in the United
States. Reprinted with permission from Wolden, C. A., et al., Journal of Vacuum Science
& Technology A 2011, 29, 030801.7
1.3. Importance of Nanostructured Solar Cells
Nanostructured thin film solar cells inspired by the DSSC design are an
exceptionally promising PV technology. While these devices are still in the fundamental
research and development phase, they provide key potential benefits, such as tunable
visible transparency, high gravimetric power (power produced per gram), and flexible
form factors.8 Further, this class of solar cells has the potential to be fabricated by low-
Katherine Roelofs Ph.D. Dissertation
6
cost, solution-process methods, and to be made with abundant, cheap materials that do
not have the high-purity requirements of silicon solar cells, the current industry standard.9
The current highest-efficiency devices in this category, perovskite solar cells,
have demonstrated 20.1 % efficiencies (not stabilized).10
The climb in these efficiencies,
which has occurred over just the past 3 years, is the most rapid increase of any
photovoltaic technology to date, due in part to the large bank of knowledge these devices
were able to borrow from the DSSC devices.10
This dissertation deals with inorganic-
absorber variants of the DSSC, where the organic dye molecules in the DSSC have been
replaced with inorganic semiconductor absorber materials. Specifically, the scope of the
work includes quantum-dot-sensitized solar cells (QDSSCs),11-14
where the absorber
material is a semiconductor nanocrystal, or quantum dot (QD), and perovskite solar cells
(PSCs),15-17
where the absorber material is a lead-iodide-based perovskite. Both designs
are shown in Figure 1-3 in the context of the scope of photovoltaic technologies. This
move towards inorganic absorbers is part of an overall push to move towards devices
composed of only inorganic materials, because organic materials are less stable under the
elevated temperatures and UV-exposure that operating solar cells experience.13, 18
Despite the promise of inorganic-absorber nanostructured solar cells, a large effort is
required to increase the device stability to achieve 25 year lifetimes and to transition lab-
scale devices into large-scale modules for commercial use. These devices fit in a general
trade-off, wherein solar cells with high-purity, hard materials like silicon are more
difficult to manufacture, but also more stable, whereas low-purity and organic materials
that can be deposited at low temperatures have the benefits of low-cost manufacturing,
but are less stable.7
Katherine Roelofs Ph.D. Dissertation
7
Figure 1-3. Overview of wafer-based photovoltaic and thin film photovoltaic
technologies. Wafer-based technologies include indirect band gap absorbers, such as Si
and Ge, that require thick films to absorb all the incident light. Thin film technologies,
which exclusively deal with direct band gap absorbers, are designated as commercial or
emerging. The emerging thin films technologies of quantum dot solar cells and
perovskite solar cells are the focus of this work. Reproduced from Ref. 8 with
permission of The Royal Society of Chemistry.
While the potential for low-cost module production has obvious benefits, these
devices still need to have high efficiencies in order to be employed as a profitable energy
source, due to the balance of systems costs. That is, low efficiencies cannot be
compensated for in a linear fashion by increasing the total solar cell area, as larger area
devices require additional wiring and mounting hardware. Indeed, there is a minimal
efficiency requirement for it to be profitable to make solar cells at all, due to these
balance of systems costs.19, 20
Further, in all potential applications, from utility-scale
power production, to residential units, to portable solar cells, space is limited. High-
efficiency devices are needed to make use of the available space. These considerations
motivate work to improve the efficiency of these low-cost, emerging PV technologies.
The ultimate industrial applications of DSSC-inspired PV technologies are still being
shaped by rapid developments in fundamental research. There is significant work needed
to increase device stability and thus long-term device efficiencies. Potential industrial
applications and the outlooks for QDSSCs and perovskite solar cells are discussed in the
concluding chapter of this dissertation.
Katherine Roelofs Ph.D. Dissertation
8
1.4. Solid-State Devices
All the quantum-dot-sensitized solar cells and perovskite solar cells in this work
were fabricated with a solid-state hole-transport material (HTM). Solid-state HTMs can
include conductive organic semiconductors, such as spiro-OMeTAD (2,2',7,7'-tetrakis-
(N,N-di-p-methoxyphenylamine)-9,9'-spirobifluorene),21
ionic species, such as CuSCN,
22 or p-type conductive metal oxides, such as NiOx.
23 In perovskite solar cells, a solid-
state HTM is standard, as the perovskite absorber will dissolve in liquid electrolytes.24, 25
In QDSSCs, both solid-state conductive organic molecules and liquid electrolytes are
used. The use of liquid electrolytes presents very real and serious problems for
commercialization: electrolyte leakage, photochemical degradation, and corrosion of the
absorber material. Leakage of the electrolyte is a safety hazard, and, as seen in the
comparable case of liquid electrolyte in disposable batteries, if low-cost sealing is
required, will ultimately lead to short device lifetimes.26
Commercial solar cells for
utility applications should have lifetimes on the order of 25 years.27
Of potential
alternative applications for QDSSCs, incorporation in power-neutral building design
would also require long lifetimes, though it is true that the use of QDSSCs in portable
technologies may not need long lifetimes. However, a prime benefit QDSSCs bring to
portable applications is their ability to be used as flexible solar cells, which makes good
sealing especially problematic. Due to the significant issues liquid electrolytes present
towards commercialization, this thesis focuses on solid-state devices.
The major drawback of solid-state QDSSCs, is that their efficiencies lag behind
those of devices with liquid electrolytes, with current record liquid-electrolyte
efficiencies at 8.6%,28
and solid-state efficiencies at only 1.5%.29
This is also seen in
DSSCs, and is due to low pore-filling of the nanoporous TiO2 substrate by a solid-state
material, and higher rates of recombination of TiO2 electrons to the solid-state HTMs.30-32
While the lower efficiencies of solid-state QDSSCs present challenges to electrical
characterization and batch-to-batch reproducibility, we believe that ultimately progress
will be made on increasing device efficiency. The success of perovskite solar cells,
which employ the same types of solid-state HTMs as QDSSCs, show that the low
QDSSC efficiencies are not due to intrinsic problems with the solid-state HTMs.
Katherine Roelofs Ph.D. Dissertation
9
1.5. Solar Cell Operating Principles
A solar cell operates by absorbing incident photons, exciting an electron from the
valence band of the absorber material to its conduction band, which is at a higher energy
level. The excited electron leaves behind a hole, which has a positive charge, in the
valence band. The excited electron is then collected while still at an elevated energy
level, while the hole travels in the opposite direction to maintain charge neutrality. The
energy difference between the conduction band and the valence band is called the band
gap of the absorber, and sets the upper limit of the voltage the solar cell can supply. The
number of electrons collected is the current. Together, the current multiplied by the
voltage gives the power output of the device.
The general requirements for a solar cell are listed below.
(1) Light absorption. The density of states distribution, i.e. the band structure, of the
absorber material primarily determines the value of the absorption coefficient
across the solar spectrum. Thin film solar cells employ absorbers with direct band
gaps, allowing for strong absorption. For single-junction solar cells, the optimal
theoretical band gap for highest efficiency devices is in the range of 1.0 to 1.2 eV.
Ideally, all photons above the band gap will be absorbed, though this depends on
the absorption coefficient of the absorber material. A higher density of states in
the conduction band should contribute to a higher absorption coefficient. The
thickness of the absorber is generally limited by the carrier diffusion length, and
the highest-efficiency solar cells are based on materials that absorb practically all
incident photons within this thickness limit.
(2) Charge separation and transport. The absorption of a photon creates an exciton –
a bound electron-hole pair. Excitons can diffuse through the absorber to be
ultimately split into the individual electron and hole at a contact that is selective
for one or the other carrier type. The electron and hole have a chance of
recombining in this bound state; often the band structure of the solar cell is
designed to split the exciton, so that charge transport occurs through movement of
the electrons and holes individually. In most thin film solar cells, e.g. CdTe,
Katherine Roelofs Ph.D. Dissertation
10
CIGS and CZTS, charge transport occurs across the absorber material so this
material should be chosen for both its absorption and charge transport properties.
However, in dye-sensitized and quantum-dot-sensitized solar cells, the absorber
material is in an ultra-thin layer sandwiched between a nanostructured bulk
heterojunction of an electron-transport material, ETM, and hole-transport
material, HTM, (see Figure 1-4 below). The exciton is split on the femtosecond
to picoseconds timescale at this heterojunction, with the electron going to the
ETM and the hole transferred to the HTM. Thus, the absorber can be optimized
for its absorption properties, and the ETM and HTM can be optimized for their
transport properties.
(3) Charge collection at elevated energy. The electrons and holes need to be
collected at highly-conductive contacts to enable lateral transport (in the plane of
the substrate) out of the solar cell. Typically, one contact is a transparent
conductive oxide (TCO) situated above the absorber to let light through, and the
other is a metal contact beneath the absorber. The larger the band gap, the higher
the energy at which the charges can be collected, but the fewer photons in the
solar spectrum with that energy or higher. The smaller the band gap, the more
photons from the solar spectrum that can be collected (since all photons with
energy levels above the band gap can potentially be absorbed), but the lower
energy the resulting charges will be collected at. This trade-off is the base of the
Shockley-Quiesser limit which establishes the maximum theoretical efficiency
that can be produced by a single band gap solar cell at a 1 sun illumination.
In quantum-dot sensitized solar cells, power conversion efficiency is largely
determined by the interfacial kinetics that are at the heart of the device operation.
QDSSCs operate as follows (Figure 1-4): an incident photon is absorbed by the
semiconductor QD, exciting an electron. This excited electron is then injected into the
nanostructured anode (typically TiO2), followed by hole transfer to the hole-transport
material (HTM), which can be a liquid electrolyte or a solid-state conductive material.
The electron and hole then travel to their respective contacts, where they are collected as
photocurrent.
Katherine Roelofs Ph.D. Dissertation
11
Figure 1-4. Operation of a quantum-dot-sensitized solar cell (QDSSC). A photon
incident on the device is absorbed by the QD, exciting an electron and creating an
exciton. The electron is transferred to the electron-transport material, TiO2, and the hole
is transferred to the hole-transport material (HTM). The charges then travel to their
respective contacts.
1.6. ALD for Interface Engineering
Atomic layer deposition is a vapor deposition method, capable of depositing metal
oxide films over a large area, conformally coating even high-aspect ratio substrates with
Ångstrom-level control of film thickness.33-35
It is also possible to deposit metal
chalcogenides36
and metals37
by ALD. During ALD a substrate is placed under vacuum
(mTorr pressure levels) on a stage that is typically heated in the range of 150 °C to 350
°C. In a single ALD cycle (Figure 1-5) a metalorganic precursor is pulsed in, followed
by a purging step in which N2 gas is flowed through the reactor, then a counter-reactant
(such as H2O, O3, or H2S) is pulsed in, followed by a final N2 purge. The precursor
chemisorbs or physisorbs on the substrate surface and excess precursor is flushed from
the reactor with the purging step. Then, the counter-reactant reacts with the ligands
attached to the metal atom in the precursor, causing the ligands to leave. In the case of
metal oxide growth, the counter-reactant delivers oxygen atoms to the surface. As these
are surface reactions, and the precursor and counter-reactant are separated temporally,
Katherine Roelofs Ph.D. Dissertation
12
less than a full lattice length is deposited with each ALD cycle, and ALD growth rates are
on the order of 1 Å/cycle. The hallmark of this process is the self-saturation behavior,
which is due to the fact that the precursor will form at most a single monolayer during
each ALD cycle.
Figure 1-5. Schematic of atomic layer deposition (ALD). A substrate is placed under
vacuum, and a single ALD cycle is as follows. (1) The metalorganic reactant is pulsed
into the chamber, followed by (2) purging of the chamber with N2 purge gas; (3) pulsing
in the counter-reactant, such as H2O, O3, or H2S; and (4) purging with N2 to remove the
leaving ligands. (5) This process is repeated for several ALD cycles to build up a thin
film. Reproduced from Ref. 35 with permission from The Royal Society of Chemistry.
The benefits of ALD in photovoltaics are clear: no other thin film deposition
technique can approach the conformality achieved by ALD on high aspect ratio
substrates, and ALD is a leading technique to produce ultra-thin, pin-hole free films.38
A
major drawback of ALD is the need for vacuum; the addition of an ALD process in solar
cell fabrication would be a major break in the mass-throughput printing setup
photovoltaic technologies aspire to. However, researchers are working on atmosphere
ALD processes,39, 40
which are feasible though require greater gas usage and attention to
contamination issues.33
The ALD process is more complicated in reality, which can introduce non-
idealities in some applications, but can be used to advantage in others. Film growth is
dependent on a variety of conditions, including substrate temperature, precursor
temperature, pulse/purge sequence, and the surface reactions of particular precursor
chemistries. For instance, metal oxide ALD on metal oxide substrates typically forms
films; this is used in our work on metal oxide barrier layers, described in Chapter 4. On
Katherine Roelofs Ph.D. Dissertation
13
the other hand, metal chalcogenide growth on metal oxide substrates may form
nanoparticles, used in this work to produce PbS QDs on TiO2 substrates, presented in
Chapter 5. Even for metal oxide growth on metal oxide substrates, the film morphology,
phase, and composition are affected by the specific growth conditions, substrate, and
precursor chemistries. These effects are explored in Chapter 7, where ALD TiO2 films of
up to 50 nm thick are grown on F:SnO2, as the compact layer in perovskite solar cells.
Examples of the interactions between the ALD material and substrate material are shown
in Figure 1-6.
Figure 1-6. ALD can produce films or nanoparticles depending on deposition conditions
and specific film material and substrate material combination. (a) Conformal metal oxide
film growth on metal oxide substrate, for ALD In2O3 on TiO2 naoparticles. Adapted with
permission from Brennan, T. P., et al., The Journal of Physical Chemistry C 2013, 117
(46), 24138-24149. Copyright 2013 American Chemical Society.41
(b) Metal
chalcogenide nanocrystal growth on metal oxide substrate, for ALD PbS on SiO2
nanowires. Adapted with permission from Dasgupta, N. P., et al., Nano Letters 2011, 11
(3), 934-940. Copyright 2011 American Chemical Society.42
(c) ALD TiO2 film
morphology and crystallinity depends on deposition conditions. Reproduced from Ref.
43 with permission of The Royal Society of Chemistry.
Katherine Roelofs Ph.D. Dissertation
14
1.7. Financial Support, Collaborations, and
Copyrights
I would like to thank the financial support that made this work possible: the
Center for Advanced Molecular Photovoltaics (Award No. KUS-C1-015-21), made by
the King Abdullah University of Science and Technology (KAUST); the Center for
Nanostructuring and Efficient Energy Conversion, an Energy Frontier Research Center
(EFRC) funded by the U.S. Department of Energy, Office of Basic Energy Sciences,
Award No. DE-SC0001060; and the Bay Area Photovoltaic Consortium, supported by
the U.S. Department of Energy, Award No. DE-EE0004946. The work that was
supported by each of these funding organizations are specified at the end of each chapter.
This dissertation is the product of work completed with extensive help from others;
collaborators are acknowledged in the end-of-chapter sections for their specific
contributions. Copyright information for studies already published in scientific journals
is noted as well.
1.8. References
1. Luther, J., Motivation for Photovoltaic Application and Development. In
Handbook of Photovoltaic Science and Engineering, Luque, A.; Hegedus, S., Eds. John
Wiley & Sons, Ltd.: New York, NY, 2003.
2. Sener, C.; Fthenakis, V., Renewable and Sustainable Energy Reviews 2014, 32
(0), 854-868.
3. Conti, J. J.; Holtberg, P. D.; Diefenderfer, J. R.; Napolitano, S. A.; Schaal, A. M.;
Turnure, J. T.; Westfall, L. D., Annual Energy Outlook 2015. Department of Energy, U.
S. E. I. A., Ed. Washington, D.C., 2015.
4. Lewis, N. S.; Crabtree, G.; Nozik, A. J.; Wasielewski, M. R.; Alivisatos, P.; Krug,
H., Basic Research Needs for Solar Energy Utilization: Report on the Basic Energy
Sciences Workshop on Solar Energy Utilization. Office of Science, U. S. D. o. E., Ed.
2005.
5. Association, A. R. T. B. Transportation FAQs.
http://www.artba.org/about/transportation-faqs/.
6. Loster, M. Total Primary Energy Supply - From Sunlight.
http://www.ez2c.de/ml/solar_land_area/index.html.
7. Wolden, C. A.; Kurtin, J.; Baxter, J. B.; Repins, I.; Shaheen, S. E.; Torvik, J. T.;
Rockett, A. A.; Fthenakis, V. M.; Aydil, E. S., Journal of Vacuum Science &
Technology A 2011, 29 (3), 030801.
Katherine Roelofs Ph.D. Dissertation
15
8. Jean, J.; Brown, P. R.; Jaffe, R. L.; Buonassisi, T.; Bulovic, V., Energy &
Environmental Science 2015, 8 (4), 1200-1219.
9. Hermes, W.; Waldmann, D.; Agari, M.; Schierle-Arndt, K.; Erk, P., Chemie
Ingenieur Technik 2015, 87 (4), 376-389.
10. NREL, N. R. E. L., Best Research-Cell Efficiencies. Energy, D. o., Ed. Golden,
CO, 2015.
11. Kamat, P. V., The Journal of Physical Chemistry Letters 2013, 4 (6), 908-918.
12. Giménez, S.; Mora-Seró, I.; Macor, L.; Guijarro, N.; Lana-Villarreal, T.; Gómez,
R.; Diguna, L. J.; Shen, Q.; Toyoda, T.; Bisquert, J., Nanotechnology 2009, 20 (29),
295204-295210.
13. Mora-Seró, I.; Bisquert, J., The Journal of Physical Chemistry Letters 2010, 1
(20), 3046-3052.
14. Rühle, S.; Shalom, M.; Zaban, A., ChemPhysChem 2010, 11 (11), 2290-2304.
15. Park, N.-G., The Journal of Physical Chemistry Letters 2012, 4 (15), 2423-2429.
16. Snaith, H. J., The Journal of Physical Chemistry Letters 2013, 4 (21), 3623-3630.
17. Gao, P.; Gratzel, M.; Nazeeruddin, M. K., Energy & Environmental Science 2014,
7 (8), 2448-2463.
18. Briscoe, J.; Dunn, S., Mater. Sci. Technol. 2011, 27 (12), 1741-1756.
19. Wang, X.; Kurdgelashvili, L.; Byrne, J.; Barnett, A., Renewable and Sustainable
Energy Reviews 2011, 15 (9), 4248-4254.
20. Fthenakis, V.; Alsema, E., Progress in Photovoltaics: Research and Applications
2006, 14 (3), 275-280.
21. Lee, H.; Leventis, H. C.; Moon, S.-J.; Chen, P.; Ito, S.; Haque, S. A.; Torres, T.;
Nüesch, F.; Geiger, T.; Zakeeruddin, S. M.; Grätzel, M.; Nazeeruddin, M. K., Advanced
Functional Materials 2009, 19 (17), 2735-2742.
22. Chang, J. A.; Im, S. H.; Lee, Y. H.; Kim, H.-J.; Lim, C.-S.; Heo, J. H.; Seok, S. I.,
Nano Letters 2012, 12 (4), 1863-1867.
23. Kim, J. H.; Liang, P.-W.; Williams, S. T.; Cho, N.; Chueh, C.-C.; Glaz, M. S.;
Ginger, D. S.; Jen, A. K. Y., Advanced Materials 2015, 27 (4), 695-701.
24. Kojima, A.; Teshima, K.; Shirai, Y.; Miyasaka, T., Journal of the American
Chemical Society 2009, 131 (17), 6050-6051.
25. Im, J.-H.; Lee, C.-R.; Lee, J.-W.; Park, S.-W.; Park, N.-G., Nanoscale 2011, 3
(10), 4088-4093.
26. Snaith, H. J., Solid-State Dye-Sensitized Solar Cells with Molecular Hole
Transporters. In Dye-sensitized solar cells, Kalyanasundaram, K., Ed. CRC Press: Boca
Raton, FL, 2010.
27. Green, M. A., J. Mater. Sci.: Mater Electron 2007, 18, S15-S19.
28. Zhao, K.; Pan, Z.; Mora-Seró, I.; Cánovas, E.; Wang, H.; Song, Y.; Gong, X.;
Wang, J.; Bonn, M.; Bisquert, J.; Zhong, X., Journal of the American Chemical Society
2015, 137 (16), 5602-5609.
29. Lee, H.; Leventis, H. C.; Moon, S. J.; Chen, P.; Ito, S.; Haque, S. A.; Torres, T.;
Nuesch, F.; Geiger, T.; Zakeeruddin, S. M.; Grätzel, M.; Nazeeruddin, M. K., Adv. Func.
Mater. 2009, 19, 2735-2742.
30. Bouclé, J.; Ackermann, J., Polymer International 2011, 61 (3), 355-373.
31. Bella, F.; Gerbaldi, C.; Barolo, C.; Gratzel, M., Chemical Society Reviews 2015,
44 (11), 3431-3473.
Katherine Roelofs Ph.D. Dissertation
16
32. Li, B.; Wang, L.; Kang, B.; Wang, P.; Qiu, Y., Solar Energy Materials and Solar
Cells 2006, 90 (5), 549-573.
33. George, S. M., Chemical Reviews 2010, 110 (1), 111-131.
34. Kim, H.; Lee, H.-B.-R.; Maeng, W. J., Thin Solid Films 2009, 517 (8), 2563-
2580.
35. Bakke, J. R.; Pickrahn, K. L.; Brennan, T. P.; Bent, S. F., Nanoscale 2011, 3 (9),
3482-3508.
36. Pore, V.; Hatanpää, T.; Ritala, M.; Leskelä, M., Journal of the American
Chemical Society 2009, 131 (10), 3478-3480.
37. Lim, B. S.; Rahtu, A.; Gordon, R. G., Nat Mater 2003, 2 (11), 749-754.
38. Bakke, J. R. Atomic Layer Deposition of Materials for Applications to
Photovoltaics. Stanford University, Stanford, CA, 2011.
39. Takahashi, N.; Yoshii, N.; Nonobe, S.; Nakamura, T.; Yoshioka, M., Journal of
Electronic Materials 2003, 32 (10), 1107-1110.
40. Takahashi, N.; Nonobe, S.; Nakamura, T., Journal of Solid State Chemistry 2004,
177 (11), 3944-3948.
41. Brennan, T. P.; Tanskanen, J. T.; Roelofs, K. E.; To, J. W. F.; Nguyen, W. H.;
Bakke, J. R.; Ding, I. K.; Hardin, B. E.; Sellinger, A.; McGehee, M. D.; Bent, S. F., The
Journal of Physical Chemistry C 2013, 117 (46), 24138-24149.
42. Dasgupta, N. P.; Jung, H. J.; Trejo, O.; McDowell, M. T.; Hryciw, A.;
Brongersma, M.; Sinclair, R.; Prinz, F. B., Nano Letters 2011, 11 (3), 934-940.
43. Guerra-Nuñez, C.; Zhang, Y.; Li, M.; Chawla, V.; Erni, R.; Michler, J.; Park, H.
G.; Utke, I., Nanoscale 2015, 7 (24), 10622-10633.
Katherine Roelofs Ph.D. Dissertation
17
Chapter 2. Interface Engineering
Strategies
Nanostructured solar cells have the potential to provide a low-cost alternative to
more traditional thin film solar cell technologies. Of particular interest are
nanostructured solar cells with inorganic semiconductor absorbers, due to their favorable
absorption properties. Such devices include quantum-dot-sensitized solar cells
(QDSSCs), extremely thin absorber solar cells (ETASCs), and colloidal quantum dot
solar cells (CQDSCs). However, these device architectures suffer from high rates of
internal recombination and other problems associated with their extensive internal surface
areas. Interfacial surface treatments have proven to be a highly effective means to
improve the electronic properties of these devices, leading to overall gains in efficiencies.
In this chapter, we focus on three types of interfacial modification: band alignment by
molecular dipole layers, improved CQD film mobilities by ligand exchange, and reduced
recombination by interfacial inorganic layers. Select examples in each of these
categories are highlighted to provide a detailed look at the underlying mechanisms. We
believe that surface modification studies in these devices – QDSSCs, ETASCs, and
CQDSCs – are of interest not only to these fields, but also to the broader photovoltaics
community.
2.1. QDSSCs, ETASCs, and CQDSCs
The introduction of high-efficiency dye-sensitized solar cells (DSSCs) has had a
large impact on the field of thin film solar cells, and has motivated further research into
nanostructured solar cells. In 1991, O’Regan and Grätzel introduced the nanoporous
anode architecture to the DSSC field, resulting in a leap in DSSC efficiencies from <1%
to 7.1%, and demonstrating that effective charge transport can occur through a 10 micron
thick film of a nanoporous anode.1 This finding was significant as it helped to resolve the
fundamental conflict faced by planar DSSCs: namely, that a thick layer of dye molecules
Katherine Roelofs Ph.D. Dissertation
18
(~50 nm)2 is required to collect the majority of incident photons, but for effective charge
transfer out of the dye layer, only a monolayer of dye molecules can be employed. With
the nanoporous TiO2 anode, which provides over a thousand-fold enhancement in internal
surface area, DSSCs could now absorb a significant fraction of incident photons while
still maintaining excellent charge transfer properties.
DSSCs fall into the broader field of nanostructured solar cells, that is, solar cells
engineered with submicron dimensions in order to benefit from nanoscale effects.
Nanostructured solar cells have several advantages when compared with first-generation
(Si) and thin film (e.g., CdTe, CIGS, and CZTS) photovoltaic technologies, such as the
potential for solution-processable, low-cost photovoltaic energy generation.3 Since the
introduction of high-efficiency DSSCs in 1991, continued development in the field of
nanostructured solar cells has led to a divergence in design. On the one hand, there are
devices that employ an organic material as the absorber; these include DSSCs, bulk
heterojunction organic photovoltaic (OPV) devices, and hybrid solar cells. On the other
hand, there are devices that employ an inorganic material as the absorber; these include
quantum-dot-sensitized solar cells (QDSSCs), extremely thin absorber solar cells
(ETASCs), and colloidal quantum dot solar cells (CQDSCs).
Until the recent advent of perovskite devices (discussed below),4 DSSCs had far
out-stripped other nanostructured solar cells in device efficiency. DSSCs have held
power conversion efficiencies (PCE’s) of over 10% since 1997,5 with the current record
device at 12.3%.6 However, DSSCs face difficulties in achieving strong broadband
absorption with a single dye species. Inorganic semiconductor absorbers are an attractive
alternative to dye molecules, as semiconductors absorb strongly at all photon energies
above the band gap. Inorganic absorbers also offer the potential for improved device
stability compared to organic absorbers (i.e., dye molecules and polymers), which are
susceptible to degradation when exposed to sunlight. Moreover, inorganic semiconductor
absorbers have comparable absorption coefficients to that of the high-performance dyes
currently used in DSSCs.7 Indeed, the promise of inorganic absorbers has been
demonstrated in just the past year with the introduction of lead-iodide-based perovskite
Katherine Roelofs Ph.D. Dissertation
19
absorbers.4, 8-12
By 2012, perovskite absorber devices have reached PCEs of over 15% in
both nanostructured12
and planar11
architectures.
2.2. Device Architectures and Efficiencies
Schematics of QDSSC, ETASC, and CQDSC architectures, shown in Figure 2-1,
highlight their similarities. We note that QDSSCs and ETASCs fall into the broader
category of semiconductor-sensitized solar cells (SSSCs), where the distinction is that the
absorber in QDSSCs exhibits size quantization. In QDSSCs and ETASCs, a nanoporous
metal oxide is used for the electron-transporting material. Common anode materials
include TiO2, SnO2, or ZnO. These wide band gap metal oxides are transparent to visible
light, in order to minimize parasitic absorption. In QDSSCs, the hole-transport material
(HTM) can either be a liquid electrolyte or a solid-state material. Common solid-state
HTMs include ionic species, such as CuSCN, or organic semiconductors, such as spiro-
OMeTAD (2,2',7,7'-tetrakis-(N,N-di-p-methoxyphenylamine)-9,9'-spirobifluorene).13
In
ETASCs, a solid-state HTM is used, since one motivation for the development of
ETASCs was the creation of an all-solid-state version of the DSSC.14
In CQDSCs, the
colloidal quantum dot (CQD) film can double as both the absorber and the HTM, making
the HTM layer optional. The nanostructured anode is also optional in CQDSC devices;
some CQDSCs employ nanostructured anodes,15
but currently, the highest-efficiency
CQDSCs do not.16
Figure 2-1. Stylized schematic of device architectures of (a) quantum dot-sensitized
solar cells (QDSSCs), (b) extremely thin absorber solar cells (ETASCs), and (c) colloidal
quantum dot solar cells (CQDSCs). For each, the electron-transporting anode is shown in
dark blue, the absorber material in yellow, and the hole-transport material (HTM) in light
blue. While a sintered network of dark blue spherical nanoparticles is shown as the
Katherine Roelofs Ph.D. Dissertation
20
anode in the QDSSCs and ETASCs above, other nanostructured photonanodes such as
nanorods or nano-pyramids have also been used in these devices.
For absorber materials, all three device architectures employ semiconductors;
common semiconductors include metal chalcogenides (such as PbS, PbSe, CdS, and
Sb2S3), which are easy to synthesize by solution methods. QDSSCs and CQDSCs both
use semiconductor nanocrystals, or quantum dots (QDs), as the absorber. Due to their
nanoscale dimensions, QDs have quantum-confined band gaps. That is, a QD’s band gap
depends upon its size; this makes QDs of great interest as absorber materials in solar
cells.7, 17, 18
In some QDSSCs, the QDs are grown in situ by solution methods, such as
successive ion layer adsorption and reaction (SILAR), or by vapor growth methods, such
as atomic layer deposition (ALD). In other QDSSCs, the QDs are synthesized ex situ by
colloidal methods, and then deposited (by electrophoretic methods, ligand-assisted
deposition, or spin-coating/drop-casting) such that the QDs penetrate the nanostructured
substrate.19
In CQDSCs, the QDs are grown ex situ by colloidal methods. Colloidal
synthesis of QDs gives precise control over the QD size distribution. For their absorber,
ETASCs employ thin (under 200 nm) semiconductor layers that conformally coat the
nanoporous anode (although conformal coating is not always achieved in practice). In
ETASCs, the thickness of the absorber is limited by the charge carrier diffusion lengths.
Following the convention of Briscoe and Dunn,14
the term ETASC is used in this chapter
to refer to solid-state nanostructured solar cells in which the inorganic absorber grains are
large enough that quantum-confined behavior is not observed, but the absorber layer is
thin enough to require a nanostructured substrate. ETASCs are also referred to as
inorganic extended-junction devices and inorganic solid-state nanostructured solar cells.2,
20
Despite the promise of inorganic absorber materials, record efficiencies of
QDSSCs, ETASCs, and CQDSCs remain low. To the authors’ best knowledge, the
current record PCEs are as follows: 6.8% in QDSSCs,21
6.3% in ETASCs,22
and 8.5% in
CQDSCs.16
However, we note that further development of these devices is being
pursued by multiple groups, and new record efficiencies are being published at a rapid
rate in the literature. In particular, in the case of QDSSCs, several groups are
Katherine Roelofs Ph.D. Dissertation
21
simultaneously pushing efficiencies past 6% (see Buhbut et al., Albero et al., and
references therein).23, 24
The 6.8% record QDSSC efficiency gives a sense of how rapidly
the field is evolving, considering that the highest-efficiency liquid QDSSCs had
efficiencies of about3% as recently as five years ago.25
Table 2-1. List of open-circuit voltage (VOC), short-circuit current density (JSC), fill
factor (FF), and efficiency (η) for top devices of various architectures as of December,
2012.§
Type Device Architecture:
photoanode/absorber/HTM
JSC (mA/cm
2)
VOC
(V) FF
η (%)
Ref.
QDSSC n-TiO2/CdTe-QD/
polysulfide electrolyte 19.6 0.61 0.57 6.8
21
Perovskite
QDSSC
n-TiO2/(CH3NH3)PbI3-QD/ I-/I
3--electrolyte
15.8 0.71 0.59 6.5* 26
Solid-state QDSSC
n-TiO2/PbS-QD/spiro-OMeTAD 4.58 0.56 0.57 1.5 27
ETASC n-TiO2/Sb2S3/
PCPDTBT & PCBM blend 16.0 0.60 0.66 6.3
22
Perovskite c-TiO2/CH3NH3PbI3-xClx/
spiro-OMeTAD 21.5 1.07 0.67 15.4
11
CQDSC c-TiO2/PbS-CQD-film 22.7 0.62 0.61 8.5 16
DSSC n-TiO2/YD2-o-C8-and-Y123-dyes/ cobalt(II)/(III)-tris(2,2′-bipyridine)-
electrolyte 17.7 0.94 0.74 12.3
6
Solid-state DSSCs
n-TiO2/N719 dye/CsSnI2.95F0.05 15.9 0.72 0.74 8.5 28
§The device architecture is listed for each, denoted by the photoanode material, the
absorber material, and the hole-transport material (HTM). Nanoporous TiO2 has been
abbreviated as n-TiO2, as distinguished from planar compact TiO2 layer (c-TiO2)
employed in the highest-efficiency CQDSC. Device architecture is indicated by the
following notation: anode-material/absorber-material/hole-transport-material, with
Katherine Roelofs Ph.D. Dissertation
22
nanoporous anodes indicated the prefix 'n' and compact, planar anodes are denoted by 'c'.
The organic molecules employed in the HTM phase include: the polymer PCPDTBT
(poly(2,6-(4,4-bis-(2-ethylhexyl)-4H-cyclopenta[2,1-b;3,4-bʹ]dithiophene)-alt-4,7(2,1,3-
benzothiadiazole)), the fullerene derivative PCBM ([6,6]-phenyl-C61-butyric acid methyl
ester), and the small molecule spiro-OMeTAD (2,2',7,7'-tetrakis-(N,N-di-p-
methoxyphenylamine)-9,9'-spirobifluorene). *This device was highly unstable, with a
lifetime of ~10 min.26
Table 2-1 lists the performance metrics for the record-efficiency devices for
QDSSCs, ETASCs, and CQDSCs as of December, 2012. For comparison, the record-
efficiency DSSCs (12.3% in liquid DSSCs6 and 8.5% in solid-state DSSCs
28) are also
listed. Devices employing solid-state HTMs are distinguished from those employing
liquid electrolytes. Solid-state DSSC efficiencies have lagged behind those of liquid
DSSCs for several years now, although solid-state DSSCs may be more commercially
viable, as they avoid problems that arise by using a liquid component, such as electrolyte
leakage or evaporation. While some QDSSC designs have employed polymer or
molecular absorbers in conjunction with the QDs to achieve even higher efficiencies than
those listed in Table 2-1,29
we have chosen to focus on QDSSCs which rely solely upon
the QDs as the light-absorbing material.
Over the past year, the introduction of Pb(I1−χXχ)3(CH3NH3)1 perovskite absorbers
has lead to a great step forward in device efficiencies. For this reason, we have chosen in
Table 2-1 to list the record-efficiency perovskite devices separately from those with
metal-chalcogenide absorbers. Perovskite-absorbers have been reported in ETASC,8
QDSSC,26
and even thin-film planar solar cell11
device architectures. The solar cell with
perovskite QDs achieved a PCE of 6.5%,26
which is on par with the 6.8% PCE record for
a metal-chalcogenide QDSSC.21
For perovskite solar cells, the thin film device (c-
TiO2/CH3NH3PbI3-xClx/spiro-OMeTAD) holds the record, at 15.4%.11
1X is used to indicate a halogen element, such as Cl or Br.
Katherine Roelofs Ph.D. Dissertation
23
2.3. Interfacial Charge Transfer Processes
Due to the extensive internal surface areas in all three device architectures –
QDSSCs, ETASCs, and CQDSCs – interfacial charge transfer processes are of critical
importance for device operation. As with DSSCs,3 interfacial charge transfer rates play
an important role in the collection of photocurrent. Nanoporous, nanocrystalline metal
oxide anodes posses a high density of defects and it is known from the DSSC literature
that TiO2 surface defects can lead to substantial losses in device efficiencies.3, 30-33
Indeed, recent reviews of inorganic-absorber nanostructured solar cells17, 18, 20, 34, 35
have
devoted much discussion to the use of interface modifications to decrease recombination,
as have recent articles focusing specifically on QDSSCs,36, 37
ETASCs,14
or CQDSCs.38,
39
Figure 2-2. Schematic of band energy levels in a QDSSC, ETASCs, and some CQDSCs.
Charge transfer steps involved in photocurrent collection are depicted by solid black
arrows for electron transfer and a hollow black arrow for hole transfer; recombination
processes are depicted by red arrows. Energetic overpotentials for electron injection and
hole transfer (regeneration of the neutral absorber species) are also shown.
Figure 2-2 shows a generalized interfacial band diagram of the
anode/absorber/HTM interface, highlighting different interfacial charge transfer
processes. The interfacial energetic levels shown are the conduction band (CB) of the
anode, the CB and valence band (VB) of the absorber, and the hole-accepting level of the
HTM. This HTM level is either the redox potential of the liquid electrolyte, or the
highest-occupied molecular orbital (HOMO) of a solid-state HTM. CQDSCs deviate
slightly from the band diagram shown in Figure 2-2; in CQDSCs the HTM is optional,
Katherine Roelofs Ph.D. Dissertation
24
and inter-QD charge transfer processes within the CQD film must also be taken into
account.
In Figure 2-2, the charge transfer steps for photocurrent collection are depicted by
black arrows. First, upon absorption of a photon, (1) electron excitation occurs. Electron
excitation is followed by (2) the injection of the electron from the absorber conduction
band (CB) into that of the anode. This electron is ultimately collected at the front
electrode of the device. Next, the absorber is regenerated to a neutral state by (3) the
transfer of a hole from the oxidized absorber to the HTM. Electron injection is proposed
here to occur prior to hole transfer, since in DSSCs, electron injection occurs on
femtosecond timescales and hole transfer on nanosecond timescales.40
The injection
overpotential, the energetic offset between the absorber and the photoanode conduction
bands, is the driving force for injection, just as hole transfer is driven by regeneration
overpotential, the offset of the absorber valence band and the relevant HTM level.
Discussion of recombination in this chapter will focus on the loss pathways depicted by
the red arrows in Figure 2-2: (4) recombination of excited electrons in the absorber CB
with holes in the absorber VB, (5) recombination of injected electrons with the oxidized
absorber layer, (6) recombination of injected electrons with the HTM, and (7)
recombination of electrons in the absorber CB with the HTM. Recombination can also
be mediated by trap states at the surface of the anode or the QD; the further the trap state
lies below the CB, the greater the expected rate of recombination. Shallow trap states in
CQDSCs can also lead to a slow trapping/detrapping electron transport process via
hopping. For further discussion of trap states in nanostructured devices, we point the
reader to the following reports.38, 41-43
Electron injection rates in QDSSCs are known to impact device efficiencies.44
However, there is still debate in the literature as to the order-of-magnitude on which
electron injection or hole transfer occurs in QDSSCs,18, 45
with some reports of electron
injection occurring at timescales as slow as 100’s of nanoseconds.46
This discrepancy is
likely due to the large variety in QDSSC fabrication processes, materials, and device
structures. In an interesting result by O'Mahony et al., it was demonstrated that hole
transfer can still occur in QDSSCs when electron transfer is prevented.47
Katherine Roelofs Ph.D. Dissertation
25
All-solid-state versions of inorganic-absorber nanostructured solar cells are
commercially attractive. Replacing the liquid electrolyte with a solid-state HTM avoids
problems associated with a liquid phase, such as leakage of the electrolyte or corrosion of
the absorber by the electrolyte. The first all-solid-state device was demonstrated in the
DSSC system in 1995, in which a CuI p-type semiconductor was used to replace the
liquid electrolyte of a DSSC.48
However, solid-state devices typically suffer from higher
rates of recombination between the metal oxide and the solid-state HTM (arrow 6, Figure
2-2) than do devices with liquid electrolytes.49
High rates of interfacial recombination are
likely an important factor in the gap of record device efficiencies between solid-state
devices and those with liquid electrolytes seen in Table 2-1.
In particular, high rates of interfacial recombination at the extensive internal
surface area lead to the low internal quantum efficiency (IQE) of nanostructured devices,
as compared to devices fabricated in more mature fields such as Si or thin film solar cells.
IQE is the fraction of absorbed photons that result in the collection of charge carriers at a
given wavelength. To increase IQE values, it is necessary to improve either charge
separation (charges separated/absorbed photons) or charge collection (charges
collected/charges separated). Here, charge separation refers to the separation of the
excited electron and hole by the injection of the excited electron into the metal oxide
electrode (arrow 2 in Figure 2-2) and hole transfer to the HTM (arrow 3 in Figure 2-2).
Charge collection processes (not shown in Figure 2-2) refer to the transport of the
electron through the anode to the front contact and the transport of the hole through the
HTM to the back contact. Interface engineering can increase IQE by decreasing losses
from recombination, increasing exciton-splitting efficiency, or improving charge
transport via increased charge carrier mobilities and lifetimes.
To achieve high IQE values, the energetic overpotentials for desired charge
transfer steps should be considered. For instance, a higher absorber CB relative to the
anode CB will lead to a higher driving force for electron injection, shown in Figure 2-2.42,
50 Likewise, a lower absorber VB relative to the hole-accepting level of the HTM will
lead to a higher driving force for hole transfer.46
However, lowering the anode CB and
increasing the HTM level will decrease the VOC, which will eventually offset the benefits
Katherine Roelofs Ph.D. Dissertation
26
of increased charge collection. Since the charge transfer rate is exponentially dependent
on the energetic offset (assuming Marcus theory kinetics) and VOC scales linearly with
changes in energy levels, large gains in improved charge collection can be realized
without a too-large impact on VOC. Increasing the absorber band gap is another way to
increase the energetic offsets for electron injection and hole transfer, but this comes at the
expense of collecting a decreased portion of the solar spectrum. Surface treatments can
be used to precisely tune band alignment for maximum benefit.
While the high internal surface area of nanostructured devices leads to high rates
of recombination, the high absorber surface area also provides the opportunity for surface
treatments to dramatically improve device efficiencies. The efficacy of surface
treatments in modifying interface band alignments, decreasing recombination, improving
stability, and increasing surface area has already been demonstrated in DSSCs,3 and these
strategies can also serve to improve performance in QDSSCs, ETASCs, and CQDSCs. It
is noteworthy that publications in the fields of QDSSCs, ETASCs, and CQDSCs
emphasize interfacial modifications as a means to improve device efficiency – though
there are many different types of surface treatments and different mechanisms by which
they improve device performance. This chapter will highlight three types of interfacial
modification: band alignment by molecular dipole layers, improved CQD film mobility
by ligand exchange, and decreased recombination by interfacial inorganic layers. We
have selected a few specific examples which we believe highlight the overall trends and
give insight into the importance of interface engineering in inorganic-absorber
nanostructured devices.
Due to the limited scope of this chapter, we direct the reader to other works for a
more comprehensive view of recent progress in QDSSCs,18, 19, 36, 43, 51
ETASCs,2, 14, 20
and
CQDSCs,35, 38, 39, 52
and to the following articles for further discussion of interfacial
charge transfer processes in QDSSCs,35, 53-57
ETASCs,14, 47
and CQDSCs.15, 16, 35, 52, 57-59
We conclude this chapter with an outlook on further developments in the field.
Katherine Roelofs Ph.D. Dissertation
27
2.4. Band Alignment by Molecular Dipole Layer
In QDSSCs, ETASCs, and CQDSCs, molecular dipole layers have been actively
pursued as a means to tune interfacial electronic properties. Molecular dipole layers are
of particular importance to devices with QD absorbers, since surface dipoles allow for the
adjustment of the interfacial band alignment without resorting to changing the band gap
of the QD. By employing such surface modifications, QDSSCs and CQDSCs retain the
ability to optimize the QD size (and thus band gap) for maximum absorption,
independent of band alignment considerations. Molecular dipole layers can be used to
tune the band offsets between the QD and the anode, as well as between the QD and the
HTM. This ability to impact the band alignment gives a high degree of control over the
energetic overpotentials for both injection and regeneration (Figure 2-2), and thus, the
overall charge collection efficiency.34, 47, 50, 58
Furthermore, fine control over the
energetic offsets at these interfaces allows researchers to push towards the highest-
possible values of VOC, since the limit of the VOC is set by the energetic offset between the
electron quasi-Fermi level in the anode and the hole quasi-Fermi level in the HTM.
A molecular dipole layer is formed by a layer of molecules chemically bonded to
a substrate, in which the molecules have a dipole moment along their backbone. For
instance, in a molecule with an electron-withdrawing functional group (the tail group),
the shift in electron density will create a physical separation of positive and negative
charge along the length of the molecule, i.e., a molecular dipole. The dipolar molecules
can be attached either directly on the QD or on the surface of the anode material. A
molecular dipole layer on a QD can cause an upward or downward shifting of the
conduction band of the QD towards vacuum, depending on the dipole orientation.
Katherine Roelofs Ph.D. Dissertation
28
Figure 2-3. Schematic of band shifting in a quantum dot (QD) due to functionalization
with a surface dipole layer. Surface dipoles have been investigated explicitly in QDSSCs
via molecular dipole layers of varying strengths, although it is also possible for inorganic
layers to create a surface dipole layer. The band energy levels of the QD prior to the
addition of the surface dipole layer are shown in black: the conduction band energy level,
E°CB, and valence band energy level E°VB. The position of the bands after the application
of the surface dipole layer are indicated by EʹCB and EʹVB.
Figure 2-3 shows schematically the impact of a molecular dipole layer on the
position of a QD’s conduction and valence bands relative to vacuum. In the example
shown, an adsorbed molecular dipole with the negative charge oriented toward the
surface (negative dipole) will lead to an upward shift of the bands.60, 61
Molecules
forming a self-assembled monolayer (SAM) can be attached by choice of a head group
that binds specifically to the absorber surface, or to the photoanode surface. For instance,
a sulfide head group will bind preferentially to metal chalcogenide QDs over the metal
oxide anode, whereas a carboxyl head group may preferentially bind to the metal oxide
anode.
In a study on benzenethiol derivatives in QDSSCs, Shalom et al. investigated the
effects of employing molecular dipoles of varying strengths in a n-TiO2/CdS-
QD/polysulfide electrolyte device architecture.61
In these experiments, a benzenethiol
molecule was chosen with the intention of the thiol group attaching selectively to the CdS
Katherine Roelofs Ph.D. Dissertation
29
QD surface. In this work, and in a follow-up study by Barea et al., the authors found that
the incident photon-to-current efficiency (IPCE) depends linearly on dipole moment; the
results are shown in Figure 4.61, 62
Enhancements in JSC from 1.2 to 2.0 mA/cm2 were
achieved with negative dipoles, which the authors concluded had shifted the conduction
band of the QDs upward (closer to vacuum level), thus increasing the electron injection
rate and device efficiency.62
In another example investigating benzenethiol derivaties,
this time in a solid-state device (n-TiO2/CdS QD/spiro-OMeTAD), Choi et al. reported a
similar result: strong negative dipoles lead to enhancements in efficiency.63
These
studies demonstrate how systematically increasing the dipole moment can increase IQE
and overall efficiencies.
Katherine Roelofs Ph.D. Dissertation
30
Figure 2-4. Control of injection in QDSSCs by molecular dipole layers. (a) Incident
photon-to-current efficiency (IPCE) measurements of QDSSCs with benzenethiol
derivatives of varying dipole moment strengths. The ordered series of aromatic
functional groups (NO2, F, CH3, OCH3) were calculated to have increasingly negative
dipole moment values (D). (b) Photovoltage spectroscopy (PVS) was employed to
measure the electron injection from the CdSe QD to TiO2. (c) The shift in PV onset to
lower energetic values (higher wavelengths) was found to correspond with increasingly
negative dipole moments. This shift in PV onset to lower energies could be caused by
Katherine Roelofs Ph.D. Dissertation
31
increased electron injection from QDs that, prior to molecular dipole treatment, had
previously had too low a CB for injection to occur efficiently into the TiO2. Adapted
with permission from Ref. 62. Copyright 2010 American Chemical Society.
Although the approach is powerful, surface functionalization via interfacial
molecular layers is limited by the number of binding sites on the anode, and so the
number and type of functional molecules attached to the surface is limited. In DSSCs,
surface attachment sites for organic functionalization are restricted as the dye molecules
themselves must bind to the anode. In inorganic-absorber devices, while the absorber
(the QD or the ETA) does consume anode surface sites in the same way as dye
molecules, the absorber at the same time creates a new surface for functionalization (the
outer QD or ETA surface, facing towards the HTM). In CQDSCs, the new surface
available for functionalization is that of the QDs in the CQD films. Thus, a benefit of
inorganic-absorber devices over organic-absorber devices is the ability to attach
molecules to the absorber surface.
2.5. Improved CQD Film Mobility by Ligand
Exchange
Another common interface modification used in inorganic-absorber
nanostructured solar cells is ligand exchange, shown schematically in Figure 2-5. Ligand
exchange of the capping ligands in colloidal quantum dots (CQDs) can be used to achieve
multiple goals. In CQDSCs, the quantum dots are synthesized by colloidal methods,
which produce a colloidal suspension of QDs in organic solvent. Capping ligands
attached to the quantum dot surface allow for suspension in the organic solvent and
prevent aggregation of the quantum dots. Typical ligands employed during synthesis are
long chain alkanes with high boiling points.64
A ligand exchange step can be undertaken
while the CQDs are still suspended in solution, or ligand exchange can be undertaken on
the CQD film itself, after the CQDs have been deposited onto a substrate. Ligands can be
chosen for fabrication benefits during film deposition, such as the mechanical stability of
the film,38, 65
or for improved electronic properties of the CQD film, such as higher CQD
film mobilities.66-68
For example, a 2008 study by Koleilat et al., provides generalized
Katherine Roelofs Ph.D. Dissertation
32
guidance for ligand choice in CQD films for improved charge mobility; from their
results, the authors conclude that in order to best assist inter-QD charge transfer, ligands
should have strongly-binding end groups, such as bidentate linkers, and short, conjugated
backbones.68
Figure 2-5. Schematic of colloidal quantum dot (CQD) film fabrication. During
colloidal synthesis of the quantum dots, long alkyl ligands are used to suspend the QDs in
the organic solvent. After the synthesis of the CQDs, the longer ligands are then
exchanged for shorter, often conjugated ligands, for better mobility properties in the final
CQD film. Due to the multi-layer processes used to grow CQD films of the desired film
thickness, ligand exchange is often conducted after CQD deposition on a substrate.
Ligand exchange at this stage also provides the change in CQD solubility required to
prevent dissolution of the CQDs back into the original solution. The exchange of longer
ligands for shorter ones results in a densification of the CQD film.
This section will focus on ligand choice for improved film mobility in CQDSCs,
as this area has seen rapid improvements over the past decade.69
The findings from
ligand exchange studies for improved CQD film mobility (QD-to-QD charge transfer)
will likely also help guide ligand choice for improved charge injection (QD-to-anode
charge transfer) in these devices. Here, it is worth noting that CQDSCs built as Schottky
devices – with a Schottky junction at the transparent front contact – actually collect holes,
not electrons, at the front contact.15
Accordingly, the generalized band diagram in Figure
2-2 does not apply to Schottky CQDSCs. On the other hand, CQDSCs with metal oxide
anodes following the DSSC design do collect electrons at the transparent front contact, in
accordance with the charge transfer processes shown in Figure 2-2; this device
architecture is also referred to as a depleted heterojunction cell.15
Improved mobility in CQD films is a key factor in increasing device efficiencies.
For complete absorption of incident light CQD film thicknesses are on the order of 500
nm, but minority carrier diffusion lengths remain on the order of tens of nanometers and
Katherine Roelofs Ph.D. Dissertation
33
require a built-in field for charge collection.70
Research on ligand exchange for improved
mobility in CQD films goes hand-in-hand with gains in fundamental knowledge of
charge transport processes in these films. There is debate in the literature as to whether a
bound electron-hole pair (i.e. exciton) diffuses through the CQD film, or whether the
charges rapidly split with the electrons and holes traveling separately through the CQD
film.38, 67
Critical ongoing questions in the field include: how quickly do excitons split,
and does the device operate by exciton diffusion or by diffusion of excited carriers?15, 38,
67 Despite the uncertainty surrounding the underlying mechanism, much progress has
been made in increasing charge mobility in CQDSCs, especially due to fundamental
advances in ligand exchange methods.38, 66, 71
Figure 2-6. Charge mobilities in PbSe CQD films were found to exponentially decay
with increased ligand length. The ligands studied were 1,2-ethanedithiol (EDT); 1,3-
propanedithiol (PDT, >98%); 1,4-butanedithiol (BuDT); 1,5-pentanedithiol (PenDT);
1,6-hexanedithiol (HDT). The electron and hole mobilities were measured in ambipolar
field-effect transistors (6.1 nm NCs). Reprinted with permission from Ref. 67.
Copyright 2010 American Chemical Society.
Figure 2-6 shows the results of a study by Liu et al. that nicely illustrates the
dependence of carrier mobility on QD size and ligand length for PbSe QDs with alkane
chain ligands.67
The electron and hole field effect mobilities (μe-FE and μh+
FE ) were found
to decrease exponentially with ligand length, with similar exponential scaling parameters:
1.10 and 1.08 Å-1
, for μe-FE and μh+
FE, respectively. These scaling parameters indicate that,
roughly speaking, the charge mobilities in the CQD film drop by a third with every
Katherine Roelofs Ph.D. Dissertation
34
additional Ångstrom of ligand length. This exponential decay is a good example of the
sensitivity of charge transfer processes to Ångstrom-scale changes in interfacial
functionalization of nanostructured devices.
Figure 2-7. CQDSCs with ligand exchange down to a single-atom ligand, studying
ethanedithiol (EDT), 1,6-hexane dithiol (HDT), and mercaptocarboxylic acid (MPA), and
the single-atom ligand of Br. (a) Time-resolved infrared (TRIR) spectroscopy
measurements of CQD films; peaks in TRIR spectra correspond to trap-to-band
transitions. In the Brˉ-capped CQD films, the TRIR peak occurs at a lower energy level,
indicating that Brˉ-capped films have shallower trap states. (b) Decay times extracted
from the absorption decay curves of the TRIR peak, tracking the rate at which trap states
are depopulated. (c) Higher CQD film mobilities were found in CQD films with lower
trap-to-band transition energies. Reprinted with permission from Macmillan Publishers
Ltd: Nature Materials (Ref. 71), copyright 2011.
Taking this inverse relationship between ligand length and film mobility one step
further, Tang et al. have provided an in-depth study of ligand exchange down to single-
atom ligands, producing the first such report of CQDSCs with single-atom ligands.71
The
authors investigate the impact ligand exchange has on charge transport in the PbS CQD
film, as well as any passivation effects of the ligand exchange on extant electronic defects
on the PbS QD surfaces. Electron mobilities of halide-treated CQD films measured by
field effect transistor methods were found to increase in the Brˉ capped CQD films, as
compared to the ethanedithiol-capped CQD films. This increase in electron mobility with
the shift to a single-atom ligand is significant since a high minority carrier mobility is
critical to achieving high CQDSC efficiencies. Indeed, the resulting atomic-capped PbS
CQDSCs achieved device efficiencies of 5%. The authors took the further step of
determining the cause of the higher mobility in the halide treated films, by time-resolved
infrared (TRIR) spectroscopy measurements shown in Figure 2-7. In the TRIR
Katherine Roelofs Ph.D. Dissertation
35
measurements, a 523 nm laser was used to excite the CQD band gap, and an IR light
source was used to probe the lower-energy trap-to-band transitions. The results (Figure
2-7a) indicated that the excited electrons are trapped in energetically shallower states in
Brˉ capped CQD films, as compared to the CQD films with organic ligands. To confirm
their interpretation of the TRIR peaks, the authors tracked the absorption decay time of
the TRIR peaks (Figure 2-7b), which indicate the rate at which the population of the trap
state is depleted due to trap-to-band transitions. As expected, shallower trap states (lower
transition energies) had faster rates of trap-to-band transitions. The authors also found
higher electron mobilities correlated with shallower trap states (Figure 2-7c). This result
can be explained by the following mechanism: for CQD films with shallower trap states,
electrons spend less time in any given trap state, and thus have a faster rate of progress
through the film. Overall, this single-atom ligand study demonstrates that ligand
exchange can be used both for improving CQD film mobilities (QD-to-QD charge
transfer rates), as well as for passivating QD surface defects by modifying the energy
levels of the trap states associated with those surface defects.
We believe that ligand choice guidelines taken from the CQDSC literature will
benefit studies in QDSSCs, such as in molecular design of organic surface
functionalization of QDSSCs. For example, while the studies of ligand exchange
discussed above were undertaken in the context of increasing film mobility in CQDSCs,
the results could also guide studies in QDSSCs for improved interfacial charge transfer.
For one, ligand design is necessarily important in QDSSCs in which the QDs are
fabricated colloidally and later infiltrated into the nanostructured anode. Moreover, a
study by Dibbel et al. has found ligand length to be a critical factor in electron injection
from the QDs into the anode material, using spectroscopic measurements of colloidal
suspensions of TiO2 nanocrystals and CdS CQDs.72
Finally, in QDSSCs functionalized
with molecular dipole layers, the molecular dipole layer can interfere with hole transfer
from the QD to the HTM; thus ligand design strategies from CQDSCs could also inform
the choice of molecules for molecular dipole layers in QDSSCs.
Katherine Roelofs Ph.D. Dissertation
36
2.6. Reduced Recombination by Interfacial
Inorganic Layers
In addition to the organic surface treatments discussed above, inorganic materials
also find use in interface modification of nanostructured solar cells. Although inorganic
surface treatments have been used for a variety of purposes, including use as a protective
capping layer on the absorber or for modification of the interfacial band alignment, the
majority of studies have explored inorganic treatments as a means to reduce interfacial
recombination. Due to their success in improving device efficiencies, inorganic
modifications have found widespread use in QDSSCs, ETASCs, and CQDSCs. Metal
oxide layers (such as Al2O3,30
HfO2,73
or ZrO274
) have been deposited at the interface by
solution deposition methods75-77
and by vapor phase growth techniques.49, 59, 77, 78
Metal
chalcogenide layers (such as CdSe, ZnS, or PbSe) have also been used in QDSSCs and
CQDSCs to coat the QD surface.79, 80
Inorganic layers can reduce recombination through chemical, electronic, or
physical mechanisms. For example, inorganic layers can decrease recombination by
electronically altering the interfacial band structure, chemically passivating surface
defects, or acting as a physical barrier between different components. Decreased
interfacial recombination can improve device efficiencies through both increased JSC
values and increased VOC values for the following reason. Interfacial recombination
represents the loss of excited charge carriers, which in turn reduces the photocurrent
(JSC). Hence, reducing interfacial recombination can help improve JSC as well as
decrease dark current (J0); both of these effects can improve VOC according to the
Katherine Roelofs Ph.D. Dissertation
37
relationships of general p-n junction theory.
Figure 2-8. Interfacial band structure depicting the potential impacts of inorganic
surface treatments in QDSSCs. Here, a nanocrystal of the nanoporous anode (i.e., a TiO2
nanocrystal) is depicted in navy, the inorganic coating layer in magenta, and the light-
absorbing quantum dot (QD) in yellow. The hole-transport material (HTM) is shown in
light blue. (a) The interfacial band structure prior to deposition of the metal oxide layer.
Only recombination pathways from the photoanode to the QD are shown, from (α) the
anode conduction band and (β) the anode density of states (DOS), although
recombination can also occur with the HTM. (b) Recombination barrier layers will
reduce both (α) and (β) recombination processes due to the tunneling effect. (c)
Recombination barriers may also alter the DOS of the anode, by passivating trap states,
leading to further reductions in the (β) recombination path. (d) The recombination barrier
layer may also act as a surface dipole, shifting the anode conduction band upward. (e)
Due to the difficulties of conformally growing Ångstrom-thick inorganic layers on a
nanoporous anode, it is possible that multiple geometric configurations of the
anode/inorganic coating/QD exist within the same device.
Figure 2-8 depicts schematically a few mechanisms by which the electronic
structure at the anode/absorber/HTM interface can be altered by the deposition of
inorganic layers: presenting a tunneling barrier to recombination of TiO2 electrons with
the absorber or the HTM (Figure 2-8b), changing the density of states (DOS) of the anode
(Figure 2-8c), or shifting the band levels of the anode (Figure 2-8d). In Figure 2-8b, if
the energetic barrier is too high for the electron to surpass by thermal or other energetic
means, the electron can quantum mechanically tunnel through the barrier. The
interpretation of the effects of inorganic layers on device performance (J-V curves) can
thus be quite complicated. Any of the changes in Figure 2-8 could lead to increases in
Katherine Roelofs Ph.D. Dissertation
38
VOC, either through decreases in the recombination rates or by an upward shift in the
anode CB. As a final consideration, it is worth noting that a given device can have
different geometric arrangements at the photoanode/QD/HTM interface (Figure 2-8e),
depending on whether the QD nucleates on the barrier layer or on the exposed
photoanode surface sites with the barrier deposited afterwards.49
Figure 2-9. The role of TiCl4 treatment in QDSSCs (n-TiO2/CdS QDs/electrolyte). (a)
J-V curves as a function of TiO2 CBD time; inset shows dark J-V curves. The highest
efficiency is achieved at 30 min. (b) Open circuit voltage decay curves under AM 1.5
illumination, show that the highest electron lifetimes (slowest rate of VOC decay) are
achieved for the 30 min samples. (c) Schematic of the authors’ proposed mechanism.
With increasing TiO2 CBD times, the resulting TiO2 coating is believed to decrease
interfacial recombination, an effect which dominates at the 30 min CBD time point. For
thicker TiO2 coatings, though, it is believed that the detrimental effects of transport
resistance lead ultimately to the lower observed electron lifetimes at 60 min. Reprinted
from Ref. 81, Copyright 2012, with permission from Elsevier.
Katherine Roelofs Ph.D. Dissertation
39
Figure 2-9 shows data from a study by Kim et al. on inorganic layer surface
modifications of liquid QDSSCs.81
In a CdS QDSSC (n-TiO2/CdS-QD/polysulfide-
electrolyte), the authors investigated TiO2 surface coatings of varying thicknesses
deposited on the TiO2 anode. Varying TiO2 thicknesses were achieved by varying
lengths of soak time in a TiCl4 aqueous solution for the chemical bath deposition of TiO2.
This work by Kim et al. is highlighted here since the results lead the authors to conclude
that the TiO2 surface coating influenced device efficiencies via two different
mechanisms, explaining the initial increase and subsequent decrease in device
efficiencies. Transient photovoltage measurements (Figure 2-9b) were used to show the
decrease in interfacial recombination achieved with increasing thicknesses of the TiO2
layer. However, at longer TiCl4 soak times, the increased thickness of the TiO2 coating
was found (by XRD measurements) to lead to lattice strain in the TiO2 nanoparticles. As
shown schematically in Figure 2-9c, the authors conclude that the observed initial
increase in device efficiency was caused by a suppression of photoanode-to-electrolyte
recombination (pathway 6, Figure 2-2), while the subsequent decrease in efficiency was a
result of increased transport resistance due to increased TiO2 lattice strain and pore-filling
issues.
In addition to metal oxides, wide band gap metal chalcogenide layers have also
been employed successfully in QDSSCs to decrease interfacial recombination. ZnS
layers were first introduced as a means to prevent photocorrosion of QDs,82
with gains in
efficiency also observed. Recently, Guijarro et al. have studied in depth the means by
which efficiency improvements are achieved with ZnS layers.83
In this work, CQDSC
devices (n-TiO2/CdSe CQD/polysulfide electrolyte) were treated with SILAR-grown ZnS
layers. The authors found evidence of both QD surface defect passivation, determined by
UV-Vis absorbance and photoluminescence, and increased charge transfer resistance (i.e.
decreased recombination), measured by electrochemical impedance spectroscopy. The
increased charge transfer resistance was attributed to increased separation between the
photoanode and the electrolyte, and led to increases in JSC. This study shows that ZnS
coatings can decrease recombination to the HTM through bare TiO2 surface regions, as
well as recombination mediated by the CdSe QDs surface defects.
Katherine Roelofs Ph.D. Dissertation
40
Figure 2-10. (a) Inorganic layers of Al2O3, BaTiO3, and MgO deposited at the interface
in ETASCs. (b) All three metal oxide layers served to improve the VOC, a good
indication of decreased interfacial recombination, though only in the MgO case did JSC
values improve as well. (c) A double layer treatment of BaTiO3 followed by MgO,
provided even better improvements in performance than either treatment alone. The
device parameters of the J-V curves are included, and have the following units: JSC
(mA/cm2), VOC (V), FF (unitless), and η (%). Reprinted from reference 75.
Inorganic layers are also used frequently in ETASCs, with the intent of decreasing
interfacial recombination. Tsujimoto et al. have studied interfacial layers of Al2O3,
BaTiO3, and MgO as recombination-blocking layers in ETASCs;75
device performances
are shown in Figure 2-10. The authors chose a device architecture of n-
TiO2/Sb2S3/CuSCN; current record-efficiency ETASCs employ Sb2S3 as the absorber
(Table 2-1)22
and the use of an inorganic, solid-state HTM (CuSCN) makes this an all-
solid-state, all-inorganic device. All three metal oxides increased the device efficiency,
primarily through increases in the VOC or the FF. As discussed above, such a result could
be expected if the barriers were acting to decrease J0, and thus increase the VOC. Then, by
combining two treatments, with a BaTiO3 layer followed by an MgO layer, the authors
were able to achieve even higher efficiencies than had been observed in either individual
treatment. This study shows the benefits of a combinatorial approach to interfacial
modifications: while the authors reported that a single material grown too thick always
led to losses in efficiency, the deposition of two thin layers of different materials
provided an additive benefit in which each outweighed any negative impact of either (that
is, the decreased JSC observed when BaTiO3 was employed on its own). We have also
observed in our work with QDSSCs (see Outlook) that the deposition of inorganic layers
at the interface leads to losses in efficiency when the inorganic layer is grown too thick.49
Katherine Roelofs Ph.D. Dissertation
41
2.7. Outlook
As with DSSCs, power generation in QDSSCs, ETASCs, and CQDSCs hinges on
interfacial charge transfer processes. By examining several examples of interfacial
modifications in this overview – namely, molecular dipole layers, capping ligands, and
inorganic layers – we have aimed to demonstrate the versatility of surface modifications
and their ability to improve the performance of QDSSCs, ETASCs, and CQDSCs. It is
tempting to approach the idea of surface modifications by employing a single
modification to achieve a single goal (e.g. adjusting band alignment, passivating surface
defects, or providing a recombination barrier). We have seen, however, that any given
surface modification has the ability to act in multiple beneficial ways. For instance, as
discussed in the previous section, interfacial inorganic layers in QDSSCs have the
potential to act both as a barrier to recombination and as a passivant of QD surface
defects.83
Considering that surface treatments inevitably have multiple effects, a better
strategy when considering interfacial modifications may be to purposefully choose
surface modifications that will have multiple beneficial effects, or to employ a combined
approach, using multiple compatible surface treatments at a given interface. Indeed, a
combined approach has been successfully pursued in all three device architectures. For
instance, in QDSSCs, there have been studies employing such combinations as F-
chemical treatments and ZnS inorganic layers,79
molecular dipole layers and ZnS
inorganic layers,62
and Al2O3 passivation layers with the super-sensitization of a light-
absorbing dye molecule.84
In addition to considering the possibility that a given surface treatment may
engender multiple effects, it is also important to identify which interface is being
modified in any given treatment. For surface modifications in QDSSCs, ETASCs, and
CQDSCs, there are two surfaces in the active layer that can be modified: the anode
surface and the absorber surface. In ETASCs, the distinction between the anode/absorber
interface and the absorber/HTM interface is clear, since these two interfaces are
physically separated. In QDSSCs, the story is more complicated, as surface treatments
on the QDs will invariably affect the metal oxide substrate as well, given the sparse
coverage of the metal oxide by the QDs.37
Thus, in the case of QDSSCs, molecular
Katherine Roelofs Ph.D. Dissertation
42
layers are an appealing tool for interfacial modification, since the head group of the
molecule can be chosen to attach specifically to either the QD surface or the anode
surface, giving a higher degree of control over interfacial engineering. While the use of
inorganic absorbers offers more opportunities to control interfacial properties, it
necessitates a careful and comprehensive sample set in order to isolate changes in device
performance. In CQDSCs, while much effort has been devoted to modifications of the
QD surfaces in the CQD film, charge transfer at the anode/CQD-film interface is also of
importance,16, 35
and CQDSCs would benefit from further research on interfacial
modifications for the optimization of the anode/CQD-film interface.
In our own work with QDSSCs, we have taken advantage of the two interfaces
available for modification, by depositing inorganic barrier layers both prior to and after
QD deposition in CdS QDSSCs49
and in PbS QDSSCs.78
Depositing the barrier layer
before QD deposition modifies the anode/absorber interface, whereas depositing the
barrier layer after QD deposition primarily modifies the absorber/HTM interface. In both
the CdS and the PbS studies, we employed ALD for the deposition of Al2O3 layers at the
interface in solid-state QDSSCs (n-TiO2/QDs/spiro-OMeTAD). ALD is a self-limiting
vapor deposition technique; at most a monolayer of material is deposited in each ALD
cycle, due to the separate introduction of precursor material in an ALD run. ALD is a
valuable technique for surface modifications in nanostructured solar cells, due to its
ability to conformally coat high aspect ratio substrates, to penetrate highly porous films
of micron-scale thicknesses, and to provide Ångstrom-scale control of film thicknesses.
Katherine Roelofs Ph.D. Dissertation
43
Figure 2-11. Inorganic surface modification of CdS QDSSCs by the deposition of ultra-
thin Al2O3 layers by atomic layer deposition (ALD). The Al2O3 layer thickness was
varied, with 0, 1, and 3 ALD cycles of Al2O3. The Al2O3 layers were deposited both after
QD deposition and before QD deposition, resulting in two different configurations: n-
TiO2/QD/Al2O3 (filled markers) and n-TiO2/Al2O3/QD (open markers). (a) Electron
lifetimes were extracted from transient photovoltage measurements, showing longer
carrier lifetimes in the case of the n-TiO2/Al2O3/QD devices. Standard deviations
represent the spread of lifetimes measured across three different batches of devices (each
data point represents ~ 6 devices in total). (b) One possible explanation for the results is
that the n-TiO2/Al2O3/QD devices could block two interfacial charge recombination
processes, whereas the n-TiO2/QD/Al2O3 devices could block only one. Reprinted from
reference 49.
For the CdS QDSSCs, while the J-V curve behavior was qualitatively similar in
comparing the two Al2O3 layer configurations (prior to QD deposition versus after QD
deposition), we found differences in the interfacial recombination processes, as tracked
by measurements of the lifetimes of excited electrons in the TiO2. Figure 2-11 illustrates
the electron lifetime results from the CdS QDSSC study, showing that devices with the
Al2O3 layers deposited prior to the QDs had higher electron lifetimes than those with
Al2O3 layers deposited after the QDs. Based on these results and others, we concluded
Katherine Roelofs Ph.D. Dissertation
44
that a likely explanation was as follows: Al2O3 layers deposited prior to the QDs could
block both recombination to the oxidized QD and the HTM, whereas Al2O3 layers
deposited after the QD could only block recombination to the HTM (Figure 2-11b). In
our work with PbS QDSSCs we also found that Al2O3 deposited prior to the QDs gave
higher electron lifetime values than that of Al2O3 deposited after the QDs.78
In this series
of studies, by experimentally separating out the different recombination losses that can
occur, we showed that while passivation of either interface can improve QDSSC
performance, the carrier losses are more severe at the TiO2/QD interfaces in these cells.
In conclusion, great benefits can be reaped in nanostructured solar cells through
surface treatments, due to the extensive interfacial areas in these devices. In QDSSC,
ETASCs, and CQDSCs, the use of inorganic absorber materials provides additional
opportunities for interfacial modification, as compared to DSSCs, due to the ability to
modify both the anode and the absorber surface. We believe that further research
exploring the exact mechanisms at work behind surface treatments will lead not only to a
deeper understanding of device operation, but also to further gains in efficiency. There is
great opportunity for cross-over in these devices; that is, that interfacial modifications
and strategies, when proved successful in one device architecture, stand a good chance of
being beneficial in another. As fundamental understanding of interfacial surface
treatments in these devices advances, there is a great potential for employing a combined
approach, of several distinct surface treatments in a given device. Overall, we believe
that in nanostructured solar cells, interfacial modifications will play a key role in moving
power conversion efficiencies into the sphere in which they are competitive with other
thin film technologies (e.g. α-Si, CdTe, and CIGS).
2.8. Financial Support, Collaborations, and
Copyrights
This chapter is adapted from work originally published in the Journal of Physical
Chemistry Letters.85
Reprinted with permission from Roelofs, K. E.; Brennan, T. P.;
Bent, S. F., The Journal of Physical Chemistry Letters 2014, 5 (2), 348-360. Copyright
Katherine Roelofs Ph.D. Dissertation
45
2014 American Chemical Society. This work was completed with help from Thomas
Brennan, and was supported as part of the Center on Nanostructuring for Efficient Energy
Conversion, an Energy Frontier Research Center funded by the U.S. Department of
Energy, Office of Science, Basic Energy Sciences under Award # DE-SC0001060.
2.9. References
1. O'Regan, B.; Grätzel, M., Nature 1991, 353 (6346), 737-740.
2. Archer, M. D.; Nozik, A. J., Series on Photoconversion of Solar Energy:
Nanostructured and Photoelectrochemical Systems for Solar Photon Conversion.
Imperial College Press: London, 2008; Vol. 3.
3. Hagfeldt, A.; Boschloo, G.; Sun, L.; Kloo, L.; Pettersson, H., Chemical Reviews
2010, 110 (11), 6595-6663.
4. Kim, H.-S.; Lee, C.-R.; Im, J.-H.; Lee, K.-B.; Moehl, T.; Marchioro, A.; Moon,
S.-J.; Humphry-Baker, R.; Yum, J.-H.; Moser, J. E.; Gratzel, M.; Park, N.-G., Scientific
Reports 2012, 2, 591.
5. K. Nazeeruddin, M.; Péchy, P.; Gratzel, M., Chemical Communications 1997,
(18), 1705-1706.
6. Yella, A.; Lee, H.-W.; Tsao, H. N.; Yi, C.; Chandiran, A. K.; Nazeeruddin, M. K.;
Diau, E. W.-G.; Yeh, C.-Y.; Zakeeruddin, S. M.; Grätzel, M., Science 2011, 334 (6056),
629-634.
7. Peter, L. M., The Journal of Physical Chemistry Letters 2011, 2 (15), 1861-1867.
8. Lee, M. M.; Teuscher, J.; Miyasaka, T.; Murakami, T. N.; Snaith, H. J., Science
2012, 338 (6107), 643-647.
9. Heo, J. H.; Im, S. H.; Noh, J. H.; Mandal, T. N.; Lim, C.-S.; Chang, J. A.; Lee, Y.
H.; Kim, H.-J.; Sarkar, A.; K., N.; Gratzel, M.; Seok, S. I., Nature Photonic 2013, 7 (6),
486-491.
10. Noh, J. H.; Im, S. H.; Heo, J. H.; Mandal, T. N.; Seok, S. I., Nano Letters 2013,
13 (4), 1764-1769.
11. Liu, M.; Johnston, M. B.; Snaith, H. J., Nature 2013, 501 (7467), 3.
12. Burschka, J.; Pellet, N.; Moon, S.-J.; Humphry-Baker, R.; Gao, P.; Nazeeruddin,
M. K.; Grätzel, M., Nature 2013, 499 (7458), 316-319.
13. Lee, H.; Leventis, H. C.; Moon, S.-J.; Chen, P.; Ito, S.; Haque, S. A.; Torres, T.;
Nüesch, F.; Geiger, T.; Zakeeruddin, S. M.; Grätzel, M.; Nazeeruddin, M. K., Advanced
Functional Materials 2009, 19 (17), 2735-2742.
14. Briscoe, J.; Dunn, S., Mater. Sci. Technol. 2011, 27 (12), 1741-1756.
15. Pattantyus-Abraham, A. G.; Kramer, I. J.; Barkhouse, A. R.; Wang, X.;
Konstantatos, G.; Debnath, R.; Levina, L.; Raabe, I.; Nazeeruddin, M. K.; Grätzel, M.;
Sargent, E. H., ACS Nano 2010, 4 (6), 3374-3380.
16. Maraghechi, P.; Labelle, A. J.; Kirmani, A. R.; Lan, X.; Adachi, M. M.; Thon, S.
M.; Hoogland, S.; Lee, A.; Ning, Z.; Fischer, A.; Amassian, A.; Sargent, E. H., ACS
Nano 2013, 7 (7), 6111-6116.
17. Hillhouse, H. W.; Beard, M. C., Current Opinion in Colloid & Interface Science
2009, 14 (4), 245-259.
Katherine Roelofs Ph.D. Dissertation
46
18. Kamat, P. V., The Journal of Physical Chemistry C 2008, 112 (48), 18737-18753.
19. Kamat, P. V., The Journal of Physical Chemistry Letters 2013, 4 (6), 908-918.
20. Dittrich, T.; Belaidi, A.; Ennaoui, A., Solar Energy Materials and Solar Cells
2011, 95 (6), 1527-1536.
21. Wang, J.; Mora-Seró, I.; Pan, Z.; Zhao, K.; Zhang, H.; Feng, Y.; Yang, G.;
Zhong, X.; Bisquert, J., Journal of the American Chemical Society 2013, 135 (42),
15913-15922.
22. Chang, J. A.; Im, S. H.; Lee, Y. H.; Kim, H.-J.; Lim, C.-S.; Heo, J. H.; Seok, S. I.,
Nano Letters 2012, 12 (4), 1863-1867.
23. Buhbut, S.; Itzhakov, S.; Hod, I.; Oron, D.; Zaban, A., Nano Letters 2013.
24. Albero, J.; Clifford, J. N.; Palomares, E., Coordination Chemistry Reviews 2013,
(0).
25. Lee, H. J.; Yum, J.-H.; Leventis, H. C.; Zakeeruddin, S. M.; Haque, S. A. Chen,
P. Seok, S. I. Grätzel, M.; Nazeeruddin, M. K., The Journal of Physical Chemistry C
2008, 112 (30), 11600-11608.
26. Im, J.-H.; Lee, C.-R.; Lee, J.-W.; Park, S.-W.; Park, N.-G., Nanoscale 2011, 3
(10), 4088-4093.
27. Lee, H.; Leventis, H. C.; Moon, S. J.; Chen, P.; Ito, S.; Haque, S. A.; Torres, T.;
Nuesch, F.; Geiger, T.; Zakeeruddin, S. M.; Grätzel, M.; Nazeeruddin, M. K., Adv. Func.
Mater. 2009, 19, 2735-2742.
28. Chung, I.; Lee, B.; He, J.; Chang, R. P. H.; Kanatzidis, M. G., Nature 2012, 485
(7399), 486-489.
29. de la Fuente, M. S.; Sánchez, R. S.; González-Pedro, V.; Boix, P. P.; Mhaisalkar,
S. G.; Rincón, M. E.; Bisquert, J.; Mora-Seró, I., The Journal of Physical Chemistry
Letters 2013, 4 (9), 1519-1525.
30. Brennan, T. P.; Bakke, J. R.; Ding, I. K.; Hardin, B. E.; Nguyen, W. H.; Mondal,
R.; Bailie, C. D.; Margulis, G. Y.; Hoke, E. T.; Sellinger, A.; McGehee, M. D.; Bent, S.
F., Physical Chemistry Chemical Physics 2012, 14 (35), 12130-12140.
31. Palomares, E.; Clifford, J. N.; Haque, S. A.; Lutz, T.; Durrant, J. R., Journal of
the American Chemical Society 2003, 125 (2), 475-482.
32. Fabregat-Santiago, F.; Garcia-Canadas, J.; Palomares, E.; Clifford, J. N.; Haque,
S. A.; Durrant, J. R.; Garcia-Belmonte, G.; Bisquert, J., Journal of Applied Physics 2004,
96 (11), 6903-6907.
33. Haque, S. A.; Palomares, E.; Cho, B. M.; Green, A. N. M.; Hirata, N.; Klug, D.
R.; Durrant, J. R., Journal of the American Chemical Society 2005, 127 (10), 3456-3462.
34. Kamat, P. V., The Journal of Physical Chemistry C 2007, 111 (7), 2834-2860.
35. Grätzel, M.; Janssen, R. A. J.; Mitzi, D. B.; Sargent, E. H., Nature 2012, 488
(7411), 304-312.
36. Rühle, S.; Shalom, M.; Zaban, A., ChemPhysChem 2010, 11 (11), 2290-2304.
37. od, I. Gonz lez-Pedro, V.; Tachan, Z. Fabregat-Santiago, F. Mora-Ser •, I.;
Bisquert, J.; Zaban, A., The Journal of Physical Chemistry Letters 2011, 2, 3032-3035.
38. Kramer, I. J.; Sargent, E. H., ACS Nano 2011, 5 (11), 8506-8514.
39. Tang, J.; Sargent, E. H., Advanced Materials 2010, 23 (1), 12-29.
40. Archer, M. D.; Nozik, A. J., Nanostructured and Photoelectrochemical Systems
for Solar Photon Conversion. Imperial College Press: London, 2008; Vol. 3.
Katherine Roelofs Ph.D. Dissertation
47
41. Bakulin, A. A.; Neutzner, S.; Bakker, H. J.; Ottaviani, L.; Barakel, D.; Chen, Z.,
ACS Nano 2013, 7 (10), 8771-8779.
42. Tvrdy, K.; Frantsuzov, P. A.; Kamat, P. V., Proceedings of the National Academy
of Sciences 2011, 108 (1), 29-34.
43. Mora-Seró, I.; Bisquert, J., The Journal of Physical Chemistry Letters 2010, 1
(20), 3046-3052.
44. Guijarro, N. s. Shen, Q. Gim nez, S. Mora-Ser , I. n. Bisquert, J. ana-
Villarreal, T. Toyoda, T. G mez, R., The Journal of Physical Chemistry C 2010, 114
(50), 22352-22360.
45. Bedja, I.; Hagfeldt, A., Advances in OptoElectronics 2011, 2011 (Article ID
424071), 5.
46. Leventis, H. C.; O'Mahony, F.; Akhtar, J.; Afzaal, M.; O'Brien, P.; Haque, S. A.,
Journal of the American Chemical Society 2010, 132 (8), 2743-2750.
47. O'Mahony, F. T. F.; Lutz, T.; Guijarro, N.; Gomez, R.; Haque, S. A., Energy &
Environmental Science 2012, 5 (12), 9760-9764.
48. Tennakone, K.; Kumara, G. R. A.; Kumarasinghe, A. R.; U., W. K. G.;
Sirimanne, P. M., Semiconductor Science and Technology 1995, 10, 1689.
49. Roelofs, K. E.; Brennan, T. P.; Dominguez, J. C.; Bailie, C. D.; Margulis, G. Y.;
Hoke, E. T.; McGehee, M. D.; Bent, S. F., The Journal of Physical Chemistry C 2013,
117 (11), 5584-5592.
50. Robel, I.; Kuno, M.; Kamat, P. V., Journal of the American Chemical Society
2007, 129 (14), 4136-4137.
51. Hetsch, F.; Xu, X.; Wang, H.; Kershaw, S. V.; Rogach, A. L., The Journal of
Physical Chemistry Letters 2011, 2 (15), 1879-1887.
52. Guyot-Sionnest, P., The Journal of Physical Chemistry Letters 2012, 3 (9), 1169-
1175.
53. Mora-Ser , I. Gim nez, S.; Fabregat-Santiago, F. G mez, R.; Shen, Q.; Toyoda,
T.; Bisquert, J., Accounts of Chemical Research 2009, 42 (11), 1848-1857.
54. Chakrapani, V.; Baker, D.; Kamat, P. V., Journal of the American Chemical
Society 2011, 133 (24), 9607-9615.
55. Zhang, Q.; Guo, X.; Huang, X.; Huang, S.; Li, D.; Luo, Y.; Shen, Q.; Toyoda, T.;
Meng, Q., Physical Chemistry Chemical Physics 2011, 13 (10), 4659-4667.
56. Kamat, P. V., Accounts of Chemical Research 2012.
57. Hines, D. A.; Kamat, P. V., The Journal of Physical Chemistry C 2013, 117 (27),
14418-14426.
58. yun, B.-R. hong, .-W. Bartnik, A. C. Sun, . Abru a, H. D.; Wise, F. W.;
Goodreau, J. D.; Matthews, J. R.; Leslie, T. M.; Borrelli, N. F., ACS Nano 2008, 2 (11),
2206-2212.
59. Kemp, K. W.; Labelle, A. J.; Thon, S. M.; Ip, A. H.; Kramer, I. J.; Hoogland, S.;
Sargent, E. H., Advanced Energy Materials 2013, 3 (7), 917-922.
60. Goh, C.; Scully, S. R.; McGehee, M. D., Journal of Applied Physics 2007, 101
(11), 114503.
61. Shalom, M. R hle, S.; Hod, I.; Yahav, S.; Zaban, A., Journal of the American
Chemical Society 2009, 131 (29), 9876-9877.
62. Barea, E. M. Shalom, M. Gim nez, S.; Hod, I.; Mora-Seró, I.; Zaban, A.;
Bisquert, J., Journal of the American Chemical Society 2010, 132 (19), 6834-6839.
Katherine Roelofs Ph.D. Dissertation
48
63. Chi, C.-F.; Chen, P.; Lee, Y.-L.; Liu, I. P.; Chou, S.-C.; Zhang, X.-L.; Bach, U.,
Journal of Materials Chemistry 2011, 21 (43), 17534-17540.
64. Talapin, D. V.; Rogach, A. L.; Kornowski, A.; Haase, M.; Weller, H., Nano
Letters 2001, 1 (4), 207-211.
65. Kramer, I. J.; Sargent, E. H., Chemical Reviews 2013.
66. Dong, A.; Ye, X.; Chen, J.; Kang, Y.; Gordon, T.; Kikkawa, J. M.; Murray, C. B.,
Journal of the American Chemical Society 2011, 133 (4), 998-1006.
67. Liu, Y.; Gibbs, M.; Puthussery, J.; Gaik, S.; Ihly, R.; Hillhouse, H. W.; Law, M.,
Nano Letters 2010, 10 (5), 1960-1969.
68. Koleilat, G. I.; Levina, L.; Shukla, H.; Myrskog, S. H.; Hinds, S.; Pattantyus-
Abraham, A. G.; Sargent, E. H., ACS Nano 2008, 2 (5), 833-840.
69. Talapin, D. V.; Murray, C. B., Science 2005, 310 (5745), 86-89.
70. Johnston, K. W.; Pattantyus-Abraham, A. G.; Clifford, J. P.; Myrskog, S. H.;
Hoogland, S.; Shukla, H.; Klem, E. J. D.; Levina, L.; Sargent, E. H., Applied Physics
Letters 2008, 92 (12), 122111.
71. Tang, J.; Kemp, K. W.; Hoogland, S.; Jeong, K. S.; Liu, H.; Levina, L.;
Furukawa, M.; Wang, X.; Debnath, R.; Cha, D.; Chou, K. W.; Fischer, A.; Amassian, A.;
Asbury, J. B.; Sargent, E. H., Nature Materials 2011, 10 (10), 6.
72. Dibbell, R. S.; Watson, D. F., Journal of Physical Chemistry C 2009, 113, 3139-
3149.
73. Ramasamy, P.; Kang, M.-S.; Cha, H.-J.; Kim, J., Materials Research Bulletin
2013, 48 (1), 79-83.
74. Kay, A.; Grätzel, M., Chemistry of Materials 2002, 14 (7), 2930-2935.
75. Tsujimoto, K.; Nguyen, D.-C.; Ito, S.; Nishino, H.; Matsuyoshi, H.; Konno, A.;
Kumara, G. R. A.; Tennakone, K., The Journal of Physical Chemistry C 2012, 116 (25),
13465-13471.
76. Giménez, S.; Mora-Seró, I.; Macor, L.; Guijarro, N.; Lana-Villarreal, T.; Gómez,
R.; Diguna, L. J.; Shen, Q.; Toyoda, T.; Bisquert, J., Nanotechnology 2009, 20 (29),
295204-295210.
77. Oja Acik, I.; Katerski, A.; Mere, A.; Aarik, J.; Aidla, A.; Dedova, T.; Krunks, M.,
Thin Solid Films 2009, 517 (7), 2443-2447.
78. Brennan, T. P.; Trejo, O.; Roelofs, K. E.; Xu, J.; Prinz, F. B.; Bent, S. F., Journal
of Materials Chemistry A 2013, 1 (26), 7566-7571.
79. Diguna, L. J.; Shen, Q.; Kobayashi, J.; Toyoda, T., Applied Physics Letters 2007,
91 (2), 023116-023113.
80. Shen, Q.; Kobayashi, J.; Diguna, L. J.; Toyoda, T., Journal of Applied Physics
2008, 103 (8), 084304-084305.
81. Kim, J.; Choi, H.; Nahm, C.; Kim, C.; Nam, S.; Kang, S.; Jung, D.-R.; Kim, J. I.;
Kang, J.; Park, B., Journal of Power Sources 2012, 220 (0), 108-113.
82. Yang, S.-m.; Huang, C.-h.; Zhai, J.; Wang, Z.-s.; Jiang, L., Journal of Materials
Chemistry 2002, 12 (5), 1459-1464.
83. Guijarro, N.; Campina, J. M.; Shen, Q.; Toyoda, T.; Lana-Villarreal, T.; Gomez,
R., Physical Chemistry Chemical Physics 2011, 13 (25), 12024-12032.
84. Choi, H.; Nicolaescu, R.; Paek, S.; Ko, J.; Kamat, P. V., ACS Nano 2011, 5 (11),
9238-9245.
Katherine Roelofs Ph.D. Dissertation
49
85. Roelofs, K. E.; Brennan, T. P.; Bent, S. F., The Journal of Physical Chemistry
Letters 2014, 5 (2), 348-360.
Katherine Roelofs Ph.D. Dissertation
50
Chapter 3. Experimental Methods
This chapter presents in detail the experimental methods employed in this work.
First, the solar cell fabrication procedure is described for both quantum-dot-sensitized
solar cells (QDSSCs) and perovskite solar cells (PSCs). Here, the goal is to set down the
fabrication process in enough detail as to allow complete replication, and to provide
guidance on which steps are important and why. Next, the material characterization
methods are described, along with a brief summary of the underlying theory where
relevant, and the specific experimental details of their use in these studies. Finally, the
electrical characterization methods are described, including the in-house set-up for
measuring external quantum efficiency, and transient photovoltage and photocurrent.
3.1. Solar Cell Fabrication
3.1.1. Substrate Preparation
F:SnO2 (FTO), Sn:In2O3 (ITO), and Al:ZnO (AZO) are transparent conductive
oxides commonly used as the substrates for solar cell fabrication (Table 3-1). FTO is
used in these studies due to its higher thermal stability, as some of the fabrication steps
require annealing at temperatures of >450°C. ITO is also of interest as its conductivity is
roughly twice that of FTO and it has a smoother surface. The surface roughness of FTO
can be on the order 200 nm, whereas some layers in the device are only tens of
nanometers thick. FTO can be easily etched in patterns with a simple masking procedure
using Scotch or Kapton© tape. We take advantage of this property to help define the
solar cell active area.
The substrate preparation procedure is shown in Figure 3-1. This procedure has
evolved through the course of this thesis work and what is listed here are the best
practices; slight deviations that were used in earlier studies are detailed in the
experimental methods of each chapter. FTO (TEC 7 or TEC 15, Hartford Glass) is first
masked with Scotch tape. Zn powder is sprinkled on the unmasked areas, and 12 vol %
Katherine Roelofs Ph.D. Dissertation
51
HCl in DI water is deposited on the unmasked areas. Bubbles of H2 gas will form in the
etchant liquid; etching is complete when the bubbling finishes, after roughly 1 to 2
minutes. The masking tape is removed under a stream of DI water to prevent the
remaining HCl from etching the unmasked areas as the tape is removed. The substrates
are then cleaned. The first cleaning step is in soap – by sonication for 15 min in either a
beaker with 1:10 laboratory-grade soap to DI H2O or in a beaker with 1:9 Extran to DI
H2O to remove any dust or water-soluble materials on the surface. Next, the substrates
are either rinsed in DI water for 5 min (with a running water source to continue to replace
the beaker volume), or sonicated in DI water for 15 min, depending on preference. The
sonication can assist with removal of all the soap from the substrate surface. The next
sonication step is carried out in organic solvents. This can either be done by a single
sonication step in ethanol for 15 min, or two sonication steps: first in acetone for 15 min
and then in isopropyl alcohol for 15 min. The substrate is then rinsed in a beaker of DI
H2O, removed, and blown dry with nitrogen. At this stage, the substrates can be left for a
few weeks in a sealed container.
Table 3-1. Electrical and material properties of common transparent conductive oxides.1
TCO material
Highest
conductivity
(S/cm)
Carrier
concentration
(cm-3
)
Mobility
(cm2/Vs)
Work
function
(eV)
Thermal
stability in
air (°C)
F: SnO2 (FTO) ~1 x 103 4 x 10
20 ~30 4.9 <700
Sn:In2O3 (ITO) ~1 x 104 ~10
21 ~40 4.7 <350
Al:ZnO (AZO) ~7 x 103 1.5 x 10
21 50-100 4.6 <400
Immediately prior to compact-TiO2 layer deposition, the substrates are cleaned by
exposure to a UV-ozone source for at least 20 min to remove any adventitious carbon or
organic contaminants. The compact TiO2 layer can be deposited by spray pyrolysis or by
atomic layer deposition (ALD); the ALD procedure is described below. The spray
pyrolysis is conducted using a spray-gun system, shown in Figure 3-2. A precursor
mixture of 1:9 volume ratio of titanium diisopropoxide bis(acetylacetonate) (Aldrich,
#325252) to ethanol is used. This solution is only stable for a week, after which the
Katherine Roelofs Ph.D. Dissertation
52
bright pale yellow color darkens to a deep yellow. The substrates are heated to 510 °C,
then the precursor solution is sprayed onto the substrates, carried by N2 gas, from a
distance of roughly 1 to 1.5 feet, with eight sprays, each 20 seconds apart. This process
should produce a ~ 50 nm thick layer of TiO2.2 To preserve the airbrush tip, the tip
should be cleaned immediately by spraying with pure ethanol into a waste container,
followed by sonication for 5 min in ethanol. It is not necessary to mask during the
deposition of the compact TiO2 layer, as an ultrasonic welder is used during the
application of solder to make contact with the FTO or ITO. The ultrasonic welder will
cut through the compact TiO2 layer.
Figure 3-1. (a) Large FTO substrate capable of producing 10 individual solar cell
substrates. (b) Masking with Scotch tape. (c) Zn and HCl etch removes the FTO from
un-masked area in the center. (d) Substrate is cleaned with soapy water and then
solvents. (e) After ozone-cleaning, the entire substrate is coated with a compact TiO2
layer, deposited by spray-pyrolysis or ALD. (f) For nanoporous architectures, a
nanoporous TiO2 layer is deposited. (g) The substrate is finished, and can be divided into
individual cells. (h) Complete device for the case of a quantum-dot-sensitized solar cell,
Katherine Roelofs Ph.D. Dissertation
53
with a hole-transport material (HTM) and top metal contact, showing the active area set
by the original FTO etch and the metal contact on top.
Figure 3-2. Spray pyrolysis set-up. (a) TiCl4 solution is a 1:9 volume ratio of titanium
diisopropoxide bis(acetylacetonate) to ethanol. The purple circles show the tip of the
solution-delivery bottle, and where it attaches to the spray-pyrolysis system. (b) The
spray-gun, with a yellow arrow indicating the N2 incoming line, a yellow circle showing
where the line attaches to the spray gun, and a red circle indicating the nob that turns the
N2 flow on or off. (c) Spray-pyrolysis onto a hot plate, using a reflective sleeve to protect
the hand and arm. Images courtesy of Colin Bailie.
For solar cells with a planar architecture, the next step is TiCl4 treatment,
followed by deposition of the absorber. For solar cells with a nanoporous architecture,
the nanoporous TiO2 layer is deposited next, followed by the TiCl4 treatment, and then
the deposition of the absorber. For the TiCl4 treatment, a 40 mM solution of TiCl4 in
ultrapure DI water is created by dilution from a 2 M TiCl4 aqueous stock solution stored
in the freezer. The substrate is placed in the diluted TiCl4 solution and either left at room
temperature overnight in the dark, or heated in an oven at 70 °C for 30 min. After rinsing
in DI water, the substrate is then annealed at 450 °C for at least 2 hrs. The TiCl4
treatment deposits an ultra-pure TiO2 layer on top of the TiO2 anode, by exposure to an
acidic solution in which the TiO2 substrate is slightly etched (and roughened in the
process), while the TiCl4 in H2O also re-deposits TiO2 on the surface by chemical bath
deposition, and has been shown to slightly lower the TiO2 conduction band.3-5
The
impact of TiCl4 treatment has been studied in dye-sensitized solar cell (DSSC), 3-6
quantum-dot-sensitized solar cell (QDSSC),7, 8
and perovskite solar cell (PSC)9-11
fields.
Nanoporous TiO2 can be deposited by doctor blading or by spin-coating. The
doctor blading procedure was used for the quantum-dot-sensitized solar cell (QDSSC)
studies, and the spin-coating procedure, which is effective in producing a thinner film,
Katherine Roelofs Ph.D. Dissertation
54
was used for the perovskite solar cell (PSC) studies. The nanoporous TiO2 paste used is
the Dyesol paste (18NR-T), which by weight is roughly 10-30% titanium dioxide, 5-15 %
ethyl cellulose, 50-70 % terpineol, and 5-20% organic plasticisers, according to Dyesol
technical customer service. The ethyl cellulose binder material helps to form the
nanoparticle sponge structure of the final fired layer, as it is burnt out at 420 °C, leaving
behind the TiO2 structure that will sinter at 520 °C. The TiO2 particles have an anatase
crystal structure and are roughly 18 nm in diameter.
For doctor blading, the as-received Dyesol paste is further diluted in a 1:1 weight
ratio with terpineol. Doctor-blading is performed on the large substrates, prior to
dividing up into individual solar cells. The doctor blading machine moves a raised blade
slowly over the surface of the sample, pushing before it a well of the TiO2 paste, and
leaving behind a fairly uniform coating. With the 1:1TiO2 paste:terpineol dilution, films
of roughly 2.2 μm in thickness are achieved, with variations in thickness from 2 μm to
2.4 μm within the central regions of the film, where the solar cell active layer is formed.
For spin-coating of the nanoporous TiO2 layer, the substrate is first divided into
individual cells that are 20 mm by 15 mm. The as-received TiO2 Dyesol paste (18NR-T)
is diluted 1:3::TiO2 paste:ethanol by weight. Spincoating is performed at 4000 rpm for
30 sec to achieve a film thickness of roughly 350 nm. Coated substrates are then heated
to 450 °C for at least 2 hrs. The nanoporous films were then TiCl4 treated, as discussed
above. Transmission electron microscopy (TEM) analysis of nanoporous TiO2 films
deposited on Si wafers which results in thicker film deposition, shown in Figure 3-3,
confirms that the films are composed of anatase TiO2 particles, as also confirmed by
XRD results shown in Chapter 5.
Katherine Roelofs Ph.D. Dissertation
55
Figure 3-3. Transmission electron microscopy (TEM) images of a nanoporous TiO2 film
completed by doctor blading on a Si wafer. (a) Cross-section of the film showing details
of the film structure, with the scale bar indicating 1 μm. (b) Close view of an anatase
TiO2 nanoparticle, with scale bar indicating 5 nm. The nanoparticle lattice spacing is
3.58 Å, which corresponds to the lattice constant a in anatase TiO2.
3.1.2. QD Deposition
QDSSCs are fabricated using the nanoporous TiO2 substrates deposited by doctor
blading, which are roughly 2.2 microns thick. Two methods were used for the deposition
of the metal sulfide quantum dots (QDs), which act as the absorber. One is successive
ion layer adsorption and reaction (SILAR), a solution based deposition process that was
first introduced by Nicolau in 1985,12
and was first used to deposit thin films of metal
sulfides.13
As depicted in Figure 3-4, a substrate is first dipped into a solution of the
desired metal cation, then a rinsing step, followed by a solution containing the counter
anion, and a final rinsing step, completing a single SILAR cycle. The growth occurs in a
layer-by-layer process, or by island-type or cluster growth,14
and as deposition is
typically conducted at room temperature, ion exchange within the deposited film is
limited, and SILAR can be used to grow multilayer films of different materials.15
SILAR
was first used to produce semiconductor nanocrystals, or QDs, in 2003 by Li and
colleagues.16
The other QD deposition method is by ALD,17-19
where H2S gas is used as
the counter-reactant to produce metal sulfide nanoparticles.
Katherine Roelofs Ph.D. Dissertation
56
Figure 3-4. Schematic of successive ion layer adsorption and reaction (SILAR), for the
deposition of CdS as an example. (a) The dip and rinse steps involved in one SILAR
cycle, and (b) the adsorption of the cadmium cation and then sulfur anion during this
process.
Cadmium sulfide and lead sulfide quantum dots were studied in this work. For
SILAR deposition of the CdS QDs, a Cd(SO4) aqueous solution is used to deliver the
cadmium cation, paired with a Na2S aqueous solution to deliver the sulfur anion. For
SILAR of PbS QDs, a Pb(NO3)2 aqueous solution is used to deliver the lead cation, and a
Na2S solution to deliver the sulfur anion. For this work, only PbS QDs were grown by
ALD, using bis(2,2,6,6-tetramethyl-3,5-heptanedionato)lead(II) (sublimated at 140 ºC)
and a gas mixture of 3.5% H2S in N2 at room temperature. The substrate temperature was
maintained at 160 °C.
Figure 3-5. CdS QDs and PbS QDs grown with varying number of SILAR cycles on
nanoporous TiO2 films on microscopy slides. The CdS QD picture was taken with back-
lighting to emphasize the change in film color with SILAR cycle, whereas an opaque
Katherine Roelofs Ph.D. Dissertation
57
background was sufficient to observe the change in PbS QD film coloration with SILAR
cycle.
3.1.3. Perovskite Deposition
Methyl ammonium lead iodide (CH3NH3PbI3) perovskite layers are deposited as
the absorber material in perovskite solar cells. These perovskite solar cells were made in
both the planar configuration, with the perovskite deposited directly on the compact TiO2
layer, and in the nanoporous configuration, with the perovskite deposited on the spin-
coated nanoporous TiO2 layer (~350 nm thick). The perovskite layer was synthesized by
the two-step deposition process put forth by Grätzel and colleagues,20
and modified by
Hagfeldt and colleagues.21
The perovskite deposition is done completely in a N2
atmosphere glovebox. First, a 1.3 M PbI2 (Sigma Aldrich) in dimethylformamide (DMF)
solution is made, by heating to 100 °C to dissolve the PbI2. A 10 mg/mL solution of
methylammonium iodide (Dyesol) in isopropyl alcohol (IPA) is made at room
temperature. The glovebox atmosphere is then purged with N2 for half an hour prior to
perovskite deposition, in order to clear out the solvent atmosphere. Deposition is
continued once the atmosphere H2O content is below 10 ppm and the O2 content is below
2 ppm.
The TiO2-coated FTO substrates are heated to above 450 °C to drive off surface
H2O, and brought into the glovebox. The substrates are pre-heated to 70 °C immediately
prior to PbI2 deposition. The PbI2 in DMF solution is maintained at 100 °C, and is
deposited on the substrates by spin-coating at 6,500 rpm for 90 s, on a spin-coater with a
plastic chuck, which helps maintain the substrate temperature. The glovebox is purged
for 5 minutes, between every couple substrates, or even after every single substrate to
remove the DMF from the atmosphere, as DMF has strong effects on the final perovskite
film.22
A rougher PbI2 film (hazier in appearance, as shown in Figure 3-6a) can be made
by either cooling the PbI2 solution, or by intentionally increasing the DMF in the
glovebox atmosphere. A rougher PbI2 film is hypothesized to promote the
methylammonium iodide (MAI) incorporation, leading to more complete conversion to
the perovskite, although a too-rough film may not be ideal for the finalized solar cell as it
could promote the formation of pinholes. The PbI2 films are then dried on a hot plate at
Katherine Roelofs Ph.D. Dissertation
58
70 °C for between 15 and 30 min, removed and allowed to come to room temperature,
and then dipped in the room temperature solution MAI in IPA for between 20 and 30
min, at which point the film becomes dark (Figure 3-6b and c). The substrates are then
rinsed in IPA for 10 s, and spin-coated dry at 4000 rpm for 30 s. The substrates are then
further dried on the hot plate at 70 °C for between 15 and 30 min. The deposition
parameter range given here is further optimized for each specific architecture, as
described in detail later in the methods section of the individual chapters.
Figure 3-6. (a) PbI2 films formed by spin-coating the substrate with a 1.3 M PbI2 in
DMF solution. The effect of increasing DMF concentration in the glovebox atmosphere
is shown, with the PbI2 film becoming noticeably rougher, resulting in a more opaque
yellow color, and ultimately brownish edges from scattered light. (b) A CH3NH3PbI3
perovskite film after dipping in MAI/IPA solution, and (c) a similar CH3NH3PbI3
perovskite film held up to a light source.
3.1.4. Hole-Transport Material and Metal Contact Deposition
The primary hole-transport material (HTM) used in this work is the small organic
molecule spiro-OMeTAD (2,2',7,7'-tetrakis-(N,N-di-p-methoxyphenylamine)-9,9'-
spirobifluorene). To deposit this layer, a 225 mg mL-1
solution of spiro-OMeTAD
(Lumtec) was dissolved in chlorobenzene, with tert-Butylpyridine added at a ratio of
1:10.3 μ :mg of spiro-OMeTAD, and lithium bis-(trifluoromethylsulfonyl)imide salt
(170 mg mL-1
in acetonitrile) added at a ratio of 1:4.8 μ :mg of spiro-OMeTAD. A small
amount of the spiro-OMeTAD solution (30 μ for 3.75 cm2 substrates) was deposited
onto the TiO2 substrates at room temperature, and spin-coated at 2000 RPM for 30 s.
For the QDSSCs, Ag metal contacts are used as the standard top contact. For the
PSCs, Au metal contacts are used. In either case of Ag or Au, a 200 nm thick layer is
Katherine Roelofs Ph.D. Dissertation
59
deposited by thermal evaporation under vacuum below 10-6
torr. For metal evaporation
on QDSSCs, an evaporator system built by George Burkhard was used, which consists of
a bell-jar system with a roughing and turbo pump attached, situated within a glove box.
Metal ingots are placed on a tungsten boat that is attached to two metal posts; voltage is
applied across the posts, running current through the boat, and heating the metal igot until
it evaporates. A quartz crystal is used to monitor the evaporation rate and total thickness.
For the work on perovskite solar cells, a thermal evaporator built by Axel Palmstrom was
used which largely replicated the previous one. Final device areas were on the order of
0.1 cm2. Devices were stored in a desiccator prior to electrical measurements. A solder
connection is made to the underlying transparent conductive oxide using an ultrasonic
welder. Flux-free solder (Cerasolzer) is used. SEM images of completed devices are
shown in Figure 3-7.
Figure 3-7. (a) Scanning electron microscope (SEM) cross-sectional image of a
completed QDSSC device, taken in a region where the FTO has been etched away. The
spiro-OMeTAD layer is too thick in this device. (b) SEM image of a device region with
FTO, shown for comparison. Adapted with permission, courtesy of I-Kang Ding.2
3.1.5. Atomic Layer Deposition of TiO2 and NiOx
ALD-grown TiO2 was used as the compact layer in these devices, as a
replacement for the TiO2 grown by spray-pyrolysis. For ALD TiO2 deposition, either
TiCl4 or titanium isopropoxide (TIPS) can be used as the precursor, with H2O as the
Katherine Roelofs Ph.D. Dissertation
60
counter-reactant. The TiCl4 process was conducted in the home-built reactor referred to
as “Nubird”,23, 24
and the TIPS process was carried out in the commercial Arradiance
GEMStar benchtop ALD. Details of the TiO2 crystal phase produced by either of these
processes, as a function of deposition parameters and substrate, are included in Chapter 7.
For each process, pulse/purge times of <1 sec/<30 sec are used for the precursor, and
pulse times of <1 sec with purge times of 25 to 45 sec are used for the H2O. The TIPS
precursor is heated to 65 °C during deposition, while the TiCl4 precursor is kept at room
temperature. In Figure 3-8, the growth characteristics of the ALD TiO2 using the TiCl4
process are shown. For TiCl4, the TiO2 growth rate is roughly 0.4 to 0.5 Å/cycle. For
TIPS, the TiO2 growth rate is roughly 0.3 Å/cycle. For ALD-grown compact TiO2 layers,
generally a 10 nm thick layer of TiO2 is used. For further details on device performance
as a function of compact TiO2 layer thickness, see Chapter 7.
Figure 3-8. Details of the compact TiO2 layer growth by TiCl4 and H2O. Film
thicknesses measured by ellipsometry on ozone-cleaned Si substrates. (a) Film thickness
as a function of ALD cycles, with varying stage temperatures. In one study, the samples
deposited at 300 °C were annealed in air at 450 °C for 2 hrs on a high-temperature hot
plate. (b) Corresponding growth rates as a function of ALD cycle, calculated by dividing
the overall film thickness by the number of ALD cycles.
In the perovskite solar cells, NiOx was explored as a hole transport material, in
both the standard n-i-p structure, with the NiOx deposited in the place of spiro-OMeTAD
on top of the perovskite, and in the inverted p-i-n structure, with the NiOx deposited on
Katherine Roelofs Ph.D. Dissertation
61
the substrate prior to perovskite deposition. The NiOx ALD process used was first
developed in our lab by Katie Nardi.25
Bis(cyclopentadienyl)nickel, also known as
nickelocene, is heated to 70 °C, and introduced into the chamber by a vapor-drawn
process, rather than by being actively carried by N2.
In the vapor-drawn process, N2 enters the nickelocene precursor line going to the
main body of the ALD chamber at a point that is after the point where nickelocene enters.
A valve in between these two points is opened, and the valve connecting the whole
precursor line to the main chamber is opened, and the nickelocene vapor that enters the
line is passively drawn into the chamber by the N2 flow. This introduces less nickelocene
precursor into the chamber than the active N2 carrier process, in which N2 is introduced
prior to the nickelocene entrance, and thus flows past the opening to the nickelocene
bubbler on its way into the main chamber. The benefit of introducing less nickelocene by
the vapor-drawn process is that we have observed too-high amounts of Ni in the finished
film creates a white, roughened film. This is thought to be due to a different morphology
that occurs with too much Ni. This hazy NiOx film appearance (most clearly seen on Si
substrates) can occur immediately after the precursor is refilled, or if the nickelocene
pulse times are too long, or if too much nickelocene is drawn actively into the chamber.
The counter-reactant is ozone. Ozone is carried into the chamber by O2 gas, with an
overall ozone concentration of roughly 20 vol %. For NiOx grown on the transparent
conductive oxide substrate, the pulse sequence used is nickelocene pulse/purge of 1 s/30
s, and ozone pulse/purge of 3 s/30 s. For NiOx grown on top of the perovskite, the ozone
pulse/purge sequence was modified to 1 s/25 s, and the ozone concentration was varied,
as described in greater detail in the experimental section of chapter 7. Characterization of
the NiOx growth is included in Appendix B, including saturation curves and studies of the
temperature window of growth.
3.2. Materials Characterization
Materials characterization techniques include a range of spectroscopy techniques
for chemical composition, diffraction techniques to determine crystal structure, and
microscopy techniques to determine morphology (as well as composition and crystal
Katherine Roelofs Ph.D. Dissertation
62
structure when paired with in-situ spectroscopy or diffraction). In addition, synchrotron
methods are used, including using synchrotron X-rays for grazing-incidence X-ray
diffraction (GIXRD) and X-ray absorption spectroscopy (XAS). Below is a description
of the specifics of how these techniques are applied to analyze the materials used in this
work, as well as notes on their underlying theory when relevant.
3.2.1. Spectroscopy, Diffraction, and Microscopy
UV-Vis Spectroscopy is used to determine the absorption properties of the metal
sulfide QDs and lead-iodide based perovskite. Through Tauc Analysis of the UV-Vis
spectra, the material band gap can be determined, and when paired with literature studies
correlating QD band gap to nanocrystal size for a given material, can be used to
determine the quantum dot crystal size. For the quantum dots, Tauc Analysis is
susceptible to interpretation as to where the tangent line is drawn, due to the extended
curvature of the absorption onset, which is caused both by the distribution in QD size
(and thus band gap) in a given substrate, as well as sub-band gap absorption from defect
states in the QDs.
X-ray photoelectron spectroscopy (XPS) is used to determine the atomic
concentration at the surface of films. XPS has an analysis depth of ~ 1.5 nm, and a spot
size of up to several microns, with the sensitivity varying exponentially with depth. XPS
detection limits are 0.1 to 1 atomic %, meaning that if over 1 atomic % of an element is
present in the film surface, it should be detected. For routine XPS measurements, the
accuracy of the atomic % values is roughly 90 to 95 % of the atomic % of the major
peaks (i.e., an element quoted as 90 atomic % could in reality be between 91 and 100
atomic %, while an element quoted as 10 atomic % could in reality be between 9 and 11
atomic %).26
The primary XPS spectral peaks used for elemental identification and
atomic concentration are the photoelectron peaks. To determine atomic concentration,
the area (#electrons/sec) of a strong photoelectron peak for a specific element is divided
by the sensitivity factor to normalize it. The normalized peak areas are then directly
proportional to the number of surface atoms of that element in the sample. Other XPS
spectral peaks include those caused by the photoelectron effect from defined electron
levels in the sample atoms, plasmon peaks which are broader peaks associated with
Katherine Roelofs Ph.D. Dissertation
63
strong photoelectron peaks, and Auger peaks that are even broader, more complex peaks
at energies characteristic of a particular element. There are also satellite peaks a few eV
higher in binding energy than photoelectron lines; these are present in certain valence
states of a few elements. Furthermore, the XPS photoelectron peaks can undergo
chemical shifts of up to a few eV, which can give information about the binding
environment of the element, in particular its oxidation state.27
Figure 3-9. Schematic of the photoelectron effect, showing an incident photon with
energy hv striking an electron in atom, causing the electron to leave the atom with a
kinetic energy, KE.
XPS is based on the photoelectron effect, in which an electron is excited by an
incident photon with an energy greater than the binding energy of that electron, causing
the electron to leave the sample with some kinetic energy. The detector measures the
kinetic energy (KE) of the escaping electron, and counts the number at a given energy
level. From KE, the binding energy (BE) of the electron can be determined by Equation
1, where hv is the energy of the incident photons.28
(1)
The rising background under a photoelectron peak is due to excited electrons that
scattered and lost energy before exiting the sample, but those with no energy loss create a
well-defined peak that can be used to track the binding energy. For the ultrathin barrier
layers of Al2O3 studied in Chapter 4 and Chapter 5, XPS can detect the aluminum in
addition to the titanium from the underlying substrate. The amount of Al detected can be
used to give a semi-quantitative measure of the thickness of these layers. To fully
quantify the amount of aluminum, a detailed study using varying incident X-ray angles
Katherine Roelofs Ph.D. Dissertation
64
would be necessary.29
In addition, XPS can be used as a qualitative check of the
conformality of ALD growth on a substrate, in addition to checking the sticking of
precursors to the substrate (i.e., if there is a nucleation delay). Here, an ALD layer is
grown just thick enough that no Si signal is seen in XPS from the Si reference wafer.
ALD growth on the substrate of interest is also checked, and provided the substrate is
flat, no signal from the underlying substrate should be seen if the nucleation delay on the
substrate matches that of the Si wafer.
Figure 3-10. Schematic of the Auger effect. (a) An incident electron collides with an
electron in an inner shell, the 1s shell as shown here, ejecting a 1s electron. (b) The hole
in the inner shell is then filled by a higher energy shell electron, from the 2s shell as
shown here. The energy released by the 2s 1s transition is absorbed by an electron in
an outer shell, the 2p shell as shown here, causing that electron to leave. This third
exiting electron is called the Auger electron. (c) A view of the transitions involved in the
process from an energy level diagram.
Auger electron spectroscopy (AES), like XPS, can give the elemental make-up of
the material.30
AES is based on the Auger effect, and for the AES machine used for this
work, the sample was hit with electrons as the source. This gives a much higher area
resolution than can be achieved with XPS, since electrons have a shorter wavelength than
X-rays, and can thus be more tightly focused. AES has an area resolution of roughly 5-
10 nm, and an analytical depth of a few nm. Figure 3-10 shows a schematic of the Auger
process due to an incident electron. An electron incident on the atom collides with an
inner-shell electron, ejecting it from the atom, and a mid-shell electron then relaxes down
to fill the hole left behind. This relaxation releases energy, which is absorbed by an
outer-shell electron termed the Auger electron, causing it to be excited above the energy
level of the vacuum, leading it to leave the atom. The kinetic energy of the Auger
Katherine Roelofs Ph.D. Dissertation
65
electron, EAuger, can be calculated from Equation 2, where ECore State is the binding energy
of the inner shell electron, EB is the binding energy of the mid-shell electron, and EC' is
the binding energy of the Auger electron, adjusted for after the removal of the mid-shell
electron.
(2)
By measuring the kinetic energy of the Auger electron, elemental assignments can thus
be made. In this work, AES is used for its small area resolution to track variations in
elemental composition across the cross-section of the solar cell (which are typically a few
microns thick).
In addition to these spectroscopic techniques, both X-ray and electron diffraction
are used to characterize the materials used in this work. Electron diffraction is conducted
in-situ during TEM analysis, and is discussed below. Powder X-ray diffraction (XRD)
can be used to determine the crystalline phase, grain size, lattice strain, chemical
composition, and crystal orientation present in a sample.31, 32
Here, the sample is exposed
to a monochromatic X-ray beam, and rotated using a 2θ or 2θ-ω scan, and the intensity of
the diffracted X-rays are measured by a detector located at a fixed angle. The resulting
diffracted X-ray intensity is plotted as a function of twice the incident angle, or 2θ. Peaks
in the diffraction data are observed when 2θ reaches a value such that Bragg’s law is
satisfied, which specifies the conditions under which the diffracted X-rays will
constructively interfering. In its simplified form, Bragg’s law is given by
(3)
where n is an integer, λ is the wavelength of the incident X-ray, dhkl is the spacing
between the planes, and θ is the angle of the incident X-ray, as seen in Figure 3-11. The
more generalized form of Bragg’s law is shown in equation 4,
(4)
where the scattering vector satisfying the Bragg condition, qB, is 2π times a reciprocal
lattice vector Ghkl. From the 2θ position at which a peak shows up, the lattice spacing
distance that caused this peak can be determined, and used to identify crystalline phases.
Katherine Roelofs Ph.D. Dissertation
66
XRD is used in this work mainly to analyze the crystal phase of ALD-grown metal
oxides, and to determine the phase composition of the CH3NH3PbI3 perovskite absorber
material.
Figure 3-11. Schematic of powder X-ray diffraction (XRD). (a) For a sample with
parallel planes of atoms, with a spacing distance dhkl, constructive interference applies
only under the conditions of Bragg’s law (Equation 3). (b) Experimental set-up of
powder XRD, where an X-ray incident on the sample at an angle θ is diffracted. The
scattering vector, q, is also shown.
Scanning electron microscopy (SEM) and transmission electron microscopy
(TEM) are used to observe film morphology and quantum dot morphology and crystal
structure. Electron microscopy has much higher resolution than light microscopy due to
the small de Broglie wavelength of electrons. The resolution of microscopes is given by
the Rayleigh criterion, in Equation 5,
(5)
where the resolution, δ, is proportional to the wavelength, λ, of the source divided by μ
the refractive index of the viewing medium and the sine of the semi-angle of collection,
β. SEM can image micron-sized areas, but has resolution limits due to the thickness of
the sample. In TEM, ultra-thin samples are used so that the transmitted electrons can be
collected to create the image.33
While TEM can produce extremely high-resolution
images, it is inherently limited as only an extremely small portion (<20 μm2) of the
sample can be analyzed at a given time. Indeed, only a few tens of nm2 can be viewed in
the high resolution TEM (HRTEM), where individual columns of atoms are resolved.
Therefore, TEM is best supplemented by a bulk technique, such as UV-Vis spectroscopy
or XRD. Care must be taking when observing PbS QDs in the TEM, as the PbS QDs can
be annealed by the electron-beam if they are over-exposed (on a timescale of roughly 30
Katherine Roelofs Ph.D. Dissertation
67
min), causing them to ball up. The CdS QDs studied in this work were much more stable
under TEM analysis.
3.2.2. Synchrotron Methods
Two techniques were used at the Stanford Synchrotron Radiation Lightsource
(SSRL) at the SLAC National Accelerator Laboratory: X-ray absorption spectroscopy
(XAS) and grazing-incidence X-ray diffraction (GIXRD). The benefit of and need for a
synchrotron source is that it provides monochromated X-rays with tunable energy and a
high flux.
X-ray absorption spectroscopy is an important technique, as it can be used to
determine local geometric and electronic structure of materials, without requiring the
presence of long-range order (as is necessary in XRD).34, 35
In addition, it can provide
information about scattered or embedded materials, which are difficult to analyze by
other techniques, such as TEM. In XAS, the sample is irradiated with X-rays, exciting
core electrons above the Fermi energy level, either to vacant states or above the vacuum
energy level, as shown in Figure 3-12. The absorption of X-rays is plotted as the incident
X-ray energy is varied. A strong rise in absorption, producing an edge, occurs when the
incident X-ray energy becomes large enough to excite an electron from a specific core
state. Past that edge, the electrons are excited above the vacuum level and have some
additional kinetic energy, with which that electron can scatter off neighboring atoms.
This scattering of the electron wave function causes constructive and destructive
interference, which leads to specific X-ray energies being able to be strongly absorbed,
while slight higher or lower X-ray energies are not as strongly absorbed, creating peaks
and troughs in the XAS spectra. In the near-edge X-ray absorption fine structure
(NEXAFS) or X-ray absorption near-edge structure (XANES) region, the kinetic energy
is low, and multiple scattering events dominate. The region around the peak edge (E0),
from energies of E0 - 10 eV to E0 + ~100 eV, is typically referred to as NEXAFS when
using soft (low-energy) X-rays, and XANES when using hard X-rays. In the extended
X-ray absorption fine structure (EXAFS) region, the kinetic energy is higher, and single
scattering by the nearest atomic neighbors normally dominates.
Katherine Roelofs Ph.D. Dissertation
68
Figure 3-12. Schematic of X-ray absorption spectroscopy (XAS), showing (a) the
sections of an XAS spectra. (b) In XAS, a X-ray is absorbed by exciting an electron.
The electron can be excited to a higher-energy vacant state, or excited above the vacuum
energy level, with some additional kinetic energy.
As the peaks in the NEXAFS and XANES region are due to the photoelectron
scattering off neighboring atoms, this region can give information about the local atomic
structure around a specific element. Analysis of these regions is still an emerging field.34,
36 However, useful information can be extracted by careful choice of control samples,
and by pairing the XAS measurements with density functional theory (DFT) models,
which can simulate XAS spectra for a given atomic structure.37
In this work, XAS
measurements were used to characterize PbS QDs grown on the TiO2 substrate, drawing
on previous studies with Pb38
and S39
XANES. XAS measurements were taken
beamlines (BLs) 2-1, 4-3, and 10-1 at SSRL. For the XAS measurements, multiple total
fluorescence yield (TFY) spectra were gathered for S K- (BL 4-3), Pb L3- (BL 4-3), O K-
(BL 10-1), and Ti L2,3-edge (BL 10-1). For each XAS measurement, a calibration spectra
was taken prior to and after that of the sample, allowing the sample spectra to be aligned.
The XAS spectra were aligned, averaged, and normalized using the SixPACK and
Athena analysis packages.
Katherine Roelofs Ph.D. Dissertation
69
GIXRD is similar to powder X-ray diffraction, where here the X-ray is incident
on the sample at a very low, grazing angle.40
However, now the X-ray source is no
longer the standard Cu K-alpha source, but rather an X-ray beam from the synchrotron
source. Accordingly, tabulated 2θ can no longer be used as a common metric for
determining d plane spacing from diffraction peaks, as the source energy is different from
that used to determine tabulated values in common crystallographic databases. GIXRD
data is collected as the diffracted X-ray intensity as a function of the scattering vector, q.
Tabulated 2θ diffraction peaks can be interconverted to the corresponding q value using
Equation 6, where θ1 and λ1 correspond to a specific X-ray energy, and θ2 and λ3
correspond to another.
(6)
GIXRD was leveraged in this work to study the crystallinity of ultra-thin films
deposited by ALD, as standard XRD methods do not have intense enough X-ray sources
to observe peaks for these films. GIXRD measurements were performed on Beamline
11-3 at the Stanford Synchrotron Radiation Lightsource (SSRL). The measurements
were taken with a MAR 345 imaging plate with 12.735 keV X-ray energy, and an
incident angle of 2°. In-situ annealing measurements were performed on Beamline 7-2 at
the SSRL. The measurements were taken with a Pilatus 100K detector with 12.398 keV
X-ray energy. The in-situ annealing chamber is constructed of aluminum and uses two
Comstat Low Voltage Cartridge Heaters (MCH2-40W-002) to heat the sample stage. A
Cryocon C24 Temperature Controller (Type K Thermocouple) is used to control the
temperature with a Sorensen power supply DLM 40-15. The chamber is hermetic with
gas flow in and out and has an integrated water cooling system. The windows are made
of 2mil (50.8um) thick Kapton Tape which is transparent to x-rays at the energies used.
The GIXRD data was analyzed with an in-house software WxWindows Diffraction
Integration Tool v1.15 (developed by Stefan Mannsfeld at SSRL). The data from the 2-D
area detectors at 11-3 or 7-2 was analyzed as follows: first it was calibrated using spectra
collected from a LaB6 crystal under identical conditions, then converted to qxy-qz,
followed by a conversion to qchi, and finally integrating over Chi to produce 1-D plot of
intensity vs. q.
Katherine Roelofs Ph.D. Dissertation
70
3.3. Electrical Characterization Techniques
Electrical measurements of deposited materials and the completed solar cells was
used to characterize solar cell performance, and gain insight into internal charge transfer,
and the working mechanisms of the device.
3.3.1. Current-Voltage Curves and Quantum Efficiency
Current-voltage (J-V) curves are the standard metric for measuring solar cell
performance, where an applied voltage is varied, while measuring the current. This
current is then divided by the active area of the solar cell to get the current density, J,
with units mA/cm2. Current multiplied by voltage has units of power, and the point along
the J-V curve that gives the largest J*V value is termed the maximum-power point, Pmax.
Figure 3-13. Current-voltage (J-V) curve of a solar cell under illumination (black) and
dark (grey dashed) conditions. The short-circuit current, JSC, is the current density with
no applied potential, the open-circuit voltage, VOC, is the voltage with zero net current,
the max power point, Pmax, is given by the current multiplied by the voltage at the
maximum power point. The fill factor, FF, is the ratio of Pmax divided by the power
value given by JSC multiplied by VOC, or the ratio of the purple square to the blue square.
The series resistance, Rseries, in the inverse of the slope at a forward bias, and the shut
resistance, Rshunt, is the inverse of the slope at reverse bias.
Katherine Roelofs Ph.D. Dissertation
71
The solar cell device metrics are shown in Figure 3-13, including the short-circuit
current (JSC), open-circuit voltage (VOC), fill factor (FF), series resistance (RS), and shunt
resistance (RSh).41
The efficiency (η) is given by Equation 7, where Pincident is the power
irradiated on the solar cell from the incident photons.
(7)
The device efficiency is the percent of power that could be produced out of the total
incident power from the sun. The JSC is a good metric to track the efficiency of charge
collection, whereas VOC is a good metric to track the maximum energy level at which
those charges can be extracted. The upper limit of VOC is set by the absorber band gap.
The fill factor is a metric to describe how square the J-V curve is, and is affected by many
things, including series and shunt resistance. RS is the resistance a charge faces when
moving across the layers in the solar cell stack, and is desired to be as low as possible.
RSh should be as high as possible, since a low shunt resistance means there is a short in
the device, such as a pinhole, allowing the top contact and bottom contact to touch.
Figure 3-13 shows a J-V curve collected under illumination (black trace) as well as under
dark conditions (grey dashed trace). The dark current provides important information
about the recombination current flow in the solar cell, absent any current generated by the
incident photons (photocurrent). The current density under illumination is the sum of the
dark current and the photocurrent. Given that the photocurrent should be fairly
monotonic with voltage, the shape of the illuminated J-V curve mirrors that of the dark
curve.
For both QDSSCs and PSCs, collecting J-V curves requires some thought as to
what is the real efficiency metric to be measured. QDSSCs show light-soaking behavior
also common in DSSCs, where the efficiency improves over time under illumination. It
is believed in DSSCs with liquid electrolytes that the light-soaking effect is due to a
downward shift in the TiO2 conduction band, which improves electron injection
efficiency.42-44
In solid-state DSSCs with spiro-OMeTAD as the HTM, it is believed that
illumination enhances the doping of spiro-OMeTAD by the commonly-added chemical
dopant lithium bis(trifluoromethanesulfonyl)imide (LiTFSI). LiTFSI does not directly
Katherine Roelofs Ph.D. Dissertation
72
oxidize, and thus dope p-type, the spiro-OMeTAD, but rather promotes the oxidation of
spiro-OMeTAD by O2 from the air in the presence of either light or thermal excitation.10,
45, 46 In DSSCs, this process typically is on the order of 10 minutes. However, in the
solid-state CdS QDSSCs discussed in Chapter 4, the enhancement in efficiency, driven
by an increase in JSC, occurs on the timescale of 100 min, as seen in the black control
cells in Figure 3-14. We attribute the 10x slower rate of performance increase to the fact
that the current in these devices is roughly 10x lower than that in DSSCs, and the doping
process is driven by the movement of photoexcited carriers in the device. Interestingly,
the devices with ALD Al2O3 barrier layers deposited after the CdS QDs (Figure 3-14a)
did not show the light-soaking effect, while those with the ALD Al2O3 barrier layers
deposited before the QDs (Figure 3-14b) did. This supports the hypothesis that the
doping of spiro-OMeTAD to spiro-OMeTAD cations is enhanced by the photoexcited
hole in the QD being transferred to spiro-OMeTAD molecules. Accordingly, device
efficiencies were reported from the stabilized values after light-soaking, with the peak
efficiencies noted as well. For perovskite solar cells, light-soaking time periods are on
the order of 5 minutes, depending on the specific perovskite processing and HTM used.9,
47, 48
Figure 3-14. Light-soaking effect in solid-state CdS QDSSCs with spiro-OMeTAD as
the solid-state HTM. In the control cells (black), the light-soaking effect occurs on a
timescale of 100 min, roughly 10 times longer than that in solid-state DSSCs made with
Katherine Roelofs Ph.D. Dissertation
73
spiro-OMeTAD. (a) Devices made with ALD Al2O3 barrier layers deposited after the
QDs, and (b) devices made with ALD Al2O3 barrier layers deposited before the QDs.
Curves are intended to guide the eye.
In perovskite solar cells, the major issue in assessing performance through J-V
curves is the hysteresis behavior of the devices. Namely, the direction of the voltage
sweep (from forward to reverse bias, or in the opposite direction) can dramatically
change the efficiency.9, 49, 50
The hysteresis effects can be mitigated by allowing enough
settling time after each voltage set point. Therefore, the scan rate of the J-V curve must be
substantially slowed, from a delay of 5 x 10-2
s in QDSSCs to a delay time of 5 s for each
10 mV step.9 The field is moving towards a consensus that this hysteresis behavior is due
to migration of iodide and methyl ammonium ions in the perovskite driven by the applied
bias.48
For the PSCs studied in this work, a settling time of 5 s was used in between each
voltage step of 10 mV, such that collection of a full J-V curve took on the order of 5 min.
External quantum efficiency (EQE) is the fraction of charges collected for every
incident photon, as a function of photon energy, and is also known as the incident-
photon-to-current efficiency (IPCE). The internal quantum efficiency (IQE) is the
fraction of charges collected for every absorbed photon, as a function of photon energy.
As EQE is based on all incident photons, some of which are parasitically absorbed or
reflected from the substrate, it will always be lower than the IQE. EQE is collected in a
home-made set up as shown in Figure 3-15.
Figure 3-15. Schematic of external quantum efficiency (EQE) measurement. A
Tungsten lamp lightsource is monochromated by an SP-150 Monocrhomator. A chopper
Katherine Roelofs Ph.D. Dissertation
74
wheel is used to chop the monochromated light, after which the light is split. One beam
goes to the device, while the other goes to a Si reference photodiode, tracked by a
current-voltage sourcemeter (Keithley 6517A). Prior to the measurement, a calibration Si
photodiode is placed in the device position, and counts the total number of photons for a
light pulse at a given photon energy. During the EQE measurement, a current-voltage
sourcemeter (Keithley 236) counts the amount of current produced by the device for a
given light pulse. The EQE is the ratio of the current produced by the device over the
number of photons counted by the Si photodiode, as adjusted by the reference diode to
account for any slight variations in the light source intensity during the measurement. If
a bias light is used during the measurement, then the sourcemeter from the device, and
the sourcemeter from the reference photodiode are connected to a lock-in amplifier such
that current is only collected from the device when it is seen on the reference photodiode
as well.
3.3.2 Transient Photovoltage and Photocurrent Measurements
Transient photovoltage (TPV) and photocurrent (TPJ) measurements were
conducted with the setup depicted in Figure 3-16; data was analyzed in accordance with
the small-perturbation model outlined by O’Regan and colleagues.51
A 1 sun bias light
by an LED array is applied to the device. Then, a function generator produces a square
pulse signal to pulse an LED of roughly 1/20th
the intensity of the bias light, in this case
roughly 0.05 sun intensity. The additional photons from the pulsed light create an
additional photocurrent, causing the J-V curve to shift outward to a new position, as
shown in Figure 3-17. For TPV measurements, the sourcemeter holds the device at a
constant current setpoint, while measuring the voltage. The light pulse excites addition
electrons in the absorber, which are transferred to the TiO2 anode. When the pulsed light
turns off, the additional photoexcited electrons recombine at the TiO2/absorber/HTM
interface, causing the voltage to decay from the steady-state voltage at 1.05 suns, V2, back
to the original steady-state voltage at 1 sun, V1 (Figure 3-17). The voltage decay is fit to
a first-order exponential, from which a time constant, τrec, is extracted, which is the
electron recombination lifetime (Equation 8). The time resolution of the measurements is
limited by the response time of the pulsed LED, and only millisecond timescales can be
resolved. The voltage decay is captured by an oscilloscope in coordination with a
current-voltage sourcemeter.
Katherine Roelofs Ph.D. Dissertation
75
(8)
Figure 3-16. Schematic of the transient photovoltage and photocurrent setup. A 1 sun
bias light is applied to the device, while a function generator (Agilent 3300) produces
pulsed signal sent to an LED held at roughly 1/20th
the intensity of the bias light. The
function generator signals an oscilloscope (Tektronix) to record voltage or current data at
the end of the pulsed light, which is measured by a current-voltage sourcemeter (Keithley
2400). In addition, the intensity of the bias LED array can be tuned by a programmable
power supply (Circuit Specialists CSI3644A). A LabVIEW software program
coordinates the interaction between these components and receives the data.
Figure 3-17. Example of a transient photovoltage measurement. (a) The solar cell is
illuminated at 1 sun bias. A pulsed light provides an additional 0.05 sun illumination,
causing the J-V curve to shift outward. The current is held constant, and when the pulsed
light turns off, the decay in voltage from V2 back to V1 is measured. (b) Voltage-decay
curves at a given current setpoint, where the light pulse is turned off at t = 0 s. The
Katherine Roelofs Ph.D. Dissertation
76
voltage decay rate changes with experimental parameters, such as in this case the
application of ALD-grown Al2O3 barrier layers of varying thickness.
For TPJ measurements, the device is held at a constant voltage set point, while the
current is measured. In this case, when the pulsed light is turned off, the decay in the
current is tracked. The time constant extracted from a first-order exponential fit of the
current decay, the extraction time constant, τext, is a combination of the transport lifetime,
τtrans, and the recombination lifetime (Equation 9 and Equation 10).
(9)
(10)
As an alternative approach to TPV and TPJ measurements, the bias light intensity can be
varied (using the programmable power supply) to determine the set point for each
measurement. The intensity of the pulsed light needs to be altered such that it still
remains roughly 1/20th
the intensity of the bias light. In the case where the bias light
intensity is used to determine the setpoint, the measurements are then all taken at VOC,
with the current held at zero.
3.4. Financial Support, Collaborations, and
Copyrights
This solar cell fabrication procedure was developed by Mike McGehee’s lab in
the Materials Science and Engineering Department at Stanford. The body of the
optimization work in the McGehee group to develop this standard procedure was
undertaken by Brian Hardin, I-Kang Ding, George Margulis, Colin Bailie, and William
Nguyen. ALD was performed in the Nubird reactor, which was originally built by
Jonathan Bakke, and maintained for the TiO2 and NiOx studies by Adrie Mackus and
Joseph Singh. UV-Vis measurements were performed in the Chemistry Optics Lab at
Stanford University. XPS, AES, and XRD measurements were conducted at the Stanford
Nanocharacterization Laboratory (SNL), and I would like to thank the SNL staff for their
Katherine Roelofs Ph.D. Dissertation
77
training and advice. The TPV and TPJ measurement set up was developed by Eric Hoke,
George Margulis, Colin Bailie, and Thomas Brennan.
3.5. References
1. van de Krol, R.; Grätzel, M., Photoelectrochemical Hydrogen Production.
Springer Science+Business Media: New York, NY, 2012.
2. Ding, I. K. Pore Filling and Light Trapping of Solid-State Dye-Sensitized Solar
Cells. Stanford University, Stanford, CA, 2011.
3. O'Regan, B. C.; Durrant, J. R.; Sommeling, P. M.; Bakker, N. J., The Journal of
Physical Chemistry C 2007, 111 (37), 14001-14010.
4. Sommeling, P. M.; O'Regan, B. C.; Haswell, R. R.; Smit, H. J. P.; Bakker, N. J.;
Smits, J. J. T.; Kroon, J. M.; van Roosmalen, J. A. M., The Journal of Physical Chemistry
B 2006, 110 (39), 19191-19197.
5. Lee, S.-W.; Ahn, K.-S.; Zhu, K.; Neale, N. R.; Frank, A. J., The Journal of
Physical Chemistry C 2012, 116 (40), 21285-21290.
6. Brennan, T. P.; Bakke, J. R.; Ding, I. K.; Hardin, B. E.; Nguyen, W. H.; Mondal,
R.; Bailie, C. D.; Margulis, G. Y.; Hoke, E. T.; Sellinger, A.; McGehee, M. D.; Bent, S.
F., Physical Chemistry Chemical Physics 2012, 14 (35), 12130-12140.
7. Samadpour, M.; Boix, P. P.; Giménez, S.; Iraji Zad, A.; Taghavinia, N.; Mora-
Seró, I.; Bisquert, J., The Journal of Physical Chemistry C 2011, 115 (29), 14400-14407.
8. Docampo, P.; Guldin, S.; Leijtens, T.; Noel, N. K.; Steiner, U.; Snaith, H. J.,
Advanced Materials 2014, 26 (24), 4013-4030.
9. Unger, E. L.; Hoke, E. T.; Bailie, C. D.; Nguyen, W. H.; Bowring, A. R.;
Heumuller, T.; Christoforo, M. G.; McGehee, M. D., Energy & Environmental Science
2014, 7 (11), 3690-3698.
10. Nguyen, W. H.; Bailie, C. D.; Unger, E. L.; McGehee, M. D., Journal of the
American Chemical Society 2014, 136 (31), 10996-11001.
11. Gratzel, M., Nat Mater 2014, 13 (9), 838-842.
12. Nicolau, Y. F., Applications of Surface Science 1985, 22-23, Part 2 (0), 1061-
1074.
13. Pathan, H. M.; Lokhande, C. D., Bull. Mater. Sci. 2004, 27, 85-111.
14. Möller, J.; Fischer, C. H.; Muffler, H. J.; Könenkamp, R.; Kaiser, I.; Kelch, C.;
Lux-Steiner, M. C., Thin Solid Films 2000, 361-362 (0), 113-117.
15. Sankapal, B. R.; Mane, R. S.; Lokhande, C. D., Materials Chemistry and Physics
2000, 63 (3), 226-229.
16. Li, J. J.; Wang, Y. A.; Guo, W.; Keay, J. C.; Mishima, T. D.; Johnson, M. B.;
Peng, X., Journal of the American Chemical Society 2003, 125 (41), 12567-12575.
17. Dasgupta, N. P.; Jung, H. J.; Trejo, O.; McDowell, M. T.; Hryciw, A.;
Brongersma, M.; Sinclair, R.; Prinz, F. B., Nano Letters 2011, 11 (3), 934-940.
18. Brennan, T. P.; Ardalan, P.; Lee, H.-B.-R.; Bakke, J. R.; Ding, I. K.; McGehee,
M. D.; Bent, S. F., Advanced Energy Materials 2011, 1 (6), 1169-1175.
19. Ardalan, P.; Brennan, T. P.; Lee, H.-B.-R.; Bakke, J. R.; Ding, I. K.; McGehee,
M. D.; Bent, S. F., ACS Nano 2011, 5 (2), 1495-1504.
Katherine Roelofs Ph.D. Dissertation
78
20. Burschka, J.; Pellet, N.; Moon, S.-J.; Humphry-Baker, R.; Gao, P.; Nazeeruddin,
M. K.; Grätzel, M., Nature 2013, 499 (7458), 316-319.
21. Bi, D.; El-Zohry, A. M.; Hagfeldt, A.; Boschloo, G., ACS Applied Materials &
Interfaces 2014, 6 (21), 18751-18757.
22. Lee, Y. H.; Luo, J.; Humphry-Baker, R.; Gao, P.; Grätzel, M.; Nazeeruddin, M.
K., Advanced Functional Materials 2015, n/a-n/a.
23. Bakke, J. R. Atomic Layer Deposition of Materials for Applications to
Photovoltaics. Stanford University, Stanford, CA, 2011.
24. Pickrahn, K. L. Atomic Layer Deposition of Earth Abundant Electrocatalysts for
the Oxygen Evolution Reaction. Stanford University, Stanford, CA., 2015.
25. Nardi, K. L.; Yang, N.; Dickens, C. F.; Strickler, A. L.; Bent, S. F., Advanced
Energy Materials 2015, n/a-n/a.
26. Seah, M. P.; Gilmore, I. S.; Spencer, S. J., Surface and Interface Analysis 2001,
31 (8), 778-795.
27. Hollas, J. M., Modern Spectroscopy. Fourth ed.; John Wiley & Sons, Ltd:
Hoboken, NJ, 2004.
28. van der Heide, P., X-ray Photoelectron Spectroscopy: An Introduction to
Principles and Practices. John Wiley & Sons, Inc.: 2011.
29. Cumpson, P. J., Journal of Electron Spectroscopy and Related Phenomena 1995,
73 (1), 25-52.
30. Shimizu, R., Japanese Journal of Applied Physics 1983, 22, 1631.
31. Langford, J. I.; Louër, D., Reports on Progress in Physics 1996, 59 (2), 131.
32. Suryanarayana, C.; Norton, G. M., X-Ray Diffraction: A Practical Approach.
Springer Science+Business Media: New York, NY, 1998.
33. Williams, D. B.; Carter, C. B., Transmission Electron Microscopy: A Textbook for
Materials Science. Springer Science+Business Media, LLC: New York, NY, 2009.
34. Rehr, J. J.; Albers, R. C., Reviews of Modern Physics 2000, 72 (3), 621-654.
35. de Groot, F., Chemical Reviews 2001, 101 (6), 1779-1808.
36. Chen, J. G., Surface Science Reports 1997, 30 (1–3), 1-152.
37. Ankudinov, A. L.; Ravel, B.; Rehr, J. J.; Conradson, S. D., Physical Review B
1998, 58 (12), 7565-7576.
38. Newville, M. īviņš, P. acoby, . Rehr, J. J. Stern, E. A., Physical
Review B 1993, 47 (21), 14126-14131.
39. Demchenko, I. N.; Chernyshova, M.; He, X.; Minikayev, R.; Syryanyy, Y.;
Derkachova, A.; Derkachov, G.; Stolte, W. C.; Liang, H., X-Ray Spectrometry 2013, 42
(4), 197-200.
40. Cristofolini, L., Current Opinion in Colloid & Interface Science 2014, 19 (3),
228-241.
41. Streetman, B. G.; Banerjee, S., Solid State Electronic Devices. Seventh ed.;
Prentice Hall: 2014.
42. Mishra, A.; Pootrakulchote, N.; Wang, M.; Moon, S.-J.; Zakeeruddin, S. M.;
Grätzel, M.; Bäuerle, P., Advanced Functional Materials 2011, 21 (5), 963-970.
43. Listorti, A.; Creager, C.; Sommeling, P.; Kroon, J.; Palomares, E.; Fornelli, A.;
Breen, B.; Barnes, P. R. F.; Durrant, J. R.; Law, C.; O'Regan, B., Energy &
Environmental Science 2011, 4 (9), 3494-3501.
Katherine Roelofs Ph.D. Dissertation
79
44. Wang, Q.; Zhang, Z.; Zakeeruddin, S. M.; Grätzel, M., The Journal of Physical
Chemistry C 2008, 112 (17), 7084-7092.
45. Abate, A.; Leijtens, T.; Pathak, S.; Teuscher, J.; Avolio, R.; Errico, M. E.;
Kirkpatrik, J.; Ball, J. M.; Docampo, P.; McPherson, I.; Snaith, H. J., Physical Chemistry
Chemical Physics 2013, 15 (7), 2572-2579.
46. Cappel, U. B.; Daeneke, T.; Bach, U., Nano Letters 2012, 12 (9), 4925-4931.
47. Zhao, C.; Chen, B.; Qiao, X.; Luan, L.; Lu, K.; Hu, B., Advanced Energy
Materials 2015, 5 (14), n/a-n/a.
48. Tress, W.; Marinova, N.; Moehl, T.; Zakeeruddin, S. M.; Nazeeruddin, M. K.;
Grätzel, M., Energy & Environmental Science 2015, 8 (3), 995-1004.
49. Snaith, H. J.; Abate, A.; Ball, J. M.; Eperon, G. E.; Leijtens, T.; Noel, N. K.;
Stranks, S. D.; Wang, J. T.-W.; Wojciechowski, K.; Zhang, W., The Journal of Physical
Chemistry Letters 2014, 5 (9), 1511-1515.
50. Kim, H.-S.; Park, N.-G., The Journal of Physical Chemistry Letters 2014, 5 (17),
2927-2934.
51. Barnes, P. R. F. Miettunen, K. i, X. Anderson, A. . Bessho, T. Grätzel, M.;
O'Regan, B. C., Advanced Materials 2013, 25 (13), 1881-1922.
Katherine Roelofs Ph.D. Dissertation
80
Chapter 4. Al2O3 Recombination
Barrier Layers by ALD in Solid-State
CdS QDSSCs
Despite the promise of quantum dots (QDs) as a light absorbing material to
replace the dye in dye-sensitized solar cells, quantum dot-sensitized solar cell (QDSSC)
efficiencies remain low, due in part to high rates of recombination. In this chapter we
demonstrate that ultra-thin recombination barrier layers of Al2O3 deposited by atomic
layer deposition (ALD) can improve the performance of cadmium sulfide (CdS) quantum
dot-sensitized solar cells (QDSSCs) with spiro-OMeTAD as the solid-state hole transport
material. We explored depositing the Al2O3 barrier layers either before or after the QDs,
resulting in TiO2/Al2O3/QD and TiO2/QD/Al2O3 configurations. The effects of barrier
layer configuration and thickness were tracked through current-voltage measurements of
device performance and transient photovoltage measurements of electron lifetimes. The
Al2O3 layers were found to suppress dark current and increase electron lifetimes with
increasing Al2O3 thickness in both configurations. For thin barrier layers, gains in open-
circuit voltage and concomitant increases in efficiency were observed, although at greater
thicknesses, losses in photocurrent caused net decreases in efficiency. A close
comparison of the electron lifetimes in TiO2 in the TiO2/Al2O3/QD and TiO2/QD/Al2O3
configurations suggests that electron transfer from TiO2 to spiro-OMeTAD is a major
source of recombination in ss-QDSSCs, though recombination of TiO2 electrons with
oxidized QDs can also limit electron lifetimes, particularly if the regeneration of oxidized
QDs is hindered by a too-thick coating of the barrier layer.
4.1. Introduction
Dye-sensitized solar cells (DSSCs) offer a compelling low-cost alternative to
conventional photovoltaic cells. The DSSC architecture consists of a mesoporous film of
Katherine Roelofs Ph.D. Dissertation
81
a wide band gap oxide, such as TiO2 or ZnO, coated with a monolayer of dye molecules.1
The pores are filled with a redox electrolyte which regenerates dye molecules that have
injected an excited electron into the metal oxide. DSSCs have recently reached power
conversion efficiencies of over 12%, employing a co-sensitization of two donor-π-bridge-
acceptor dyes.2 In the search for new approaches to increase efficiency and device
stability, a number of research groups have investigated replacing the sensitizing dye
with semiconductor quantum dots (QDs), creating quantum dot-sensitized solar cells
(QDSSCs).3-8
The size quantization of QDs9 allows for precise control over the band gap
for optimal absorption and over band offsets for optimal charge transfer. In particular,
QDs can be tuned to absorb in the near-IR, which is difficult to achieve with dyes, and
QDs can exhibit higher absorption cross sections than organic or metal-organic dyes over
a broad spectral range.10, 11
QDs can be grown directly on the mesoporous TiO2 by
chemical bath deposition,12, 13
successive ion layer adsorption and reaction (SILAR),14, 15
electrodeposition,16
or atomic layer deposition (ALD).17, 18
Commercialization of DSSC technology has generated interest in employing
solid-state hole-transport materials (HTMs), such as the commonly used spiro-OMeTAD
(2,2',7,7'-tetrakis-(N,N-di-p-methoxyphenylamine)-9,9'-spirobifluorene), to replace the
liquid electrolyte.19
The use of solid-state HTMs avoids the problem of electrolyte
leakage and corrosion of metal contacts, and aims to improve long-term stability.
Unfortunately, the recombination rate of electrons in TiO2 with holes in spiro-OMeTAD
is higher than the analogous pathway with the standard I-/I3
- liquid electrolyte.
20 The high
recombination rate limits the active layer thickness in solid-state devices to ~2 μm, due to
the consequently low charge carrier diffusion lengths as well as further increases in
recombination rate at greater thicknesses from poor pore-filling by spiro-OMeTAD.21-23
Active layers of TiO2 coated with dye molecules must reach thicknesses of ~10 μm to
absorb all incident light, so solid-state DSSC efficiencies are limited by insufficient light
absorption. The high absorption cross section of QDs, with the potential to absorb
strongly in a limited thickness, makes QDs especially suitable for use in solid-state
devices.
Katherine Roelofs Ph.D. Dissertation
82
To reduce recombination in DSSCs, the deposition of thin layers of metal oxides
at the TiO2 interface has been explored. For the current work, ALD was chosen to deposit
barrier layers of Al2O3, due to the ability of ALD to introduce high-quality, conformal
films on high aspect ratio surfaces, with angstrom-level control of film thickness, and
excellent uniformity over large areas.24
In DSSCs, metal oxide barrier layers are
deposited on TiO2 before the dye, and in this position can detrimentally interfere with
charge injection from the dye into the TiO2.25
In QDSSCs, there is the possibility to
deposit barrier layers after the QDs, due to the higher temperature and physical stability
of QDs compared to dyes. While there has been recent progress in developing low
temperature ALD for deposition of barrier layers after dye molecules in DSSCs,26
the
temperature constraints still greatly limit material choice. Previous reports of surface
coatings in QD-sensitized devices were conducted in cells with liquid electrolytes, with
the intent of protecting the QDs from corrosion by the liquid electrolyte, passivating QD
surface defects, or increasing the adsorption of co-sensitized dyes.27-30
Figure 4-1. Schematic of barrier layer configurations (not to scale) available in quantum
dot-sensitized solar cells: (a) TiO2/Al2O3/QD and (b) TiO2/QD/Al2O3, resulting
respectively from deposition of the Al2O3 layer before and after the CdS QDs. Spiro-
OMeTAD is employed as the hole-transport material (HTM). Arrows indicate
undesirable recombination pathways; pathways that may be blocked by the Al2O3 barrier
layer are shown by dashed arrows.
Katherine Roelofs Ph.D. Dissertation
83
In this work, Al2O3 barrier layers of varying thicknesses are deposited in solid-
state CdS QDSSCs before and after the QDs (Figure 4-1), with spiro-OMeTAD as the
HTM. CdS was chosen as the QD material due to its ease of deposition by SILAR and as
the large band gap of CdS ensures electron injection into the TiO2 conduction band for
any QD size. This study takes advantage of the flexibility of barrier placement that the
QDSSC system offers to optimize device performance and better understand the
recombination processes limiting QDSSC device efficiency. As shown in the energy
diagrams of Figure 4-1, in both the resulting TiO2/Al2O3/QD and TiO2/QD/Al2O3
configurations, Al2O3 blocks recombination from TiO2 to spiro-OMeTAD, and in the
TiO2/Al2O3/QD configuration Al2O3 also blocks recombination from TiO2 to oxidized
QDs. In both configurations, however, the barrier layer can also interfere with charge
separation steps necessary for photocurrent collection: either electron injection into TiO2
or hole transfer to spiro-OMeTAD. Thus, optimization rests on the concept of ‘kinetic
redundancy’, in which the desirable charge transfer step is much faster than the
competing undesirable step, such as in the case of electron injection into TiO2, which
occurs at a much faster rate than the competing undesirable decay of the excited electron
within the QD.31
In such a situation, the insertion of a barrier layer could slow electron
injection without appreciably affecting the electron injection yield, allowing for devices
to benefit from decreased recombination without detrimental effects on photocurrent
collection. In the QDSSC system, there is the opportunity to place the barrier layer such
that it slows whichever charge separation step has the greatest kinetic redundancy,
electron injection or hole transfer to spiro-OMeTAD.
The effects of barrier layer thickness and position were tracked through current-
voltage measurements of device performance and transient photovoltage measurements
of electron lifetimes. For the Al2O3 ALD system employed in this work, less than a
monolayer of Al2O3 is deposited per cycle on the TiO2 surface.25, 32, 33
Thus, while ALD
allows for the growth of ultra-thin Al2O3 layers, it is worth nothing that one Al2O3 ALD
cycle results in incomplete coverage of the TiO2 surface, and that for the low cycle
numbers used in this study (≤5), the spatial variation in film thickness is considerable
compared to the total film thickness.25
Katherine Roelofs Ph.D. Dissertation
84
4.2. Experimental Section
Mesoporous TiO2 Substrates. Except for the sensitization steps, solid-state
QDSSCs were fabricated similarly to the procedure for solid-state DSSCs described
elsewhere.34
For the transparent electrode, glass substrates coated with fluorine-doped tin
oxide (15 Ω/□, Pilkington) were patterned by etching with 4 M HCl and Zn. Substrates
were then coated with a thin (~50 nm) compact layer of TiO2 by aerosol spray pyrolysis
at 450 °C using air as a carrier gas. The mesoporous TiO2 layer was then deposited by
doctor-blading a commercial paste of 20 nm diameter anatase TiO2 particles (Dyesol 18-
NRT) diluted with terpineol. The films were annealed at 450 °C, resulting in a film
thickness of ~2.2 μm as measured by a Dektak profilometer. The mesoporous TiO2 films
were immersed overnight in a 0.02 M aqueous TiCl4 solution, and then annealed again at
450 °C.
SILAR of CdS QDs. CdS QDs were grown on mesoporous TiO2 substrates by
the SILAR process at room temperature. To complete a SILAR cycle, substrates were
first dipped in a 0.1 M CdSO4 aqueous solution for 5 min. They were then rinsed in DI
water and dipped in a 0.1 M Na2S aqueous solution for 5 min.14, 15
We found previously
that maximum device efficiency was achieved when CdS QD growth was stopped at 6
SILAR cycles,15
so all devices reported in this work were fabricated with 6 CdS SILAR
cycles. As characterized in previous work from Tauc Analysis of UV-Vis absorption
spectra of QD-sensitized mesoporous TiO2 substrates, CdS QDs deposited by 6 SILAR
cycles exhibit band gaps of 2.5‒2.8 eV.15
By comparison with known values in the
literature, these band gap values correspond to CdS QD sizes of 3-5 nm in diameter.35
For
devices where the Al2O3 barrier layer was deposited before the QDs, a buffered Na2S
solution was used, to avoid etching of the Al2O3 layer. Specifically, the 0.1 M Na2S
aqueous solution was buffered with an ammonium chloride/ammonium hydroxide
solution to reach a pH of 9. UV-Visible spectroscopy confirmed that CdS QDs produced
by the buffered SILAR process were optically identical to those produced by the
unbuffered process.
Katherine Roelofs Ph.D. Dissertation
85
ALD of Al2O3. ALD of Al2O3 was performed in a custom-built, travelling-wave,
hot wall, tube furnace-type reactor with five radially-oriented precursor manifolds. The
substrate temperature was maintained at 175 °C during ALD. The precursors,
trimethylaluminum (TMA) (Sigma-Aldrich) and water, were held at room temperature.
N2 was used as a carrier and purge gas. To ensure that the precursors were able to
sufficiently penetrate the mesoporous substrate, a soak step (when the reaction chamber
is isolated from the vacuum pump and the precursor manifold valve is closed) was
employed, similar to that adopted by Lin et al.33
First, TMA was pulsed for 5 s and
allowed to soak for an additional 40 s; after this the chamber was evacuated for 70 s
(purge step). Next, H2O was pulsed for 3 s and allowed to soak for an additional 40 s,
followed by a 70 s purge step. The growth rate of this process over 100 cycles on Si (100)
wafer (Silicon Quest) with a native oxide as measured by a spectroscopic ellipsometer
(Woollam Alpha SE) was ~1.6 Å/cycle. This process was previously shown, by Auger
electron spectroscopy of TiO2 layer cross-sections, to deposit Al2O3 through the entire
thickness of the mesoporous TiO2 substrate.25
The stability of the Al2O3 layer in the buffered SILAR process was confirmed by
tracking the etching rate of a thick planar film (100 ALD cycles) of Al2O3 deposited on a
planar Si (100) wafer, submersed in the buffered Na2S solution. The etching rate was
determined by spectroscopic ellipsometry to be 0.2 ± 0.1 Å/hr. As the Al2O3-coated
mesoporous TiO2 substrates were submersed in the Na2S solution for less than 30 min
throughout the entire SILAR process, a negligible amount of the Al2O3 layer will be
etched.
Active Layer Characterization. A Cary 6000i UV-Visible spectrometer
(Varian) was used to characterize the optical properties of the active layer. For these
studies, mesoporous TiO2 was deposited on glass microscope slides, and treated with the
desired combination of SILAR and ALD cycles.
To study the growth of Al2O3 on TiO2 and CdS surfaces, X-ray photoelectron
spectroscopy measurements (XPS) were taken with a PHI 500 VersaProbe Scanning XPS
Microprobe equipped with an Al Kα 1486 eV radiation source at a pressure of 6x10-10
Torr. To fabricate the TiO2 surfaces, thick planar films (~8 nm) of anatase TiO2 were
Katherine Roelofs Ph.D. Dissertation
86
deposited by ALD on Si (100) wafers, as described in previous work.15
To make the CdS
surface, thick planar films of CdS were deposited by SILAR (150 cycles) on Si (100)
wafers. Al2O3 was then grown by ALD on the TiO2 and the CdS films. The Al2O3 growth
was tracked with high-resolution XPS scans of the Al (2p), Ti (2p), and Cd(3d) peaks,
taken at a pass energy of 23.5 eV and a resolution of 0.1 eV/step, at multiple spots on the
surface.
Device Fabrication. For the TiO2/Al2O3/QD devices, mesoporous TiO2
substrates were subjected first to ALD of Al2O3, then to the buffered SILAR process. For
the TiO2/QD/Al2O3 devices, mesoporous TiO2 substrates first underwent the unbuffered
SILAR process, then underwent ALD of Al2O3. The QD-sensitized control substrates (0
ALD cycles) used for comparison with the TiO2/QD/Al2O3 devices were heated to 175
°C to account for any effects the heat treatment involved in the ALD process might have
on the QDs. As described elsewhere,34
the solution of the solid-state hole-transporting
material was composed of 225 mg mL-1
of spiro-OMeTAD (Lumtec) dissolved in
chlorobenzene, with tert-Butylpyridine added at a ratio of 1:10.3 μ :mg of spiro-
OMeTAD, and lithium bis-(trifluoromethylsulfonyl)imide salt (170 mg mL-1
in
acetonitrile) added at a ratio of 1:4.8 μ :mg of spiro-OMeTAD. A small amount of the
spiro-OMeTAD solution (30 μ for 3.75 cm2 substrates) was deposited onto the TiO2
substrates at room temperature, and spin-coated at 2000 RPM for 30 s. Finally, 200 nm
thick Ag counter electrodes were deposited by thermal evaporation under vacuum below
10-6
torr. Final device areas were on the order of 0.1 cm2. Devices were stored inside a
desiccator prior to electrical measurements.
Electrical Measurements. For photovoltaic measurements, an AM 1.5 solar
simulator (Oriel 91160) equipped with a 300 W ozone-free Xe arc lamp (6258) was used.
The lamp was calibrated to 1 sun (100 mW cm-2
) using a reference NREL calibrated Si
photodiode equipped with an IR cutoff filter. Current-voltage (J-V) curves were collected
with a Keithley 2400 SourceMeter, with a sweep delay of 40 ms. Devices were light-
soaked until maximum efficiencies were reached (up to 1 hr). Transient photovoltage
measurements were conducted using a setup described in greater detail elsewhere.23, 36
Briefly, devices were white-light-biased at 1 sun with an array of white LEDs (Lumiled)
Katherine Roelofs Ph.D. Dissertation
87
and pulsed with a square wave from a white LED (~0.05 suns). A given device was held
at constant current with a 2400 Keithley SourceMeter, biasing the device to a specific
point on its J-V curve, and the decay of the increased photovoltage from the light pulse
was collected at various bias points. The ~0.05 sun light pulse length was set to 50 ms,
100 ms, or 500 ms, depending on the time necessary to sufficiently capture the
photovoltage decay. The mono-exponential photovoltage decay was fit to extract a
recombination lifetime.
4.3. Results and Discussion
Solid-state quantum dot-sensitized solar cells (ss-QDSSCs) were fabricated with
spiro-OMeTAD as the solid-state HTM and CdS QDs grown by 6 SILAR deposition
cycles as the sensitizer, as optimized in previous work.15
4.3.1. TiO2/Al2O3/QD Device Performance
One factor limiting efficiencies of QDSSCs is low surface coverage of QDs on
the TiO2,37, 38
which leaves bare TiO2 surface areas at which recombination can easily
occur via direct contact with spiro-OMeTAD. To reduce this major recombination
pathway, barrier layers of Al2O3 were deposited on the TiO2 surface prior to QD
deposition, in an analogous configuration to that employed in DSSCs. The J-V curves of
the resulting TiO2/Al2O3/QD devices under illumination and in the dark are shown in
Figure 4-2. The Al2O3 layer delays the onset of dark current, as evidenced by the shift
with increasing barrier thickness to higher voltage required to extract a given amount of
current (Figure 4-2b). As discussed below in the analysis of the role of the Al2O3 layer,
the suppression of the dark current is attributed to the Al2O3 layer providing resistance to
electron transfer from TiO2 to spiro-OMeTAD. Since open-circuit voltage (VOC) is
determined by the point at which dark current fully cancels out photocurrent, the
suppression of the dark current with increasing ALD cycles leads to increases in VOC.
However, with increasing Al2O3 thickness, the photocurrent is attenuated, as seen by the
decreasing current density after more than 1 ALD cycle in Figure 4-2a. At 5 ALD cycles
the loss of photocurrent is so severe that it actually causes a decrease in VOC relative to 3
cycles.
Katherine Roelofs Ph.D. Dissertation
88
Figure 4-2. Current density-voltage curves of representative devices in the
TiO2/Al2O3/QD configuration with increasing ALD cycles of Al2O3 (a) under 1 sun of
illumination and (b) in the dark. Al2O3 barrier layers are effective at suppressing the onset
of dark current, leading to increases in VOC. However, when more than 1 ALD cycle is
performed, the Al2O3 layer hinders electron injection thereby reducing JSC.
For one ALD cycle, the increase in short-circuit current (JSC) seen in Figure 4-2a
is attributed to improved charge collection efficiency, as electrons diffusing through TiO2
have a better chance of reaching the electrode before recombining. The increases in JSC
and VOC at 1 ALD cycle produce a net increase in power conversion efficiency (η), from
an average efficiency of 0.18% at 0 ALD cycles to 0.23% at 1 ALD cycle. However, for
devices with thicker barrier layers, JSC drops significantly (by 61% after 5 ALD cycles),
causing a net decrease in efficiency despite the overall gains in VOC. Part of the drop in
current is due to a slight decrease in QD deposition on Al2O3-coated TiO2. UV-Vis
absorption measurements shown in Figure 4-3a indicate a ~20% decrease (calculated
from the Beer-Lambert law) in the amount of CdS material deposited on TiO2 coated
with 5 Al2O3 ALD cycles, but this does not sufficiently explain the 61% drop in JSC. The
high conduction band of Al2O3 can inhibit electron injection from the excited QD into
TiO2 if the barrier layer is too thick, causing a drop in current as decay of the excited
electron to the ground state in the QD competes with injection. This inhibition of
injection accounts for the substantial decreases in JSC at 3 and 5 ALD cycles, which
Katherine Roelofs Ph.D. Dissertation
89
correspond to barrier thicknesses of ~0.5 and ~0.8 nm. In DSSCs, barrier layers of similar
thicknesses have been shown to interfere with electron injection.25, 33
Figure 4-3. UV-Vis absorption spectra of CdS QDs (which begin absorbing at 510 nm)
deposited on mesoporous TiO2 coated with varying ALD cycles of Al2O3. For the
TiO2/Al2O3/QD configuration (a), the presence of Al2O3 reduces CdS growth, but the
reduction is not sufficient to fully explain the drops in JSC observed. For the
TiO2/QD/Al2O3 configuration (b), no reduction in QD absorption is observed.
4.3.2. TiO2/QD/Al2O3 Device Performance
Taking advantage of the opportunity in QDSSCs to deposit barrier layers after the
TiO2 has been coated with the sensitizer, TiO2/QD/Al2O3 devices were fabricated. The
results (Figure 4-4) show a similar suppression of the dark current with increasing barrier
thickness as observed in TiO2/Al2O3/QD devices, leading to initial gains in VOC.
Katherine Roelofs Ph.D. Dissertation
90
Figure 4-4. Current density-voltage curves of representative devices in the
TiO2/QD/Al2O3 configuration increasing ALD cycles of Al2O3 (a) under 1 sun of
illumination and (b) in the dark. As with the TiO2/Al2O3/QD configuration, the Al2O3
layer effectively suppresses dark current but reduces JSC if too thick.
Notably, a drop in JSC is again observed for barriers thicker than one ALD cycle
(Figure 4-4a). Part of the incentive for the TiO2/QD/Al2O3 configuration is to avoid
hindering electron injection into TiO2, so that gains in VOC can be achieved without
decreases in current. Since the drop in JSC cannot be attributed to blocking of electron
injection into TiO2, here other possible explanations are considered. First, UV-Vis
absorption measurements (Figure 4-3b) show that when deposited after the QDs, the
Al2O3 layers have no effect on QD absorption, ruling out oxidation or desorption of QDs
as potential sources of current loss. Second, it is important to determine whether Al2O3
ALD coats the CdS QD surfaces as well as the TiO2. X-ray photoelectron spectroscopy
(XPS) of Al2O3 ALD on CdS and TiO2 films (Figure 4-5) confirm that Al2O3 grows as
readily on CdS as on TiO2 surfaces, suggesting that the Al2O3 is indeed capping the CdS
QDs in the mesoporous TiO2. Thus, the decrease in JSC is attributed to the Al2O3 layer
preventing regeneration of oxidized QDs; that is, the low valence band of Al2O3 acts as a
tunneling barrier to the transfer of holes from oxidized CdS QDs to spiro-OMeTAD. This
conclusion is consistent with recent reports that suggest that regeneration of oxidized
QDs is relatively slow compared to dye regeneration in DSSCs, and may be a critical
factor in achieving higher efficiencies.39, 40
Katherine Roelofs Ph.D. Dissertation
91
Figure 4-5. High resolution X-ray photoelectron spectroscopy scans of the Al 2p peak of
Al2O3 deposited by ALD on thick CdS films (grown by SILAR) or on thick TiO2 films
(grown by ALD) as a function of the number of Al2O3 ALD cycles. The data indicate that
Al2O3 readily grows on both TiO2 and CdS surfaces.
4.3.3. Comparison of Layer Placement and Role of Al2O3 Layer
Figure 4-6 analyzes batch-to-batch variability and Figure 4-7 provides a
comparison of the effect of Al2O3 layers on average device parameters in both
TiO2/Al2O3/QD and TiO2/QD/Al2O3 devices (values are given in Table 4-1). The
variation in JSC for the control
Table 4-1. Device Parameters for TiO2/Al2O3/QD and TiO2/QD/Al2O3 Configurations.†
Barrier Layer Configuration
# ALD cycles
JSC
(mA cm-2)
VOC
(V) Fill Factor Efficiency
(%)
TiO2/Al2O3/QD 0 0.59 0.59 0.52 0.18
1 0.75 0.62 0.50 0.23
3 0.38 0.73 0.45 0.12
5 0.23 0.70 0.33 0.05
TiO2/QD/Al2O3 0 0.84 0.59 0.53 0.26
1 0.84 0.70 0.60 0.35
3 0.29 0.69 0.60 0.12
5 0.07 0.62 0.52 0.02
Katherine Roelofs Ph.D. Dissertation
92
†These values for barrier configurations with varying ALD cycles of Al2O3, plotted in
Figure 5, were collected under 1 sun of illumination. Parameters are the average of the
top 50% of devices (ranked by efficiency); the set of the top 50% range in number from
6-8 devices.
Figure 4-6. Percent change of device parameters for the TiO2/QD/Al2O3 configuration
averaged over three batches of devices, relative to the average value at 0 ALD cycles of a
given batch. Specifically, within a batch of n devices, the percent change of a parameter
for a given device (pn) was calculated using the value for that device (vn) and the average
value at 0 ALD cycles for devices in that batch:
. The
values of pn were then binned for all three batches and the average and standard deviation
(error bars) of this total set are plotted above.
cells (0 ALD cycles), reflects the batch-to-batch variations that cause offsets in JSC, and
thus efficiency, for the entire batch of devices (0 to 5 ALD cycles of a given
Katherine Roelofs Ph.D. Dissertation
93
configuration). However, the relative dependence of JSC on barrier layer thickness within
a batch are not affected by these batch-to-batch variations (see Figure 4-6), hence the
trends are reliable. The absolute value of JSC, and thus efficiency, varied from batch to
batch due to batch-dependent variations in incompletely controlled process parameters
such as the thickness of the compact TiO2 layer. However, relative changes in device
performance with barrier thickness within a given batch were consistent. Thus, when the
device parameters for a given batch are baselined relative to that batch’s performance at 0
ALD cycles (by taking percent changes), then averaging data across all three batches
produces a clear trend in device performance with varying ALD cycles (shown in Figure
4-6). This analysis shows that while scatter exists in the absolute performance of different
batches of devices, as seen by the difference in efficiency at 0 ALD cycles in Figure 4-7,
changes in devices parameters due to varying ALD cycles within a given batch are
significant.
Figure 4-7 shows that qualitatively similar trends in device performance are
observed when depositing the Al2O3 layer before or after the QDs. Both configurations
result in efficiency improvements at one ALD cycle, but at greater cycle numbers, the
efficiency decreases, driven by decreases in JSC. These results suggest that charge
separation steps (electron injection to TiO2 and hole transfer to spiro-OMeTAD) are
highly dependent on barrier layer thickness. The decrease in JSC at 3 ALD cycles
indicates the barrier layer is thick enough so that there is no longer any kinetic
redundancy in the electron injection step, or in the hole transfer to spiro-OMeTAD. In
other words, the electron injection or hole transfer steps cannot be slowed down any
further without affecting charge collection efficiency.
Katherine Roelofs Ph.D. Dissertation
94
Figure 4-7. Comparison of device parameters for TiO2/Al2O3/QD and TiO2/QD/Al2O3
configurations under 1 sun of illumination with varying ALD cycles of Al2O3. Parameters
are the average of the top 50% of devices (ranked by efficiency); the set of the top 50%
range in number from 6-8 devices, and error bars indicate standard deviations. The
corresponding values of device parameters are given in Table 4-1. In both configurations,
device efficiency improves after 1 ALD cycle of Al2O3 but drops thereafter due to
decreases in JSC.
The efficiency improvements observed for both configurations at one ALD cycle
motivated the fabrication of devices in which the mesoporous TiO2 was treated with 1
Al2O3 ALD cycle, followed by CdS QD deposition, and a final 1 Al2O3 ALD cycle after
the QDs. The resulting TiO2/Al2O3(1)/QD/Al2O3(1) devices, where (1) indicates 1 Al2O3
ALD cycle, had an average VOC of 0.69 V, which is within error of the VOC enhancement
achieved in TiO2/QD/Al2O3(1) devices (0.70 V). An average JSC of 0.81 mA cm-2
was
observed, leading to an average efficiency of 0.36%. The efficiency of the
Katherine Roelofs Ph.D. Dissertation
95
TiO2/Al2O3(1)/QD/Al2O3(1) devices did not exceed those of the TiO2/QD/Al2O3(1) or
TiO2/Al2O3(1)/QD devices fabricated within the same batch, although it is comparable to
the 0.35% efficiency achieved for TiO2/QD/Al2O3(1) devices in the highest-performing
TiO2/QD/Al2O3 batch, shown in Figure 4-7.
In the discussion thus far, we have referred to the Al2O3 layers as ‘barrier layers’,
implying that the Al2O3 layer acts as a tunneling barrier to recombination. The large
increases in electron lifetimes with increasing thickness of the Al2O3 layer, discussed
below, suggest that the Al2O3 layer acts primarily as a tunneling barrier, although there
may be small contributions from a TiO2 band shift or passivation of surface defects. We
have considered these two other mechanisms by which coating the TiO2 surface with
Al2O3 might lead to gains in VOC. In one mechanism, Al2O3 acts to passivate the TiO2
surface, decreasing recombination mediated by TiO2 surface defects. The successive
suppression of the dark current at 3 and 5 ALD cycles (Figure 4-8) make a purely
surface-mediated effect less likely, although not impossible, as less than a monolayer of
Al2O3 is deposited in each ALD cycle. Furthermore, in this work, the mesoporous TiO2
electrodes were coated with a thin layer of TiO2 deposited by TiCl4 chemical bath
treatment, which has been found to improve charge separation in DSSCs, potentially by
passivating surface defects on the mesoporous TiO2.41
In the second mechanism, the high
conduction band of Al2O3 could cause an upward shift in the TiO2 conduction band,
increasing the offset between TiO2 and the highest occupied molecular orbital (HOMO)
of spiro-OMeTAD. However, the deposition of Al2O3 on mesoporous TiO2 has been
shown to have only minor effects on TiO2 conduction band, insufficient to cause the
~100 mV gains in VOC that we observed.42-44
In addition, the increases in electron lifetime
measured in this work (Figure 4-9) are too large to be explained by a conduction band
shift.
Katherine Roelofs Ph.D. Dissertation
96
Figure 4-8. Dark current onset for devices with TiO2/Al2O3/QD and TiO2/QD/Al2O3
configurations, with varying ALD cycles of Al2O3. The quantified dark current onset is
the voltage at which the current of a dark J-V curve reaches a threshold of 0.2 mA/cm2.
Error bars indicate standard deviation. The two configurations show similar suppression
of the dark current at thicker Al2O3 barrier layers.
A comparison of dark current measurements in TiO2/Al2O3/QD vs.
TiO2/QD/Al2O3 devices provides further insight into the role of the Al2O3 layer (Figure
4-8). The suppression of the dark current was quantified by taking the voltage at which
the current in a dark J-V curve exceeds a threshold of 0.2 mA cm-2
, termed the quantified
dark current onset.25
The current threshold was set relatively low to avoid any effects of
variation in series resistance. Unlike VOC, the quantified dark current onset is independent
of photocurrent, and thus provides a good metric for comparing the resistance the barrier
layer presents to recombination at the interface. Dark current consists of electrons
traveling through TiO2 from the FTO electrode towards the interface with spiro-
OMeTAD. In dark conditions the QDs are not oxidized by light absorption, so electrons
leaving TiO2 recombine at the interface with holes traveling through the spiro-OMeTAD
phase from the Ag electrode. The more competitive pathway for electrons leaving TiO2 is
through completely bare TiO2 or TiO2 surface regions only covered by Al2O3, rather than
TiO2 surface areas covered by a QD and an Al2O3 layer. This is because tunneling
probability decays exponentially with barrier thickness and a surface coating of a QD and
an Al2O3 layer would be thicker than just an Al2O3 layer. Consequently, in the
TiO2/Al2O3/QD and TiO2/QD/Al2O3 devices, the electrons leaving TiO2 face identical
barriers, as they pass through only an Al2O3 layer, rather than through the Al2O3/QD or
Katherine Roelofs Ph.D. Dissertation
97
QD/Al2O3 band structures (which would present different barriers to tunneling). Thus, it
is expected that in the dark, the Al2O3 layer suppresses dark current identically for the
TiO2/Al2O3/QD and TiO2/QD/Al2O3 configurations, as is indeed observed at 3 and 5
ALD cycles (Figure 4-8). The greater suppression of dark current for the TiO2/QD/Al2O3
devices at the very first ALD cycle, which leads to greater increases in VOC (Figure 4-7),
could be explained by a surface effect in the ALD process. Namely, the SILAR process
for the deposition of QDs could potentially increase surface roughness of the TiO2,
leading to increased nucleation of Al2O3 for the first ALD cycle in TiO2/QD/Al2O3
devices, and hence a slightly thicker Al2O3 recombination layer.
4.3.4. Effect of Al2O3 on Electron Lifetimes
Decreases in the rate of recombination at the TiO2 interface should result in
increased electron lifetimes (τn) in the TiO2. To test this, the impact of Al2O3 layers on
electron lifetime was determined by transient photovoltage measurements of
TiO2/Al2O3/QD and TiO2/QD/Al2O3 devices (Figure 4-9). The observed decrease of
electron lifetimes with applied bias is well known for DSSCs, and can be understood by
the strong positive dependence of the recombination rate constant on electron
concentration in TiO2, and the increase in electron concentration with applied bias.44, 45
The electron lifetime values for the control case of 0 ALD cycles are similar to those
previously reported in 2 μm thick solid-state DSSCs employing spiro-OMeTAD as the
HTM.46
As shown in Figure 4-9, we found that increasing the Al2O3 layer thickness leads
to substantial improvements in electron lifetimes, supporting the conclusion that the
Al2O3 layer acts as a tunneling barrier to electron recombination.
Katherine Roelofs Ph.D. Dissertation
98
Figure 4-9. Electron lifetimes for the TiO2/Al2O3/QD (open symbols) and
TiO2/QD/Al2O3 (closed symbols) as determined via transient photovoltage measurements
for varying ALD cycles of Al2O3. The larger lifetime improvements achieved with Al2O3
barrier layers in the TiO2/Al2O3/QD configuration are attributed to the fact that in that
configuration, recombination to both spiro-OMeTAD and oxidized quantum dots is
suppressed.
The difference between electron lifetime improvements in TiO2/Al2O3/QD versus
TiO2/QD/Al2O3 devices indicates that, under illumination (transient photovoltage
measurements are taken at 1 sun illumination), the CdS QDs do participate in
recombination. The TiO2/Al2O3/QD configuration results in the greatest enhancement of
electron lifetimes. This is attributed to the ability of TiO2/Al2O3/QD devices to block both
recombination to oxidized QDs and to spiro-OMeTAD, while the TiO2/QD/Al2O3
configuration only blocks recombination to spiro-OMeTAD (shown schematically in
Katherine Roelofs Ph.D. Dissertation
99
Figure 4-1). In addition, the Al2O3 layer in TiO2/QD/Al2O3 devices may interfere with
regeneration of oxidized QDs by spiro-OMeTAD, leading to an increase in the
concentration of oxidized QDs at a given time. This in turn could increase recombination
to oxidized QDs, and thus partially offset the benefits (i.e., increased electron lifetimes)
of decreased recombination with spiro-OMeTAD.
The slope of log(τn) vs. voltage depends on the exponential coefficient of the
density of localized states in TiO2 (which increases exponentially with energy level), as
well as the exact dependence of the recombination rate constant, kn = 1/τn, on the electron
concentration.44, 47
Therefore, the higher slope observed in TiO2/Al2O3/QD devices
compared to the TiO2/QD/Al2O3 devices could suggest a difference in the rate constant’s
dependence on electron density, consistent with different recombination pathways at
work in the different configurations. However, the difference in slope could also be due a
lower capacitance at the interface in the TiO2/Al2O3/QD devices.
That the TiO2/Al2O3/QD configuration yields the greatest improvements in
electron lifetimes further suggests that recombination of TiO2 electrons through defects
on the QD outer surface is not the limiting factor in achieving high electron lifetimes.
Here, defects on the QD outer surface refers to sites on the QD surface facing spiro-
OMeTAD, not the QD interface with TiO2. If the dominant recombination pathway were
defects on the QD outer surface, we would expect TiO2/QD/Al2O3 devices with the Al2O3
layer passivating the QD surface to show the greatest increases in electron lifetimes. This
assessment is based on the assumption that any impact the TiO2/Al2O3/QD barrier
configuration might have on this QD outer surface defect recombination (by increasing
the separation between the TiO2 and QDs) would be minimal compared to the
TiO2/QD/Al2O3 barrier configuration, which directly passivates QD surface defects.
In summary, these results demonstrate the ability of Al2O3 barrier layers to greatly
enhance electron lifetimes, and suggest that recombination to both spiro-OMeTAD and to
oxidized QDs play a role in limiting electron lifetimes. Overall, the effectiveness of
Al2O3 layers as a barrier to recombination leads us to propose a route to capturing the
benefits of enhanced VOC without losses in current. Specifically, deposition of barrier
layers after QDs could be modified to allow for selective deposition only on bare TiO2
Katherine Roelofs Ph.D. Dissertation
100
surfaces, such that the layer is not coating the QD surface. This could be achieved by
material choice of a barrier layer that grows selectively on the TiO2, or by masking the
QD surface with a removable organic surface layer. In this geometry, the barrier layer
would prevent the recombination of electrons in TiO2 to holes in spiro-OMeTAD,
without obstructing electron injection into TiO2 or regeneration of the oxidized QD by
hole transfer to spiro-OMeTAD. Decreasing the rate of recombination (thus increasing
electron lifetimes and diffusion lengths) enables the use of thicker active layers in solid-
state devices and the potential for substantial gains in efficiency from increased
absorption. Furthermore, solving the problem of recombination in QD-sensitized devices
is particularly attractive, as the higher temperature stability of QDs, compared to dyes,
provides the opportunity to deposit the hole transport material by melt infiltration, thus
addressing the other challenge of moving to thicker active layers, that of insufficient
pore-filling of the solid-state HTM.22, 34
4.4. Conclusions
Solid-state CdS QDSSCs were fabricated with Al2O3 barrier layers deposited via
ALD. The Al2O3 layer was found to act primarily as a tunneling barrier to electron
recombination, resulting in a suppression of the dark current and substantial increases in
electron lifetimes in TiO2. Both TiO2/Al2O3/QD and TiO2/QD/Al2O3 configurations
resulted in the same qualitative trends in device performance with increasing barrier layer
thickness. Namely, VOC initially increases with a concomitant increase in efficiency
before the barrier layer begins at greater thicknesses to interfere with charge transfer
steps necessary for photocurrent collection, resulting in drops in efficiency. That drops in
JSC with increased barrier thickness were observed in both barrier configurations suggests
that both electron injection to TiO2 and hole transfer to spiro-OMeTAD are sensitive to
the presence of tunneling barriers. A quantitative comparison of the TiO2/Al2O3/QD and
TiO2/QD/Al2O3 barrier layer configurations’ impact on electron lifetimes indicates that
electron transfer from TiO2 to spiro-OMeTAD is a major source of recombination,
though the back transfer of TiO2 electrons to oxidized QDs also has the potential to limit
electron lifetimes, particularly when regeneration of oxidized QDs is inhibited by a
barrier layer capping the QD, as in the TiO2/QD/Al2O3 configuration.
Katherine Roelofs Ph.D. Dissertation
101
4.5. Financial Support, Collaborations, and
Copyrights
This chapter is adapted from work originally published in The Journal of Physical
Chemistry C.48
Reprinted with permission from Roelofs, K. E., et al., The Journal of
Physical Chemistry C 2013, 117 (11), 5584-5592. Copyright 2013 American Chemical
Society. This work was supported by the Center for Advanced Molecular Photovoltaics
(Award No. KUS-C1-015-21), made by King Abdullah University of Science and
Technology (KAUST). We would like to thank the Stanford Nanocharacterization
Laboratory (SNL) staff for their support. The ALD reactor and process development for
these studies were carried out with support by the Center for Nanostructuring for
Efficient Energy Conversion, an Energy Frontier Research Center (EFRC) funded by the
U.S. Department of Energy, Office of Basic Energy Sciences, Award No. DE-
SC0001060.
4.6. References
1. O'Regan, B.; Grätzel, M., Nature 1991, 353 (6346), 737-740.
2. Yella, A.; Lee, H.-W.; Tsao, H. N.; Yi, C.; Chandiran, A. K.; Nazeeruddin, M. K.;
Diau, E. W.-G.; Yeh, C.-Y.; Zakeeruddin, S. M.; Grätzel, M., Science 2011, 334 (6056),
629-634.
3. Nozik, A. J., Physica E: Low-dimensional Systems and Nanostructures 2002, 14,
115-120.
4. Kamat, P. V., The Journal of Physical Chemistry C 2008, 112 (48), 18737-18753.
5. Mora-Seró, I.; Bisquert, J., The Journal of Physical Chemistry Letters 2010, 1
(20), 3046-3052.
6. Rühle, S.; Shalom, M.; Zaban, A., ChemPhysChem 2010, 11 (11), 2290-2304.
7. Plass, R.; Pelet, S.; Krueger, J.; Grätzel, M.; Bach, U., The Journal of Physical
Chemistry B 2002, 106 (31), 7578-7580.
8. Shen, Q.; Arae, D.; Toyoda, T., Journal of Photochemistry and Photobiology A:
Chemistry 2004, 164, 75-80.
9. Weller, H., Advanced Materials 1993, 5 (2), 88-95.
10. Peter, L. M., The Journal of Physical Chemistry Letters 2011, 2 (15), 1861-1867.
11. Moreels, I.; Lambert, K.; Smeets, D.; De Muynck, D.; Nollet, T.; Martins, J.;
Vanhaecke, F.; Vantomme, A.; Delerue, C.; Allan, G.; Hens, Z., ACS Nano 2009, 3 (10),
3023-3030.
Katherine Roelofs Ph.D. Dissertation
102
12. Diguna, L. J.; Shen, Q.; Kobayashi, J.; Toyoda, T., Applied Physics Letters 2007,
91 (2), 023116-023113.
13. Niitsoo, O.; Sarkar, S. K.; Pejoux, C.; Rühle, S.; Cahen, D.; Hodes, G., Journal of
Photochemistry and Photobiology A: Chemistry 2006, 181, 306-313.
14. Baker, D. R.; Kamat, P. V., Advanced Functional Materials 2009, 19 (5), 805-
811.
15. Ardalan, P.; Brennan, T. P.; Lee, H.-B.-R.; Bakke, J. R.; Ding, I. K.; McGehee,
M. D.; Bent, S. F., ACS Nano 2011, 5 (2), 1495-1504.
16. Lévy-Clément, C.; Tena-Zaera, R.; Ryan, M. A.; Katty, A.; Hodes, G., Advanced
Materials 2005, 17 (12), 1512-1515.
17. Brennan, T. P.; Ardalan, P.; Lee, H.-B.-R.; Bakke, J. R.; Ding, I. K.; McGehee,
M. D.; Bent, S. F., Advanced Energy Materials 2011, 1 (6), 1169-1175.
18. Dasgupta, N. P.; Jung, H. J.; Trejo, O.; McDowell, M. T.; Hryciw, A.;
Brongersma, M.; Sinclair, R.; Prinz, F. B., Nano Letters 2011, 11 (3), 934-940.
19. Bach, U.; Lupo, D.; Comte, P.; Moser, J. E.; Weissortel, F.; Salbeck, J.; Spreitzer,
H.; Grätzel, M., Nature 1998, 395 (6702), 583-585.
20. Kron, G.; Egerter, T.; Werner, J. H.; Rau, U., The Journal of Physical Chemistry
B 2003, 107 (15), 3556-3564.
21. Snaith, H. J.; Schmidt-Mende, L., Advanced Materials 2007, 19 (20), 3187-3200.
22. Ding, I. K.; Tétreault, N.; Brillet, J.; Hardin, B. E.; Smith, E. H.; Rosenthal, S. J.;
Sauvage, F.; Grätzel, M.; McGehee, M. D., Advanced Functional Materials 2009, 19
(15), 2431-2436.
23. Melas-Kyriazi, J.; Ding, I. K.; Marchioro, A.; Punzi, A.; Hardin, B. E.; Burkhard,
G. F.; Tétreault, N.; Grätzel, M.; Moser, J.-E.; McGehee, M. D., Advanced Energy
Materials 2011, 1 (3), 407-414.
24. Bakke, J. R.; Pickrahn, K. L.; Brennan, T. P.; Bent, S. F., Nanoscale 2011, 3 (9),
3482-3508.
25. Brennan, T. P.; Bakke, J. R.; Ding, I. K.; Hardin, B. E.; Nguyen, W. H.; Mondal,
R.; Bailie, C. D.; Margulis, G. Y.; Hoke, E. T.; Sellinger, A.; McGehee, M. D.; Bent, S.
F., Physical Chemistry Chemical Physics 2012, 14 (35), 12130-12140.
26. Son, H.-J.; Wang, X.; Prasittichai, C.; Jeong, N. C.; Aaltonen, T.; Gordon, R. G.;
Hupp, J. T., Journal of the American Chemical Society 2012, 134 (23), 9537-9540.
27. od, I. Gonz lez-Pedro, V.; Tachan, Z.; Fabregat-Santiago, F. Mora-Ser •, I.
n.; Bisquert, J.; Zaban, A., The Journal of Physical Chemistry Letters 2011, 2, 3032-
3035.
28. Shalom, M. Dor, S. R hle, S.; Grinis, L.; Zaban, A., The Journal of Physical
Chemistry C 2009, 113 (9), 3895-3898.
29. Shen, Q.; Kobayashi, J.; Diguna, L. J.; Toyoda, T., Journal of Applied Physics
2008, 103 (8), 084304-084305.
30. Choi, H.; Nicolaescu, R.; Paek, S.; Ko, J.; Kamat, P. V., ACS Nano 2011, 5 (11),
9238-9245.
31. Haque, S. A.; Palomares, E.; Cho, B. M.; Green, A. N. M.; Hirata, N.; Klug, D.
R.; Durrant, J. R., Journal of the American Chemical Society 2005, 127 (10), 3456-3462.
32. Mäkinen, V. onkala, K. äkkinen, H., The Journal of Physical Chemistry C
2011, 115 (18), 9250-9259.
Katherine Roelofs Ph.D. Dissertation
103
33. Lin, C.; Tsai, F.-Y.; Lee, M.-H.; Lee, C.-H.; Tien, T.-C.; Wang, L.-P.; Tsai, S.-Y.,
Journal of Materials Chemistry 2009, 19 (19), 2999-3003.
34. Leijtens, T.; Ding, I. K.; Giovenzana, T.; Bloking, J. T.; McGehee, M. D.;
Sellinger, A., ACS Nano 2012, 6 (2), 1455-1462.
35. Spanhel, L.; Haase, M.; Weller, H.; Henglein, A., Journal of the American
Chemical Society 1987, 109 (19), 5649-5655.
36. Snaith, H. J.; Humphry-Baker, R.; Chen, P.; Cesar, I.; Zakeeruddin, S. M.;
Grätzel, M., Nanotechnology 2008, 19 (42), 424003.
37. Shen, H.; Lin, H.; Liu, Y.; Li, J.; Oron, D., The Journal of Physical Chemistry C
2012, 116 (29), 15185-15191.
38. ee, . Wang, M. Chen, P. Gamelin, D. R. akeeruddin, S. M. Grätzel, M.;
Nazeeruddin, M. K., Nano Letters 2009, 9 (12), 4221-4227.
39. Zewdu, T.; Clifford, J. N.; Hernandez, J. P.; Palomares, E., Energy &
Environmental Science 2011, 4 (11), 4633-4638.
40. Martinez-Ferrero, E.; Sero, I. M.; Albero, J.; Gimenez, S.; Bisquert, J.; Palomares,
E., Physical Chemistry Chemical Physics 2010, 12 (12), 2819-2821.
41. O'Regan, B. C.; Durrant, J. R.; Sommeling, P. M.; Bakker, N. J., The Journal of
Physical Chemistry C 2007, 111 (37), 14001-14010.
42. Ondersma, J. W.; Hamann, T. W., The Journal of Physical Chemistry C 2010,
114 (1), 638-645.
43. Antila, . J. eikkilä, M. J. Mäkinen, V. umalamäki, N. aitinen, M. inko,
V. Jalkanen, P. Toppari, J. Aumanen, V. Kemell, M. Myllyperki , P. onkala, K.
äkkinen, . eskelä, M.; Korppi-Tommola, J. E. I., The Journal of Physical Chemistry
C 2011, 115 (33), 16720-16729.
44. O'Regan, B. C.; Scully, S.; Mayer, A. C.; Palomares, E.; Durrant, J., The Journal
of Physical Chemistry B 2005, 109 (10), 4616-4623.
45. van de Lagemaat, J.; Park, N. G.; Frank, A. J., The Journal of Physical Chemistry
B 2000, 104 (9), 2044-2052.
46. Snaith, H. J.; Grätzel, M., Advanced Materials 2007, 19 (21), 3643-3647.
47. Bisquert, J. Fabregat-Santiago, F. Mora-Ser , I. Garcia-Belmonte, G. Gim nez,
S., The Journal of Physical Chemistry C 2009, 113 (40), 17278-17290.
48. Roelofs, K. E.; Brennan, T. P.; Dominguez, J. C.; Bailie, C. D.; Margulis, G. Y.;
Hoke, E. T.; McGehee, M. D.; Bent, S. F., The Journal of Physical Chemistry C 2013,
117 (11), 5584-5592.
Katherine Roelofs Ph.D. Dissertation
104
Chapter 5. PbS QD Synthesis by ALD
This chapter compiles the results of several studies on atomic layer deposition
(ALD) of quantum dots. We begin by describing the deposition of lead sulfide quantum
dots by ALD, and compare PbS QDs grown by ALD to those grown by the solution-
based deposition process, successive ion layer adsorption and reaction (SILAR). Several
attempts were made to improve the QD nucleation by SILAR. This is followed by details
of the fabrication of quantum-dot-sensitized solar cells (QDSSCs) with ALD-grown PbS
QDs and ALD-grown metal oxide barrier layers; notably here the QD and barrier layer
interface structure is fabricated in a single reactor, without breaking vacuum.1 Both
Al2O3 and TiO2 barrier layers are investigated. Al2O3 barrier layers grown prior to the
quantum dots resulted in a near-doubling of device efficiency (0.30% to 0.57%) whereas
Al2O3 barrier layers grown after quantum dots did not improve efficiency, indicating the
importance of quantum dots in recombination processes. Interestingly, for the TiO2
barrier layers, only layers deposited after the QDs improved device efficiencies, to
0.41%. Finally, a summary is given of X-ray absorption near-edge spectroscopy
(XANES) characterization of the interface structure of ALD QDs grown on nanoporous
TiO2 films, with the full description of the results included in Appendix A.
5.1. Introduction
We have seen in Chapter 4 that bare areas on the TiO2 surface, uncovered by
QDs, are regions through which recombination of TiO2 electrons with the hole-transport
material (HTM) can occur. As discussed, our first approach to decreasing this
recombination was to employ ultra-thin, insulating metal oxide barrier layers to coat the
TiO2 surface. However, we found that such barrier layers also introduce a resistance to
the charge transfer across the TiO2/QD/HTM interface that is necessary for photocurrent
collection.2 Another approach to limiting the bare areas on the TiO2 surface is to improve
QD loading, such that QDs cover a higher fraction of the TiO2 surface area. In the ideal
structure, QDs would form a complete monolayer coating the TiO2 surface, with each QD
slightly separated to maintain the quantum confinement effect of individual QDs.
Katherine Roelofs Ph.D. Dissertation
105
Accordingly, it is of great interest to understand how QDs nucleate and grow on the TiO2
surface, in order to improve QD deposition.
Atomic layer deposition has been portrayed in previous chapters as a method for
the deposition of ultra-thin, conformal films. However, this is not always the case:
island-type growth can occur depending on the specifics of the deposition conditions and
substrate. It is common for an island-type growth regime to occur at the beginning of
ALD film growth, during the period in which the ALD material is still growing on the
substrate itself.3 Ultimately, the ALD material coats the substrate, and subsequent ALD
layers are deposited on these initial coatings, which often leads to the transition to a true
ALD growth regime in which each layer conformally coats the substrate. For the metal
oxide substrates of interest in this dissertation – that is, the TiO2 anode of devices inspired
by dye-sensitized solar cells (DSSCs) – conformal film growth is readily achieved for
ALD films of metal oxides. For instance, with Al2O3 ALD films grown on TiO2, film
roughness features are on the order of less than 5 Å (~ 5 monolayers).4 For ALD films of
metals5, 6
or metal chalcogenides7-10
grown on metal oxide substrates, there can be a
pronounced regime where individual particles on the order of a few nanometers in
diameter are grown, due in part to disfavored surface energetics of ‘wetting’ the metal
oxide substrate. In this study, we leverage this non-ideality to grow metal chalcogenide
quantum dots by ALD. The high degree of control offered by ALD provides a good
system to study QD nucleation and growth.
The growth of QDs by ALD is also of interest in that, as a vapor phase deposition
method, ALD may produce cleaner material interfaces than a solution phase deposition
method like SILAR. Material interfaces in solar cells are critical to the collection of
photoexcited carriers.11, 12
In QDSSCs, the anode/QD/hole-transport material interface is
used to split the exciton, or bound electron-hole pair, that is generated in the QD upon the
absorption of a photon. The prominent anode material in QDSSCs is TiO2. The
energetic offset of the QD conduction band (CB) and the TiO2 CB leads to electron
injection into the TiO2, while the energetic offset of the QD valence band (VB) and the
hole-accepting level of the hole-transport material (HTM) leads to hole transfer. Voids or
Katherine Roelofs Ph.D. Dissertation
106
defects at the TiO2/QD interface could interfere with the initial electron injection and
ultimately the charge collection efficiency.
In addition, it is of interest to compare the QD precursor penetration into a
nanoporous film with the ALD process as compared to the SILAR process. In QDSSCs
with the QDs fabricated in-situ, the QDs nucleate and grow directly on the nanoporous
TiO2 anode. The nanoporous TiO2 anodes are typically between 1 to 10 μm in thickness,
with pore diameters on the order of 50 nm. A common problem limiting high QD
loading is the pore-blocking effect.13-16
Namely, QDs grow more quickly at the film
surface than the interior, due to the relative ease of access of the precursors, ultimately
leading to the larger QDs at the surface blocking pore channels into the interior of the
film. This both prevents further QD growth in the interior, as well as prevents infiltration
of the HTM, leading to a poor TiO2/QD/HTM heterojunction. It is expected that the
ALD precursors will be better able to infiltrate the nanoporous TiO2 substrates for more
uniform QD growth throughout the film thickness.
Finally, as improvements in efficiency were seen in our previous work with a
single ALD cycle of Al2O3 metal oxide films, we will also combine the ALD-grown QD
studies with ALD barrier layers. We explore both TiO2 and Al2O3 barrier layers. ALD-
TiO2 is expected to have electronic properties and band positions matching that of the
nanoporous TiO2, making it a less insulating barrier layer than Al2O3. Al2O3 has also
been observed in a variety of ALD applications to act as a seeding layer; that is,
depositing a small amount of Al2O3 (e.g. 1-2 ALD cycles) prior to ALD of a second
material tends to enhance the nucleation and subsequent growth of the second material.17-
19 This potential two-fold benefit from ALD of Al2O3 prior to PbS QD growth therefore
motivated the current investigation. The ability to deposit both the barrier layer and QD
by ALD in a single reaction chamber without breaking vacuum opens up new avenues for
studying the intricacies of substrate-absorber-barrier interaction.
We also explore the nucleation and growth of ALD QDs on the nanoporous TiO2
substrates. In this, we take advantage of the ability to grow modify the nanoporous TiO2
surface with metal oxide layers and deposit ALD QDs in the same reactor, without
breaking vacuum. The particular focus is the atomic structure of the TiO2/PbS-QD
Katherine Roelofs Ph.D. Dissertation
107
interface, which is studied by X-ray and electron diffraction, as well as X-ray absorption
near edge structure (XANES) spectroscopy in combination with density functional theory
(DFT) modeling. We find that the crystallinity of the TiO2 substrate affects the PbS QD
crystal structure, which has implications for charge collection.
5.2. Experimental Details
Fabrication of Nanostructured TiO2 Substrates. As described elsewhere,20
fluorine-doped tin oxide (FTO)-coated glass (Pilkington, 15 Ω/□) was etched using 4 M
HCl and Zn powder in order to create the desired device pattern. A compact layer of
TiO2 ~50 nm thick (to prevent contact between the FTO and the hole-transport material
(HTM)) was deposited on the substrates at 450 °C by spray pyrolysis of titanium
diisopropoxide bis(acetylacetonate) (Aldrich) in ethanol (1:9 v/v) using air as the carrier
gas. Roughly 2 µm-thick nanoporous TiO2 films were made by doctor-blading a 1:1
(w/w) ratio of terpineol and a commercial paste (Dyesol 18NR-T) atop the compact TiO2
layer followed by an anneal at 450 °C. The substrates were then left overnight in a bath
of aqueous 0.02 M TiCl4 and subsequently annealed again at 450 °C for thirty minutes.
Successive Ion Layer Adsorption and Reaction of PbS. For the standard
process, lead sulfide quantum dots were grown by SILAR employing 0.02 M aqueous
Pb(NO3)2 as the cation solution (pH 4.3), and 0.02 M aqueous Na2S as the anion solution
(pH 12.3). Nanoporous TiO2 films on FTO-coated glass were dipped in the cation
solution for 2 min, rinsed in DI water for 1 min, then dipped in the anion solution for 2
min, followed by a final dip in DI water for 1 min. Variants on this process were
explored, including varying the cation and anion solution concentrations, and varying the
dipping time and the rinsing time.
Atomic Layer Deposition of PbS, TiO2, and Al2O3. The PbS quantum dots
were deposited by using a low number of ALD cycles on nanoporous TiO2 substrates.
Bis(2,2,6,6-tetramethyl-3,5-heptanedionato)lead(II) (Pb(tmhd)2) (Strem Chemicals, Inc.),
sublimated at 140 °C, and 3.5% H2S in N2 were the deposition reagents. At a temperature
of 160 °C, the reactions took place in a customized ALD station that is H2S compatible.21
Pulse/purge times of 0.5/45 s for Pb(tmhd)2 and 0.1/40 s for H2S were used. With the
Katherine Roelofs Ph.D. Dissertation
108
same reactor and conditions, TiO2 was deposited at 160 °C using the precursors
tetrakis(dimethylamido)titanium(IV) TDMA, heated at 70 ºC, and DI water at room
temperature. Pulse/purge times of 2/45 s were used for both the TDMA and the H2O.
Likewise, in the same reactor, Al2O3 was deposited at 160 °C using the precursors
trimethylaluminum (TMA) (Aldrich) and DI water, both held at room temperature.
Pulse/purge times of 0.1/40 s were used for both TMA and H2O. Promptly after
deposition, the samples were stored in an inert environment awaiting further processing
and characterization.
QDSSC Device Fabrication. Spiro-OMeTAD (2,2',7,7'-tetrakis-(N,N-di-p-
methoxyphenylamine)-9,9'-spirobifluorene) ( umtec) was dissolved in chlorobenzene at
a concentration of 225 mL-1
. Tert-butylpyridine was added to the solution at the ratio of
1:10.3 µL:mg of spiro-OMeTAD, and lithium bis-(trifluoromethylsulfonyl)imide salt
(170 mg mL-1
in acetonitrile) was added at the ratio of 1:4.8 µL:mg of spiro-OMeTAD.
The resultant solution was deposited onto QD-coated TiO2 substrates (~30 µL per 3 cm2
substrates) and then spin-coated for 30 s at 2000 RPM. 200 nm-thick silver electrodes
were then evaporated onto the substrates under vacuum at 10-6
torr yielding device areas
~9-10 mm2.
Device Testing. Solar cell performance was measured as described elsewhere.20
Current-voltage (J-V) curves were obtained using a Xe lamp AM 1.5 solar simulator
calibrated to 1 sun. Devices were light-soaked until maximum efficiencies were obtained
(~30 minutes). Recombination lifetimes were obtained via transient photovoltage
experiments.22, 23
Devices were biased at 1 sun using a white LED array. A Keithley
2400 SourceMeter was used to keep the device current constant at a value corresponding
to a given photovoltage (the photovoltage was swept in 0.05 V increments). A white
LED (~0.05 sun) was then pulsed (50 ms square-wave) at the samples, yielding a
transient increase and subsequent in the device photovoltage. Recombination lifetimes
were obtained by fitting the exponential decay of the photovoltage. External quantum
efficiency (EQE) measurements were taken at short circuit using monochromated white
light from a 100 W tungsten lamp, which was focused through a monochromator. The
monochromated illumination, chopped at 40 Hz, was superimposed on top of a
Katherine Roelofs Ph.D. Dissertation
109
continuous-wave white light illumination (~1 sun) from a white LED array incident on
the device. The photocurrent action spectrum of the device was acquired through a lock-
in amplifier and EQE was calculated by referencing the device current to a NIST
traceable calibration photodiode. While we have not performed rigorous long-term
device stability studies, we note that device performance (and behavior, for example, in
recombination measurements) is reproducible over several-day time intervals.
Characterization via Transmission Electron Microscopy (TEM) and UV-Vis
Spectroscopy. Plain-view TEM micrographs of the samples were taken using a FEI
Tecnai G2 F20-X-TWIN at an accelerating voltage of 200 kV. Before TEM
characterization, the PbS QD/nanoporous TiO2 composite films were scratched off from
the glass substrates using a sterile razor. Afterwards, the residue was sonicated in 1 mL of
ethanol for 5 mins. A fresh ultrathin carbon on copper TEM grid (Ted Pella, Inc.) was
introduced into the solution to extract PbS QD/nanoporous TiO2 clusters.
UV-Vis measurements were obtained using a PerkinElmer Lambda 1050
UV/VIS/NIR Spectrometer. The samples were measured as deposited after the ALD step
(i.e. lacking both spiro-OMeTAD and silver electrodes). Additional spectra were
collected from complete devices to calculate the EQE. These measurements were
collected using the same tungsten lamp and monochromator as the EQE measurements
and were measured using an integrating sphere with an attached silicon photodiode.
XANES Procedures and Interpretation. The samples were characterized at
beamlines 2-1, 4-3, and 10-1 at the Stanford Synchrotron Radiation Lightsource (SSRL).
For the XAS measurements, multiple total fluorescence yield (TFY) spectra were
gathered for S K- (BL 4-3), Pb L3- (BL 4-3), O K- (BL 10-1), and Ti L2,3-edge (BL 10-1).
The Ti L2,3 edge showed similar trends as the O K-edge (Supplementary Figure 4). Total
electron yield (TEY) spectra were also gathered for oxygen and agreed with the TFY data
sets. The XAS spectra were aligned, averaged, and normalized using the SixPACK and
Athena analysis packages. Sample loading condition can be found in the Supplementary
Information. The XRD measurements were gathered at BL 2-1 using 12.5 keV x-rays and
a point detector.
Katherine Roelofs Ph.D. Dissertation
110
The multiple-scattering simulations were done using the FEF9 software package.
As shown in the supplementary information, the S K-edge was for PbS converged by
varying the cluster, self-consistent field (SCF), and scattering sphere sizes. Geometric
parameters for the distorted PbS lattice structured were taken directly from a previous
study30
with no further modifications. The S K-edge of the modified PbS clusters were
calculated using the converged parameters of the rocksalt clusters. We investigated the
modified α-GeTe and TlI structure types, along with rocksalt structures of varying lattice
constants. The models assume that there is a local distortion similar to that of the
modified lattices but do not imply that the QDs of interest have such lattices throughout.
5.3. Results and Discussion
5.3.1. PbS QD growth by ALD and SILAR
PbS quantum dots grown by ALD on nanoporous anatase TiO2 films were studied
using TEM. Figure 5-1 shows the conformal and dispersed PbS QD coverage with (a)
and without (b, c) an ALD cycle of Al2O3 between the PbS QDs and the TiO2
nanoparticles. This TEM study strongly suggests that the PbS QD coverage is not
affected by the Al2O3 surface treatment. It is important to note that it is difficult to
capture the as-deposited shape of the PbS QDs since they are sensitive to electron-beam
annealing. Annealing of the PbS QDs usually results in the dome-shaped clusters
observed in Figure 5-1, which may not be the morphology of the PbS QDs in a working
solar cell. Nevertheless, the TEM images provide key insight on the surface coverage and
dimensions of the PbS QDs. Most of the PbS QDs are between 2.8 nm and 4.8 nm in
diameter, similar in size to those employed elsewhere in TiO2-based QDSSCs.24, 25
As
shown in the inset of Figure 5-1b, the lattice spacing of a PbS QD captured by TEM has a
measured lattice constant of 5.9 Å, which agrees with the known bulk lattice constant of
PbS. TEM-EDAX analysis of the samples (Table 5-1) verifies the presence of O, S, Ti,
and Pb.
Katherine Roelofs Ph.D. Dissertation
111
Figure 5-1. Transmission electron micrographs of PbS QDs on nanoporous TiO2 (a) with
(a) and (b,c) without an intermediate Al2O3 cycle. As seen in (a) and (b), the coverage of
the QDs is not affected by Al2O3. (c) Lattice diffraction of the rock salt PbS QDs is
captured with a measured lattice constant of 5.9 Å.
Katherine Roelofs Ph.D. Dissertation
112
Table 5-1. Average atomic concentrations measured using TEM-EDS. The atomic shells
used for each element are denoted inside parentheses.
Atomic Species (shell) 10 cyc PbS 1 cyc Al2O3 + 10 cyc PbS
C (K) 54.3 50.0
O (K) 15.3 24.4
Al (K) - 0.8
S (K) 1.9 0.3
Ti (K) 8.1 10.9
Pb (L) 2.6 0.6
PbS QDs were also grown by SILAR for comparison. Figure 5-2 shows the TEM
analysis of SILAR-grown PbS QDs on anatase TiO2 nanoparticles. We observe similar
density of QDs on the TiO2 surface for deposition cycles giving the same QD size for
ALD and SILAR-grown QDs. Specifically, 6 SILAR cycles gave QDs of similar
diameter to 10 ALD cycles, and for these deposition conditions, both substrates had a
surface concentration of roughly 11,500 QDs per square micrometer. We note that the
ALD gives a slightly narrower QD size distribution as compared to the SILAR process.
A more significant difference is seen in the QD penetration into the nanoporous TiO2
substrate. Figure 5-3 shows the Auger cross-section line scan of the atomic concentration
of PbS QDs grown by SILAR as compared to ALD. We see that the SILAR and ALD
PbS QDs penetrate throughout the thickness of the TiO2 film. While it is difficult to
directly compare the total amount of PbS present in each film as the ALD PbS QD film
has a higher O and lower Ti content, it is clear that the ALD PbS film shows a marked
increase in Pb content near the film surface. To directly compare the two films, the ratio
of the atomic concentrations of Pb:Ti are plotted in Figure 5-3c with the top of each film
aligned at an x-scale position of 2.55 μm. We observe a higher Pb and S content
throughout the film for ALD PbS QDs. The large increase in Pb at the film surface in the
ALD case is not matched by a corresponding increase in S, which could be explained if
Katherine Roelofs Ph.D. Dissertation
113
the Pb near the film surface is oxidized to PbO. We note the non-stoichiometric ratio of
Ti:O (1:3) observed in the ALD case suggests some of the oxygen atoms are indeed
forming lead oxide species; in addition there is a slight decrease in the Ti:O ratio near the
film surface for the ALD case, which supports the conclusion that additional oxygen is
attached to Pb atoms, displacing sulfur.
Figure 5-2. TEM images of TiO2 nanoparticles coated with PbS QDs grown by 6 SILAR
cycles, taken from the same film. (a), and (b) show PbS QD coverages typical of higher-
coverage regions observed, while (c) shows an example from a low QD-coverage region.
Katherine Roelofs Ph.D. Dissertation
114
Figure 5-3. Auger electron spectroscopy cross-sectional line scans with SEM images of
nanoporous TiO2 films sensitized with PbS QDs. On the linescan position axis, 0 marks
the boundary between the substrate and TiO2 film, and ~2.5 μm marks the surface of the
TiO2 film. The variation in the background coloration (dark on the left and brighter on
the right) is due to variations in the signal collected from the film for the SEM, as one
TiO2 film was deposited on FTO and one on Si. (a) 6 SILAR cycles of PbS on
nanoporous TiO2 on F:SnO2 coated glass. (b) 10 ALD cycles of PbS on nanoporous TiO2
on a Si wafer. These deposition cycles of SILAR and ALD were chosen to give similar
sized QDs. (c) A comparison of the Pb:Ti and S:Ti ratios for the ALD-grown QDs and
SILAR-grown QDs.
Katherine Roelofs Ph.D. Dissertation
115
In conclusion, these results suggest that, contrary to expectations, PbS QDs can
penetration throughout a 2 micron nanoporous TiO2 film, whether deposited by SILAR or
by ALD. For the 6 SILAR cycles of PbS, no gradient in PbS is observed, which matches
visual observations that the additional darkening of the outer-edges of a film, taken to be
a sign of pore-blocking, typically occur at 10 SILAR cycles or greater. A higher surface
concentration of Pb is observed for the 10 ALD cycles of PbS, though it is unclear if the
QDs are large enough at this point that they are indeed blocking pores, or if the ALD
nucleation process lends itself to more rapid growth on the surface of a nanoporous film.
The even QD distribution on the TiO2 nanoparticles as seen in the TEM images of
both ALD and SILAR-grown QDs implies either a randomized nucleation process, for
instance nucleation on randomly distributed defect sites, or a growth mechanism that
maintains fairly even distances between growing PbS QDs on the TiO2 anode surface.
That PbS QD coverage is not affected by pre-coating the TiO2 anode with a single cycle
of Al2O3 or TiO2, or even 60 cycles of TiO2 (see the XANES study discussed below), is
evidence against QD nucleation being dependent on defect sites on the TiO2 surface, as
these pre-treatments would be expected to passivate those defects.
5.3.2. SILAR studies to increase PbS nucleation
More in-depth studies were undertaken with the SILAR process to increase the
PbS nucleation. In these studies, another benefit of the ALD process over the SILAR
process can be seen. A key component of a well-defined ALD process is saturation
behavior, where the precursor pulse time is increased to the point where a longer pulse
time will not further increase the growth rate. Physically, the amount of precursor
adsorbed onto the substrate surface has reached saturation, and further pulsing of the
precursor into the reactor will not change the precursor adsorption. The SILAR process
is quite similar to ALD, in that the substrate is exposed sequentially to the two
precursors, such that the precursors never interact directly. Accordingly, for the SILAR
process as well, a saturation point should be able to be reached, where past a given
amount of dipping time in the SILAR cation solution or SILAR anion solution, the
growth rate will not increase. Figure 5-4a shows UV-Vis spectra for PbS SILAR at
varying dipping times, with the Tauc analysis shown in Figure 5-4b. We observe that
Katherine Roelofs Ph.D. Dissertation
116
past 5 min, increased SILAR dipping time will actually decrease the total deposition,
likely due to the acidity of the Pb(NO3)2 cation solution etching away the QDs. In
addition, 5 min rinsing time in DI H2O leads to degradation of the QDs, as compared to 1
min rinsing time. From this work, we conclude that the ALD process is more
reproducible than SILAR, which is supported by narrow QD size distribution achieved by
Prinz and coworkers for PbS QD growth by ALD.9
Figure 5-4. (a) UV-visible spectra of PbS QDs grown with 0.02 M Pb(NO3)2 aqueous
solution and 0.02 M Na2S aqueous solution for 8 SILAR cycles. (a) Dip times in the
Pb(NO3)2 and Na2S solutions, as well as the rinse times in the DI H2O solution were
varied. (b) Tauc Analysis of the UV-vis spectra, with an upper and lower tangent line to
represent the range of band gaps (QD diameters) in each sample.
We explored several approaches to increase PbS QD deposition by the SILAR
process, while maintaining a narrower size distribution. First, we attempted to lower the
solution concentrations to narrow the size distribution and have a slower growth rate,
Katherine Roelofs Ph.D. Dissertation
117
which could allow for more SILAR deposition cycles before the adverse affects of pore-
blocking are observed. We did see a narrowing of the size distribution when decreasing
the cation and anion concentrations from 1 M to 0.02 M. However, increased QD density
in the TiO2 film was not achieved. The second approach was to modify the SILAR
process to have high concentration cation and anion solutions for the first initial dips for
high initial nucleation, and then decrease concentration for subsequent cycles to achieve a
slow growth period. However, the QD growth achieved with this high-nucleation/slow-
growth was identical to that achieved with just the low concentration for slow growth
throughout. A third approach was to sulfidize the TiO2 surface by heating the TiO2 in a
tube furnace with an H2/H2S environment (H2:H2S was 90:10 volume %) to create a
surface layer of TiS2. It is believed that while PbO will readily form on the TiO2 surface,
TiS2 is more difficult to form, as motivated by a previous report of sulfidization of ZnO
anodes improving CdS growth by chemical bath deposition.26
While S was observed by
XPS on the TiO2 substrates after sulfidization, the TiO2 was visibly darkened from a
brownish yellow at 350 °C to a dark brown at 450 °C. This could be either due to the
formation of TiS2 itself, which is a golden yellow color, or due to organic contaminants
on the glass of the tube furnance. This study was halted, as the TiO2 should be optically
transparent, however, further work is needed to determine if this approach could be
successful. A fourth approach was to apply a negative bias to the TiO2 anode while
immersed in the Pb(NO3)2 solution. In this study, we did succeed in increasing PbS
deposition, however, the devices produced were shorted. A subsequent report has
appeared in the literature of such potential-induced ionic layer adsorption and reaction
(PILAR), in which 4.3% efficient CdS QDSSCs are achieved. In contrast to our work
where the potential was applied continuously during the cation dip step, in this study the
authors cycled the applied potential from negative to positive bias in order to prevent
build-up of the metal cation on the TiO2 surface,13
which could explain why our devices
shorted.
A final approach was to grow PbS QDs to the near-optimal size, then cap them
with an ALD coating of TiO2, and continue PbS deposition for additional SILAR cycles.
As seen in the schematic for this approach in Figure 5-5, it was hoped that the TiO2
capping would halt further growth of the initial QDs, and that subsequent SILAR cycles
Katherine Roelofs Ph.D. Dissertation
118
would nucleate new QDs. Both 3x-PbS/2x-TiO2/3x-PbS and 5x-PbS/2x-TiO2/5x-PbS
configurations were tried, where x stands for either the number of PbS SILAR cycles or
TiO2 ALD cycles. Unfortunately, devices fabricated in this manner suffered from higher
series resistance resulting in lower efficiencies than those with either 3 or 6 PbS SILAR
cycles, or 5 or 10 PbS SILAR cycles, respectively. We believe these studies are
inherently limited by the need for a capping layer that is thick enough to prevent further
growth of the first round of QDs deposited, but still thin enough to allow for charge
transfer. Our previous reports suggest that even 1 ALD cycle of Al2O3 capping the QDs
can interfere with hole transfer to the HTM;2 however, we had hoped since TiO2 is less
insulating the Al2O3 that this would not be the case. Indeed, our studies with barrier
layers with ALD-grown PbS QDs, discussed below, indicate that 1 ALD cycle of TiO2
does not interfere with charge transfer. We did try a capping layer of 5 ALD cycles of
TiO2, in case the TiO2 layer was too thin to fully cap the first round of QDs deposited, but
these devices had ~0% efficiencies, indicating that, as expected, thicker TiO2 layers
interfere with charge collection.
Figure 5-5. (a) Schematic of approach of capping QDs with a TiO2 layer, with
subsequent SILAR deposition to nucleate new QDs. (b) J-V curves in light of three
device configurations: 3 SILAR cycles of PbS, 3 SILAR cycles of PbS capped with 2
ALD cycles of TiO2 followed by an additional 3 SILAR cycles of PbS, and 6 SILAR
cycles of PbS. (c) Corresponding J-V dark curves of the devices in (b).
Katherine Roelofs Ph.D. Dissertation
119
5.3.3. Device Performance
Figure 5-6 compares the device performance of solid-state QDSSCs with SILAR-
grown and ALD-grown QDs. We see that similar power conversion efficiencies are
observed in either case. However, the SILAR-grown QDs give higher JSC, and increasing
photocurrent when going to reverse bias. That SILAR-grown QDs have identical
absorption to ALD-grown QDs, but give higher JSC’s suggests that SI AR-grown QDs
have higher charge collection efficiencies. EQE measurements of these devices, shown
in Figure 5-6b, show similar EQE peak values for both, with a slightly higher EQE for
devices with SILAR-grown QDs at higher wavelengths.
Figure 5-6. (a) Current-voltage curve of solid-state QDSSCs with SILAR-grown or
ALD-grown PbS QDs. Shown are devices with the number of deposition cycles giving
the highest efficiency for each process. Device metrics are shown in the inset table. (b)
External quantum efficiency (EQE) of same samples, showing a peak EQE of 10%.
In DSSCs the majority of research on recombination suppression focuses on the
photoanode-HTM pathway because due to the rapid regeneration of dye molecules by
commonly-used HTMs (both liquid and solid), recombination from the photoanode to the
oxidized dye molecule is not significant. Studies of QDSSCs have shown, however, that
QD regeneration by spiro-OMeTAD is slower than dye regeneration, which means that
the photoanode-QD recombination pathway is relevant.27, 28
Another difference between
DSSC and QDSSC recombination, is that in QDSSCs recombination is restricted to QDs
Katherine Roelofs Ph.D. Dissertation
120
of materials with small bulk band gaps, like PbS: when the QDs reach a large enough
size, the QD HOMO energy level exceeds the HTM HOMO energy level and the QDs
then serve as mediators of photoanode-HTM recombination.27
In order to investigate
these various recombination scenarios, QDSSCs with TiO2 or Al2O3 barrier layers grown
after PbS were compared to those with barrier layers grown before PbS.
Starting from TiCl4-treated nanoporous TiO2 substrates (n-TiO2), five different
sample types were created in order to study light absorption and device performance: 10
cycles of PbS ALD (n-TiO2/PbS) (the control sample); 1 cycle of TiO2 ALD followed by
10 cycles of PbS ALD (n-TiO2/TiO2/PbS); 1 cycle of Al2O3 ALD followed by 10 cycles
of PbS ALD (n-TiO2/Al2O3/PbS); 10 cycles of PbS ALD followed by 1 cycle of TiO2
ALD (n-TiO2/PbS/TiO2); and 10 cycles of PbS ALD followed by 1 cycle of Al2O3 ALD
(n-TiO2/PbS/Al2O3); (schematics of these configurations Figure 5-7). We note that a
single ALD cycle of Al2O3 covers approximately 50-75% of the TiO2 surface with a
single Al2O3 layer.29
Figure 5-7. Schematics of the configurations studied. (a) Control devices with PbS
quantum dots grown on TiO2 nanoparticles, surrounded by the spiro-OMeTAD hole
transport material (HTM). (b) TiO2 deposited prior to QD growth. (c) Al2O3 deposited
prior to QD growth. In this configuration the TiO2 or Al2O3 barrier layer is expected to
slow recombination involving electrons in TiO2 and holes in the HTM. (d) TiO2
deposited after growth of QDs. (e) Al2O3 deposited after growth of QDs. In this
configuration the TiO2 or Al2O3 barrier layer is expected to slow recombination involving
electrons in TiO2 and holes in both the HTM and oxidized quantum dots. Configuration
(c) yielded the highest device efficiencies.
Katherine Roelofs Ph.D. Dissertation
121
UV-Vis spectroscopy of the PbS absorption PbS absorption indicates that ALD
TiO2 or Al2O3 deposited prior to PbS ALD does not increase PbS nucleation: absorption
levels are unchanged and there is no qualitative change in the shape of the absorption
spectrum that would indicate either ALD pre-coating alters the size distribution of
subsequent PbS QDs (Figure 5-8). This is in keeping with the TEM results in Figure 5-1.
In all configurations the onset of PbS absorption occurs near 800 nm, which suggests the
larger PbS QDs have optical band gaps around 1.6 eV and significant deviation from the
blank TiO2 substrates begins after 650 nm. Such large band gaps are needed for
favorable energetic alignment of the PbS and TiO2 conduction bands. The deviation of
the absorption onset energy from the bulk PbS band gap of 0.37 eV clearly indicate that
10 ALD cycles of PbS result in the deposition of quantum-confined structures on the
TiO2 surface with favorable absorption properties.
Figure 5-8. UV-Vis spectra of full devices. No significant changes in light absorption
were observed when a single ALD cycle of TiO2 or Al2O3 preceded PbS deposition,
indicating that neither enhanced PbS nucleation. All spectra exhibit interference fringes
common in thin films. The UV-Vis spectra represent data averaged from multiple
devices.
In contrast to the absorption data, device performance was dramatically different
for the TiO2/Al2O3/PbS configuration versus the other two: a single cycle of Al2O3 ALD
Katherine Roelofs Ph.D. Dissertation
122
improved device efficiency by nearly a factor of two (from 0.30% to 0.57%) whereas no
appreciable change was observed for the TiO2/PbS/Al2O3 configuration (Table 5-2,
Figure 5-9 and Figure 5-10b). The improved performance in the TiO2/Al2O3/PbS
configuration arises from a much higher JSC and a modestly improved VOC. We note that
the 0.57% efficiency for the TiO2/Al2O3/PbS configuration is on par with that reported by
Lee et al. (0.72%) in a device otherwise identical but for the fact that SILAR was used to
grow the PbS QDs.27
Lee et al. also observed that using larger TiO2 nanoparticles for the
active layer (which results in larger pores, and hence, more facile HTM pore-filling)
doubled device efficiency, a strategy that may also prove fruitful with ALD-grown PbS.27
A single ALD cycle of TiO2 deposited prior to the PbS QDs had no significant
change on device efficiency (Table 5-2, Figure 5-9 and Figure 5-10b). However, an ALD
cycle of TiO2 deposited after the PbS QDs did lead to a slight increase in efficiency, to
0.41%, largely through an increase in JSC. We note that no corresponding increase is seen
in the EQE spectra of these n-TiO2/PbS/TiO2 devices (Figure 5-10c), which indicates that
one of these measurements is not capturing the full picture – see below for further
discussion of possible inconsistencies in the EQE measurement system.
Table 5-2. Average device parameters for the five configurations tested. A substantial
improvement in efficiency was achieved by performing a single ALD cycle of Al2O3
prior to PbS growth whereas no improvement was observed with a single Al2O3 ALD
cycle after PbS growth. The efficiency improvement is driven by a large increase in JSC
and a smaller increase in VOC. Values were obtained by taking the average of the top two
of four devices (~10 mm2
in area) on a given substrate and then averaging over the four
substrates tested with each configuration.
JSC
(mA/cm2) VOC (V) Fill Factor Efficiency
(%)
n-TiO2/PbS 0.89 0.51 0.69 0.30
n-TiO2/TiO2/PbS 1.21 0.51 0.56 0.31
n-TiO2/Al2O3/PbS 1.73 0.58 0.58 0.57
n-TiO2/PbS/TiO2 1.50 0.48 0.59 0.41
n-TiO2/PbS/Al2O3 0.93 0.49 0.57 0.26
Katherine Roelofs Ph.D. Dissertation
123
Figure 5-9. Average device parameters for the five configurations tested, listed in Table
5-2, showing the standard deviations calculated from the best two devices of four
substrates tested, eight devices in total, for each configuration.
The fact that JSC improved dramatically with a pre-PbS ALD cycle of Al2O3
despite the absence of changes to the QDs in quantity or size suggests that the barrier
layer aspect of Al2O3 is significant. In order to understand the effect of Al2O3 on
recombination we collected dark current-voltage (J-V) curves and external quantum
efficiency (EQE) data and performed transient photovoltage analysis,30
which is
particularly well-suited for measuring recombination with the spiro-OMeTAD HTM.
EQE results (Figure 5-10c) indicate that the superior performance of the TiO2/Al2O3/PbS
configuration is due to enhanced efficiency across a wide spectral range, which, coupled
with the lack of change in light absorption observed in the UV-Vis data (Figure 5-8a),
suggests an improvement in internal quantum efficiency (IQE) in the TiO2/Al2O3/PbS
configuration. A decrease relative to control devices larger than that observed in the J-V
measurements is apparent in the EQE data for the TiO2/PbS/Al2O3 configuration; we
attribute this unusually low EQE to the fact that the bias light source used in EQE
measurements may alter the effects of light-soaking as compared to the AM1.5G light
spectrum used in J-V measurements.
Katherine Roelofs Ph.D. Dissertation
124
Figure 5-10. (a) Dark current-voltage (J-V) curves for typical devices of the five
configurations studied. Whether deposited before or after the PbS quantum dots, a single
cycle of Al2O3 suppressed the onset of dark current to a similar degree. The value in
parentheses corresponds to a current-threshold voltage, that is, the voltage at which the
dark current in the devices exceeded an arbitrary value (here, 0.4 mA cm-2
). These
current threshold values are averaged over all the devices for that particular
configuration. (b) Illuminated current-voltage curves for typical devices of the five
configurations studied. (c) External quantum efficiency (EQE) spectra of devices in each
of the device configurations collected at zero bias, data is averaged from multiple
Katherine Roelofs Ph.D. Dissertation
125
devices. When Al2O3 precedes PbS QD growth, a substantial improvement in both EQE
and JSC is observed.
Dark J-V curves (Figure 5-10a) indicate that Al2O3 reduces dark current whether
it is deposited before or after the PbS. The transient photovoltage studies show that in
both instances Al2O3 increases recombination lifetime relative to control cells (Figure
5-11), but the enhancement is most apparent at higher photovoltages in the n-
TiO2/Al2O3/PbS configuration and at lower photovoltages in the n-TiO2/PbS/Al2O3
configuration. We attribute the lifetime enhancement in the n-TiO2/Al2O3/PbS
configuration to suppression of recombination with (or mediated by) oxidized PbS QDs.
The recombination suppression in the n-TiO2/PbS/Al2O3 configuration is attributed to
blocking of recombination with holes in spiro-OMeTAD (we note that the recombination
pathway from the PbS excited state to spiro-OMeTAD, which should also be blocked in
this configuration, is not detectable via transient photovoltage measurements). We
speculate that the reason recombination to oxidized QDs appears to dominate at higher
photovoltages and recombination to spiro-OMeTAD dominates at lower photovoltages is
due to the different routes and energetics for each process (the model described by Hod et
al. wherein TiO2-QD interface states mediate recombination is particularly illustrative
here31
). We note that the overall higher lifetimes observed in with Al2O3 deposited
before the QDs matches our previous results with Al2O3 ALD barrier layers in CdS
QDSSCs.2
Given the slight lifetime enhancement observed in the TiO2/PbS/Al2O3
configuration, it is therefore surprising that JSC and EQE did not increase as in the
TiO2/Al2O3/PbS configuration: this could be due to the Al2O3 on PbS blocking
regeneration of the PbS QDs, which has been observed previously with CdS QDs and
Al2O3.2 This is consistent with the observed increase in JSC in the TiO2/Al2O3/PbS
configuration but it is also possible that Al2O3 increases JSC in this cell configuration by
improving charge injection efficiency by lowering the TiO2 conduction band, raising the
PbS conduction band, or reducing surface states at the PbS-TiO2 interface that participate
in trap-mediated recombination. Further investigation is needed to probe this
Katherine Roelofs Ph.D. Dissertation
126
mechanism, which is an intriguing contrast to the detrimental effects of Al2O3 on
injection efficiency observed in DSSCs.4, 29
For the TiO2 layers, the TiO2 ALD cycle deposited before the PbS QD gives
similar dark current results as the control of just the PbS QDs, and the transient
photovoltage studies show that the recombination lifetimes are quite close to the control
as well. This matches the results of the J-V performance under light. While such a
coating of TiO2 might be expected to passivate dangling surface bonds, potential sources
of recombination, previous work has shown that in these systems, the standard TiCl4
treatment of the nanoporous TiO2 prior to absorber deposition plays a similar role in
reducing surface recombination.4 We believe the TiCl4 treatment applied to the
nanoporous TiO2 films prior to ALD is why the ALD TiO2 deposited before the PbS QDs
has no impact in this study, unlike the ALD Al2O3, which is more insulating than TiO2
and better able to act as a barrier layer at the interface. A single TiO2 ALD cycle
deposited after the PbS QDs lead to a more rapid onset in the dark current, in both the
dark J-V curves and the illuminated J-V curves. This is in contrast to the Al2O3 results,
where a post-QD Al2O3 coating actually suppressed the dark current. Likely, the Al2O3,
when deposited before or after the QDs, is blocking recombination from the TiO2 anode
to the spiro-OMeTAD due to its insulating nature. However, as seen in the n-
TiO2/TiO2/PbS case, the TiO2 ALD layer after the QD does not have any beneficial effect
of decreasing TiO2 anode-to-HTM recombination. The negative effect that is newly
observed in the n-TiO2/PbS/TiO2 case could be due to the TiO2 layer (or the residual
ligands from the TDMA precursor) blocking hole transfer from the PbS QD to the spiro-
OMeTAD.
Katherine Roelofs Ph.D. Dissertation
127
Figure 5-11. Electron lifetimes for the five different device configurations as determined
by fitting the photovoltage decay obtained from transient photovoltage experiments. At
higher photovoltages the recombination suppression from depositing Al2O3 prior to PbS
growth is readily apparent and is superior to Al2O3 deposited after PbS, although
recombination is also reduced in that configuration, particularly at lower photovoltages.
Error bars indicate standard deviations of the several devices tested.
5.4. Quantifying Geometric Strain at the ALD PbS
QD/ TiO2 Interface
ALD growth of PbS QDs also provides a unique opportunity to investigate the
TiO2/PbS-QD interface in a highly controlled system. Here, we give a summary of the
results from this study, with the full report in Appendix A. In these studies, PbS QDs
were grown directly on the crystalline anatase TiO2 nanoporous substrates (referred to as
the ‘AN’ set), as well as on nanoporous TiO2 substrates coated with amorphous ALD
TiO2 (referred to as the ‘AM’ set). The exact nature of such surface modifications is
often difficult to determine because X-ray based techniques such as diffraction,
reflectivity, and scattering provide only basic information about bulk crystallinity and
Katherine Roelofs Ph.D. Dissertation
128
long-range order. Further, inherently local techniques, such as transmission electron
microscopy (TEM), scanning tunneling microscopy (STM) and extended x-ray
absorption fine structure (EXAFS), are limited when determining atomic structure of
scattered or buried nanomaterials.32
Due to its high sensitivity to interatomic distances,
geometry, and oxidation state,33
and its ability to characterize subsurface materials,34, 35
X-ray absorption spectroscopy (XAS) is an indispensable tool to address atomic-scale
engineering challenges, in particular, close analysis of the X-ray absorption near edge
structure (XANES) and near-edge X-ray absorption fine structure (NEXAFS) spectral
regions remains an emerging technique.36-40
In this work, we exploit electronic and geometric information contained in sulfur
and oxygen K-edge XANES spectra to resolve interfacial structure of a metal-sulfide
nanoparticle grown on a nanostructured-metal-oxide surface (i.e., PbS QDs on TiO2
nanoparticles). Of particular interest is how the TiO2 crystal structure affects the
nucleation of PbS nanocrystals; accordingly, we studied PbS QD growth on a film of
anatase TiO2 nanoparticles, as compared to PbS QD growth on an identical substrate
precoated with a layer of amorphous TiO2. Lattice mismatch between metal-sulfide
sensitizer and the TiO2 anode produces bonding distortions, whereas an amorphous TiO2
layer between the anode and QD is found to help preserving the original crystal structure
of QD. The lattice disruption in PbS, observed in X-ray diffraction (XRD) and selected
area electron diffraction (SAED) can be further quantified as bonding angle distortions
through interpretation of XAS spectra with density-functional theory (DFT) calculations.
The differences in bonding environment and the resulting interfacial electronic band
structure suggest potential impacts on device performance.
Atomic layer deposition (ALD) was employed to deposit the QD and metal oxide
layers on the TiO2 anode, due to its self-limiting, layer-by-layer nature. The use of ALD
to create highly controlled nanoscale architectures in combination with XAS allows us to
probe properties of atomically engineered interfaces. We use this approach in the present
work to characterize the coordination environment at the interface between PbS QDs and
TiO2 nanoparticles (NPs) in a quantum-dot-sensitized solar cell architecture. Schematics
of our sample architectures are shown in Figure 5-12. We analyze two sets of samples for
Katherine Roelofs Ph.D. Dissertation
129
comparison: the first, labeled AN for “anatase”, consists of PbS QDs of various sizes
deposited by ALD directly on anatase TiO2 NPs, while the second, labeled AM for
“amorphous”, introduced an amorphous 60-cycle ALD TiO2 film between anatase NPs
and PbS QDs. In both cases, QDs were created in separate samples with 10, 20, and 40
ALD cycles of PbS.41
Figure 5-12. Schematics of sample architectures. The sample architectures for the AN (a)
and AM (b) sample sets are shown. For both sets, 10, 20, and 40 ALD cycles (10x, 20x, and 40x)
of PbS were deposited on the nanoporous TiO2 substrates. The resulting PbS QDs were capped
with 60 cycles of ALD TiO2, which is amorphous.
To determine the interfacial structure, we performed X-ray and electron
diffraction measurements on both sample sets. Figure 5-13a shows X-ray diffraction
(XRD) data for each sample and anatase TiO2 NPs for reference. All samples from the
AM group show PbS crystallinity, as seen by the appearance of (111) and (200) rocksalt
PbS peaks in Figure 5-13a. However, in the AN samples there is no evidence of
crystallinity in the QDs until 40x PbS, in agreement with the emergence of peak A4 in
the S K-edge measurements that will be presented later. Selected area electron diffraction
(SAED) measurements match the XRD results (Figure 5-13b-e). For the AM set, SAED
Katherine Roelofs Ph.D. Dissertation
130
patterns exhibit diffraction rings which are assigned to (111), (200) and (220) crystal
planes of rocksalt PbS QDs. Larger diffraction spots in SAED of 40x PbS (AM) indicates
the increase in the size of the PbS QDs in the samples formed with higher number of
ALD cycles. For the 10x PbS (AN) no crystalline phase of PbS was observed from the
SAED pattern. Some crystallinity of PbS QDs is observed in the 40x PbS (AN) sample.
Moreover, SAED allowed us to locally explore the possibility of an alternate crystalline
phase in the AN PbS QDs. No evidence of an alternate crystal structure was found
(Figure 5-13d,e), suggesting that the disorder is non-crystalline and amorphous (further
discussion in Supporting Information). Simulated SAED patterns of anatase TiO2 (white)
and rocksalt PbS (yellow) are shown for comparison. The SAED regions for the 10x PbS
AN and AM samples are shown in Figure 5-13f and g, respectively. Overall, the
diffraction results suggest that the interface between the PbS QDs and anatase TiO2 NPs
prevents the formation of a consistent PbS lattice spacing; which could be due to bonding
strain between the anatase TiO2 and the rocksalt PbS.
Katherine Roelofs Ph.D. Dissertation
131
Figure 5-13. Diffraction and TEM characterization of PbS QDs on TiO2 nanoparticles.
(a) XRD measurements show the presence of rocksalt PbS (111) and (200) peaks in the
AM set while there is no evidence of crystallinity other than the substrate for the AN set.
SAED measurements were done on 10x PbS and 40x PbS for the AN (b,c) and AM (d,e)
sets. BF-TEM images of (f) 10x PbS AN and (g) 10x PbS AM show SAED regions with
the embedded PbS QDs, respectively.
Katherine Roelofs Ph.D. Dissertation
132
To further quantify the amorphous structures as a result of strain at the interface,
XAS spectra were collected. The sulfur K-edge XAS spectrum of crystalline PbS was
benchmarked using reference samples consisting of commercial PbS powder, colloidal
PbS QDs (5 nm), and 200x ALD PbS grown on TiO2 NPs. As shown in Figure 5-14a,
four main features in the total fluorescence yield sulfur K-edge XAS spectrum, labeled
A1-A4, are seen in each of the reference samples. Features A1 and A2, in the XANES
region, correspond to electronic excitations from S 1s states to unoccupied states in the
PbS conduction band, which is composed of the bonding environment between Pb 6p and
S 3p orbitals.42
Features A3 and A4, in the near-edge X-ray absorption fine structure
(NEXAFS) region result from short and long range multiple-scattering effects arising
from geometric arrangements of Pb and S atoms in PbS.42-44
Spectra in Figure 5-14b correspond to experimental QDs from the AN and AM
sets. The AN spectra deviate from reference samples in features A3 and A4 of the sulfur
K-edge, suggesting that long—range order or local atomic arrangement does not
correspond to rocksalt PbS. Note that a lack of feature energy shifts indicate no change in
oxidation state, which remains -2. Thus, changes in spectral feature A3 and A4 do not
suggest significant change in Pb and S bonding but point to a structural arising from
measurable deviation from the rocksalt octahedral coordination. At 40x PbS, the AN
spectrum starts to resemble the reference samples, indicating the ability of larger QDs to
overcome the interfacial effect causing strain in the PbS structure and realigning into an
octahedral coordination.
In contrast to QDs grown on anatase TiO2, each sample in the AM set preserves
the reference spectral features (Figure 5-14b, top), demonstrating that QDs deposited on
an amorphous TiO2 layer exhibit rocksalt crystallinity even at smaller sizes. These results
are consistent with the SAED and XRD findings, discussed above. This indicates that the
bonding environment of an amorphous TiO2 surface allows the energetically-favorable
rocksalt arrangement of Pb and S atoms. Similar trends are observed for Pb L3 XAS in
the AN and AM sets (Appendix A).
To probe the bonding environment of TiO2, reference O K-edge spectra were
gathered for anatase TiO2 NPs in nanostructured TiO2 film substrates, an anatase TiO2
Katherine Roelofs Ph.D. Dissertation
133
ALD film, and an amorphous TiO2 ALD film. Figure 5-14c has five characteristic
features of the O K-edge. Features B1 and B2 correspond to transitions from O 1s states
to unoccupied hybridized orbitals containing dominantly Ti 3d and O 2p character,45
while hybridization between Ti 4p and O 2p creates unoccupied states for the B3 and B4
transitions.45-47
Feature B5 results from multiple scattering of the photoexcited wave and
is only pronounced in the anatase samples. Broadening of all features is observed in the
amorphous ALD TiO2 film, which is due to non-periodic bonding environments.48, 49
The
O K-edge features for the AN and AM sets in Figure 5-14d are primarily probing the
amorphous ALD TiO2 layers (the 60x outer-coating of ALD TiO2 deposited on each
sample to protect the PbS QDs from oxidiation prior to XAS), and very little variation in
the spectra are observed as expected. The Ti L2,3 edge showed similar trends as the O K-
edge (Appendix A).
The amorphous TiO2 data indicate that O 2p/3p and Ti 3d/4p orbitals are not
periodically bound as in anatase, allowing small distortions in individual Ti and O orbital
mixing and bonding. Combining this result with the S K-edge XAS indicates that non-
periodicity in amorphous TiO2 accomodates the sharp transition to periodic bonds in PbS
without significant interfacial strain between TiO2 and PbS. The (101) orienation is the
prevalent surface of anatase TiO2 NPs,50
which results in a significant lattice mismatch
between the anatase (a = 3.78 Å , b = 9.52 Å) and rocksalt (a = 5.94 Å) structures. Thus,
the interaction energy at the PbS/anatase NP interface is greater than the energy penalty
to distort the rocksalt structure in PbS.
Katherine Roelofs Ph.D. Dissertation
134
Figure 5-14. X-ray absorption characterization of S K and O K edges. The sulfur and
oxygen XANES K-edges are shown for the reference samples (a, c) and the PbS QD
sample sets (b, d). (a) The reference samples for the S K-edge depict four features that are
due to electronic transitions into unoccupied orbitals (peaks A1 and A2) and features
resulting from multiple-scattering of photoelectrons (peaks A3 and A4). (b) Similarly, the
O K-edge can be analyzed from orbital contributions (peaks B1-B4) and multiple-
scattering effects (peak B5). The broadening of the features in amorphous TiO2 are due to
Katherine Roelofs Ph.D. Dissertation
135
perturbations to Ti and O bonding interaction. The S-K XANES for the AM set look
similar to the references suggesting a rocksalt atomic environment (the features become
more pronounced from 10x to 40x PbS because as the QDs grow the number of sulfur
atoms found in the octahedral environment increases), while there are clear distinctions in
the AN set, specifically in the A4 peak (starred). (d) The O-K XANES for the AM and
AN sets have features that confirm that ALD TiO2 on anatase TiO2 and PbS QDs,
respectively, is amorphous.
Overall, we have demonstrated the combined use of XANES and ALD to
understand and manipulate the atomic arrangement of quantum dots in a quantum-dot-
sensitized solar cell architecture. From local structure measurements and quantum
simulations, we found that the periodic bonds at the anatase TiO2 surface compromise the
crystal structure in bottom-up deposited PbS QDs by inducing asymmetric Pb—S bonds,
whereas a non-periodic TiO2 surface structure does not distort the rocksalt structure in
PbS. Moreover, DFT modeling of the system indicate that the geometric distortions alter
the electronic structure of the PbS QDs by increasing their energy gap at the interface.
Quantum simulations and DFT modeling are included in Appendix A. We believe our
approach lays out a novel framework for how XAS can be exploited to provide geometric
and electronic information over the interfacial bonding environment of nanomaterials for
the development of nanostructured devices.
5.5. Conclusions
In conclusion, PbS QDs grown by ALD and SILAR appear to have similar QD
loading on the nanoporous TiO2 film substrates, when compared by TEM and UV-Vis.
Auger electron spectroscopy elemental linescans of film cross-sections indicate in both
cases that the PbS QDs are able to completely penetrate a ~2.5 micron TiO2 film. The
AES linescans indicate that the ALD-grown PbS QDs have a gradient of increased
growth at the top of the film, and slightly higher growth throughout the film as compared
to the SILAR-grown QDs. The surface gradient in with the ALD-grown QDs suggests
that higher ALD QD loading could be achieved throughout the film, perhaps by applying
a pump blocking procedure used in other ALD studies of nanoporous substrates.
However, the AES results are not conclusive due to the limited sample size, as just one
Katherine Roelofs Ph.D. Dissertation
136
film was studied in each case. We believe the UV-Vis data, supplemented by the TEM
results, is a more robust indicator of QD loading. It is quite surprising that this data
indicates similar QD loading for ALD-grown and SILAR-grown QDs, despite the
different precursor chemistries, and that one is a vapor-phase process under vacuum
conditions and another is a solution process. Moreover, we have also observed no
significant impact on PbS QD growth when coating the anatase TiO2 anode with 1 ALD
cycle of TiO2 or Al2O3, or even 60 TiO2 ALD cycles giving a quite thick amorphous TiO2
layers (in the XANES study). This shows that PbS QD growth does not depend on
nucleation via defect sites on the TiO2 surface, as these ALD treatments are expected to
passivate such defects. Further, the even spacing of QDs in the TEM results indicates
that the QD nucleation and growth occurs through a mechanism that maintains a fairly
even distribution of PbS QDs on the TiO2 anode surface, with the number of QDs not
significantly changing with increased deposition cycles (see Chapter 6). For instance, the
QDs could nucleate by a random process, leading to an even distribution across the
surface, and once QDs are nucleated on the surface, further nucleation could be
suppressed due to favorable deposition of Pb and S atoms on pre-existing QDs.
With the barrier layer studies, we have demonstrated the application of ALD to
grow lead sulfide QD absorbers for use in a solid-state QDSSC. Metal oxide barrier
layers were deposited in the same reactor without a vacuum break allowing for rigorous
control over device interfaces. A single ALD cycle of Al2O3 prior to PbS growth
improved efficiency nearly twofold and the resultant devices were on par with spiro-
based cells in the literature utilizing the SILAR growth method. The improvement in
device efficiency observed in the TiO2/Al2O3/PbS configuration as compared to the
TiO2/PbS/Al2O3 configuration was attributed to the greater importance of blocking
recombination at the TiO2-PbS interface than the TiO2-HTM interface as indicated by the
large improvement in EQE observed in the TiO2/Al2O3/PbS configuration. This result
highlights the importance of recombination with oxidized QDs and not just with the hole-
transport material. The growth scheme employed in this work is highly modular and
further work is underway to optimize performance by adjusting the number of ALD
cycles and material choices for absorber, barrier, and passivation layers. We believe
Katherine Roelofs Ph.D. Dissertation
137
fundamental interfacial studies enabled by this single-chamber deposition method will
further understanding and guide future progress in QDSSCs.
In addition, we have demonstrated the combined use of XANES and ALD to
understand and manipulate the atomic arrangement of quantum dots in a quantum-dot-
sensitized solar cell (QDSSC) architecture. From local structure measurements and
quantum simulations, we found that the periodic bonds at the anatase TiO2 surface
compromise the crystal structure in bottom-up deposited PbS QDs by inducing
asymmetric Pb—S bonds, whereas a non-periodic TiO2 surface structure does not distort
the rocksalt structure in PbS. Moreover, DFT modeling of the system indicate that the
geometric distortions alter the electronic structure of the PbS QDs by increasing their
band gap at the interface (see Appendix A). We believe our approach lays out a novel
framework for how XANES can be exploited to provide geometric and electronic
information over the interfacial bonding environment of nanomaterials for the
development of nanostructured devices.
5.6. Financial Support, Collaborations, and
Copyrights
This chapter represents the fruition of a collaborative effort between Orlando
Trejo and John Xu from Fritz Prinz’s research group in Mechanical Engineering at
Stanford, and Thomas Brennan and myself from Stacey Bent’s group in Chemical
Engineering. We would like to thank Michael McGehee for the generous use of lab
space, device fabrication procedures, and Colin Bailie, Eric Hoke, and George Margulis
for training on transient photovoltage measurements.
The work in Sections 5.3.1 and 5.3.2 on Al2O3 barrier layers with ALD-grown
PbS QDs in this chapter is adapted from work originally published in the Journal of
Materials Chemistry A with Tom Brennan and Orlando Trejo contributing equally as the
leading authors. Adapted from Ref. 1 with permission from The Royal Society of
Chemistry. The details of that study are expanded and included in full length here. The
Katherine Roelofs Ph.D. Dissertation
138
work on TiO2 barrier layers with ALD-grown PbS QDs, and the comparison of ALD and
SILAR growth of PbS QDs is published for the first time in this dissertation. The work
in Section 5.3.2 on ALD-grown versus SILAR-grown PbS QDs was conducted by
Katherine Roelofs; Janny Cahatol and Susan Cooper also contributed to the SILAR
studies.
The work in Section 5.4 on the XANES studies of the TiO2-PbS QD interface has
been submitted for publication, with Orlando Trejo as the leading author and Katherine
Roelofs as the second author.51
Orlando Trejo lead the analysis of the XANES through
the DFT simulation of XANES spectra (Appendix A), while Katherine Roelofs lead the
sample fabrication; both worked to collect the XANES data.
This work was supported by the Center for Advanced Molecular Photovoltaics
(Award No. KUS-C1-015-21), made by the King Abdullah University of Science and
Technology (KAUST).
5.7. References
1. Brennan, T. P.; Trejo, O.; Roelofs, K. E.; Xu, J.; Prinz, F. B.; Bent, S. F., J. Mat.
Chem. A 2013, 1 (26), 7566-7571.
2. Roelofs, K. E.; Brennan, T. P.; Dominguez, J. C.; Bailie, C. D.; Margulis, G. Y.;
Hoke, E. T.; McGehee, M. D.; Bent, S. F., The Journal of Physical Chemistry C 2013,
117 (11), 5584-5592.
3. George, S. M., Chemical Reviews 2010, 110 (1), 111-131.
4. Brennan, T. P.; Bakke, J. R.; Ding, I. K.; Hardin, B. E.; Nguyen, W. H.; Mondal,
R.; Bailie, C. D.; Margulis, G. Y.; Hoke, E. T.; Sellinger, A., et al., PCCP 2012, 14 (35),
12130-12140.
5. Hämäläinen, J.; Ritala, M.; Leskelä, M., Chem. Mater. 2014, 26 (1), 786-801.
6. Lee, H.-B.-R.; Baeck, S. H.; Jaramillo, T. F.; Bent, S. F., Nano Lett. 2013, 13 (2),
457-463.
7. Bakke, J. R.; Jung, H. J.; Tanskanen, J. T.; Sinclair, R.; Bent, S. F., Chem. Mater.
2010, 22, 4669-4678.
8. Bakke, J. R.; King, J. S.; Jung, H. J.; Sinclair, R.; Bent, S. F., Thin Sol. Films
2010, 518, 5400-5408.
9. Dasgupta, N. P.; Jung, H. J.; Trejo, O.; McDowell, M. T.; Hryciw, A.;
Brongersma, M.; Sinclair, R.; Prinz, F. B., Nano Lett. 2011, 11 (3), 934-940.
10. Dasgupta, N. P.; Meng, X.; Elam, J. W.; Martinson, A. B. F., Acc. Chem. Res.
2015, 48 (2), 341-348.
Katherine Roelofs Ph.D. Dissertation
139
11. Bakke, J. R.; Pickrahn, K. L.; Brennan, T. P.; Bent, S. F., Nanoscale 2011, 3 (9),
3482-3508.
12. Roelofs, K. E.; Brennan, T. P.; Bent, S. F., J. Phys. Chem. Lett. 2014, 5 (2), 348-
360.
13. Liu, I. P.; Chang, C.-W.; Teng, H.; Lee, Y.-L., ACS Appl. Mater. Inter. 2014, 6
(21), 19378-19384.
14. Yun, H. J.; Paik, T.; Edley, M. E.; Baxter, J. B.; Murray, C. B., ACS Appl. Mater.
Inter. 2014, 6 (5), 3721-3728.
15. Song, X.; Wang, M.; Xing, T.; Deng, J.; Ding, J.; Yang, Z.; Zhang, X., J. Power
Sources 2013, 253 (0), 17-26.
16. Jin-nouchi, Y.; Naya, S.-i.; Tada, H., J. Phys. Chem. C 2010, 114 (39), 16837-
16842.
17. Wilson, C. A.; Grubbs, R. K.; George, S. M., Chemistry of Materials 2005, 17
(23), 5625-5634.
18. Elam, J. W.; Libera, J. A.; Pellin, M. J.; Zinovev, A. V.; Greene, J. P.; Nolen, J.
A., Applied Physics Letters 2006, 89 (5), 053124-3.
19. Standridge, S. D.; Schatz, G. C.; Hupp, J. T., Langmuir 2009, 25 (5), 2596-2600.
20. Leijtens, T.; Ding, I. K.; Giovenzana, T.; Bloking, J. T.; McGehee, M. D.;
Sellinger, A., ACS Nano 2012, 6 (2), 1455-1462.
21. Dasgupta, N. P.; Mack, J. F.; Langston, M. C.; Bousetta, A.; Prinz, F. B., Review
of Scientific Instruments 2010, 81 (4), 044102-6.
22. O'Regan, B. C.; Lenzmann, F., The Journal of Physical Chemistry B 2004, 108
(14), 4342-4350.
23. O'Regan, B. C.; Durrant, J. R.; Sommeling, P. M.; Bakker, N. J., The Journal of
Physical Chemistry C 2007, 111 (37), 14001-14010.
24. Hyun, B.-R. hong, .-W. Bartnik, A. C. Sun, . Abru a, H. D.; Wise, F. W.;
Goodreau, J. D.; Matthews, J. R.; Leslie, T. M.; Borrelli, N. F., ACS Nano 2008, 2 (11),
2206-2212.
25. Snaith, H. J.; Stavrinadis, A.; Docampo, P.; Watt, A. A. R., Solar Energy 2011,
85 (6), 1283-1290.
26. Edri, E.; Rabinovich, E.; Niitsoo, O.; Cohen, H.; Bendikov, T.; Hodes, G., J.
Phys. Chem. C 2010, 114 (30), 13092-13097.
27. Lee, H.; Leventis, H. C.; Moon, S.-J.; Chen, P.; Ito, S.; Haque, S. A.; Torres, T.;
Nüesch, F.; Geiger, T.; Zakeeruddin, S. M., et al., Advanced Functional Materials 2009,
19 (17), 2735-2742.
28. Leventis, H. C.; O'Mahony, F.; Akhtar, J.; Afzaal, M.; O'Brien, P.; Haque, S. A.,
Journal of the American Chemical Society 2010, 132 (8), 2743-2750.
29. Antila, L. J. eikkilä, M. J. Mäkinen, V. umalamäki, N.; Laitinen, M.; Linko,
V.; Jalkanen, P.; Toppari, J.; Aumanen, V.; Kemell, M., et al., The Journal of Physical
Chemistry C 2011, 115 (33), 16720-16729.
30. Barnes, P. R. F.; Miettunen, K.; Li, X.; Anderson, A. Y.; Bessho, T.; Grätzel, M.;
O'Regan, B. C., Advanced Materials 2013, 25 (13), 1881-1922.
31. Hod, I.; González-Pedro, V.; Tachan, Z.; Fabregat-Santiago, F.; Mora-Seró, I.;
Bisquert, J.; Zaban, A., The Journal of Physical Chemistry Letters 2011, 2 (24), 3032-
3035.
32. Billinge, S. J. L.; Levin, I., Science 2007, 316, 561-565.
Katherine Roelofs Ph.D. Dissertation
140
33. Levina, A.; Armstrong, R. S.; Lay, P. A., Coordination Chemistry Reviews 2005,
249, 141-160.
34. Tröger, L.; Arvanitis, D.; Rabus, H.; Wenzel, L.; Baberschke, K., Physical Review
B 1990, 41 (10), 7297-7300.
35. Becker, M. A.; Radich, J. G.; Bunker, B. A.; Kamat, P. V., J. Phys. Chem. Lett.
2014, 5 (9), 1575-1582.
36. Rehr, J. J.; Ankudinov, A. L., Coordination Chemistry Reviews 2005, 249, 131-
140.
37. Koide, A.; Fujikawa, T.; Ichikuni, N., Journal of Electron Spectroscopy and
Related Phenomena 2014, 195, 375-381.
38. Kuzmin, A.; Chaboy, J., IUCrJ 2014, 1 (Pt 6), 571-89.
39. Henderson, G. S.; de Groot, F. M. F.; Moulton, B. J. A., Reviews in Mineralogy
and Geochemistry 78 (1), 75-138.
40. Li, Z.; Dervishi, E.; Saini, V.; Zheng, L.; Yan, W.; Wei, S.; Xu, Y.; Biris, A. S.,
Particulate Science and Technology 2010, 28 (2), 95-131.
41. Dasgupta, N. P.; Jung, H. J.; Trejo, O.; McDowell, M. T.; Hryciw, A.;
Brongersma, M.; Sinclair, R.; Prinz, F. B., Nano Lett. 2011, 11, 934-940.
42. Oertzen, G. U. v.; Jones, R. T.; Gerson, A. R., Phys Chem Minerals 2005, 32,
255-268.
43. Soldatov, A. V.; Ivanchenko, T. S.; Bianconi, A., Solid-State Spectroscopy 1994,
3, 372-374.
44. Lavrentyev, A. A.; Gabrelian, B. V.; Nikiforov, I. Y.; Rehr, J. J., Journal of
Physics and Chemistry of Solids 1999, 60, 787-790.
45. de Groot, F.; Grioni, M.; Fuggle, J.; Ghijsen, J.; Sawatzky, G.; Petersen, H., Phys.
Rev. B 1989, 48, 5715-5723.
46. de Groot, F.; Faber, J.; Michiels, J.; Czyżyk, M.; Abbate, M.; Fuggle, J., Phys.
Rev. B 1993, 48, 2074-2080.
47. Wu, Z. Y.; Ouvrard, G.; Gressier, P.; Natoli, C. R., Phys. Rev. B 1997, 55, 10382-
10391.
48. Kucheyev, S. O.; van Buuren, T.; Baumann, T. F., Phys. Rev. B 2004, 69,
245102-245102.
49. Zhang, H.; Chen, B.; Banfield, J. F.; Waychunas, G. A., Physical Review B 2008,
78 (21), 214106-214106.
50. Gong, X. Q.; Selloni, A.; Batzill, M.; Diebold, U., Nat. Mater. 2006, 5, 665-670.
51. Trejo, O.; Roelofs, K. E.; Xu, J.; Logar, M.; Sarangi, R.; Norlund, D.; Dadlani,
A.; Kravec, R.; Dasgupta, N. P.; Bent, S. F., et al., Submitted to Nano Letters 2015.
Katherine Roelofs Ph.D. Dissertation
141
Chapter 6. Increased QD Loading by
pH Control Reduces Interfacial
Recombination in QDSSCs
The power conversion efficiency of quantum-dot-sensitized solar cells (QDSSCs)
hinges on interfacial charge transfer. Increasing quantum dot (QD) loading on the TiO2
anode has been proposed as a means to block recombination of electrons in the TiO2 to
the hole transport material; however, it is not known whether a corresponding increase in
QD-mediated recombination processes might lead to an overall higher rate of
recombination. In this work, a three-fold increase in PbS QD loading was achieved by
the addition of an aqueous base to negatively charge the TiO2 surface during Pb cation
deposition. Increased QD loading improved QDSSC device efficiencies through both
increased light absorption and an overall reduction in recombination. Unexpectedly, we
also found increased QD size had the detrimental effect of increasing recombination.
Kinetic modeling of the effect of QD size on interfacial charge transfer processes
provided qualitative agreement with the observed variation in recombination lifetimes.
These results demonstrate a robust method of improving QD loading, identify the specific
mechanisms by which increased QD deposition impacts device performance, and provide
a framework for future efforts optimizing the device architecture of QDSSCs.
6.1. Introduction
Dye-sensitized solar cells (DSSCs) have a highly modular architecture,1
composed of a nanostructured metal oxide anode, sensitized with a light-absorbing dye,
and infiltrated with a hole-transport material (HTM). The DSSC device structure has
served as the basis for other solar cell architectures due to the ease of exchanging any
given component, such as the electron-transporting anode, the absorber, or the HTM.2
Quantum-dot-sensitized solar cells (QDSSCs) are an attractive variation on DSSCs, in
Katherine Roelofs Ph.D. Dissertation
142
which the monolayer of dye molecules is replaced by a single layer of quantum dots
(QDs) as the absorber.3-6
QDs, which are semiconductor nanocrystals with sizes small
enough to fall in the quantum-confined regime, have drawn interest as a next-generation
absorber material.7 One benefit is that the QD band gap varies with size, allowing the
absorption onset to be tuned while employing the same absorber material, simply by
changing the nanocrystal size. QDs have higher absorption coefficients and potentially
greater stability than high-performance dye molecules.7, 8
Due to their size-dependent
band gap, QDs of semiconductor materials with bulk band gaps below 1 eV can have the
nanocrystal band gap tuned by quantum confinement to harvest the optimal range of
photons from the solar spectrum, whereas it remains difficult to synthesize dye molecules
with low-energy absorption onsets, since extensive conjugation is required.7 Typically,
the nanostructured metal oxide anode employed in QDSSCs is TiO2, and promising metal
chalcogenide QDs for high-efficiency QDSSCs include CdS,9 CdSe,
10 CdTe,
11 PbS,
12
CuInSexS2-x,13, 14
In2S3,15
and Sb2S3.16
Despite the promising optical properties of QDs,
QDSSCs have only reached record efficiencies of 8.6%,17
still lagging behind the
efficiencies of 12.3% in DSSCs.18
QDSSCs are part of an overall trend in the field towards inorganic-absorber, all-
solid-state devices.2, 19
Replacing the organic dye molecules with inorganic absorbers in
DSSC-inspired architectures has lead to the development of QDSSCs, extremely-thin
absorber (ETA) solar cells,20
and most recently, perovskite solar cells.21-24
A solid-state
device architecture refers to the replacement of the traditional liquid electrolyte with
solid-state HTMs, e.g. ionic species such as CuSCN25
or organic semiconductors such as
spiro-OMeTAD (2,2',7,7'-tetrakis-(N,N-di-p-methoxyphenylamine)-9,9'-
spirobifluorene).26
Employing solid-state HTMs avoids problems of leakage or corrosion
of the absorber that can occur with liquid electrolytes.27
The use of liquid electrolytes is
a real and significant barrier to commercialization in these devices.28
PV technologies for
utility power generation need lifetimes approaching 25 years to compete with the cost of
grid electricity; even for technologies like QDSSCs with potentially lower materials
costs, the balance of systems costs is still high.29, 30
As can be seen from the example of
disposable batteries, the issues associated with liquid electrolytes limit device lifetimes.31
Moreover, non-utility applications of QDSSCs, such as portable or flexible solar cells,
Katherine Roelofs Ph.D. Dissertation
143
typically involve direct use by consumers for which the safety concerns of liquid
electrolytes are a problem. This is especially true for flexible solar cells, where low-cost
encapsulation techniques will be a challenge.32
However, the incorporation of a solid-state HTM introduces new challenges; a
major issue is reducing charge recombination at the anode/QD/HTM interface, and
improving charge transport to the electrodes. These effects are especially critical in
devices with solid-state HTMs, which have more severe rates of recombination relative to
liquid electrolytes,27, 33, 34
and lower mobilities.35
Due to these issues, solid-state
QDSSCs have reached device efficiencies of only 1.5%,36
and the majority of research in
the field is on QDSSCs with liquid electrolytes. We believe that for commercialization
of QDSSCs, the issues in solid-state devices need to be solved, and that one of the first
challenges to tackle is to decrease interfacial recombination in these devices. A common
approach used to decrease recombination is to modify the QD or TiO2 anode, e.g. through
increased QD loading of the TiO2 surface, doping of the QDs to optimize interfacial band
alignment, or deposition of additional materials such as insulating metal oxide layers at
the TiO2 surface.2
Increasing QD loading strikes at the heart of a primary problem in the QDSSC
design: QDSSCs struggle to achieve high QD loading.37-39
While DSSCs realize nearly
complete dye coverage of the TiO2,40-43
QD coverage of 15% of the TiO2 surface is high
for QDSSCs,44
with, for example, only 6% coverage achieved in the highest-efficiency
ss-QDSSCs.36
Increased QD deposition has the clear benefit of increased absorption; a
previous studies on in-situ45, 46
and ex-situ44
synthesized QDs indicate that increased QD
deposition increases external quantum efficiency (EQE), which the authors attribute to
increased adsorption. Bare regions of the TiO2 surface, not covered by an absorber,
contribute significantly to recombination due to the close proximity with the HTM.47
Increasing QD loading could prevent recombination by limiting exposed TiO2 surface. In
the DSSC system, increases in dye coverage have been found to have a positive effect on
electron lifetimes, through the suggested mechanism of blocking recombination from the
TiO2 to the HTM.48
In QDSSCs, such a beneficial effect of increased absorber coverage
cannot be assumed, as interfacial charge transfer is expected to be significantly
Katherine Roelofs Ph.D. Dissertation
144
different.39
For instance, in some of the highest-performing dye molecules, after electron
transfer the HOMO orbital of the dye molecule (where the hole is localized) is shifted
away from the TiO2 surface, thus helping to decrease recombination of TiO2 electrons
with the dye cation.49
Conversely, QDs introduce multiple new recombination pathways,
due to defects and trap states at the TiO2/QD interface, through which TiO2 electrons can
recombine with holes in the QD or in the HTM. In fact, studies have shown in a few
systems that the rate of QD-mediated recombination is higher than that of TiO2 electrons
directly recombining with the HTM (i.e., through bare regions on the TiO2 surface).47, 50
Other studies suggest that while recombination to the HTM dominates initially, there is a
regime change to favor QD-mediated recombination at higher deposition cycles of the
QD.51
Accordingly, it is not obvious if increasing QD loading will decrease or increase
recombination.
QDSSCs can be fabricated by the ex situ synthesis of QDs via colloidal methods,
or by the in situ growth of QDs directly on the anode surface. QDs synthesized by ex-situ
methods can be infiltrated into the nanoporous TiO2 substrate by simply soaking in a
suspension of colloidal QDs,52, 53
by employing linker molecules to attach the QDs to the
TiO2,54-56
or by electrophoretic deposition.57, 58
Despite this variety of infiltration
techniques, the ex situ method gives lower QD loading than in situ QD growth, because it
relies on the infiltration of pre-formed QDs into the nanostructured anode, rather than the
infiltration of the chemical precursors for QD growth.6 In situ QD growth techniques
include chemical bath deposition (CBD),59, 60
atomic layer deposition (ALD),61-63
and,
perhaps the most common technique, successive ion layer and adsorption reaction
(SILAR).36, 64
In one SILAR deposition cycle, the anode is first dipped in a solution
containing the metal cation precursor, followed by a rinsing step, and then the anion
precursor solution, with a final rinse. Additional SILAR cycles can introduce two types
of deposition: the continued growth of existing nanocrystals, or the nucleation of new
nanocrystals.64, 65
A fairly broad distribution of QD sizes is characteristic of the SILAR
process.36, 64, 66
Several approaches have been taken to improve QD nucleation and growth on the
TiO2 anode during SILAR. An intrinsic difficulty with SILAR deposition on nanoporous
Katherine Roelofs Ph.D. Dissertation
145
substrates is that the precursor ions easily reach and deposit on the outer region of the
nanoporous film, but have a more difficult time diffusing into the interior region of the
film. Thus, at higher SILAR cycles, QDs growing on the outer region of the TiO2 film
can actually block the pore channels to the inner regions, hindering any further QD
nucleation on the interior.67, 68
This phenomenon is referred to here as pore-blocking.
Strategies to improve the ion transport to and reaction with the anode surface include
potential-induced solution deposition techniques,69
improving the wetting of precursor
solutions on the anode surface,36
and sulfidizing the anode (for better growth of metal
sulfide QDs).70
A recent study by Park and colleagues focused on Hg-doped PbS QDs,
and found that adding a base to the Pb+ cation solution increased QD loading.
71 The
authors hypothesize that the lower pH of the cation precursor solution causes the TiO2
surface to be negatively charged, encouraging the adsorption of Pb cations, and leading to
the observed increase in QD loading.
In this work we use base-assisted SILAR deposition to investigate the impact of
increased PbS QD loading on device performance, with particular focus on the impact of
QD coverage of the TiO2 surface on interfacial recombination. We demonstrate a robust
method of improving initial QD loading through the use of three different bases, NaOH,
ethylenediamine (ED), and triethanolamine (TEA), and examine the mechanism by which
the addition of a base to the cation precursor solution influences QD deposition. The
QD-coated films are used to fabricate solid-state QDSSCs with spiro-OMeTAD as the
HTM, as in solid-state devices, decreasing interfacial recombination is of even greater
importance than devices with liquid electrolytes. Increased QD coverage is shown to
increase recombination lifetimes by over an order of magnitude, showing the beneficial
impact of higher QD loading. The effects of increased QD deposition are complicated by
the variation in both the number of QDs and QD size with increased SILAR cycles;
increased QD size is found to decrease recombination lifetimes. The effect of changing
QD size, i.e. QD band gap, is modeled based on the expected change in rate constants
with shifts in the QD band energy levels, providing a potential explanation for the
observed trend in recombination lifetimes. While previous studies have reported
recombination in CdSe51, 72
and PbS73
QDSSCs as a function of SILAR cycle, this is the
first work to combine recombination measurements with a technique to significantly
Katherine Roelofs Ph.D. Dissertation
146
enhance QD loading without further deposition cycles. This has provided the new insight
that two opposing effects are at play with QD deposition: that higher QD loading reduces
recombination, while increased QD size can actually increase interfacial recombination.
6.2. Experimental Methods
TiO2 Film Deposition. Solid-state QDSSCs were fabricating according to a
previous procedure.74
For the transparent electrode, glass substrates coated with fluorine-
doped tin oxide (15 Ω/□, Pilkington) were patterned by etching with 4 M Cl and n
powder. Substrates were then coated with a thin (~50 nm) compact layer of TiO2 by
aerosol spray pyrolysis at 450 °C using air as a carrier gas, with titanium diisopropoxide
bis(acetylacetonate) (Sigma 325252) diluted in ethanol as the precursor. The nanoporous
TiO2 layer was then deposited by doctor-blading a commercial paste of 20 nm diameter
anatase TiO2 particles in ethyl cellulose and terpineol (Dyesol 18-NRT), which was
diluted with additional terpineol at a 1:1 weight ratio of the commercial paste to
terpineol. The films were annealed at 450 °C, resulting in a film thickness of ~2.2 μm as
measured by a Dektak profilometer. The nanoporous TiO2 films were immersed
overnight in a 0.02 M aqueous TiCl4 solution at room temperature, and then annealed
again at 450 °C, to complete the TiCl4 treatment which is standard in DSSC devices.
QD Growth. Lead sulfide quantum dots were grown by SILAR employing 0.02
M aqueous Pb(NO3)2 (Sigma-Aldrich 11520) as the cation solution (pH 4.3), and 0.02 M
aqueous Na2S (Sigma-Aldrich 407410) as the anion solution (pH 12.3). Nanoporous
TiO2 films were dipped in the cation solution for 2 min, rinsed in DI water for 1 min, then
dipped in the anion solution for 2 min, followed by a final dip in DI water for 1 min. For
the base-assisted QD growth, NaOH, ethylenediamine (Sigma-Aldrich 03550), and
triethanolamine (Sigma 90279) were added to separate solutions of the 0.02 M aqueous
Pb(NO3)2. For NaOH and ED, each was added dropwise to the Pb(NO3)2 solution to
avoid precipitation of PbO, until a pH of 9 was reached. In the case of TEA, a final
concentration of 1 M TEA was aimed for, giving a slightly higher pH of 9.2. Care was
taken to maintain these pH values (as measured by pH meter) throughout sequential
SILAR deposition cycles, through the addition of further base if necessary.
Katherine Roelofs Ph.D. Dissertation
147
QD Characterization. A Cary 6000i UV-visible spectrophotometer (Varian) was
used to characterize the optical properties of the nanoporous TiO2 film sensitized with
PbS QDs. For this study, the nanoporous TiO2 film was deposited on glass microscope
slides. The QDs were also analyzed by transmission electron microscopy (TEM; 200 kV
FEI Tecnai G2 F20 X-TWIN). To prepare the TEM samples, material was scraped from
QD-covered nanoporous TiO2 films with an acetone solvent and dispersed on an ultrathin
carbon coating on a holey carbon film supported by a 300 mesh copper grid (Ted Pella
Inc., #01824). To calculate the average QD size and QD surface coverage of the TiO2
TEM images were analyzed using ImageJ software. Regions of non-overlapping TiO2
nanocrystals were used to calculate surface coverage, by summing the area of all the PbS
QDs on that nanocrystal, and dividing that by the doubled area of the TiO2 nanocrystal,
as QDs visible in the TEM can be attached to the front or back of the TiO2 nanocrystal.
Overall, at least six TEM images were analyzed for each sample, corresponding to
roughly 0.03 μm2 of TiO2 surface area and 300 QDs.
Device Fabrication. PbS QDs were deposited on nanoporous TiO2 substrates by
SILAR. As described elsewhere,35
the solution of the solid-state hole-transporting
material was composed of 225 mg mL-1
of spiro-OMeTAD (Lumtec LT-S922) dissolved
in chlorobenzene, with tert-butylpyridine added at a ratio of 1:10.3 μ :mg of spiro-
OMeTAD, and lithium bis-(trifluoromethylsulfonyl)imide salt (170 mg mL-1
in
acetonitrile) added at a ratio of 1:4.8 μ :mg of spiro-OMeTAD. A small amount of the
spiro-OMeTAD solution (30 μ for 3.75 cm2 substrates) was deposited onto the TiO2
substrates at room temperature, and spin-coated at 2000 RPM for 30 s. Finally, 200 nm
thick Ag counter electrodes were deposited by thermal evaporation under vacuum below
10-6
torr. Device areas were defined by a combination of an FTO etch line and the metal
top-contact, and the top metal contact. Final device areas were on the order of 0.1 cm2,
varying slightly from device to device based on the exact position of the FTO etch line
with respect to the top contact. The area for each device was measured by an optical
microscope. Devices were stored inside a desiccator prior to electrical measurements.
Electrical Measurements. For J-V measurements, an AM 1.5 solar simulator
(Oriel 91160) equipped with a 300 W ozone-free Xe arc lamp (6258) was used. The
Katherine Roelofs Ph.D. Dissertation
148
lamp was calibrated to 1 sun (100 mW cm-2
) using a reference NREL calibrated Si
photodiode equipped with an IR cutoff filter. Current-voltage (J-V) curves were
collected with a Keithley 2400 SourceMeter, with a sweep delay of 40 ms. Devices were
light-soaked until maximum efficiencies were reached (up to 1 hr), as was previously
found necessary.63, 74
The increase in efficiency with exposure to light is due to doping of
the spiro-OMeTAD by oxygen molecules from the atmosphere, and is induced by
current-flow in the device.75, 76
The lower JSC values of these devices explain why the
process is slower than the analogous one in DSSCs employing spiro-OMeTAD, which
occurs on the timescale of 10 minutes.77, 78
Representative light-soaking data is shown in
Figure 6-1. Dark curves were measured after the light-soaking process.
Figure 6-1. Example of light-soaking behavior, shown for the control QDSSCs, with 2,
4, 6, and 8 SILAR deposition cycles of PbS QDs. The cell is exposed continuously to
Katherine Roelofs Ph.D. Dissertation
149
light, and a J-V curve is measured at 10 min intervals, with each measurement taking
roughly 1 minute. Plotted are the resulting parameters (a) short-circuit current, (b) fill
factor, (c) open-circuit voltage, and (d) efficiency.
External quantum efficiency (EQE) measurements were taken at short circuit
using monochromated white light from a 100 W tungsten lamp, which was focus through
a monochromator. The monochromated illumination, chopped at 40 Hz, was applied in
addition to a constant bias light illumination from a white LED array. The photocurrent
of the device at each wavelength was acquired through a lock-in amplifier and the EQE
was calculated by referencing the photocurrent from the device to the current from a
NIST traceable calibration photodiode.
Transient photovoltage measurements, developed by O’Regan et al.,79, 80
were
performed at open-circuit voltage (VOC) conditions, following a previously used
procedure.78
A Keithley 2400 source meter was used to maintain the device at zero
current. Measurements were taken at different VOC values for a single device by
changing the incident bias light. A programmable power supply was used to adjust the
white LED array bias light from 0.01 to ~1 sun. A white LED pulse light driven by a
function generator (Agilent) was adjusted to approximately 5% of the bias light intensity.
The decay of the VOC when the pulsed light turned off was tracked by an oscilloscope
(Tektronix). The pulse length was varied from 50 ms to 5 s as necessary to capture the
full decay of the VOC. The decay in the VOC was fitted to a single exponential curve to
extract a time constant, which is the recombination lifetime.
6.3. Results and Discussion
6.3.1. QD Characterization
We first characterized the lead (II) sulfide QD growth as a result of adding
different bases to the cation solution during SILAR; NaOH, ED, and TEA were explored.
In each case, the concentration of the base in the Pb(NO3)2 aqueous solution was adjusted
to reach a pH of 9 (for TEA the pH was 9.2). Figure 6-2 plots the UV-vis absorption
measurements of QD-coated nanoporous TiO2 films for various SILAR cycles and
Katherine Roelofs Ph.D. Dissertation
150
growth conditions. For clarity of viewing, Figure 6-3a replots 2 SILAR cycles of QDs
are grown with the addition of NaOH, ED, or TEA, or by the standard process. In terms
of the absorption response, all three bases were found to increase PbS deposition to a
similar extent.
Figure 6-2. UV-Vis absorption of PbS QD-coated nanoporous TiO2 substrates. QDs
were grown for 2, 4, 6, and 8 SILAR cycles, with base-assisted QD growth employing
NaOH and ED. TEA-assisted QD growth is also plotted for comparison.
Katherine Roelofs Ph.D. Dissertation
151
Figure 6-3. UV-vis spectroscopy of nanoporous TiO2 films sensitized by PbS QDs.
Shown are (a) blank TiO2 films and 2 SILAR cycles of QDs grown by the standard
process (solid line) and in the presence of NaOH, ED, and TEA (dashed lines). Also
shown (b) is the trend with 2, 4, 6, and 8 SILAR cycles comparing the standard process
and TEA-assisted QD growth, with the corresponding Tauc analysis curves shown in the
inset.
Figure 6-3b tracks the progression in the UV-vis spectra with increasing SILAR
cycles, comparing the standard process to the case of TEA. In all cases, increasing the
Katherine Roelofs Ph.D. Dissertation
152
number of SILAR cycles leads to absorption onsets at longer wavelengths. This
demonstrates the quantum effect that increasing the QD size decreases the band gap, until
bulk conditions are approached. Although we note that higher QD loading causing
stronger absorption, or aggregation of the QDs, can also contribute to the apparent shift in
absorption onset. PbS is a desirable QD absorber material as it has a large Bohr exciton
radius (18 nm) and confinement occurs up to a relatively large QD radius, permitting
greater control over the band gap for a given change in radius. The bulk band gap of PbS
is 0.37 eV and confinement increases the band gap through the optimal point of interest
for capturing the solar spectrum (~1.1 eV).
The UV-vis results show, at each SILAR cycle, an increased absorption for QDs
grown with TEA compared to those grown without. To isolate the effect of QD size, the
inset in Figure 6-3b plots the Tauc analysis of the UV-vis spectra, from which the QD
band gap was determined for each condition (Table 6-1). We note that the Tauc analysis
of these substrates is susceptible to interpretation as to where the tangent line is drawn,
due to the extended curvature of the absorption onset, which is caused both by the
distribution in QD size (and thus band gap) in a given substrate, as well as sub-band gap
absorption from defect states in the QDs. Therefore the band gaps extracted from the
Tauc analysis should be taken as nominal values indicative of trends rather than absolute
values. The QD diameters expected for these band gaps, as calculated through band-
gap/size calibration curves for PbS QDs found in the literature, can be found in
Supporting Information (Table 6-1).81
Higher band gaps are observed with TEA,
indicating that the increased absorption at a given SILAR cycle is due at least in part to a
shift towards an earlier absorption onset from larger QDs. A greater number of QDs may
also be present, but this cannot be determined conclusively from the UV-vis results.
Table 6-1 compiles the relevant parameters characterizing the PbS QDs, extracted
from both the UV-Vis absorption data, and from the TEM images. To check if the band
gap values from the Tauc analysis make sense, the QD diameters expected for these band
gaps are listed (Tauc: diameter), and a comparison with the average QD diameters
observed by TEM shows fairly good agreement. The QD valence band positions, which
Katherine Roelofs Ph.D. Dissertation
153
are used in the modeling section below, were determined from literature reports of band
positions at a given QD size. The QD diameter from TEM was used for determining the
Table 6-1. Characterization of PbS QDs deposited under varying SILAR deposition
conditions. Listed are the QD band gap extracted from Tauc analysis of the UV-Vis
spectra, and the QD diameter expected from the Tauc band gaps, using a literature
calibration of PbS QD band gap and size.81
Also listed are the average QD diameter and
percent of the TiO2 surface area covered by QDs obtained from TEM, and the valence
band position expected from the TEM diameter, using a literature calibration of PbS QD
size and valence band position.82
Six TEM were analyzed for each sample,
corresponding to roughly 0.03 μm2 of TiO2 surface area and 300 QDs; error bars show
the standard deviation.
Tauc:
band gap
(eV)
Tauc: diameter
(nm)
TEM:
avg. diameter
(nm)
TEM: surface coverage
(% area) Valence band position (eV)
2 cyc 2.66 1.9 1.6 ± 0.3 2.5 ± 0.2 1.42
2 cyc NaOH 2.54 2.0 2.1 ± 0.3 7.0 ± 1.2 3.26
2 cyc ED 2.55 2.0 2.2 ± 0.3 6.5 ± 1.3 3.44
2 cyc TEA 2.56 2.0 2.0 ± 0.3 6.8 ± 1.1 3.49
4 cyc 2.10 2.4 2.6 ± 0.9 8.0 ± 2.0 2.77
4 cyc NaOH 1.95 2.6 2.6 ± 0.6 10 ± 2.2 3.23
4 cyc ED 1.95 2.6 2.6 ± 0.7 10.5 ± 2 3.49
4 cyc TEA 1.96 2.6 2.6 ± 0.7 10.5 ± 2.3 3.90
6 cyc 1.88 2.7 2.9 ± 0.6 7.8 ± 2.0 2.91
6 cyc NaOH 1.76 2.9 3.0 ± 1.0 10.6 ± 2.4 3.23
6 cyc ED 1.74 2.9 2.6 ± 0.9 10.8 ± 2.5 3.29
6 cyc TEA 1.75 2.9 2.8 ± 0.8 10.0 ± 2.2 3.60
8 cyc 1.72 3.0 3.0 ± 0.7 8.4 ± 2.5 2.56
8 cyc NaOH 1.63 3.2 4.3 ± 0.8 10.8 ± 2.4 3.26
8 cyc ED 1.63 3.2 3.2 ± 0.8 11 ± 2.0 3.38
8 cyc TEA 1.62 3.2 3.4 ± 1.1 11.1 ± 2.5 3.66
valence band positions, as it is a more direct measurement than the diameters from the
Tauc Analysis.
Katherine Roelofs Ph.D. Dissertation
154
To directly observe the impact of high-pH cation solution on QD nucleation and
growth, samples from QD-sensitized nanoporous TiO2 films were examined by TEM.
Representative TEM images are shown in Figure 6-4. The PbS QDs are shown by
HRTEM (Figure 6-5) to be in the rocksalt crystal structure, and that near-stoichiometric
PbS is indeed deposited is further confirmed by Auger electron spectroscopy analysis,
presented below. A rough calculation of absorbance from the TEM loading data matches
the absorbance measured in the UV-vis spectra within an order of magnitude. For this
back of the envelope calculation, we used a rough value of the extinction coefficient, 3 x
108 M
-1cm
-1, based on the extinction coefficients reported for PbS QDs in the literature:
ε400 ~ 2.3 x 108 M
-1cm
-1,83
and ε555~3.5 x 108 M
-1cm
-1.84
The internal surface area for
Dyesol 18NR-T TiO2 films has been found to be roughly 7 m2/cm
3.85
The surface density
of PbS QDs, as measured in the TEM, was roughly 4 x 1016
cm-3
or 7.8 x 10-5
M. From
these values, knowing our film thicknesses were 2 microns, we were able to calculate an
absorbance value of 5. This is on the same order of magnitude as the absorbance
measured in the UV-vis spectra in the 400 nm to 550 nm range.
Figure 6-4. Representative TEM images of PbS QDs grown on TiO2 nanocrystals, at the
same magnification. Shown are 2 SILAR cycles comparing (a) the standard process with
that with (b) NaOH, (c) ED, and (d) TEA. Also shown are higher SILAR cycles by (e),
(g), & (i) the standard process, and (f), (h), & (j) that with TEA.
Katherine Roelofs Ph.D. Dissertation
155
Figure 6-5. HRTEM images showing rocksalt PbS QDs grown by the standard SILAR
process, on anatase TiO2 nanocrystals. In (a) the lattice spacing along the red line is 3.53
Å, fitting the PbS rocksalt d111=3.44 Å, and the lattice spacing along the yellow line is
3.02 Å, matching PbS rocksalt d200=2.98 Å. (b) A higher-resolution image, with a lattice
spacing of 2.91 Å along the red line, matching PbS rocksalt d200=2.98 Å.
At 2 SILAR cycles, increased QD loading and larger QDs were observed for
NaOH, ED, and TEA, as compared to the standard process. TEM images were analyzed
to determine the portion of the TiO2 surface area covered by QDs, as well as the average
QD diameter for each condition, shown in Figure 6-3. There is a fair match between the
QD size extracted from the Tauc analysis of the UV-vis spectra and the average QD size
observed in TEM (Table 6-1), though we note that TEM measurements of QD size are
more reliable that band gaps extracted from UV-vis spectra due to the intrinsic
difficulties of Tauc analysis of nanocrystals, discussed above. However, it is
encouraging that the bulk measurement technique of UV-vis spectroscopy and the
inherently local sampling technique of TEM are in agreement here. Tracking the
continued deposition in the case of TEA, higher QD loading is still apparent at 4, 6, and 8
SILAR cycles, although the effect diminishes with increasing SILAR cycles and is close
to the range of error by 6 cycles. The diminished gains in coverage with TEA at higher
SILAR cycles (Figure 6-6) seem to indicate a self-saturating process, in which QD
deposition has an upper limit that is approached by the TEA-assisted growth sooner than
by the standard process. The self-saturation in QD growth could be due to pore-blocking,
Katherine Roelofs Ph.D. Dissertation
156
as discussed above, or due to epitaxial growth of the QDs, where the nanocrystal size
limit is ultimately determined by any lattice mismatch between the TiO2 substrate and the
QD.73
Another possible explanation for the saturation of QD growth is that once some
QDs are nucleated in the first cycle or two, additional Pb2+
and S2-
adsorption occurs on
the already-nucleated QDs, which may be the case if PbS strongly prefers to grow on
itself over TiO2. Overall, the largest gains in QD coverage of the TiO2 surface with base-
assisted growth were seen initially at 2 SILAR cycles, and each of the three bases
investigated produced similar gains in the number and size of QDs.
Figure 6-6. PbS QD size and loading calculated from TEM images, showing the average
QD diameter and percent of the TiO2 surface area covered by QDs. QDs were deposited
for varying SILAR cycles, and by the standard process and with NaOH, ED, and TEA-
assisted growth. At each SILAR cycle the points are staggered for visibility. Six TEM
images were analyzed for each sample, corresponding to roughly 0.03 μm2 of TiO2
surface area and 300 QDs; error bars show the standard deviation.
Figure 6-7 shows Auger electron spectroscopy (AES) line scans of the cross-
section of TiO2 films infiltrated with PbS QDs grown by the standard SILAR process as
compared to that with TEA-assisted growth; the raw intensity data is included at the
bottom. The AES line scans confirm the TEM and UV-vis results that base-assisted QD
growth gives a higher QD loading, with roughly twice as much Pb and S as in the
standard process. In both cases, higher amounts of Pb can be observed near the TiO2 film
surface, indicating QD growth is indeed faster at the top of the film.
Katherine Roelofs Ph.D. Dissertation
157
Figure 6-7. Auger electron spectroscopy line scan of sample in cross-section for
nanoporous TiO2 films on FTO, sensitized with 6 SILAR cycles of PbS QDs grown by
(a) the standard process, and (b) with TEA. Shown are the atomic concentrations with
the raw intensities at the bottom.
The mechanism by which the addition of a base increases the nucleation and
growth of QDs on TiO2 is not obvious. Lee et al. propose that TEA could increase Pb2+
adsorption by negatively charging the TiO2 surface.71
The 0.02 M Pb(NO3)2 aqueous
solutions have a pH of 4.3, while the pristine point of zero charge of anatase TiO2 in
aqueous solutions occurs at a pH of ~6.86
Therefore adjusting the cation solution to a pH
Katherine Roelofs Ph.D. Dissertation
158
to 9 should change the TiO2 surface charge from positive to negative. In addition to
affecting the TiO2 surface charge, TEA is a chelating agent that can complex with the
Pb2+
ions in the aqueous Pb(NO3)2 solution. Thus, TEA can lower the concentration of
free Pb2+
in solution (though this would be expected to decrease, not increase, PbS
deposition). Further, TEA can be chemisorbed or physisorbed on the surface of TiO2,
which could affect the transfer of Pb2+
from the bulk solution to Helmholtz layers
immediately above the TiO2 surface. Employing NaOH and ED bases allows us to
separate out these effects. NaOH will not act as a chelating agent to complex with Pb2+
,
nor adsorb to the TiO2 surface, so it is expected to purely affect the TiO2 surface charge.
Like TEA, ED can adsorb on the TiO2 surface, but ED is only a bidentate chelating agent,
while TEA is a tridentate chelating agent. From the UV-vis results (Figure 6-2 and
Figure 6-3) and TEM results (Figure 6-4 and Figure 6-6), it can be seen that the use of
NaOH or ED to adjust the Pb(NO3)2 solution to a pH of 9 leads to similar gains QD
nucleation and growth as is achieved with TEA. The common factor between these three
bases is the shift in the solution pH, therefore we conclude that the increased PbS
deposition is due electrostatic effects rather than side chemical reactions. In addition, we
have demonstrated a robust method to achieving increased QD loading through the use of
different bases. This method is applicable to the deposition of other QD absorbers, such
as Sb2S3 and In2S3, in which, depending on the particular cation salt used, the cation
solutions have pH < 6.64
6.3.2. Device Performance
Figure 6-8 plots the photovoltaic performance (J-V curves) of champion solid-
state QDSSCs fabricated with PbS QDs deposited under varying conditions; J-V curves
for higher SILAR cycles with only the control and TEA are shown in Figure 6-9b-d for
clarity of viewing. These J-V curves are reported for the maximum efficiency achieved
over the course of light-soaking. The average performance metrics for each condition are
shown in Figure 6-10 and listed in Table 6-2. At 2 SILAR cycles (Figure 6-9a), there are
dramatic gains in power conversion efficiency (PCE) for the case of NaOH, ED, and
TEA, through gains in both short-circuit current (JSC) and open-circuit voltage (VOC).
Comparing NaOH, ED, and TEA to the standard deposition process at 2 SILAR cycles,
Katherine Roelofs Ph.D. Dissertation
159
the 3-3.5 fold gain in JSC’s observed with base-assisted growth is on the same order as the
2.6-2.8 fold gains in QD coverage of TiO2 (Figure 6-6), suggesting that the increased
light absorption from higher QD loading could be a significant contributor to the
increased current collected. Though increased absorption does not ensure the charges are
collected, our recombination studies (see below) suggest charge collection improves as
well. While the device efficiency for the standard process peaks at 6 SILAR cycles, the
highest efficiency for NaOH and TEA were observed at 2 cycles, and for ED at 4 cycles
(though 2 cyc ED had efficiencies close to 4 cyc ED). At a given deposition cycle,
devices with base-assisted growth had dark current curves shifted outward to higher
voltages, and higher
Figure 6-8. J-V curves of the highest-efficiency solid state QDSSCs at each condition,
plotting curves collected under 1 sun illumination and in the dark. Shown are devices
grown with (a) 2, (b) 4, (c) 6, and (d) 8 SILAR cycles of PbS QDs with NaOH and ED-
assisted growth.
Figure 6-9. Selected J-V curves from Figure 6-8 showing only the control versus TEA-
assisted growth at higher SILAR cycles.
Katherine Roelofs Ph.D. Dissertation
160
Table 6-2. Values of average ss-QDSSC metrics as plotted in Figure 6-8, for varying
QD deposition conditions. Device metrics were averaged from 10 devices (the two best
cells from five different substrates), and error bars represent standard deviations.
JSC
(mA/cm2) VOC (V)
Fill Factor PCE (%)
2 cyc 0.66 ± 0.13 0.440 ± 0.043 0.33 ± 0.03 0.09 ± 0.01
2 cyc NaOH 1.86 ± 0.17 0.537 ± 0.025 0.60 ± 0.03 0.60 ± 0.09
2 cyc ED 2.00 ± 0.18 0.552 ± 0.023 0.57 ± 0.04 0.63 ± 0.10
2 cyc TEA 2.24 ± 0.16 0.553 ± 0.026 0.51 ± 0.04 0.64 ± 0.09
4 cyc 1.04 ± 0.17 0.494 ± 0.013 0.50 ± 0.05 0.26 ± 0.03
4 cyc NaOH 1.93 ± 0.13 0.542 ± 0.021 0.59 ± 0.03 0.62 ± 0.05
4 cyc ED 1.83 ± 0.16 0.551 ± 0.024 0.53 ± 0.03 0.53 ± 0.05
4 cyc TEA 2.03 ± 0.16 0.540 ± 0.023 0.48 ± 0.03 0.53 ± 0.05
6 cyc 1.54 ± 0.31 0.499 ± 0.009 0.43 ± 0.06 0.32 ± 0.04
6 cyc NaOH 1.19 ± 0.16 0.531 ± 0.019 0.51 ± 0.02 0.32 ± 0.04
6 cyc ED 1.24 ± 0.13 0.538 ± 0.018 0.45 ± 0.03 0.30 ± 0.03
6 cyc TEA 1.04 ± 0.13 0.532 ± 0.022 0.44 ± 0.01 0.24 ± 0.05
8 cyc 0.88 ± 0.06 0.484 ± 0.024 0.51 ± 0.02 0.22 ± 0.01
8 cyc NaOH 0.81 ± 0.02 0.518 ± 0.017 0.48 ± 0.02 0.20 ± 0.01
8 cyc ED 0.81 ± 0.03 0.532 ± 0.017 0.49 ± 0.02 0.21 ± 0.01
8 cyc TEA 0.82 ± 0.02 0.525 ± 0.014 0.48 ± 0.01 0.20 ± 0.02
VOC values than the standard devices. Outward-shifted dark curves and higher VOC values
are consistent with decreased recombination. The efficiency for our best device, 2 cycle
TEA, is 0.65%, while the record device efficiency for solid-state QDSSCs is 1.5%.36
To better understand the impact of variations in QD deposition on device
performance, we measured the devices’ EQE. EQE measurements can shed light on the
contribution of different-sized QDs to the JSC in a given device. Figure 6-11 plots the
EQE spectra for increasing SILAR cycles of the standard devices compared to TEA. The
EQE data match the trend in JSC, although integration of the EQE spectra gives expected
JSC values that are roughly 20% lower than the JSC values from the J-V curves (Table
6-3). This discrepancy could be due to the light-soaking effects observed in devices with
Katherine Roelofs Ph.D. Dissertation
161
spiro-OMeTAD.75, 76
For the solid-state QDSSCs in this work and previous studies,63, 74
device efficiency initially improves over exposure to 1 sun illumination, after which it
Figure 6-10. Average device parameters—power conversion efficiency (PCE), short-
circuit current (JSC), open circuit voltage (VOC) and fill factor (FF)—for varying QD
SILAR deposition cycles, by the standard process (control), as well as NaOH, ED, and
TEA-assisted growth. At each SILAR cycle, the points are staggered for visibility.
Device parameters were averaged from 10 devices (the two best cells from five different
substrates), and error bars represent standard deviations. Corresponding values are listed
in Table S2.
levels off and eventually begins to decay. Figure 6-1 shows a representative set of
control devices, at varied PbS SILAR cycles to show the effects at varied efficiencies. It
can be seen that most of the changes in efficiency are driven by changes in JSC. For the
highest efficiency devices, the initial increase happens quite rapidly, and then there is a
decay in efficiency, though for lower-efficiency devices, the light-soaking process is
slower. As this work does not deal with stability issues, we report the efficiency at the
highest point. We note that the relative efficiency differences between devices are
maintained over the course of light-soaking, validating the trends in the maximum
Katherine Roelofs Ph.D. Dissertation
162
efficiencies discussed in the main text. Thus the difference in JSC values could be due to
the difference in the light soaking due to the difference in light source for the J-V
Figure 6-11. External quantum efficiency (EQE) spectra measured at short circuit of the
highest-efficiency devices with increasing SILAR cycles of PbS QDs compared for the
standard process (solid lines), and that with TEA (dashed lines).
measurements, which are taken using a Xe lamp AM 1.5 solar simulator, and the EQE
measurements, which are obtained while the device is illuminated by a bias light (~1 sun)
from a white LED array. Another possibility is differences in masking. J-V curves were
taken with no aperture defining the illuminated area; instead device area was set by the
size of the metal contact, so additional photoexcited carriers immediately outside the
device area could diffuse into the area and be collected, inflating the JSC. The EQE
measurements were taken with a controlled illuminated area, smaller than the metal
contact area.
The main observation from the EQE data is that, although there are some
differences due to the broadening of the QD size distribution, the improvements in the JSC
with TEA appear to be dominated by the increase in the number of QDs at a given size.
For instance, at 2 SILAR cycles, the EQE spectra for the TEA device has a similar
distribution over wavelength as the standard device, just with higher values, indicating
Katherine Roelofs Ph.D. Dissertation
163
that QDs of a similar size range are contributing in each case, with a greater number of
QDs present with TEA.
Table 6-3. JSC values determined by integrating the EQE spectra from champion-
efficiency ss-QDSSC devices, plotted in Figure 6, along with JSC values of the same
devices as determined from the J-V curves.
JSC
(mA/cm2) from EQE
JSC (mA/cm2) from J-V
curve
2 cyc 0.44 0.63
2 cyc TEA 1.97 2.27
4 cyc 0.84 1.13
4 cyc TEA 1.73 2.01
6 cyc 1.31 1.55
6 cyc TEA 1.03 1.07
8 cyc 0.50 0.88
8 cyc TEA 0.56 0.82
It might be expected that larger QDs would shift the EQE spectra towards longer
wavelengths. However, even comparing the 8 cycle and 2 cycle standard devices, there
is only a slight skewing of the EQE spectra towards longer wavelengths for 8 cycles.
Most strikingly, the EQE spectra remains quite narrow at higher SILAR cycles, despite
the shift in QD absorption onset to longer wavelengths observed in the UV-vis spectra.
Thus, while there is strong absorption from QDs with band gaps 2 eV (absorption
onsets 600 nm), for instance in the 8 cycle standard device as seen in Figure 6-3, these
QDs are not contributing to charge collection. It could be that at less than 2 eV the PbS
QD band gap is not large enough to properly align the QD conduction band with that of
TiO2 for transfer of the excited electron and the QD valence band with the spiro-
OMeTAD HOMO level for hole transfer, or it could be that there is not enough of an
energetic overpotential for efficient electron injection into the TiO287
or hole transfer to
the spiro-OMeTAD.88
. This result is similar to a previous report in a TiO2/PbS QD/spiro-
OMeTAD device, that band gaps greater than 1.55 eV were required for effective charge
Katherine Roelofs Ph.D. Dissertation
164
splitting.36
Therefore, increased device efficiency hinges on increased initial QD
nucleation, such that many small QDs can be introduced with the correct band gap level.
6.3.3. Interfacial Recombination
We are interested in determining the effect of increased QD coverage of the TiO2
surface on recombination at the TiO2/HTM interface. Transient photovoltage
measurements allow us to measure the recombination lifetimes at the interface; longer
electron recombination lifetimes ( ) correspond to decreased rates of recombination.
Figure 6-13 plots the recombination lifetimes measured for the same set of devices in
Figure 6-9. Recombination lifetimes are plotted against VOC, the open-circuit voltage
values at which they were measured. Representative raw transient photovoltage decay
curves are shown in the Supporting Information Figure S6. We confirmed that there is no
shift in the TiO2 conduction band level between different devices by measuring the
transient photocurrent decay for the same light pulse flux densities as the transient
photovoltage decay set points. By integrating the photocurrent decay, the charge density
shift corresponding to the voltage shift in the transient photovoltage measurements can be
found. For the set of devices in this study, the charge density versus voltage plots were
identical. Therefore, TiO2 conduction band shifts can be ruled out as a cause of the
change in recombination lifetimes.
Katherine Roelofs Ph.D. Dissertation
165
Figure 6-12. Representative transient photovoltage decay curves, shown at the highest
light intensity collected ~1.1 sun, along with the lowest light intensity reported ~0.09 sun.
Plotted are the decay curves for 2 SILAR cycles of the control QDSSCs at (a) ~1.1 sun
and (b) ~0.09 sun, and 2 SILAR cycles with TEA-assisted QD growth at (c) ~1.1 sun and
(d) ~0.09 sun.
In Figure 6-13a, the recombination lifetimes at 2 SILAR cycles for the standard
process are compared to QDs grown with NaOH, ED, and TEA. We observe an order of
magnitude change toward longer recombination lifetimes with each of the bases, as
compared to the standard process. When considering the physical differences in the QD
between these samples, this data set highlights the large change in surface coverage at 2
SILAR cycles, with 2.5% QD coverage of the TiO2 surface for 2 SILAR cycles by the
standard deposition, and coverages of 7.0%, 6.5%, and 6.8% achieved with NaOH, ED,
and TEA, respectively. We note that effects of increased coverage are convoluted with
any effects due to the shifting QD band gap, as the base-assisted growth also gives larger
QDs (from 1.6 nm diameter QDs by the standard process to 2.1, 2.2, and 2.0 nm with
Katherine Roelofs Ph.D. Dissertation
166
Figure 6-13. Recombination lifetimes comparing (a) 2 SILAR cycles of QDs grown by
the standard process and in the presence of NaOH, ED, and TEA. (b) Recombination
lifetimes of 2, 4, 6, and 8 SILAR cycles comparing the standard process and TEA-
assisted QD growth. To guide the eye, the purple band highlights the standard process
(control), and the light blue band highlights the TEA data points.
NaOH, ED, and TEA, respectively). However, we observe in Figure 6-13b that
increasing QD size is correlated with shorter recombination lifetimes (see discussion
below). Accordingly, we attribute the longer recombination lifetimes with base-assisted
QD growth in Figure 6-13a to the increased QD loading. At higher QD loading, the
Katherine Roelofs Ph.D. Dissertation
167
overall effect is longer lifetimes, demonstrating that the beneficial effects of the QD, such
as blocking TiO2-to-HTM recombination, outweigh any harmful effects, such as
participation in QD-mediated recombination processes. We note that studies reporting
higher rates of QD-mediated recombination than TiO2-to-HTM recombination (which
suggests that more QDs would have an overall harmful effect) were conducted with
liquid electrolytes.47, 50, 51
Together, the longer recombination lifetimes, outward shift in
the dark current J-V curves, and higher VOC’s all indicate decreased recombination in the
devices with base-assisted QD growth, particularly at 2 SILAR cycles. Our results
highlight that in QDSSCs with solid-state HTMs, which have faster rates of TiO2-to-
HTM recombination than liquid electrolytes,27, 33, 34
increasing the QD loading leads to
overall benefits in decreased recombination.
It is difficult to separate the impact of increased coverage of the TiO2 surface
from that of increased QD size, because changing deposition parameters normally affects
both. In this work, we are able to substantially change QD surface coverage with base-
assisted QD growth, and compare those results to the standard SILAR growth process.
However, due to the slight differences in QD size between base-assisted growth and
standard growth (Figure 3), there were only a few data points that had identical QD sizes
but different surface coverages. However, holding surface coverage constant, we were
able to see the effects of varied QD size by comparing samples grown with TEA at 4, 6,
and 8 SILAR cycles, where QD diameter increases substantially (from 2.6 to 3.2 nm) but
surface coverage remains roughly constant (10.5 to 11.1 %). From Figure 6-13b, it can
be seen that moving from 4 to 6 to 8 SILAR cycles with TEA leads to shorter
recombination lifetimes. From this, we conclude that increasing QD size increases the
rate of recombination. Nevertheless, for each number of SILAR cycles studied, use of
TEA consistently helps increase recombination lifetimes and, as discussed above, we
attribute this effect to blocking of TiO2-to-HTM recombination by the QD layer. A
possible mechanism by which increasing QD size could decrease recombination lifetimes
is that the rise of the QD valence band that occurs as the band gap shrinks could hinder
hole transfer from the oxidized QD to the spiro-OMeTAD. Hindering this hole transfer,
which regenerates the QDs to a neutral charge state, would leave more QDs in the
oxidized state and increase the rate of recombination (which is dependent on the
Katherine Roelofs Ph.D. Dissertation
168
recombination rate constant and the concentration of oxidized QDs) from the TiO2
conduction band to oxidized QDs. The rate of hole transfer for QD regeneration has
recently been shown to have a significant impact on device performance. Regeneration
rates have been measured at microsecond time scales,36
while the initial electron injection
in QDSSCs with in-situ grown QDs occurs on the picosecond timescale or faster,3, 89-91
though we note that there are recent reports of QDSSCs with ex-situ grown QDs in which
electron injection has been reported on the nanosecond timescale.92
Due to the sensitivity
of device performance to the QD regeneration step, we suggest slow hole transfer for QD
regeneration as a likely mechanism by which decreasing the QD band gap increases the
rate of recombination. This proposed mechanism is further supported by previous
literature experiments showing a decrease in the yield of hole transfer from PbS QDs to
spiro-OMeTAD for PbS QDs larger than 2.5 nm; the authors also observe a faster decay
of the spiro-OMeTAD cation with increasing QD size, leading them to conclude that the
hole can be back-transferred from the spiro-OMeTAD cation to the QD if the QD valence
band shifts to high enough energy levels.36
Our observed result of shorter recombination lifetimes with increased SILAR
cycles runs contrary to previous reports on recombination in CdSe51, 72
and PbS73
QDSSCs. In the report on PbS QDs on SnO2 substrates, Cánovas and colleagues found
that recombination lifetimes increased from 2.6 to 8.1 nanoseconds from 1 to 3 SILAR
cycles,73
using optical pump-terahertz probe spectroscopy that focuses on the short-range
recombination processes the authors attribute to recombination to oxidized QDs. Our
results are not directly comparable, as we are measuring small-perturbation
recombination processes with resolution limited to the microsecond timescale;
recombination lifetimes at the TiO2/spiro-OMeTAD interface determined by transient
photovoltage measurements are typically 100 to 1000 microseconds.93
In addition, we
studied a larger size range of the PbS QDs, from 2 to 8 SILAR cycles, and on the
standard TiO2 substrate. For CdSe QDs, Bisquert and colleagues have found that
recombination lifetimes are longer with increasing SILAR cycles, which the authors
attribute to the QDs blocking recombination from the TiO2 to the HTM. They further
report that recombination lifetimes do not significantly change with increasing SILAR
cycles of CdS QDs.51
The difference in our results could be due to our use of PbS QDs;
Katherine Roelofs Ph.D. Dissertation
169
increasing QD size approaches the PbS bulk band gap of 0.37 eV, whereas CdSe has a
bulk band gap of 1.7 eV. Accordingly, the valence band edge of PbS QDs can rise quite
significantly, compared to CdSe or CdS QDs, making hole transfer from the QD to the
HTM more difficult in the case of PbS QDs, in keeping with the mechanism proposed
above. However, a range of QD materials with varying bulk band gaps would need to be
studied to determine if this accounts for the difference.
Because our result of shorter recombination lifetimes with higher SILAR cycles
was unexpected, we have further investigated this effect with a model, to see if the
proposed mechanism could produce the observed trend in recombination lifetimes. We
modeled the dependence of the recombination lifetime on the QD valence band level.
The measured recombination lifetime, , has components due to two main sources of
recombination at the interface: electrons in the TiO2 conduction band recombining with
the oxidized QD giving a lifetime , or with holes in the spiro-OMeTAD giving a
lifetime (eqn 1). This model is based on the treatment put forth by O’Regan,
Miettunen, Grätzel and colleagues in their review on optoelectronic measurements in
DSSCs.80
The transient photovoltage measurements above only track the recombination
of photoelectrons, already injected into the TiO2, that are now leaving the TiO2 via
recombination. As such, the ultimate recombination destinations of the TiO2 electrons in
QDSSCs are the same as DSSCs: the electron will recombine with a hole in the absorber
(the QD) or a hole in the TM. owever, the approach put forth by O’Regan et al. for
DSSCs is further complicated when applied to QDSSCs, due to the presence of defects at
the TiO2/QD interface, which can act as trap states, delaying or facilitating electrons
leaving the TiO2 to ultimately reach the QD or the HTM. Therefore, for simplicity of
approach, we assume that the additional complications of trap states can be incorporated
into the recombination time constants for recombination to the QD or the HTM.
(2)
(3)
Katherine Roelofs Ph.D. Dissertation
170
For this model, we focus on , which will have the strongest dependence on QD size.
is a function of the corresponding rate constant, , and the concentration of
oxidized quantum dots, , as seen in eqn 2. will depend on the rate constant of the
regeneration step . Hole transfer for regeneration is treated as having an Arrhenius
dependence with an activation energy, , associated with the transition state
energy level μ that the hole has to overcome when moving from the valence band level of
the QD, , to the spiro-OMeTAD HOMO level, as shown in eqn 3. Here, kB is the
Boltzmann constant, T is the temperature, and A is the frequency factor.
(4)
Equation (3) explains the decrease in recombination with increasing QD size (raising
levels) if we assume that the transition state energy level μ is constant, so that an increase
in does translate to decreased , and decreased amount of oxidized QDs, , to
recombine to. As we do not know the absolute values of the frequency factor or the
activation energy, we can only determine shifts in recombination lifetimes relative to a
known change in recombination parameters, in order to cancel out both A and μ. The
change in recombination lifetime of the case of interest, , relative to the base case, ,
is normalized by the change in the recombination lifetime from another known point, ,
relative to the base case, as shown in eqn 4. This normalization allows to be
simplified to depend on relative shifts in and known parameters of , the
Boltzmann constant, and , the temperature.
(5)
Katherine Roelofs Ph.D. Dissertation
171
Figure 6-14. Predicted dependence of recombination lifetimes from the valence band
model, equation (4). Experimental recombination lifetimes at VOC of 0.37 V are plotted
against the calculated QD valence band level. Included samples are for 4, 6, and 8
SILAR deposition cycles with base-assisted growth (NaOH, ED, and TEA). These
samples had similar QD coverages of the TiO2 surface (coverage ~ 10%), but varied in
QD size, and thus valence band position.
The full derivation is included below, and assumes that in equation (1) is
unchanging. Figure 6-14 plots the experimental recombination lifetimes versus for
the NaOH, ED, and TEA devices with 4, 6, or 8 SILAR deposition cycles, which were
found to have approximately the same QD coverage (~10.5%). The values (listed in
Supporting Information Table S2) were determined from reports of PbS QD band
positions at a given size,82
using the average QD diameters measured from TEM. As
these devices all had the same coverage, we are interested to see if, assuming the shift in
recombination lifetime are only due to changes in the QD EVB level, the proposed model
can predict the relative shifts in lifetime values. Accordingly, the shortest lifetime value
(8 cyc TEA) was chosen as and the longest lifetime value (4 cyc TEA) was chosen as
. With the high and low points of the model thus pinned, we look at whether the
intermediate values follow the dependency predicted by the model. It can be seen that
this model well describes the dependence of recombination lifetime on , leading to
the conclusion that the increased rates of recombination observed with increased QD size
could indeed be due to the upward shift of the QD VB that slows regeneration of the QD
by hole transfer to spiro-OMeTAD.
Katherine Roelofs Ph.D. Dissertation
172
One question remains – whether the longer recombination lifetimes could be
caused in the case of ED and TEA by the adsorption of these bases onto the TiO2 surface.
ED or TEA molecules attached to the TiO2 surface could act to passivate dangling bonds
on the TiO2, reducing trapping of electrons at TiO2 surface states; in addition, the
presence of ED or TEA on the TiO2 surface could inhibit electrons leaving the TiO2 to
recombine with spiro-OMeTAD. To determine whether adsorption of the base onto the
TiO2 surface was helping to lengthen recombination lifetimes, substrates were prepared
with QDs grown by the standard deposition process, and then dipped in 1 M TEA
aqueous solution after QD growth was complete (referred to as a post-growth TEA
treatment). Data is presented at 6 SILAR cycles, as this gives the optimal device
efficiency for the standard deposition process, and if TEA adsorbed on the TiO2 surface
is acting to reduce recombination, a post-growth TEA treatment should further improve
the performance of the standard process. Figure 6-15 compares the recombination
lifetimes of the standard process, TEA-assisted QD growth, and post-growth TEA
treatment. The corresponding J-V curves and EQE spectra are shown in Figure 6-16. It
can be seen that performance of the post-growth TEA treatment is almost identical to that
of the standard process, in the J-V, EQE, and recombination lifetime performance. Yet,
devices with TEA-assisted QD growth had longer recombination lifetimes. This shows
that the longer recombination lifetimes observed with TEA-assisted QD growth is a result
of changes in the QD deposition due to TEA, rather than changes due to the presence of
TEA on the TiO2 surface. This conclusion is further supported by the fact that the longer
recombination lifetimes observed with ED and TEA in Figure 6-13a were also observed
with NaOH-assisted QD growth. While the presence of NaOH during QD growth led to
quantitatively similar increases in QD deposition compared to ED and TEA, NaOH does
not introduce ligands into the system which can bind to the TiO2 surface.
Katherine Roelofs Ph.D. Dissertation
173
Figure 6-15. Recombination lifetimes of QDs grown for 6 SILAR cycles by the standard
process (6 cyc), by TEA-assisted QD growth (6 cyc TEA), as compared to QDs grown by
the standard process with a post-QD-growth TEA treatment (6 cyc post-growth TEA).
Figure 6-16. (a) J-V curves under 1 sun (AM 1.5) illumination and in the dark, and (b)
the EQE spectra for QDs grown for 6 SILAR cycles by the standard process (6 cyc), by
TEA-assisted QD growth (6 cyc TEA), as compared to QDs grown by the standard
process with a post-growth TEA treatment (6cyc post-growth TEA).
6.4. Conclusions
The loading of PbS QDs on nanoporous TiO2 substrates is significantly enhanced
by controlling the pH of the cation solution during SILAR deposition. A roughly three-
Katherine Roelofs Ph.D. Dissertation
174
fold increase in QD loading is achieved by growing QDs at pH 9 when using three
different bases – NaOH, ethylenediamine (ED), or triethanolamine (TEA) – indicating
that the changes in QD deposition are due primarily to the change in pH, rather than side
chemical reactions involving the ligands on ED or TEA. The higher QD loading leads to
increased device efficiency of solid-state QDSSCs, doubling the efficiency from an
average of 0.32% by the standard process to an average of 0.64% with base-assisted QD
growth. We found that higher QD loading lead to recombination lifetimes that were
longer by over an order of magnitude. The possibility of surface-adsorbed base
molecules impacting recombination was ruled out, leading us to conclude that the
improved recombination lifetimes with each base are due to the higher QD coverage of
the TiO2 surface effectively blocking TiO2 electrons from recombining with the HTM.
The net result of higher QD loading is increased light absorption and decreased
interfacial recombination, both of which can contribute to higher efficiencies. The QDs
also grow in size with increased deposition, which, for cases where QD coverage was
constant, was found to give higher interfacial recombination. Our proposed mechanism
for the increased recombination with QD size is that the upward shift in the QD valence
band in larger QDs inhibits hole transfer from the oxidized QD to the HTM, thus
boosting the concentration of oxidized QDs and increasing the recombination of TiO2
electrons to oxidized QDs. A kinetic model of the effects of QD size is proposed, and
provides a good explanation of the dependence of recombination lifetimes on size.
Despite this indication that QD-mediated recombination processes (e.g. electrons
from TiO2 to oxidized QDs) can be significant, the net effect of increased QD loading is
decreased interfacial recombination, due to reduced TiO2-HTM contact, and thus higher
charge collection efficiencies. To the authors’ knowledge, this constitutes the first report
separating the impact on recombination rates of QD loading and QD size in QDSSCs,
two variables which usually increase simultaneously during QD deposition. These results
identify the mechanisms by which increased QD deposition impacts device performance,
and clarify that the decline in device efficiencies at higher QD deposition is due to
changes in QD size rather than an increase in the number of QDs present. If QD size can
be carefully controlled while pushing to near-complete coverage of the TiO2 surface, we
Katherine Roelofs Ph.D. Dissertation
175
believe QDSSC device efficiencies can be increased into a new regime and emerge as a
competitive thin film photovoltaic technology.
6.5. Derivation: Modeling Recombination Lifetime
Here we show the derivation of the models applied to the recombination lifetime
data. We start with the electron recombination dependence as outlined in a recent review
by Grätzel, O’Regan and co-workers.80
Namely, there are two main recombination
routes at the interface, recombination of electrons in the TiO2 ( ) with the oxidized
quantum dot ( ), giving a recombination rate , or recombination of electrons in the
TiO2 with the oxidized spiro-OMeTAD ( ), giving a recombination rate . Assuming
first order reactions, these recombination rates can be written
(6)
(7)
where and are the respective rate constants.
(8)
The pseudo first order electron lifetime, , can be defined by
(9)
As described in detail by Grätzel, O’Regan and co-workers,80
the recombination lifetime
determined by small perturbation photovoltage transient measurements employed in our
studies is the small perturbation electron lifetime , which is related to by the
thermodynamic factor, (see equation 27 in ref 80
)
(10)
From the previous relations, we can write
(11)
Katherine Roelofs Ph.D. Dissertation
176
which is the basis for our recombination lifetime model.
Our model aims to describe the decrease in recombination lifetime with
increasing QD size, through the mechanism proposed in the main text. Namely, that with
increasing QD size or decreasing QD band gap, the QD valence band level ( ) rises.
The rise in hinders hole transfer from the oxidized QD to the spiro-OMeTAD,
boosting the concentration of oxidized QDs. This increases the rate of recombination
from electrons in the TiO2 to oxidized QDs, and thus decreases . With QD size
assumed to only affect in equation (11) through this mechanism, (i.e., , , and
are constant) we can write
(12)
Next, the overall rate equation is considered to find the dependence of on , the
rate constant for the transfer of holes from oxidized QDs to spiro-OMeTAD, with
concentration .
(13)
Here, is rate of electron injection into TiO2 from the photoexcited QD, with
concentration . Under steady state conditions,
and
(14)
As this model assumes the impact on with changing QD band gap is dominated by
changes in , can be linearized through a Taylor Expansion with respect to
(15)
The rate constant is modeled as having an Arrhenius dependence on an activation
energy that the hole has to overcome when moving from the QD valence band to the
spiro-OMeTAD HOMO level
Katherine Roelofs Ph.D. Dissertation
177
(16)
where A is the frequency factor, is the Boltzmann constant, and is an energy level of
the transition state for hole transfer. The activation energy for hole transfer is .
From equations (12), (15), and (16), we find
(17)
Equation (17) is used as the valence band model plotted in Figure 8, to describe the
dependence of on .
6.6. Financial Support, Collaborations, and
Copyrights
This chapter is adapted from work published in ACS Nano.94
Reprinted with
permission from Roelofs, K. E.; Herron, S. M.; Bent, S. F., ACS Nano 2015,
10.1021/acsnano.5b02853. Copyright 2015 American Chemical Society. The PbS QD
growth and device fabrication were supported by the Center on Nanostructuring for
Efficient Energy Conversion (CNEEC) at Stanford University, an Energy Frontier
Research Center funded by the US Department of Energy, Office of Science, Office of
Basic Energy Sciences under Award No. DE-SC0001060. Advanced materials
characterization of the PbS QDs (TEM) and advanced device characterization (EQE and
Transient Photovoltage Measurements) were supported by the US Department of Energy
through the Bay Area Photovoltaic Consortium under Award No. DE-EE0004946. We
would like to thank Mike McGehee for his generous sharing of lab equipment and solar
cell fabrication proceedures, and in particular Thomas Brennan, Colin Bailie, Eric Hoke,
and George Margulis for help setting up the transient photovoltage and photocurrent
measurement system. We further thank Pralay Santra and Axel Palmstrom for their input
on interfacial recombination processes.
Katherine Roelofs Ph.D. Dissertation
178
6.7. References
1. O'Regan, B.; Grätzel, M., Nature 1991, 353 (6346), 737-740.
2. Roelofs, K. E.; Brennan, T. P.; Bent, S. F., The Journal of Physical Chemistry
Letters 2014, 5 (2), 348-360.
3. Kamat, P. V., The Journal of Physical Chemistry Letters 2013, 4 (6), 908-918.
4. Rühle, S.; Shalom, M.; Zaban, A., ChemPhysChem 2010, 11 (11), 2290-2304.
5. Mora-Seró, I.; Bisquert, J., The Journal of Physical Chemistry Letters 2010, 1
(20), 3046-3052.
6. Hetsch, F.; Xu, X.; Wang, H.; Kershaw, S. V.; Rogach, A. L., The Journal of
Physical Chemistry Letters 2011, 2 (15), 1879-1887.
7. Peter, L. M., The Journal of Physical Chemistry Letters 2011, 2 (15), 1861-1867.
8. Yu, W. W.; Qu, L.; Guo, W.; Peng, X., Chemistry of Materials 2003, 15 (14),
2854-2860.
9. Meng, K.; Surolia, P. K.; Byrne, O.; Thampi, K. R., Journal of Power Sources
2014, 248 (0), 218-223.
10. Lee, H.; Wang, M.; Chen, P.; Gamelin, D. R.; Zakeeruddin, S. M.; Grätzel, M.;
Nazeeruddin, M. K., Nano Letters 2009, 9, 4221-4227.
11. Wang, J.; Mora-Seró, I.; Pan, Z.; Zhao, K.; Zhang, H.; Feng, Y.; Yang, G.;
Zhong, X.; Bisquert, J., Journal of the American Chemical Society 2012, 135 (42),
15913-15922.
12. Sung, S. D.; Lim, I.; Kang, P.; Lee, C.; Lee, W. I., Chemical Communications
2013, 49 (54), 6054-6056.
13. McDaniel, H.; Fuke, N.; Makarov, N. S.; Pietryga, J. M.; Klimov, V. I., Nat
Commun 2013, 4, 2887.
14. Pan, Z.; Mora-Seró, I.; Shen, Q.; Zhang, H.; Li, Y.; Zhao, K.; Wang, J.; Zhong,
X.; Bisquert, J., Journal of the American Chemical Society 2014, 136 (25), 9203-9210.
15. Duan, J.; Tang, Q.; He, B.; Yu, L., Electrochimica Acta 2014, 139 (0), 381-385.
16. O'Mahony, F. T. F.; Lutz, T.; Guijarro, N.; Gomez, R.; Haque, S. A., Energy &
Environmental Science 2012, 5 (12), 9760-9764.
17. Zhao, K.; Pan, Z.; Mora-Seró, I.; Cánovas, E.; Wang, H.; Song, Y.; Gong, X.;
Wang, J.; Bonn, M.; Bisquert, J.; Zhong, X., Journal of the American Chemical Society
2015, 137 (16), 5602-5609.
18. Yella, A.; Lee, H.-W.; Tsao, H. N.; Yi, C.; Chandiran, A. K.; Nazeeruddin, M. K.;
Diau, E. W.-G.; Yeh, C.-Y.; Zakeeruddin, S. M.; Grätzel, M., Science 2011, 334 (6056),
629-634.
19. Jean, J.; Brown, P. R.; Jaffe, R. L.; Buonassisi, T.; Bulovic, V., Energy &
Environmental Science 2015, 8 (4), 1200-1219.
20. Briscoe, J.; Dunn, S., Mater. Sci. Technol. 2011, 27 (12), 1741-1756.
21. Lee, M. M.; Teuscher, J.; Miyasaka, T.; Murakami, T. N.; Snaith, H. J., Science
2012, 338 (6107), 643-647.
22. Heo, J. H.; Im, S. H.; Noh, J. H.; Mandal, T. N.; Lim, C.-S.; Chang, J. A.; Lee, Y.
H.; Kim, H.-J.; Sarkar, A.; K., N.; Gratzel, M.; Seok, S. I., Nature Photonic 2013, 7 (6),
486-491.
23. Park, N.-G., The Journal of Physical Chemistry Letters 2013, 4 (15), 2423-2429.
Katherine Roelofs Ph.D. Dissertation
179
24. Snaith, H. J., The Journal of Physical Chemistry Letters 2013, 4 (21), 3623-3630.
25. Boix, P. P.; Larramona, G.; Jacob, A.; Delatouche, B.; Mora-Seró, I.; Bisquert, J.,
The Journal of Physical Chemistry C 2012, 116 (1), 1579-1587.
26. Bach, U.; Lupo, D.; Comte, P.; Moser, J. E.; Weissortel, F.; Salbeck, J.; Spreitzer,
H.; Grätzel, M., Nature 1998, 395 (6702), 583-585.
27. Snaith, H. J.; Schmidt-Mende, L., Advanced Materials 2007, 19 (20), 3187-3200.
28. Wolden, C. A.; Kurtin, J.; Baxter, J. B.; Repins, I.; Shaheen, S. E.; Torvik, J. T.;
Rockett, A. A.; Fthenakis, V. M.; Aydil, E. S., Journal of Vacuum Science &
Technology A 2011, 29 (3), 030801.
29. Wang, X.; Kurdgelashvili, L.; Byrne, J.; Barnett, A., Renewable and Sustainable
Energy Reviews 2011, 15 (9), 4248-4254.
30. Fthenakis, V.; Alsema, E., Progress in Photovoltaics: Research and Applications
2006, 14 (3), 275-280.
31. Snaith, H. J., Solid-State Dye-Sensitized Solar Cells with Molecular Hole
Transporters. In Dye-sensitized solar cells, Kalyanasundaram, K., Ed. CRC Press: Boca
Raton, FL, 2010.
32. Pagliaro, M.; Ciriminna, R.; Palmisano, G., ChemSusChem 2008, 1 (11), 880-891.
33. Docampo, P.; Tiwana, P.; Sakai, N.; Miura, H.; Herz, L.; Murakami, T.; Snaith,
H. J., The Journal of Physical Chemistry C 2012, 116 (43), 22840-22846.
34. Li, T. C.; Goes, M. S.; Fabregat-Santiago, F.; Bisquert, J.; Bueno, P. R.;
Prasittichai, C.; Hupp, J. T.; Marks, T. J., Journal of Physical Chemistry C 2009, 113
(42), 18385-18390.
35. Leijtens, T.; Ding, I. K.; Giovenzana, T.; Bloking, J. T.; McGehee, M. D.;
Sellinger, A., ACS Nano 2012, 6 (2), 1455-1462.
36. Lee, H.; Leventis, H. C.; Moon, S. J.; Chen, P.; Ito, S.; Haque, S. A.; Torres, T.;
Nuesch, F.; Geiger, T.; Zakeeruddin, S. M.; Grätzel, M.; Nazeeruddin, M. K., Adv. Func.
Mater. 2009, 19, 2735-2742.
37. Jun, H. K.; Careem, M. A.; Arof, A. K., Renewable and Sustainable Energy
Reviews 2013, 22 (0), 148-167.
38. Mora-Ser , I. Gim nez, S. Fabregat-Santiago, F. G mez, R.; Shen, Q.; Toyoda,
T.; Bisquert, J., Accounts of Chemical Research 2009, 42 (11), 1848-1857.
39. Hodes, G., The Journal of Physical Chemistry C 2008, 112, 17778-17787.
40. Handa, S.; Wietasch, H.; Thelakkat, M.; Durrant, J. R.; Haque, S. A., Chemical
Communications 2007, (17), 1725-1727.
41. Palomares, E.; Martinez-Diaz, M. V.; Haque, S. A.; Torres, T.; Durrant, J. R.,
Chemical Communications 2004, (18), 2112-2113.
42. Pellegrin, Y.; Le Pleux, L.; Blart, E.; Renaud, A.; Chavillon, B.; Szuwarski, N.;
Boujtita, M.; Cario, L.; Jobic, S.; Jacquemin, D.; Odobel, F., Journal of Photochemistry
and Photobiology A: Chemistry 2011, 219 (2–3), 235-242.
43. Hagberg, D. P.; Yum, J.-H.; Lee, H.; De Angelis, F.; Marinado, T.; Karlsson, K.
M.; Humphry-Baker, R.; Sun, L.; Hagfeldt, A.; Grätzel, M.; Nazeeruddin, M. K., Journal
of the American Chemical Society 2008, 130 (19), 6259-6266.
44. Guijarro, N. ana-Villarreal, T. Mora-Ser , I. n. Bisquert, J. G •mez, R., The
Journal of Physical Chemistry C 2009, 113 (10), 4208-4214.
45. Vogel, R.; Pohl, K.; Weller, H., Chemical Physics Letters 1990, 174, 241-245.
46. Toyoda, T.; Kobayashi, J.; Shen, Q., Thin Solid Films 2008, 516 (9), 2426-2431.
Katherine Roelofs Ph.D. Dissertation
180
47. od, I. Gonz lez-Pedro, V. Tachan, . Fabregat-Santiago, F. Mora-Ser •, I.;
Bisquert, J.; Zaban, A., The Journal of Physical Chemistry Letters 2011, 2, 3032-3035.
48. Sewvandi, G. A.; Chen, C.; Ishii, T.; Kusunose, T.; Tanaka, Y.; Nakanishi, S.;
Feng, Q., The Journal of Physical Chemistry C 2014, 118 (35), 20184-20192.
49. Tachibana, Y.; Haque, S. A.; Mercer, I. P.; Durrant, J. R.; Klug, D. R., The
Journal of Physical Chemistry B 2000, 104 (6), 1198-1205.
50. Guijarro, N.; Campina, J. M.; Shen, Q.; Toyoda, T.; Lana-Villarreal, T.; Gomez,
R., Physical Chemistry Chemical Physics 2011, 13 (25), 12024-12032.
51. González-Pedro, V.; Xu, X.; Mora-Seró, I.; Bisquert, J., ACS Nano 2010, 4 (10),
5783-5790.
52. McDaniel, H.; Fuke, N.; Pietryga, J. M.; Klimov, V. I., The Journal of Physical
Chemistry Letters 2013, 4 (3), 355-361.
53. Markus, T. Z.; Itzhakov, S.; Alkotzer, Y. I.; Cahen, D.; Hodes, G.; Oron, D.;
Naaman, R., The Journal of Physical Chemistry C 2011, 115 (27), 13236-13241.
54. Watson, D. F., The Journal of Physical Chemistry Letters 2010, 1 (15), 2299-
2309.
55. Mora-Seró, I.; Giménez, S.; Moehl, T.; Fabregat-Santiago, F.; Lana-Villareal, T.;
Gómez, R.; Bisquert, J., Nanotechnology 2008, 19 (42), 424007.
56. Wang, H.; McNellis, E. R.; Kinge, S.; Bonn, M.; Cánovas, E., Nano Letters 2013,
13 (11), 5311-5315.
57. Salant, A.; Shalom, M.; Hod, I.; Faust, A.; Zaban, A.; Banin, U., ACS Nano 2010,
4 (10), 5962-5968.
58. Salant, A.; Shalom, M.; Tachan, Z.; Buhbut, S.; Zaban, A.; Banin, U., Nano
Letters 2012, 12 (4), 2095-2100.
59. Chang, J. A.; Im, S. H.; Lee, Y. H.; Kim, H.-J.; Lim, C.-S.; Heo, J. H.; Seok, S. I.,
Nano Letters 2012, 12 (4), 1863-1867.
60. Shalom, M. Dor, S. R hle, S.; Grinis, L.; Zaban, A., The Journal of Physical
Chemistry C 2009, 113 (9), 3895-3898.
61. Sarkar, S. K.; Kim, J. Y.; Goldstein, D. N.; Neale, N. R.; Zhu, K.; Elliott, C. M.;
Frank, A. J.; George, S. M., The Journal of Physical Chemistry C 2010, 114 (17), 8032-
8039.
62. Brennan, T. P.; Ardalan, P.; Lee, H.-B.-R.; Bakke, J. R.; Ding, I. K.; McGehee,
M. D.; Bent, S. F., Advanced Energy Materials 2011, 1 (6), 1169-1175.
63. Brennan, T. P.; Trejo, O.; Roelofs, K. E.; Xu, J.; Prinz, F. B.; Bent, S. F., Journal
of Materials Chemistry A 2013, 1 (26), 7566-7571.
64. Pathan, H. M.; Lokhande, C. D., Bull. Mater. Sci. 2004, 27, 85-111.
65. Baker, D. R.; Kamat, P. V., Advanced Functional Materials 2009, 19 (5), 805-
811.
66. Zeng, Q.; Kong, X.; Sun, Y.; Zhang, Y.; Tu, L.; Zhao, J.; Zhang, H., The Journal
of Physical Chemistry C 2008, 112 (23), 8587-8593.
67. Chi, C.-F.; Chen, P.; Lee, Y.-L.; Liu, I. P.; Chou, S.-C.; Zhang, X.-L.; Bach, U.,
Journal of Materials Chemistry 2011, 21 (43), 17534-17540.
68. Jun, H. K.; Careem, M. A.; Arof, A. K., International Journal of Photoenergy
2014, 2014, 14.
69. Liu, I. P.; Chang, C.-W.; Teng, H.; Lee, Y.-L., ACS Applied Materials &
Interfaces 2014, 6 (21), 19378-19384.
Katherine Roelofs Ph.D. Dissertation
181
70. Edri, E.; Rabinovich, E.; Niitsoo, O.; Cohen, H.; Bendikov, T.; Hodes, G., The
Journal of Physical Chemistry C 2010, 114 (30), 13092-13097.
71. Lee, J.-W.; Hong, J.-D.; Park, N.-G., Chemical Communications 2013, 49 (57),
6448-6450.
72. Guijarro, N.; Lana-Villarreal, T.; Shen, Q.; Toyoda, T.; Gómez, R., The Journal
of Physical Chemistry C 2010, 114 (50), 21928-21937.
73. Wang, H.; Barceló, I.; Lana-Villarreal, T.; Gómez, R.; Bonn, M.; Cánovas, E.,
Nano Letters 2014, 14 (10), 5780-5786.
74. Roelofs, K. E.; Brennan, T. P.; Dominguez, J. C.; Bailie, C. D.; Margulis, G. Y.;
Hoke, E. T.; McGehee, M. D.; Bent, S. F., The Journal of Physical Chemistry C 2013,
117 (11), 5584-5592.
75. Nguyen, W. H.; Bailie, C. D.; Unger, E. L.; McGehee, M. D., Journal of the
American Chemical Society 2014, 136 (31), 10996-11001.
76. Yang, L.; Xu, B.; Bi, D.; Tian, H.; Boschloo, G.; Sun, L.; Hagfeldt, A.;
Johansson, E. M. J., Journal of the American Chemical Society 2013, 135 (19), 7378-
7385.
77. Brennan, T. P.; Bakke, J. R.; Ding, I. K.; Hardin, B. E.; Nguyen, W. H.; Mondal,
R.; Bailie, C. D.; Margulis, G. Y.; Hoke, E. T.; Sellinger, A.; McGehee, M. D.; Bent, S.
F., Physical Chemistry Chemical Physics 2012, 14 (35), 12130-12140.
78. Brennan, T. P.; Tanskanen, J. T.; Roelofs, K. E.; To, J. W. F.; Nguyen, W. H.;
Bakke, J. R.; Ding, I. K.; Hardin, B. E.; Sellinger, A.; McGehee, M. D.; Bent, S. F., The
Journal of Physical Chemistry C 2013, 117 (46), 24138-24149.
79. O'Regan, B. C.; Lenzmann, F., The Journal of Physical Chemistry B 2004, 108
(14), 4342-4350.
80. Barnes, P. R. F. Miettunen, K. i, X. Anderson, A. . Bessho, T. Grätzel, M.;
O'Regan, B. C., Advanced Materials 2013, 25 (13), 1881-1922.
81. Thielsch, R.; Böhme, T.; Reiche, R.; Schläfer, D.; Bauer, H. D.; Böttcher, H.,
Nanostructured Materials 1998, 10 (2), 131-149.
82. yun, B.-R. hong, .-W. Bartnik, A. C. Sun, . Abru a, H. D.; Wise, F. W.;
Goodreau, J. D.; Matthews, J. R.; Leslie, T. M.; Borrelli, N. F., ACS Nano 2008, 2 (11),
2206-2212.
83. Moreels, I.; Lambert, K.; Smeets, D.; De Muynck, D.; Nollet, T.; Martins, J. C.;
Vanhaecke, F.; Vantomme, A.; Delerue, C.; Allan, G.; Hens, Z., ACS Nano 2009, 3 (10),
3023-3030.
84. Cademartiri, L.; Montanari, E.; Calestani, G.; Migliori, A.; Guagliardi, A.; Ozin,
G. A., Journal of the American Chemical Society 2006, 128 (31), 10337-10346.
85. Trang Pham, T. T.; Bessho, T.; Mathews, N.; Zakeeruddin, S. M.; Lam, Y. M.;
Mhaisalkar, S.; Gratzel, M., Journal of Materials Chemistry 2012, 22 (32), 16201-16204.
86. Kosmulski, M., Advances in Colloid and Interface Science 2002, 99 (3), 255-264.
87. Robel, I.; Kuno, M.; Kamat, P. V., Journal of the American Chemical Society
2007, 129 (14), 4136-4137.
88. Leventis, H. C.; O'Mahony, F.; Akhtar, J.; Afzaal, M.; O'Brien, P.; Haque, S. A.,
Journal of the American Chemical Society 2010, 132 (8), 2743-2750.
89. Yang, Y.; Rodríguez-Córdoba, W.; Xiang, X.; Lian, T., Nano Letters 2012, 12
(1), 303-309.
Katherine Roelofs Ph.D. Dissertation
182
90. Pijpers, J. J. H.; Koole, R.; Evers, W. H.; Houtepen, A. J.; Boehme, S.; de Mello
Donegá, C.; Vanmaekelbergh, D.; Bonn, M., The Journal of Physical Chemistry C 2010,
114 (44), 18866-18873.
91. Guijarro, N.; Shen, Q.; Giménez, S.; Mora-Seró, I.; Bisquert, J.; Lana-Villarreal,
T.; Toyoda, T.; Gómez, R., The Journal of Physical Chemistry C 2010, 114 (50), 22352-
22360.
92. Makarov, N. S.; McDaniel, H.; Fuke, N.; Robel, I.; Klimov, V. I., The Journal of
Physical Chemistry Letters 2014, 5 (1), 111-118.
93. O'Regan, B. C.; Scully, S.; Mayer, A. C.; Palomares, E.; Durrant, J., The Journal
of Physical Chemistry B 2005, 109 (10), 4616-4623.
94. Roelofs, K. E.; Herron, S. M.; Bent, S. F., ACS Nano 2015,
10.1021/acsnano.5b02853.
Katherine Roelofs Ph.D. Dissertation
183
Chapter 7. ALD Transport Layers in
Perovskite Solar Cells
The high efficiency of lead-based perovskite solar cells (PSCs) has generated
great interest in both fundamental research and commercialization of this thin film
photovoltaic (PV) technology. However these devices still suffer from stability issues,
hysteresis effects, and processing variability. In this chapter, we report on the
morphology, structure, and electrical properties of compact TiO2 layers in perovskite
solar cells, which act to collect electrons and block holes. We demonstrate the
fabrication of compact TiO2 layers by atomic layer deposition (ALD) in nanoporous
perovskite solar cells, and show that a 3 nm thick ALD TiO2 film can out-perform the
conventionally-used 50 nm spray-pyrolysis TiO2 layer. The ALD-grown TiO2 is found to
exist in the rutile phase, and the details of the variation in crystallinity with film thickness
and annealing time are studied by grazing-incidence X-ray diffraction (GIXRD) with a
synchrotron X-ray source and an in-situ annealing chamber.
We further explore the use of ALD metal oxide capping layers deposited above
the perovskite layer in the device stack to act as a passivation and/or encapsulation layer.
We first explore the stability of the perovskite layer to ALD films deposited directly atop
the perovskite, as the ALD process itself could degrade the perovskite due to the use of
H2O or ozone as a counter-reactant. We further examine the ability of ALD metal oxide
layers to protect the perovskite from oxidizing conditions. Finally, we look at the use of
ALD layers in wide band gap perovskite solar cells as a passivating layer to push the
device to high VOC values. Such a device, a wide band gap perovskite cell encapsulated
with a metal oxide layer, could potentially be used as the top solar cell in a tandem
photoelectrocatalysis device.
Katherine Roelofs Ph.D. Dissertation
184
7.1. Experimental Methods
TiO2 Film Deposition. Perovskite solar cells were fabricated according to
previous procedures.1 For the transparent electrode, glass substrates coated with fluorine-
doped tin oxide (15 Ω/□, Pilkington) were patterned by etching with 4 M Cl and n
powder. The substrate was cleaned by sonicating for 15 min in DI water with lab grade
soap, followed by sonication for 15 min in ethanol.
The compact TiO2 layer was deposited by spray pyrolysis or ALD. For aerosol
spray pyrolysis, substrates were heated to 450 °C, and then coated with a thin (~50 nm)
compact layer of TiO2 using air as a carrier gas, with titanium diisopropoxide
bis(acetylacetonate) (Sigma 325252) diluted in ethanol as the precursor. For ALD TiO2
deposition, TiCl4 was used as the precursor, with H2O as the counter-reactant. The TiCl4
process was conducted in the home-built reactor, as described in previous work.2 A
pulse/purge times of 1 sec/30 sec are used for the TiCl4 and the H2O. For ALD, the TiO2
growth rate is roughly 0.4‒0.5 Å/cycle. TiO2 thicknesses were determined by TiO2
growth on a reference Si wafer, using a J.A. Woollam M2000 Variable Angle
Spectroscopic Ellipsometer (VASE) at 65, 70 and 75° angles of incidence and
wavelengths ranging from 210 to 1688 nm.
For nanoporous devices, the nanoporous TiO2 layer was then deposited by spin-
coating a commercial paste of 20 nm diameter anatase TiO2 particles in ethyl cellulose
and terpineol (Dyesol 18-NRT), which was diluted with additional terpineol at a 1:3
weight ratio of the commercial paste to ethanol. The films were annealed at 450 °C,
resulting in a film thickness of ~350 nm as observed by scanning electron microscopy
(SEM). For the TiCl4 treatment, the nanoporous TiO2 films were immersed overnight in a
0.02 M aqueous TiCl4 solution at room temperature, and then annealed again at 450 °C.
NiOx ALD. For the inverted wide band gap perovskite solar cells, NiOx films are
used as an inorganic hole transport layer underneath the perovskite solar cells. The NiOx
films are deposited on In:SnO2 (ITO), due to the lower roughness of ITO as compared to
FTO, making ITO preferable for planar architectures. The ALD process used
nickelocene [bis(cyclopentadienyl)nickel(II), Sigma Aldrich] as the precursor, and ozone
Katherine Roelofs Ph.D. Dissertation
185
as the counter-reactant. The ALD process conditions were adapted from those developed
by Katie Nardi.3 The UV-ozone source was delivered with an O2 carrier gas, with an
ozone concentration of roughly 200 to 250 grams/normal-m3. The pulse/purge time of
the nickelocene was 1s/30s, while ozone was 3s/30s. The NiOx ALD process was
conducted at a stage temperature of 180 °C, with the nickelocene bubbler heated to 70 °C
and the nickelocene delivery line heated to 110 °C. The growth rate was found to be
~0.43 Å/cycle, with a roughly 20 cycle nucleation delay. The resulting NiOx films on
ITO were annealed at 250 °C in air prior to use in devices, to improve the conductivity of
the NiOx layer.
Grazing incidence X-ray diffraction. GIXRD measurements were performed
on Beamline 11-3 at the Stanford Synchrotron Radiation Lightsource (SSRL). The
measurements were taken with a MAR 345 imaging plate with 12.735 keV X-ray energy,
and an incident angle of 2°. In-situ annealing measurements were performed on
Beamline 7-2 at the SSRL. The measurements were taken with a Pilatus 100K detector
with 12.398 keV X-ray energy. The in-situ annealing chamber is constructed of
aluminum and uses two Comstat Low Voltage Cartridge Heaters (MCH2-40W-002) to
heat the sample stage. A Cryocon C24 Temperature Controller (Type K Thermocouple)
is used to control the temperature with a Sorensen power supply DLM 40-15. The
chamber is hermetic with gas flow in and out and has an integrated water cooling system.
The windows are made of 2mil (50.8um) thick Kapton Tape which is transparent to x-
rays at the energies used. The GIXRD data was analyzed with an in-house software
WxWindows Diffraction Integration Tool v1.15 (developed by Stefan Mannsfeld at
SSRL). The data from the 2-D area detectors at 11-3 or 7-2 was analyzed as follows: first
it was calibrated using spectra collected from a LaB6 crystal under identical conditions,
then converted to qxy-qz, followed by a conversion to qchi, and finally integrated over Chi
to produce 1-D plot of intensity vs. q.
Perovskite deposition. Methyl ammonium lead iodide (CH3NH3PbI3) perovskite
layers are deposited as the absorber material. These perovskite solar cells were made in
both the planar configuration, with the perovskite deposited directly on the compact TiO2
layer, and in the nanoporous configuration, with the perovskite deposited on the spin-
Katherine Roelofs Ph.D. Dissertation
186
coated nanoporous TiO2 layer (~350 nm thick). The perovskite layer was synthesized by
the two-step deposition process put forth by Grätzel and colleagues,4 and modified by
Hagfeldt and colleagues.5 The perovskite deposition is done completely in a N2
atmosphere glovebox. First, a 1.3 M PbI2 (Sigma Aldrich) in dimethylformamide (DMF)
solution is made, by heating to 100 °C to dissolve the PbI2. A 10 mg/mL solution of
methylammonium iodide (Dyesol) in isopropyl alcohol (IPA) is made at room
temperature.
The TiO2-coated FTO substrates are heated to above 450 °C to drive of surface
H2O, and brought into the glovebox. The substrates are pre-heated to 70 °C immediately
prior to PbI2 deposition. The PbI2 in DMF solution is maintained at 100 °C, and, for the
standard deposition process, is spin-coated at 6,500 rpm for 90 s. Section 3 above
discusses the impact of different spin-coating speeds. The glovebox is purged for 5
minutes, between every couple substrates, or even after every single substrate to remove
the DMF from the atmosphere. The PbI2 films are then dried on a hot plate at 70 °C for
between 15 and 30 min, removed and allowed to come to room temperature, and then
dipped in the room temperature solution MAI in IPA for between 20 and 30 min. The
substrates are then rinsed in IPA for 10 s, and spin-coated dry at 4000 rpm for 30 s. The
substrates are then further dried on the hot plate at 70 °C for between 15 and 30 min. The
deposition parameter range given here is further optimized for each specific architecture.
Perovskite Characterization. A Cary 6000i UV-Visible spectrometer (Varian)
was used to characterize the optical properties of the deposited perovskite. The
perovskite was also analyzed by powder X-ray diffraction (XRD), using the using the
PANalytical X'Pert PRO system in parallel beam mode with Cu Kα radiation at 45 kV
and 40 mA.
Device Fabrication. For the standard n-i-p structure, the perovskite films on
TiO2/FTO substrates were coated with spiro-OMeTAD and a top contact of Au. The
spiro-OMeTAD layer was deposited as follows. As described elsewhere,6 the solution of
the solid-state hole-transporting material was composed of 225 mg mL-1
of spiro-
OMeTAD (Lumtec LT-S922) dissolved in chlorobenzene, with tert-butylpyridine added
at a ratio of 1:10.3 μ :mg of spiro-OMeTAD, and lithium bis-
Katherine Roelofs Ph.D. Dissertation
187
(trifluoromethylsulfonyl)imide salt (170 mg mL-1
in acetonitrile) added at a ratio of 1:4.8
μ :mg of spiro-OMeTAD. A small amount of the spiro-OMeTAD solution (30 μ for
3.75 cm2 substrates) was deposited onto the TiO2 substrates at room temperature, and
spin-coated at 2000 RPM for 30 s. Finally, 200 nm thick Au counter electrodes were
deposited by thermal evaporation under vacuum below 10-6
torr.
For the inverted p-i-n structure, the perovskite films on NiOx/ITO substrates were
coated with PCBM (phenyl-C61-butyric acid methyl ester) and ALD TiO2 layers,
followed by a top contact of Ag. The PCBM layer was deposited following the
procedure developed by Bai et al.7 A solution of 20 mg/mL PCBM in chlorobenzene was
dissolved overnight at 50 °C, as well as a separate solution of 6 mg/mL polystyrene in
chlorobenzene at 60 °C. The polystyrene was then mixed with the PCBM by 1.7 wt% (or
2.4 wt%) of the polystyrene solution to the PCBM solution. The resulting solution was
spin-coated onto the perovskite substrate at 1500 rpm for 60 s and 2000 rpm for 5 s.
Finally, 200 nm thick Ag counter electrodes were deposited by thermal evaporation under
vacuum below 10-6
torr. Final device areas were on the order of 0.1 cm2. Devices were
stored inside a N2 environment prior to electrical measurements.
Electrical Measurements. For J-V measurements, an AM 1.5 solar simulator
(Oriel 91160) equipped with a 300 W ozone-free Xe arc lamp (6258) was used. The lamp
was calibrated to 1 sun (100 mW cm-2
) using a reference NREL calibrated Si photodiode
equipped with an IR cutoff filter. Current-voltage (J-V) curves were collected with a
Keithley 2400 SourceMeter. For optimization of device performance, J-V curves were
collected at a sweep rate of 0.1 s delay per every 0.1 V step. To collected the ‘settled’,
non-hysteretic device performance, J-V curves were collected at a sweep rate of 4 s delay
per every 0.1 V step.1 Devices were light-soaked until maximum efficiencies were
reached (up to 10 min), as was previously found necessary.1 Dark curves were measured
after the maximum efficiency was reached, i.e., after the light-soaking process.
7.2. Conformal TiO2 Blocking Layers by ALD
Perovskite solar cells have recently surpassed efficiencies of 20%,8 emerging as a
competitive thin film photovoltaic technology. In PSCs, the absorber has the
Katherine Roelofs Ph.D. Dissertation
188
composition ABX3 and a perovskite crystal structure; here A is an organic cation such as
CH3NH3+, B is almost always Pb
2+ as the divalent cation, and X is a halide anion,
commonly I-. Lead-based perovskite absorbers were first incorporated in solar cells by
Miyasaka et al. in 2009 as perovskite-nanocrystal-sensitized solar cells with efficiencies
of 3.8%.9 This design was further improved by Park and coworkers to 6.8% efficiencies
although the use of a liquid electrolyte in that work caused the rapid degradation of the
device.10
Then in 2012 Grätzel, Park and colleagues replaced the liquid electrolyte with a
solid-state hole-transport material (HTM), producing perovskite solar cells with
efficiencies of 9.7%.11
This was followed closely by Snaith and coworkers reporting
10.9% efficiencies.12
The high efficiencies of perovskite solar cells have since been attributed to several
properties of the perovskite: strong absorption13
due to a high density of states in the
conduction band,14
electron and hole diffusion lengths of over 100 nm,15
and a suggestion
based on initial results that intrinsic defects and grain boundary defects induce minimal
recombination.14
The long diffusion lengths of both carriers far surpass that of most
solution-processed materials, and this property together with other measurements16
indicate that lead-based perovskite are intrinsic (i-type) conductors. The promising
properties of perovskite solar cells have generated great interest in their
commercialization, yet there remain several critical challenges to address. These include
the toxicity of Pb, the hysteresis behavior17
that is thought to be due to ion migration in
the perovskite film,18
and the stability issues of the perovskite absorber,19
which is
susceptible to degradation when exposed to oxygen or moisture.
Perovskite solar cells borrow from the dye-sensitized solar cell (DSSC) and
quantum-dot-sensitized solar cell (QDSSC) architectures, yet there are substantial
differences. For one, perovskite solar cells perform well in both the nanostructured
architecture of the sensitized design and in the planar architecture where the perovskite is
deposited as a film atop a planar substrate. Due to such differences, the selective contacts
for electron collection and hole collection need to be re-optimized from that of the DSSC
and QDSSC designs to work effectively in perovskite solar cells.20
For the standard n-i-p
architecture of the perovskite solar cell, a transparent conductive oxide (TCO) contact is
Katherine Roelofs Ph.D. Dissertation
189
coated with the electron-transport material, upon which the perovskite layer is deposited,
followed by the hole-transport material. Due to processing constraints, the HTMs are
typically a solid-state organic material such as spiro-OMeTAD. For the n-type electron-
transport layer, a compact TiO2 film is commonly employed. A nanoporous TiO2 film is
deposited atop the compact TiO2 layer to produce the nanoporous architecture, or the
perovskite is deposited directly atop the compact TiO2 to produce the planar architecture.
In the nanoporous design, the compact TiO2 prevents shorting that could otherwise occur
if the HTM penetrates the nanoporous layer, touching the TCO. Thus the compact TiO2
layer is also referred to as a hole-blocking layer, or an electron-collecting selective
contact.
In perovskite solar cells, the morphology of the compact TiO2 layer is important
in creating a good junction for charge collection. For use as an effective blocking layer,
the compact TiO2 should be as thin as possible while still achieving complete coverage
and good blocking properties. The deposition of a uniform, conformal layer is critical for
minimizing overall film thickness. In addition, the TiO2/perovskite interface has
attracted significant interest in perovskite solar cells, because the TiO2 has been found to
impact perovskite film morphology, perovskite stability due to TiO2 photocatalytic-
induced degradation processes,19
and the hysteresis behavior of the device by reducing
the internal electric field via surface capacitance.18, 21, 22
Finally, initial reports suggest
that the morphology of the perovskite layer is impacted by the roughness and/or thickness
of the underlying TiO2 layer and possibly also the morphology of the underlying TCO.23
Processing limitations of the standard spray-pyrolysis fabrication of the compact
TiO2 has been implicated in batch variability in devices,24
although for perovskite solar
cells the processing of the perovskite layer is also a strong contributor to the variability.
The compact TiO2 layer in perovskite solar cells are adapted from that of DSSCs, and are
primarily deposited by aerosol spray-pyrolysis,25-29
though other methods are also used,
e.g. spin-coating,30
electrostatic spray deposition,31
chemical vapor deposition,32
thermal
oxidation,33
electrochemical techniques,34
and ALD.30, 35
Despite the use of compact TiO2 in DSSCs for over two decades, the impact of
the phase and crystallinity of the compact TiO2 layer has not been extensively studied,26,
Katherine Roelofs Ph.D. Dissertation
190
36 partially due to the difficulty of collecting XRD signal from the <50 nm thick films
using standard X-ray sources. A few initial reports have suggested that rutile TiO2
performs similarly to anatase TiO2,37
despite the differences in band positions.38
There
have been studies of the impact of c-TiO2 film thickness in perovskite solar cells often
along with comparisons of different deposition techniques.30, 39, 40
Previous reports on
compact TiO2 layers deposited on FTO show the TiO2 is in the anatase phase for use in
DSSC41-44
or perovskite solar cells45
though the rutile TiO2 phase has been observed as
well.46
In this work, we explore how phase, thickness, and crystallinity of the TiO2
compact film will affect perovskite solar cell performance. We apply ALD as a proof of
concept method to achieve conformal ultra-thin TiO2 layers with controlled crystallinity
to explore the impact on device performance.
7.3. Results and Discussion - TiO2 Blocking
Layers
A synchrotron X-ray source was used to conduct GIXRD analysis of thin TiO2
blocking layers grown by spray pyrolysis and by ALD on FTO and silicon substrates.
The spray-pyrolysis TiO2 was deposited at 500 °C, while the ALD TiO2 was deposited at
300 °C – a temperature optimized in our previous studies for TiO2 ALD on Si wafers.2
The spray-pyrolysis TiO2 film thickness of 50 nm was chosen based on optimization in
the DSSC system over the previous decade25-29
and is also used widely in perovskite solar
cells.47
The spray-pyrolysis deposition parameters were taken from our previous work on
DSSCs.48
Figure 7-1 shows the GIXRD data for 50 nm ALD TiO2 films on an FTO
substrate compared to a Si substrate, with both samples annealed at 450 °C in air to
replicate the heat treatment used in device fabrication. Interestingly, the annealed ALD
TiO2 films were in the rutile phase when deposited on FTO, but the anatase phase on the
Si substrate. Prior to annealing, substrate specificity was also observed: for the films that
were thick enough to be crystalline without annealing, the TiO2 grew in the rutile phase
on FTO and the anatase phase on Si. The growth of rutile TiO2 by ALD is attributed to
texturing by the rutile FTO substrate. There are reports in the literature of high-
temperature deposition processes producing rutile TiO2 films instead of anatase on SnO2
Katherine Roelofs Ph.D. Dissertation
191
substrates,37, 49
though reports of growth of rutile TiO2 nanorods/nanowires50-52
or
branched nanocrystals53, 54
on SnO2 are more common. Spray-pyrolysis on FTO was
found to produce amorphous TiO2, even after annealing at 450 °C. While epitaxial
growth of rutile TiO2 on rutile SnO2 have been reported,49
the area pattern of the GIXRD
in Figure 7-1b is not consistent with epitaxial growth. We note that in Figure 7-1a and b
the rutile TiO2 (200) peak is obscured by the rutile SnO2 (111) peak.
Katherine Roelofs Ph.D. Dissertation
192
Figure 7-1. (a) GIXRD data of 1700 ALD cycles of TiO2 (roughly 50 nm), initially
deposited at 300 °C and subsequently annealed at 450 °C for 1 hr. Shown are ALD TiO2
films deposited on a Si substrate, producing the anatase phase, and on an FTO substrate,
producing the rutile phase. Also shown are the GIXRD of a Si blank and FTO blank.
Anatase TiO2 peaks are labeled with circles, rutile TiO2 with red squares, and rutile SnO2
with diamonds. (b) GIXRD area pattern of 1700 ALD cycles of TiO2 on FTO, initially
Katherine Roelofs Ph.D. Dissertation
193
deposited at 300 °C and subsequently annealed at 450 °C for 1 hr scale bar shown
ranging from high (red) to low (blue) intensity.
ALD TiO2 film thicknesses were approximated based on ellipsometry
measurements of ALD TiO2 grown on reference Si wafers that were included in each
ALD run. From these measurements the ALD TiO2 growth rate was found to be between
0.4 to 0.5 Å per ALD cycle, which is consistent with previous reports in the literature for
the TiCl4 precursor with H2O as the counter-reactant.2, 55
The growth on FTO is
estimated to occur at a similar rate, since 1700 TiO2 ALD cycles on FTO was measured
to be roughly 50 nm thick by SEM cross-sectional images. We note that ALD TiO2
deposited on FTO could have different initial growth behavior, such as a nucleation delay
that would overestimate the thickness of low-ALD cycle films on FTO, due to the fact
that the ALD TiO2 growth on Si has a minimal if any nucleation delay.2
We explored the crystallographic structure as a function of thickness for ALD
TiO2 films deposited at 300 °C. The GIXRD data only showed diffraction peaks for TiO2
films grown thicker than ~200 ALD cycles (~6 nm), all in the rutile phase. The thinner
TiO2 films of 50 and 100 ALD cycle TiO2 films were either amorphous or too thin to
detect a signal. Due to the unexpected observation of rutile TiO2 phase grown by ALD,
the effect of deposition temperature was also studied. It was found that ALD TiO2 films
grown at 100 °C were amorphous for all thicknesses studied (up to 1700 ALD cycles).
At 350 °C, the temperature limit of the reactor, the behavior was similar to that at 300 °C,
with films above ~6 nm showing crystallinity. At intermediate temperatures, the film
crystallinity was dependent on film thickness, with greater crystallinity observed at
higher temperatures and thicker films.
It is of interest to determine whether the ALD process itself leads to the formation
of rutile TiO2. To explore this, we conducted in-situ GIXRD annealing of TiO2 films
initially deposited by ALD at 100 °C. As discussed above, ALD of TiO2 films at 100 °C
on FTO were found to be amorphous. The aim of this annealing study was to learn
whether the rutile TiO2 observed at higher ALD temperatures is formed due to the gas-
phase ALD precursors orienting to the underlying rutile SnO2 surface, or by a texturing
process that only occurs at the higher ALD temperatures. If the former, it may be
Katherine Roelofs Ph.D. Dissertation
194
expected that the amorphous films produced by the low-temperature ALD process at 100
°C will crystallize into the anatase phase when annealed. 1700 cyc ALD TiO2 films
deposited at 100 °C on FTO were subsequently annealed to 500 °C in an in-situ chamber
with GIXRD data collected intermittently during the annealing process. The substrate
was heated at a rate of 5 °C/min, from room temperature to 500 °C and then held at 500
°C for 2 hr., following a cooling to room temperature that occurred over the course of
roughly 1 hr. The spectra shown in Figure 7-2 were collected at time intervals during the
2 hr hold at 500 °C, where 0 min refers to the point where the substrate has finished the
initial ramp up. As shown in Figure 7-2, the initially amorphous films crystallized into
the rutile phase on FTO, indicating that the rutile phase forms at the TiO2/FTO interface
at elevated temperatures.
Figure 7-2. In-situ GIXRD taken during annealing of 1700 ALD cycle TiO2 film grown
on FTO at 100 °C. From the initially-amorphous TiO2 film, the (101) rutile TiO2 peak at
Q = 2.5 Å-1
is observed to form during annealing at 500 °C for 2 hrs. Significant heating
occurred during the temperature ramp-down, furthering the crystallization.
Figure 7-3 shows scanning electron microscopy (SEM) images of ALD TiO2 of
varied thicknesses on FTO compared to spray-pyrolysis TiO2 on FTO. First, comparing
Figure 7-3a (collected at 100 °C) to Figure 7-3b (300 °C), the effect of temperature can
be seen. Based on previous studies, and the GIXRD results of 100 cycle ALD TiO2 films
Katherine Roelofs Ph.D. Dissertation
195
grown at 100 °C, we know that the morphology shown in Figure 7-3a is that of
amorphous TiO2. The amorphous TiO2 is seen to grow in domains roughly 200 nm
across. However, the 100 cycle ALD TiO2 grown at 300 °C in Figure 7-3b forms small
~10 nm particles. Then by 1700 cycles (Figure 7-3c), these TiO2 particles converge to
completely cover the Si substrate; according to the GIXRD results, the TiO2 under these
conditions is crystalline and in the anatase phase.
The TiO2 deposited by spray-pyrolysis on FTO at 450 °C (Figure 7-3d) is
amorphous, and can be seen to roughen the underlying FTO nanocrystals (~100 nm in
diameter) with smaller features (~20 nm in diameter). Figure 7-3d shows 100 cyc ALD
TiO2 at 300 °C, annealed at 450 °C to match device processing temperatures. Here, the
ALD TiO2 layer is not visible, and the sample is indistinguishable from an uncoated FTO
substrate. When the ALD layer is grown thicker on the FTO, it becomes visible. Figure
7-3e shows 1700 cycle ALD TiO2 deposited on FTO with identical temperature
processing. The improved conformality of 1700 ALD cycles (~50 nm) as compared to
the 50 nm spray-pyrolysis layer can be seen.
Figure 7-3. Scanning electron microscopy (SEM) images of TiO2 films at the same
magnification, with scale bar of 200 nm shown. ALD TiO2 grown on Si: (a) 100 cyc at
100 °C, (b) 100 cyc at 300 °C, and (c) 1700 cyc at 300 °C. Also shown are TiO2 films
grown on FTO: (d) 50 nm spray-pyrolysis TiO2 at 450 °C as well as (e) 100 cyc ALD
Katherine Roelofs Ph.D. Dissertation
196
TiO2 at 300 °C and (f) 1700 cyc ALD TiO2 at 300 °C. Films (e) and (f) were annealed at
450 °C to match device fabrication conditions. 1700 ALD cycles of TiO2 give a ~50 nm
thick film.
Nanoporous perovskite solar cells were fabricated with 50 nm thick spray-
pyrolysis TiO2 as well as ALD TiO2 of varied thicknesses. The settled J-V curves are
shown in Figure 7-4a. These curves were collected with a delay of 4 sec between each
0.1 V step, at a rate that previous studies have found to give the ‘settled’ device
performance, avoiding hysteresis effects. Indeed, we found in our devices nominal
hysteresis behavior at this scan rate, though faster scan rates gave higher efficiencies and
greater hysteresis. We found that cells made with 100 cycle ALD TiO2 films (~3 nm
thick) rival the performance of cells consisting of a 50 nm thick spray-pyrolysis TiO2
layer. Interestingly, a 1700 cycle ALD TiO2 film (~50 nm thick) produced near-infinite
series resistance resulting in almost 0 % efficinecy. These results speak to the extreme
conformality of the ALD TiO2 films, for the following reason. The spray-pyrolysis
method has been optimized at 50 nm, likely as that is the thickness required to ensure
there are no shunts in the compact TiO2 layer, i.e. that at the thinnest points of the film
the TiO2 still coats the FTO. We conclude that current is collected through the spray-
pyrolysis TiO2 films via the thinner regions while the 50 nm thick regions are insulating
and prevent charge collection. On the other hand, the 1700 cycle ALD TiO2 film is
uniformly 50 nm thick and insulating throughout, causing the high series resistance in the
J-V curve, and preventing any current collection.
Katherine Roelofs Ph.D. Dissertation
197
Figure 7-4. Performance of nanoporous perovskite solar cells with 50 nm thick compact
TiO2 layers deposited by spray-pyrolysis, compared to those made with varied
thicknesses of ALD TiO2 from 100 ALD cycles (~3 nm thick) to 1700 ALD cycles (~50
nm thick). Shown are the J-V performance under (a) 1 sun illumination and (b) dark
conditions, with an inset enlarging the low-current region.
The dark J-V curves for the nanoporous perovskite solar cells are shown in Figure
7-4b, with an enlarging inset. With increasing ALD cycles, the dark current is further
suppressed, pushing the onset to higher voltages. This suppression of the dark current is
also reflected in the high VOC values observed initially with increased ALD cycles in the
illuminated J-V curves, before the high series resistance drives the efficiency to zero.
Interestingly, a closer look at the dark current curves (inset) shows that the ALD TiO2
films in the range of thicknesses tested are better than spray-pyrolysis TiO2 films at
suppressing the dark current for voltages below the onset voltage. These results support
the conclusion that the ALD layers are highly insulating, and that ultra-thin ALD TiO2
layers can perform as efficiently as the standard spray-pyrolysis TiO2 layers.
Katherine Roelofs Ph.D. Dissertation
198
Table 7-1. Series resistance of the different compact TiO2 layers, measured by
impedance spectroscopy of full perovskite solar cells under dark conditions. These
values are the average series resistance across the different voltage setpoints.
Series
Resistance (Ω)
spray pyrolysis 9.7
200 cycles 10.2
800 cycles 11.1
1000 cycles 11.3
Impedance spectroscopy measurements were collected of completed devices to
analyze the impact of the compact TiO2 layer on device performance. Of particular
interest is the series resistance of the various TiO2 films. Listed in Table 1 are the overall
series resistance of the perovskite solar cells with 50 nm thick spray-pyrolysis TiO2
compared to varied thicknesses of ALD TiO2. As shown in Figure 7-5, the series
resistance values do not change with the voltage setpoint during the impedance
spectroscopy measurements. The contribution of the compact TiO2 layers to the series
resistance of the overall cell cannot be separated out, but the relative variations in the
overall series resistance are attributed to the compact TiO2 layers. We found the spray
pyrolysis TiO2 layers to have the lowest series resistance values, suggesting that their
thinnest regions are likely thinner than that of the 200 cyc (~6 nm) ALD TiO2 layers.
Increasing the thickness of the ALD TiO2 leads to slight increases in series resistance.
These values for series resistance are in keeping with those from previous studies of
perovskite solar cells.56, 57
The full impedance spectroscopy plots for a voltage setpoint
of 0.7 V are shown in Figure 7-6.
Katherine Roelofs Ph.D. Dissertation
199
Figure 7-5. Series resistance of compact TiO2 layers, measured by impedance
spectroscopy of perovskite solar cells under dark conditions at different voltage setpoints.
Figure 7-6. Impedance spectroscopy measurements of full perovskite solar cells with
varied compact TiO2 layers taken at a voltage setpoint of 0.7 V in dark conditions. Shown
are (a) an enlarged region near the origin, and (b) the full results. Traces for multiple
devices are shown for each condition, to give an idea of the variability of the
measurement.
Katherine Roelofs Ph.D. Dissertation
200
7.4. Wide Band Gap Perovskite Solar Cells
We have also explored capping ALD layers in wide band gap perovskite solar
cells for application as a high-voltage transparent top cell in a tandem device. This work
was completed with Dowon Bae, from the Center for Individual Nanoparticle
Functionality (CINF), Department of Physics, Technical University of Denmark, and
Axel Palmstrom, Department of Chemical Engineering, Stanford University.
Wide band gap perovskite can be synthesized by introducing Br as the halide ion
in the perovskite crystal structure, creating CH3NH3PbIxBr(3-x).58, 59
The ratio between
iodide and bromide ions controls the band gap of the perovskite, with more bromide
leading to wider band gaps. Figure 7-7 is a picture of the standard perovskite film next to
the wide band gap perovskite. Perovskite solar cells were fabricated in both the
conventional and inverted design, as shown schematically in Figure 7-8.
Figure 7-7. CH3NH3PbI3 perovskite films (left) along with wide band gap
CH3NH3PbIxBr(3-x) perovskite films (right).
Katherine Roelofs Ph.D. Dissertation
201
Figure 7-8. (a) Conventional n-i-p perovskite solar cell stack, along with (b) the inverted
p-i-n perovskite solar cell stack used in this work.
For the conventional n-i-p design, a few key features were determined essential.
We observed that for the planar structure, TiCl4 treatment of the ALD TiO2 was critical
for producing a high VOC. Shown in Figure 7-9 are the highest-efficiency J-V curves
collected from devices without the TiCl4 treatment, as compared to those fabricated with
the TiCl4 treatment. Average device parameters from the batch of 6 devices without and
8 devices with TiCl4 treatment are shown in Figure 7-10. We conclude that TiCl4 gives
improved photovoltage and photocurrent by the diffusion of Cl- ions into the perovskite
layer. It was also found that TiCl4 reduces the series resistance, perhaps due to increased
conductivity of the TiO2 layer allowing electrons to be collected at the FTO more easily,
but at the same time reduces the shunt resistance as the TiO2 layer is no longer as
insulating.
Katherine Roelofs Ph.D. Dissertation
202
Figure 7-9. Conventional n-i-p perovskite solar cells in the planar architecture, showing
(a) the J-V curve for the best solar cell without and (b) with TiCl4 treatment.
Figure 7-10. Statistics of devices fabricated without TiCl4 treatment as compared to
those fabricated with TiCl4, showing parameters determined from the J-V curves under 1
sun illumination, including (a) the VOC, (b) JSC, (c) RShunt, and (d) RSeries.
Katherine Roelofs Ph.D. Dissertation
203
Figure 7-11. Optimization of standard n-i-p structure with wide band gap perovskite
solar cells fabricated by two-step deposition process. Best efficiency solar cells, with
each of the following modifications (a) two times the concentration of spiro-OMeTAD,
with also (b) two times the concentration of PbI2, and the further addition of (c) a
mesoporous TiO2 layer.
Further experiments were conducted to optimize the conventional n-i-p structure
with wide band gap perovskite fabricated by the two-step deposition procedure.5 In the
two-step deposition procedure, as detailed more completely in Chapter 3, a layer of PbI2
is initially deposited by spin-coating a PbI2 solution in DMF. Then, the PbI2 layer is
soaked in a solution of methylammonium iodide in IPA. As shown in Figure 7-11, it was
found that doubling the concentration of the spiro-OMeTAD solution in chlorobenzene,
bringing the concentration of spiro-OMeTAD to 450 mg/mL and the Li-TFSI up to 0.016
mg/mL, almost doubled the device efficiency as compared to the device in Figure 7-9b.
This could in part be due to a thicker deposited layer of the spiro-OMeTAD more fully
coating the perovskite, though the original cell did not appear to have problems with
shorting. Another possible explanation is a higher conductivity of spiro-OMeTAD layer,
as literature reports have found increased conductivity at higher spiro-OMeTAD
concentration60
or higher LiTFSI dopant concentrations.61
Spin-coating the PbI2 solution
down a second time atop the initial PbI2 spin-coated layer was also found to improve
efficiencies to 1.86% and VOC to 1.0 V, largely through a dramatically improved fill
Katherine Roelofs Ph.D. Dissertation
204
factor. Though a loss in JSC occurred, the higher VOC is critical for applications of the
wide band gap perovskite. Finally, the best planar device resulting from these changes
(TiCl4 treatment, doubled spiro-OMeTAD-Li concentration, 2x PbI2 coatings) were
compared to a device with a nanoporous TiO2 layer using the same fabrication recipe –
the nanoporous device reached an efficiency of 3.72% and a VOC of 1.1 V.
Figure 7-12. Scanning electron microscopy (SEM) images of perovskite (PSK) films
fabricated by conversion of PbI2 layers deposited by different spin speeds. The film
morphology of (a) spin-coating the PbI2 at 2000 rpm followed by a 6000 rpm deposition
of PbI2, while (b) shows spin-coating of 6000 rpm PbI2 followed by a 2000 rpm
deposition where the lower layer was found to be PbI2-rich.
Katherine Roelofs Ph.D. Dissertation
205
We further explored the effect of spin-coating speeds of the two PbI2 spin-coating
steps of this modified version of the two-step deposition process. We found that slower
spin speeds (2000 rpm) produced a lower-density, larger-crystal PbI2 layer, while faster
spin speeds (6000 rpm) produced a higher-density, smaller-crystal PbI2 layer. These spin
speeds were tried in different combinations by the two spin-coat depositions of PbI2.
Devices fabricated by a 6000 rpm followed by 2000 rpm deposition of PbI2 performed
much more poorly than a 2000 rpm followed by a 6000 rpm deposition. Scanning
electron microscopy (SEM) images of film morphologies of the converted perovskite
from these films are shown in Figure 7-12. The proposed mechanism is shown in Figure
7-13. Here, the larger perovskite crystals formed at 2000 rpm lead to a lower-density
perovskite film with the second deposition at 6000 rpm, allowing the CH3NH3Br to
effectively permeate the film and fully convert to the perovskite phase (Figure 7-13a).
On the other hand, the initial 6000 rpm deposition on the substrate will produce a highly-
compact lead iodide layer, allowing the subsequent 2000 rpm deposition to form a large-
crystal PbI2 film that is also very dense (Figure 7-13b). Upon the drop-casting of
CH3NH3Br, only the top layer of this film converts to the perovskite phase, leaving the
bottom dense PbI2 layer not fully converted.
Figure 7-13. Schematic of proposed mechanism for the effect of spin-coating speed of
PbI2 layers, showing (a) 2000 rpm followed by 6000 rpm producing a fully converted
perovskite film, while (b) 6000 rpm followed by 2000 rpm only converts the top layer of
the PbI2 to perovskite. (c) J-V curves of the completed devices.
Katherine Roelofs Ph.D. Dissertation
206
7.5. ALD Encapsulation Layers
An inverted architecture was fabricated to test the use of ALD layers to achieve
high VOC’s in wide band gap perovskite solar cells. A NiOx hole transport layer was
deposited by ALD on In:SnO2 (ITO) substrates, following the NiOx ALD process
developed our previous work using nickelocene and ozone.3 The deposition temperature
was set to 180 °C to protect the ITO layer, though it was found the NiOx-coated ITO
required a post-deposition annealing step at 300 °C in air to produce working cells.
Presumably the annealing step increases the conductivity of the NiOx layer by altering the
Ni-O ratio producing an intrinsically-doped film, without degrading the ITO too much. It
is also possible that indium from the underlying ITO is diffusing into the NiOx.
For the electron-transport layer, the fullerene-derivative PCBM ([6,6]-phenyl-C61-
butyric acid methyl ester) was used. A small amount (2.4 wt%) of polystyrene (PS) was
added to the PCBM in chlorobenzene mixture to improve the conformality of the
PCBM:PS layer to achieve good coating of the perovskite. ALD TiO2 layers of varied
thickness were deposited atop the PCBM:PS layer, showing slight improvements in
device performance, particularly in the VOC. The device schematic and J-V curves are
shown in Figure 7-14; note that the J-V curves for these initial optimization studies are all
collected at a rate of 0.5 s per 0.1 V step, whereas a rate of 5 s per 0.1 V step is required
to observed the settled, non-hysteretic device behavior for high-efficiency perovskite
solar cells.1 This result represents the first incorporation of ALD layers into a perovskite
solar cell leading to improved initial J-V performance; previous reports of Al2O3 ALD
layers in perovskite solar cells showed enhanced stability but lower initial performance.62
The TiO2-capped perovskite solar cell could be used in photoelectrocatalysis of the
hydrogen evolution reaction, for instance with Pt particles atop the TiO2 layer to catalyze
the reaction.
Katherine Roelofs Ph.D. Dissertation
207
Figure 7-14. Initial results showing ALD TiO2 capping layers in the inverted perovskite
solar cell. (a) Structure of the inverted device, (b) schematic of the energy levels in the
device, not to scale, (c) J-V curves collected under illumination collected at 0.5 s delay
between each 0.1 V step, and (d) performance metrics of the devices.
The inverted wide band gap perovskite solar cell performance was further
optimized as shown in Figure 7-15. Performance was improved by lowering the
polystyrene weight ratio to 1.7% in the PCBM:PS layer (Figure 7-15a), while keeping the
TiO2 thickness of 3 nm. We conclude that here, as has been previously reported,63
too-
high concentrations of polystyrene increase the series resistance and decrease the electron
mobility of the PCBM:PS layer. These devices still had a 10 nm thick NiOx layer. We
found, however that a 5 nm NiOx layer gave better performance, as seen in the red trace
in Figure 7-15b which was fabricated with the new optimized PCBM:PS weight ratio of
1.7% and the optimized thickness of 3 nm ALD TiO2. Also shown in Figure 7-15b is an
identical device fabricated with 6 nm ALD TiO2 capping layer, to check if greater
thicknesses of the TiO2 layer would be more effective. Through these series of changes,
we were able to produce devices with 1.2% efficiencies and VOC’s of 1.06 V (red trace,
Figure 7-15b).
Katherine Roelofs Ph.D. Dissertation
208
Figure 7-15. Optimization of the inverted wide band gap perovskite solar cells. (a) J-V
performance of devices with 10 nm thick NiOx layers, 3 nm thick TiO2 capping layers,
and varied PCBM:PS weight ratios of polystyrene. (b) J-V performance of devices with
5 nm thick NiOx layers, the optimized PCBM:PS ratio of 1.7 wt% polystyrene, and varied
TiO2 ALD capping layer thicknesses.
These results are very promising in producing a perovskite solar cell capped with
an inorganic layer, to protect the perovskite against exposure to aqueous solutions for use
in photoelectrocatalysis of the hydrogen evolution reaction, for example if Pt particles are
deposited on the TiO2 surface. This is shown schematically in a tandem design in Figure
7-16.
Katherine Roelofs Ph.D. Dissertation
209
Figure 7-16. Schematic of tandem design where an inverted wide band gap perovskite
solar cell can be paired with a Si solar cell to perform photocatalysis.
7.6. Conclusions
In conclusion, compact TiO2 layers deposited by ALD were incorporated into
nanoporous perovskite solar cells. Unexpectedly, ALD growth of thin TiO2 films on
FTO substrates formed the rutile TiO2 phase, instead of the more common anatase phase.
The formation of rutile TiO2 is attributed to texturing by the underlying FTO, in which
the fluorine-doped SnO2 is in the rutile phase. In contreast, ALD TiO2 deposited on Si
wafers formed the anatase TiO2 phase. The 50 nm thick spray-pyrolysis TiO2 deposited
on FTO was amorphous. It was observed that ALD TiO2 with a 100 °C deposition
temperature was amorphous regardless of thickness, while ALD TiO2 at a 300 °C
deposition temperature formed crystalline rutile TiO2 at thicknesses above ~6 nm. Using
Katherine Roelofs Ph.D. Dissertation
210
an in-situ annealing chamber, we observed that even for TiO2 layers deposited initially in
the amorphous phase at 100 °C, annealing at 500 °C lead to the formation of rutile TiO2
films. This demonstrates the ability to tune the crystallinity of the TiO2 layer, which
could be of use in future studies for optimization of perovskite solar cells, or further
exploration of the mechanisms of TiO2-initiated degradation of the perovskite layer by
the photocatalytic properties of TiO2.
For the wide band gap perovskite solar cells, we were able to optimize both the
conventional and inverted designs to produce high VOC solar cells. The champion device
had a VOC of 1.10 V in the standard design and 1.08 V in the inverted design. More work
is required to fully test these devices, particularly to determine the scan rate required to
report the settled device performance for these lower-efficiency wide band gap perovskite
solar cells. We are also highly interested in the stability behavior of these ALD-capped
devices in atmospheric conditions, and when exposed to aqueous solutions. If the solar
cell surface is indeed successfully stabilized to aqueous solutions, they could be applied
to photoelectrocatalysis of hydrogen evolution or other reactions.
The ultimate goal of these studies, which remains for future work, is the
development of ALD transport layers that can be deposited directly atop the perovskite to
work as both a transport layer and an encapsulation layer. We are particularly excited for
further development of the NiOx process in this manner. The use of ozone as the counter-
reactant in the current process prevents its employment as the top-coating, since the
ozone converts the perovskite back to the PbI2 phase. Nickel oxide ALD processes with
water as the counter-reactant have been demonstrated using the Ni(dmamp)2 precursor,
and it is of interest to see if these can be successfully applied to coat the perovskite.64
7.7. Financial Support, Collaborations, and
Copyrights
This work has not yet been submitted for publication. These studies were
completed with help from several collaborators. Section 7.3 was completed through
Katherine Roelofs Ph.D. Dissertation
211
collaboration with Vanessa Poole and Michael Toney, who helped with the analysis of
the GIXRD data; in particular Vanessa Poole was instrumental in collecting the GIXRD
data and helped build the in-situ annealing chamber. Dara Bobb-Semple, Axel
Palmstrom, and Pan Chuen also contributed to the work in Section 7.3. Sections 7.4 and
7.5 were driven by a collaboration with Dowon Bae, who fabricated all the wide band-
gap perovskite films, and Axel Palmstrom, who conducted materials characterization
such as the SEM analysis. This work was supported by the US Department of Energy
through the Bay Area Photovoltaic Consortium under Award No. DE-EE0004946.
7.8. References
1. Unger, E. L.; Hoke, E. T.; Bailie, C. D.; Nguyen, W. H.; Bowring, A. R.;
Heumuller, T.; Christoforo, M. G.; McGehee, M. D., Energy & Environmental Science
2014, 7 (11), 3690-3698.
2. Lee, H.-B.-R.; Bent, S. F., Chemistry of Materials 2012, 24 (2), 279-286.
3. Nardi, K. L.; Yang, N.; Dickens, C. F.; Strickler, A. L.; Bent, S. F., Advanced
Energy Materials 2015, n/a-n/a.
4. Burschka, J.; Pellet, N.; Moon, S.-J.; Humphry-Baker, R.; Gao, P.; Nazeeruddin,
M. K.; Grätzel, M., Nature 2013, 499 (7458), 316-319.
5. Bi, D.; El-Zohry, A. M.; Hagfeldt, A.; Boschloo, G., ACS Applied Materials &
Interfaces 2014, 6 (21), 18751-18757.
6. Leijtens, T.; Ding, I. K.; Giovenzana, T.; Bloking, J. T.; McGehee, M. D.;
Sellinger, A., ACS Nano 2012, 6 (2), 1455-1462.
7. Bai, Y.; Yu, H.; Zhu, Z.; Jiang, K.; Zhang, T.; Zhao, N.; Yang, S.; Yan, H.,
Journal of Materials Chemistry A 2014, 3 (17), 9098-9102.
8. National Center for Photovoltaics, N. R. E. L. Best Research-Cell Efficiencies.
http://www.nrel.gov/ncpv/.
9. Kojima, A.; Teshima, K.; Shirai, Y.; Miyasaka, T., Journal of the American
Chemical Society 2009, 131 (17), 6050-6051.
10. Im, J.-H.; Lee, C.-R.; Lee, J.-W.; Park, S.-W.; Park, N.-G., Nanoscale 2011, 3
(10), 4088-4093.
11. Kim, H.-S.; Lee, C.-R.; Im, J.-H.; Lee, K.-B.; Moehl, T.; Marchioro, A.; Moon,
S.-J.; Humphry-Baker, R.; Yum, J.-H.; Moser, J. E.; Gratzel, M.; Park, N.-G., Scientific
Reports 2012, 2, 591.
12. Lee, M. M.; Teuscher, J.; Miyasaka, T.; Murakami, T. N.; Snaith, H. J., Science
2012, 338 (6107), 643-647.
13. Green, M. A.; Ho-Baillie, A.; Snaith, H. J., Nat Photon 2014, 8 (7), 506-514.
14. Yin, W.-J.; Yang, J.-H.; Kang, J.; Yan, Y.; Wei, S.-H., Journal of Materials
Chemistry A 2015, 3 (17), 8926-8942.
15. Xing, G.; Mathews, N.; Sun, S.; Lim, S. S.; Lam, Y. M.; Grätzel, M.; Mhaisalkar,
S.; Sum, T. C., Science 2013, 342 (6156), 344-347.
Katherine Roelofs Ph.D. Dissertation
212
16. Edri, E.; Kirmayer, S.; Mukhopadhyay, S.; Gartsman, K.; Hodes, G.; Cahen, D.,
Nat Commun 2013, 5.
17. Snaith, H. J.; Abate, A.; Ball, J. M.; Eperon, G. E.; Leijtens, T.; Noel, N. K.;
Stranks, S. D.; Wang, J. T.-W.; Wojciechowski, K.; Zhang, W., The Journal of Physical
Chemistry Letters 2014, 5 (9), 1511-1515.
18. Tress, W.; Marinova, N.; Moehl, T.; Zakeeruddin, S. M.; Nazeeruddin, M. K.;
Grätzel, M., Energy & Environmental Science 2015, 8 (3), 995-1004.
19. Niu, G.; Guo, X.; Wang, L., Journal of Materials Chemistry A 2015, 3 (17), 8970-
8980.
20. Gao, P.; Grätzel, M.; Nazeeruddin, M. K., Energy & Environmental Science 2014,
7 (8), 2448-2463.
21. Fakharuddin, A.; Di Giacomo, F.; Ahmed, I.; Wali, Q.; Brown, T. M.; Jose, R.,
Journal of Power Sources 2015, 283 (0), 61-67.
22. Zhang, Y.; Liu, M.; Eperon, G. E.; Leijtens, T. C.; McMeekin, D.; Saliba, M.;
Zhang, W.; de Bastiani, M.; Petrozza, A.; Herz, L. M.; Johnston, M. B.; Lin, H.; Snaith,
H. J., Materials Horizons 2015, 2 (3), 315-322.
23. Eperon, G. E.; Burlakov, V. M.; Docampo, P.; Goriely, A.; Snaith, H. J.,
Advanced Functional Materials 2014, 24 (1), 151-157.
24. Oja Acik, I.; Junolainen, A.; Mikli, V.; Danilson, M.; Krunks, M., Applied
Surface Science 2009, 256 (5), 1391-1394.
25. Kavan, L.; Grätzel, M., Electrochimica Acta 1995, 40 (5), 643-652.
26. Peng, B.; Jungmann, G.; Jäger, C.; Haarer, D.; Schmidt, H.-W.; Thelakkat, M.,
Coordination Chemistry Reviews 2004, 248 (13-14), 1479-1489.
27. Unger, E. L.; Spadavecchia, F.; Nonomura, K.; Palmgren, P. l.; Cappelletti, G.;
Hagfeldt, A.; Johansson, E. M. J.; Boschloo, G., ACS Applied Materials & Interfaces
2012, 4 (11), 5997-6004.
28. Kavan, L.; Tétreault, N.; Moehl, T.; Grätzel, M., The Journal of Physical
Chemistry C 2014, 118 (30), 16408-16418.
29. Cameron, P. J.; Peter, L. M., The Journal of Physical Chemistry B 2003, 107 (51),
14394-14400.
30. Ke, W.; Fang, G.; Wang, J.; Qin, P.; Tao, H.; Lei, H.; Liu, Q.; Dai, X.; Zhao, X.,
ACS Applied Materials & Interfaces 2014, 6 (18), 15959-15965.
31. Sudhagar, P.; Asokan, K.; Jung, J.; Lee, Y.-G.; Park, S.; Kang, Y., Nanoscale
Research Letters C7 - 30 2011, 6 (1), 1-7.
32. Thelakkat, M.; Schmitz, C.; Schmidt, H. W., Advanced Materials 2002, 14 (8),
577-581.
33. Xia, J.; Masaki, N.; Jiang, K.; Yanagida, S., The Journal of Physical Chemistry B
2006, 110 (50), 25222-25228.
34. Kavan, L.; O'Regan, B.; Kay, A.; Grätzel, M., Journal of Electroanalytical
Chemistry 1993, 346 (1-2), 291-307.
35. Chandiran, A. K.; Yella, A.; Stefik, M.; Heiniger, L.-P.; Comte, P.; Nazeeruddin,
M. K.; Grätzel, M., ACS Applied Materials & Interfaces 2013, 5 (8), 3487-3493.
36. Burke, A.; Ito, S.; Snaith, H.; Bach, U.; Kwiatkowski, J.; Grätzel, M., Nano
Letters 2008, 8 (4), 977-981.
37. Park, N. G.; van de Lagemaat, J.; Frank, A. J., The Journal of Physical Chemistry
B 2000, 104 (38), 8989-8994.
Katherine Roelofs Ph.D. Dissertation
213
38. Scanlon, D. O.; Dunnill, C. W.; Buckeridge, J.; Shevlin, S. A.; Logsdail, A. J.;
Woodley, S. M.; Catlow, C. R. A.; Powell, M. J.; Palgrave, R. G.; Parkin, I. P.; Watson,
G. W.; Keal, T. W.; Sherwood, P.; Walsh, A.; Sokol, A. A., Nat Mater 2013, 12 (9), 798-
801.
39. Wang, X.; Fang, Y.; He, L.; Wang, Q.; Wu, T., 2014, 27, 569-576.
40. Wu, Y.; Yang, X.; Chen, H.; Zhang, K.; Qin, C.; Liu, J.; Peng, W.; Islam, A.; Bi,
E.; Ye, F.; Yin, M.; Peng, Z.; Han, L., Applied Physics Express 2014, 7 (5), 052301.
41. Yu, H.; Zhang, S.; Zhao, H.; Will, G.; Liu, P., Electrochimica Acta 2009, 54 (4),
1319-1324.
42. Chen, L.-C.; Chen, C.-C.; Tseng, B.-S., Journal of Nanomaterials 2010, 2010,
374052.
43. Kim, H.-J.; Jeon, J.-D.; Kim, D. Y.; Lee, J.-J.; Kwak, S.-Y., Journal of Industrial
and Engineering Chemistry 2012, 18 (5), 1807-1812.
44. Abdullah, M. H.; Rusop, M., Ceramics International 2014, 40 (1, Part A), 967-
974.
45. Chen, H.; Wei, Z.; Yan, K.; Yi, Y.; Wang, J.; Yang, S., Faraday Discussions
2014, 176 (0), 271-286.
46. Yella, A.; Heiniger, L.-P.; Gao, P.; Nazeeruddin, M. K.; Grätzel, M., Nano Letters
2014, 14 (5), 2591-2596.
47. Jung, H. S.; Park, N.-G., Small 2014, 11 (1), 10-25.
48. Brennan, T. P.; Bakke, J. R.; Ding, I. K.; Hardin, B. E.; Nguyen, W. H.; Mondal,
R.; Bailie, C. D.; Margulis, G. Y.; Hoke, E. T.; Sellinger, A.; McGehee, M. D.; Bent, S.
F., Physical Chemistry Chemical Physics 2012, 14 (35), 12130-12140.
49. Wanga, H.; Xub, S.; Gordonb, R. G., Electrochemical and Solid-State Letters
2010, 13 (9), G75.
50. Kumar, A.; Madaria, A. R.; Zhou, C., The Journal of Physical Chemistry C 2010,
114 (17), 7787-7792.
51. Liu, B.; Aydil, E. S., Journal of the American Chemical Society 2009, 131 (11),
3985-3990.
52. Feng, X.; Zhu, K.; Frank, A. J.; Grimes, C. A.; Mallouk, T. E., Angewandte
Chemie International Edition 2012, 51 (11), 2727-2730.
53. Wang, H.; Bai, Y.; Wu, Q.; Zhou, W.; Zhang, H.; Li, J.; Guo, L., Physical
Chemistry Chemical Physics 2011, 13 (15), 7008-7013.
54. Ye, M.; Liu, H.-Y.; Lin, C.; Lin, Z., Small 2012, 9 (2), 312-321.
55. Ferguson, J. D.; Yoder, A. R.; Weimer, A. W.; George, S. M., Applied Surface
Science 2004, 226 (4), 393-404.
56. Juarez-Perez, E. J. Wuβler, M. Fabregat-Santiago, F.; Lakus-Wollny, K.;
Mankel, E.; Mayer, T.; Jaegermann, W.; Mora-Sero, I., The Journal of Physical
Chemistry Letters 2014, 5 (4), 680-685.
57. Kim, H.-S.; Mora-Sero, I.; Gonzalez-Pedro, V.; Fabregat-Santiago, F.; Juarez-
Perez, E. J.; Park, N.-G.; Bisquert, J., Nat Commun 2013, 4.
58. Singh, S. P.; Nagarjuna, P., Dalton Transactions 2014, 43 (14), 5247-5251.
59. Hoke, E. T.; Slotcavage, D. J.; Dohner, E. R.; Bowring, A. R.; Karunadasa, H. I.;
McGehee, M. D., Chemical Science 2015, 6 (1), 613-617.
Katherine Roelofs Ph.D. Dissertation
214
60. Abate, A.; Leijtens, T.; Pathak, S.; Teuscher, J.; Avolio, R.; Errico, M. E.;
Kirkpatrik, J.; Ball, J. M.; Docampo, P.; McPherson, I.; Snaith, H. J., Physical Chemistry
Chemical Physics 2013, 15 (7), 2572-2579.
61. Nguyen, W. H.; Bailie, C. D.; Unger, E. L.; McGehee, M. D., Journal of the
American Chemical Society 2014, 136 (31), 10996-11001.
62. Dong, X.; Fang, X.; Lv, M.; Lin, B.; Zhang, S.; Ding, J.; Yuan, N., Journal of
Materials Chemistry A 2014, 3 (10), 5360-5367.
63. Bai, Y.; Yu, H.; Zhu, Z.; Jiang, K.; Zhang, T.; Zhao, N.; Yang, S.; Yan, H.,
Journal of Materials Chemistry A 2015, 3 (17), 9098-9102.
64. Chung, T.-M.; Lee, S. S.; Cho, W.; Kim, M.; Lee, Y. K.; Hwang, J.-H.; An, K.-S.;
Kim, C. G., Bulletin of the Korean Chemical Society 2011, 32 (3), 783.
Katherine Roelofs Ph.D. Dissertation
215
Chapter 8. Conclusion and Outlook
8.1. Summary of Work and Future Studies
This work focused on interface engineering in thin film PV technologies inspired
by the dye-sensitized solar cell (DSSC) design. As presented in Chapter 2, inorganic-
absorber nanostructured solar cells are receiving growing attention, and they are part of a
broader effort to move towards an all-solid-state all-inorganic version of the DSSC.1
This thesis presented a series of studies on engineering the internal interfaces in such
nanostructured solar cells, specifically quantum-dot-sensitized solar cells (QDSSCs) and
perovskite solar cells (PSCs). The experimental methods of device fabrication, interface
modification techniques, and materials and electronic characterization for these studies
were detailed in Chapter 3.
Atomic layer deposition (ALD) is one technique used for interface engineering,
through surface modification, e.g. coating with ultra-thin (<5 nm) passivation or barrier
layers, or through the deposition of thin films (<100 nm) as charge transport layers. A
drawback of ALD in the context of solution-deposited, large-scale solar cells is that ALD
is a vacuum deposition method. This does not prohibit the use of ALD in high-volume
manufacturing, in particular high-margin products in the semiconductor industry where
ALD can be used to achieve further reductions in feature size. We further note that the
use of vacuum is not an intrinsic part of the ALD mechanism and researchers are
currently seeking means to make atmospheric-pressure ALD feasible.2, 3
The high
control and reproducibility of ALD made it ideal for the fundamental studies in this
thesis, and we have used it to gain fundamental knowledge about the working of
QDSSCs and PSCs, and to also learn about the ‘ideal’ device architecture. The A D-
modified devices in this work have demonstrated potential gains in efficiency, helped
identify critical properties of interfacial modification, and motivated future work on low-
cost, solution-processable methods to achieve similar results.
Katherine Roelofs Ph.D. Dissertation
216
Chapters 4 – 6 presented interfacial modification of QDSSCs. QDSSCs face
significant challenges, as record efficiencies in solid-state devices are only 1.5%,4 with
the record device efficiency using a liquid electrolyte currently at 8.6%.5 We believe that
improving the efficiency of solid-state QDSSCs is critical for the commercialization of
these devices, and that the challenges facing solid-state QDSSCs center around
optimizing the device interface to minimize recombination.
As such, in Chapter 4, we presented the use of ultra-thin, insulating ALD Al2O3
barrier layers which succeeded in achieving an order of magnitude longer recombination
lifetimes in solid-state CdS QDSSCs.6 We showed that the Al2O3 barrier layers act as a
tunneling barrier, perhaps in addition to imparting a passivation effect, as continued
improvements in lifetime are seen past the first Al2O3 ALD cycle. However, the decrease
in JSC past the first ALD cycle – for barrier layers deposited either before or after QD
deposition – indicated that the Al2O3 layers also limit the charge transfer processes
necessary for photocurrent collection. In addition, we found that thickness control of the
barrier layer is critical. This motivates future work on slower-growing barrier materials,
to give a wider parameter space to work with. Another promising direction for future
work is the deposition of barrier layers selectively on the bare TiO2 (so photocurrent
collection through the QD is not blocked). This could be achieved by depositing the QDs
first, and then developing an ALD process for an insulating material that deposits
selectively on the TiO2. An alternative approach to achieve selective ALD is applying a
self-assembled monolayer (SAM) of organic molecules with attachment groups specific
to the sulfide surface of the QD which block ALD film growth on the QD. Such SAM
layers could potentially then be removed by chemical or heat treatments, leaving the
surface of the QD exposed again.7
However, the limitations found in this initial work on Al2O3 barrier layers
motivated a return to the fundamental problem: too-high rates of recombination at the
interface of QDSSCs. DSSCs do not suffer from such high rates of recombination, in
part because the dye can achieve near-complete (over 80%) surface coverage of the
nanostructured TiO2,8 while QDs struggle to reach coverages of 15%.
9 For this reason,
we pursued a strategy of increasing QD loading to decrease interfacial recombination, as
Katherine Roelofs Ph.D. Dissertation
217
reported in Chapter 5.10
At this point, our system of interest shifted from CdS QDs (bulk
band gap 2.4 eV) to PbS QDs (bulk band gap 0.37 eV). For our preliminary results, the
high band gap of CdS QDs ensured a greater chance of collecting charges generated by
absorbed photons, allowing for ease of study, particularly under higher intensity
illumination. On the other hand, the band gap of the PbS QDs can be tuned to reach the
optimal value for high-efficiency single band gap solar cells, ~1.2 eV. Hence they are
arguably a more interesting semiconductor QD for solar cells.
In the study reported in Chapter 5, QD loading was increased by controlling the
TiO2 surface charge to assist Pb2+
ion deposition during QD growth. We found that QD
loading has a significant impact on performance, almost doubling the device efficiencies,
reaching 0.6%. This improvement was due to both increased light absorption and overall
reduced recombination. We observed two opposing effects with increased QD
deposition: higher QD surface coverage lengthened recombination lifetimes, but at higher
SILAR cycle numbers recombination increased. We attributed this effect to the presence
of larger QDs which remain in an oxidized state and incite additional recombination, due
to poor hole transfer from the QD to the hole-transport material. Notably, we found that
larger QDs with band gaps below ~2 eV did not contribute to photocurrent collection.
From this we concluded that an optimal device structure would have the nanostructured
TiO2 substrate completely coated with small QDs of a uniform size.
In Chapter 6 we presented the results of fundamental studies on the nucleation
and growth of PbS QDs on nanoporous TiO2, which was motivated by the need to further
increase initial QD nucleation to achieve complete QD coverage of the TiO2 surface.
Here, PbS QDs were grown by ALD, and using X-ray absorption spectroscopy we
studied PbS QD growth on anatase TiO2 nanoparticles versus amorphous TiO2 deposited
by ALD. We found that the anatase TiO2 surface can disrupt the rocksalt PbS crystal
structure.11
In addition, QDs grown by ALD were compared to those grown by the
solution-based process (successive ion layer adsorption and reaction, or SILAR) used in
previous chapters, showing comparable QD growth by each technique.
Finally, Chapter 7 described our work on incorporating ALD thin film layers in
perovskite solar cells. First, ALD is used to grow the electron-transport layer, TiO2,
Katherine Roelofs Ph.D. Dissertation
218
which coats the transparent conductive oxide, rutile F:SnO2 (FTO).12
Unexpectedly,
ALD-grown TiO2 formed rutile TiO2 phase when grown on FTO substrates, but formed
the more common anatase TiO2 phase when grown on Si substrates, suggesting a
texturing effect from the FTO. We found that perovskite solar cells with ultra-thin ALD
TiO2 layers (< 3 nm thick) could perform comparably to 50 nm of TiO2 deposited by the
standard technique, spray-pyrolysis. Encapsulation layers of TiO2 also showed promise
in an inverted design (p-i-n) to push wide band gap perovskite solar cells to the highest
VOC values for potential applications in photocatalytic water splitting. Of great interest
for future work is to evaluate if such inorganic encapsulation layers have any stabilizing
benefits, since initial results showed some protection against oxidizing conditions. Next,
an ALD process for NiOx was used to develop an inorganic hole-transport material
(HTM). A direction of future work is the deposition of ALD NiOx layers atop the
perovskite, which could act as both an inorganic HTM and as an encapsulating inorganic
layer to protect the perovskite from oxygen and moisture from the environment.
8.2. Industrial Applications
The industrial prospects of QDSSCs and PSCs are still being shaped by
fundamental research in the field. For wide band gap QDSSCs and PSCs, one can look to
the current use of DSSCs as translucent solar cells. Both QDSSCs and PSCs can be
made translucent, and by tuning the absorber band gap, their visible appearance can be
tailored through a range of colors. Translucency and color are appealing properties for
use in building materials, such as tinted glass that can provide power to off-set a
building’s energy use.13, 14
Although both of these modifications – translucency and non-
black coloration – will decrease device efficiencies, such variants are ideal for
applications where the aesthetic appeal is the primary concern. An additional benefit of
inorganic-absorber nanostructured solar cells is that they could be of use in reclaiming
building energy: these devices work relatively well under low-light, or glancing
illumination conditions,15
while Si and planar thin film solar cells require strong, direct
sunlight to reach their top power conversion efficiencies.
Katherine Roelofs Ph.D. Dissertation
219
Another promising direction is the possibility of using inorganic-absorber
nanostructured solar cells as the top layer in a tandem device with a commercial solar cell
as the bottom layer.16
Both the low-temperature processing of these devices, and in
particular the potential of perovskite solar cells for high VOC’s, motivates their use as the
top layer in a tandem device. A final direction to consider is flexible photovoltaics,
where the processing temperature needs to be limited to less than 150 °C to allow for the
use of plastic substrates.
Still, both technologies face significant challenges to realize their full potential.
Foremost are the stability issues, along with safety concerns. A major drawback of top-
performing QDSSCs is the use of liquid electrolytes, as well as toxic elements for the QD
absorber such as Pb or Cd. In addition, the QD absorber can oxidize or degrade, giving
current QDSSCs lifetimes on the order of days or weeks. Likewise, for perovskite solar
cells, the perovskite absorber can degrade under ambient conditions, according to:
CH3NH3PbI3 PbI2 (s) + CH3NH3I (aq). Degradation factors include oxygen, moisture,
UV light, temperature, and applied voltages;17, 18
these can lead to significant losses in
efficiency within days. Moreover, the perovskite is soluble in water, which is a safety
concern if the external encapsulation fails, due to the toxicity of lead. This contrasts with
CdTe solar cells, which have been successfully commercialized by First Solar, where the
stability of the CdTe keeps the Cd localized to the module even if the encapsulation fails.
We believe that proper encapsulation can overcome these toxicity concerns, and think
there is room for the benefits of ALD capping layers built into the device stack itself, in
addition to more traditional external plastic encapsulation layers. To successfully
commercialize these technologies, further fundamental research on improving device
stability is critical.
8.3. Future Directions and Outlook
For quantum-dot-sensitized solar cells, the overarching goal is the replacement of
the dye in DSSCs with an inorganic absorber – while still maintaining the benefits of
DSSCs. Such benefits include the use of solution-processable, low-purity, low-cost
materials, and the deposition of the absorber using low processing temperatures, opening
Katherine Roelofs Ph.D. Dissertation
220
up potential deposition on plastic substrates for flexible PV applications. We note that
while the nanoporous TiO2 substrate in DSSCs and QDSSCs does require a high-
temperature (~500 °C) processing step, there are low-temperature alternative processing
steps, and high-temperature processing is not an intrinsic requirement. This contrasts with
existing thin film PV technologies with inorganic absorbers, such as amorphous Si,
CdTe, and CIGS, which require high-purity crystalline absorbers fabricated via high
processing temperatures.19
The nanostructured interface in DSSCs and QDSSCs is
critical in allowing for the use of low-purity materials, as both electrons and holes are not
transported through a bulk absorber layer, but rather are quickly separated into an n-type
or p-type transport layer.
Perovskite solar cells fulfill the goal sought by QDSSCs, in creating a high-
efficiency inorganic-absorber alternative to DSSCs. However, while some perovskite
solar cells are fabricated with the nanoporous structure, the nanostructured interface is
not required as the perovskite absorber can conduct both electrons and holes. Indeed, for
a low-temperature solution-processed material, the lead-based perovskites have
remarkably long electron and hole diffusion lengths of over 100 nm.20
Thus, it is
possible that a solar cell with the DSSC benefits of solution-processable, low-temp, low-
purity materials can be achieved without the nanostructured interface, begging the
question of whether such structures are still desired. It would seem that if a
nanostructured interface can be avoided, it should be, as it adds unnecessary complexity
to the design. However, there are some initial studies in the literature suggesting that a
nanostructured or microstructured substrate might be necessary to stabilize the perovskite
crystals, particularly for solution-deposited perovskite layers.21, 22
To develop the best
planar perovskite solar cells, recent studies used higher-temperature (still <150 °C) vapor
deposition to deposit the perovskite,23, 24
moving away from the low-temp solution
processing and closer towards fabrication procedures used for CdTe and CIGS. Thus, it
may be that for low-purity perovskite layers, the nanostructured interface is still
beneficial, while for high-purity perovskite layers, the planar device design is optimal.
To successfully commercialize QDSSCs and PSCs, fundamental research on the
working mechanisms of these devices, as well as material engineering, is critical. We
Katherine Roelofs Ph.D. Dissertation
221
believe a focus on interface engineering will be rewarding, as charge separation and
transfer across interfaces is a critical component of device operation. Further, moving
these technologies towards all-inorganic, all-solid-state devices will have great appeal for
commercial applications and will have benefits in improved device stability and safety.
8.4. References
1. Roelofs, K. E.; Brennan, T. P.; Bent, S. F., The Journal of Physical Chemistry
Letters 2014, 5 (2), 348-360.
2. George, S. M., Chemical Reviews 2010, 110 (1), 111-131.
3. Munoz-Rojas, D.; MacManus-Driscoll, J., Materials Horizons 2014, 1 (3), 314-
320.
4. Lee, H.; Leventis, H. C.; Moon, S.-J.; Chen, P.; Ito, S.; Haque, S. A.; Torres, T.;
Nüesch, F.; Geiger, T.; Zakeeruddin, S. M.; Grätzel, M.; Nazeeruddin, M. K., Advanced
Functional Materials 2009, 19 (17), 2735-2742.
5. Zhao, K.; Pan, Z.; Mora-Seró, I.; Cánovas, E.; Wang, H.; Song, Y.; Gong, X.;
Wang, J.; Bonn, M.; Bisquert, J.; Zhong, X., Journal of the American Chemical Society
2015, 137 (16), 5602-5609.
6. Roelofs, K. E.; Brennan, T. P.; Dominguez, J. C.; Bailie, C. D.; Margulis, G. Y.;
Hoke, E. T.; McGehee, M. D.; Bent, S. F., The Journal of Physical Chemistry C 2013,
117 (11), 5584-5592.
7. Hashemi, F. S. M.; Prasittichai, C.; Bent, S. F., The Journal of Physical Chemistry
C 2014, 118 (20), 10957-10962.
8. Pellegrin, Y.; Le Pleux, L.; Blart, E.; Renaud, A.; Chavillon, B.; Szuwarski, N.;
Boujtita, M.; Cario, L.; Jobic, S.; Jacquemin, D.; Odobel, F., Journal of Photochemistry
and Photobiology A: Chemistry 2011, 219 (2-3), 235-242.
9. Guijarro, N. ana-Villarreal, T. Mora-Ser , I. n. Bisquert, J. G •mez, R., The
Journal of Physical Chemistry C 2009, 113 (10), 4208-4214.
10. Roelofs, K. E.; Herron, S. M.; Bent, S. F., ACS Nano 2015, submitted.
11. Trejo, O.; Roelofs, K. E.; Xu, J.; Logar, M.; Sarangi, R.; Norlund, D.; Dadlani,
A.; Kravec, R.; Dasgupta, N. P.; Bent, S. F.; Prinz, F. B., Submitted to Nano Letters 2015.
12. Roelofs, K. E.; Poole, V.; Toney, M.; Bent, S. F., In Preparation 2015.
13. Eperon, G. E.; Burlakov, V. M.; Goriely, A.; Snaith, H. J., ACS Nano 2014, 8 (1),
591-598.
14. Skandalos, N.; Karamanis, D., Renewable and Sustainable Energy Reviews 2015,
49 (0), 306-322.
15. Hagfeldt, A.; Boschloo, G.; Sun, L.; Kloo, L.; Pettersson, H., Chemical Reviews
2010, 110 (11), 6595-6663.
16. Bailie, C. D.; Christoforo, M. G.; Mailoa, J. P.; Bowring, A. R.; Unger, E. L.;
Nguyen, W. H.; Burschka, J.; Pellet, N.; Lee, J. Z.; Gratzel, M.; Noufi, R.; Buonassisi, T.;
Salleo, A.; McGehee, M. D., Energy & Environmental Science 2015, 8 (3), 956-963.
17. Niu, G.; Guo, X.; Wang, L., Journal of Materials Chemistry A 2015, 3 (17), 8970-
8980.
Katherine Roelofs Ph.D. Dissertation
222
18. Tress, W.; Marinova, N.; Moehl, T.; Zakeeruddin, S. M.; Nazeeruddin, M. K.;
Grätzel, M., Energy & Environmental Science 2015, 8 (3), 995-1004.
19. Jean, J.; Brown, P. R.; Jaffe, R. L.; Buonassisi, T.; Bulovic, V., Energy &
Environmental Science 2015, 8 (4), 1200-1219.
20. Xing, G.; Mathews, N.; Sun, S.; Lim, S. S.; Lam, Y. M.; Grätzel, M.; Mhaisalkar,
S.; Sum, T. C., Science 2013, 342 (6156), 344-347.
21. Fakharuddin, A.; Di Giacomo, F.; Ahmed, I.; Wali, Q.; Brown, T. M.; Jose, R.,
Journal of Power Sources 2015, 283 (0), 61-67.
22. Zhang, Y.; Liu, M.; Eperon, G. E.; Leijtens, T. C.; McMeekin, D.; Saliba, M.;
Zhang, W.; de Bastiani, M.; Petrozza, A.; Herz, L. M.; Johnston, M. B.; Lin, H.; Snaith,
H. J., Materials Horizons 2015, 2 (3), 315-322.
23. Liu, M.; Johnston, M. B.; Snaith, H. J., Nature 2013, 501 (7467), 3.
24. Leyden, M. R.; Ono, L. K.; Raga, S. R.; Kato, Y.; Wang, S.; Qi, Y., Journal of
Materials Chemistry A 2014, 2 (44), 18742-18745.
Katherine Roelofs Ph.D. Dissertation
223
Appendix A. Quantifying Geometric Strain at the ALD PbS QD/ TiO2 Interface
This Appendix contains the full details of the work presented in Chapter 5,
Section 4 and are taken from a manuscript submitted for publication by Trejo, Roelofs,
et. al.1
I. ALD PbS QD/ TiO2 Sample Preparation
Samples were fabricated to mimic the TiO2/QD interface in quantum-dot-
sensitized solar cells (QDSSCs), devices based on the Grätzel solar cell in which dye
molecules are replaced by QDs as the light-absorbing material. A commercial organic
paste with 20 nm diameter anatase TiO2 NPs (Dyesol, 18NR-T) was diluted with terpinol
in a 1:1 weight ratio, doctor-bladed onto glass slides and sintered at 450 ºC, resulting in a
nanoporous TiO2 film with a thickness of ∼2.2 μm. The substrates were then TiCl4-
treated, in keeping with common practice in the QDSSC and dye-sensitized solar cell
field. For this, the nanoporous TiO2 substrates were immersed overnight in a 0.02 M
aqueous TiCl4 solution and then annealed again at 450 ºC, resulting in substrates ready
for ALD. The substrates were loaded into a customized flow-type ALD system designed
for the deposition of PbS thin films. The substrate temperature was maintained at 160 ºC.
Bis(2,2,6,6-tetramethyl-3,5-heptanedionato)lead(II) (sublimated at 140 ºC) and
tetrakis(dimethylamido)titanium(IV) (heated at 70 ºC), distilled H2O, and a gas mixture
of 3.5% H2S in N2 were used as precursors to deposit PbS and TiO2. The ALD materials
were deposited sequentially and without breaking vacuum to prevent contamination or
surface restructuring due to air exposure.
During ALD deposition, all of the precursors were given a residence time in the
reaction chamber of 2 s to ensure their uniform infiltration of the nanoporous TiO2 film.
After each precursor pulse, 45 s were allotted for byproduct evacuation. The ALD
materials were deposited sequentially and without breaking vacuum to prevent
Katherine Roelofs Ph.D. Dissertation
224
contamination or surface restructuring due to air exposure. The reference ALD TiO2
films were about 10 nm thick (200 cycles) on a silicon wafer. To induce anatase
crystallinity, amorphous ALD TiO2 films were annealed at 400 ◦C under 8.1 x 10
-7 Torr
of pressure for 3 hr.
Figure A - 1a, b show schematics of the two sample sets. Figure A - 1c,d shows
UV-vis absorption spectra confirming size trends of the ALD-deposited QDs. For
improved air stability and passivation, the PbS QDs were subsequently encapsulated by
an additional 60 ALD cycles of amorphous TiO2.
Figure A - 1. Schematics of the sample architectures. The sample architectures for the
AN (a) and AM (b) sample sets are shown. For both sets, 10, 20, and 40 ALD cycles
(10x, 20x, and 40x) of PbS were deposited on the nanoporous TiO2 substrates. The
resulting PbS QDs were capped with 60 cycles of ALD TiO2, which is amorphous.
Absorption measurements capture changes in the PbS QDs average size. 40x PbS result
in larger QDs and have a smaller gap as seen by the absorptions edge shifts for the AN
Katherine Roelofs Ph.D. Dissertation
225
(c) and AM (d) sets. The differences in absorption between both sample sets are due to
distinct electronic structures, which arise from different atomic arrangements.
II. TEM Characterization
Samples for TEM characterization were prepared by scratching the substrate and
dispersing the residue in acetone. The solution was dropped casted on Formwar-coated
carbon TEM grids. TEM imaging and selected area electron diffraction (SAED) were
performed in a FEI Titan scanning transmission electron microscope (STEM) operated at
300 kV, equipped with a Cs image corrector operated at 300 kV. TEM images were taken
using an Ultrascan 1000 CCD camera at binning 2 (1024 × 1024 pixels) and an exposure
time of 0.5 s per image. The SAED patterns were recorded over the same area of AM and
AN sets of samples. The diffraction aperture of 100 nm diameter was used to obtain the
experimental diffraction rings. The chemical stability of ALD PbS on a TiO2 surface was
characterized with cross-sectional STEM and energy dispersive x-ray spectroscopy
(Figure A - 2).
Katherine Roelofs Ph.D. Dissertation
226
Figure A - 2. (a) HRTEM image of an ALD PbS QD encapsulated in TiO2 shows that
there is a clear interface at the PbS/TiO2 interface. (b) FFT of PbS QD shown in (a) fitted
by simulated diffraction patter for fcc crystal modification for [101] zone axis. (c, d) EDX
line scans were performed as shown in the schematic and corresponding DF-STEM
image. (e) EDX line scan over the cross-sectional area of PbS QD encapsulated in TiO2.
(d) EDX line scan over the sandwiched PbS layer in TiO2 as shown in the schematic and
corresponding DF-STEM image. The EDX analysis shows that ALD TiO2 does not
intermix with ALD PbS in either deposition sequence. This shows that TiO2 is an inert
capping layer for ALD PbS QDs.
Katherine Roelofs Ph.D. Dissertation
227
III. XANES Procedures and Interpretation
The samples were characterized at beamlines (BLs) 2-1, 4-3, and 10-1 at the
Stanford Synchrotron Radiation Lightsource (SSRL). For the XAS measurements,
multiple total fluorescence yield (TFY) spectra were gathered for S K- (BL 4-3), Pb L3-
(BL 4-3), O K- (BL 10-1), and Ti L2,3-edge (BL 10-1). Total electron yield (TEY)
spectra were also gathered for oxygen and agreed with the TFY data sets. The XAS
spectra were aligned, averaged, and normalized using the SixPACK and Athena analysis
packages. The XRD measurements were gathered at BL 2-1 using 12.5 keV x-rays and a
point detector. The sulfur K-edge and oxygen K-edge XAS spectra are shown in Chapter
5, section 4. The Pb L3-edge and Ti L2,3-edge are shown below in Figure A - 3 and
Figure A - 4 respectively.
Figure A - 3. (a) Pb L3-edge XANES for references and PbS QDs in the AN set with
three distinct peaks labeled. (b) Pb L3-edge XANES for the AM set. The references share
similar feature shapes and definition with the AM set PbS QDs. The features for the PbS
QDs in the AN set are more broad relative to the AM set and references. This is in line
with the trends observed in the S K-edge XANES.
Katherine Roelofs Ph.D. Dissertation
228
Figure A - 4. (a) Ti L2,L3-edge XANES for references and the AN set with five distinct
peaks labeled. (b) Ti L2,L3-edge XANES for the AM set. The anatase spectra have the
shoulder (feature 3) well defined, while broadening throughout the amorphous spectra
conceals features 3. Like in the O K-edge XANES, broadening of the features in the Ti
L2,L3-edge XANES attributed to the randomly-oriented orbitals of an amorphous TiO2.
The multiple-scattering simulations were done using the FEF9 software package.
As shown in the supplementary information, the S K-edge was for PbS converged by
varying the cluster, self-consistent field (SCF), and scattering sphere sizes. Geometric
parameters for the distorted PbS lattice structured were taken directly from a previous
study2 with no further modifications. The S K-edge of the modified PbS clusters were
calculated using the converged parameters of the rocksalt clusters. We investigated the
modified α-GeTe and TlI structure types, along with rocksalt structures of varying lattice
constants. The models assume that there is a local distortion similar to that of the
modified lattices but do not imply that the QDs of interest have such lattices throughout.
We demonstrate the nature of bonding distortions in PbS through multiple-
scattering simulations of PbS clusters. Theoretical XAS spectra are simulated using the
FEFF9 code,3, 4
based on Green’s function multiple-scattering theory. We first obtained
spectral agreement between simulated and experimental rocksalt PbS (Figure A - 5). As a
basis for simulating differences observed in the S K-edge XAS spectra of the AN
Katherine Roelofs Ph.D. Dissertation
229
samples, structural data from a comprehensive theoretical study on low-energy
distortions of rocksalt PbS were used.2
Figure A - 5. (a) Simulated S K-edge XANES spectra of PbS QDs with varying size. (b)
A cluster of 257 atoms, ~ 2.4 nm in diameter, was used to simulate the XANES curves
from the electronic structure and photoexcited electron scattering in a 1.2 nm and a 2 nm
QD. (c) A cluster of 1309 atoms, ~ 3.6 nm in diameter, was used to simulate the XANES
curves from the electronic structure and photoexcited electron scattering of 2 nm, 2.4 nm,
and 3.2 nm QDs. The cluster diameters were larger than the simulated QD diameter in
order to prevent edge effects in the scattering contributions to the spectra.
We found that the simulated XAS spectra for the Pb and S atoms arranged in a
modified α-GeTe structure, had the closest resemblance to the AN samples. Figure A - 6a
shows the simulated XAS spectra for rocksalt PbS and PbS in α-GeTe type structure,
which were computed in Ref. 28 with CRYSTAL under a density functional theory
(DFT) approach and the Hartree-Fock method. Other atomic configurations that were
similar to the rocksalt structure in ground state energy were considered but their
differences in coordination number resulted in significantly different XAS spectra, which
suggests their atomic arrangements are not present in our PbS QDs. As shown Figure A -
Katherine Roelofs Ph.D. Dissertation
230
6a-d, a larger deviation in Pb–S bond lengths and angles results in a weaker A4 intensity.
This trend suggests that the decrease in the A4 peak observed for the PbS QDs grown
directly on anatase TiO2, especially for the smaller QDs, stems from distortion in bond
angles and asymmetry in bond lengths.
Figure A - 6. Full multiple scattering simulations of S K-edge in distorted atomic
arrangements. The calculated S K-edge XAS spectra are plotted for PbS in a rocksalt
structure, an α-GeTe modified structure calculated from a density functional theory
approach (B3LYP)3 and an α-GeTe modified structure calculated using the Hartree-Fock
(HF)3 method (a); for comparison, the local bonding environment is detailed for each
structure (b-d), respectively. The starred feature (peak A4) of the S-K edge XAS is
dependent on the symmetry of the local bonding environment.
IV. Discussion on lattice distortions and the α-GeTe phase
In the XANES S K-edge spectra (Chapter 5, Section 4), the decrease in the A4
peak observed for PbS QDs grown directly on anatase TiO2 is attributed to distortions in
Katherine Roelofs Ph.D. Dissertation
231
the PbS rocksalt crystal structure due to interaction with the lattice structure of the
underlying anatase (101) TiO2 surface. The simulation results (Figure A - 6) suggest that
the AN PbS QD samples could be distorted towards the α-GeTe crystal structure.
However, the XRD and SAD results for the low-cycle-number AN PbS QD samples
(Chapter 5, Section 4) do not show any peaks attributable to PbS (only peaks from the
TiO2 substrate are observed) why are no diffraction peaks for α-GeTe PbS observed?
The XANES result – the decrease in the A4 peak – only definitively shows that the long-
range order of the PbS NaCl structure breaks down for low-cycle-number AN PbS QD
samples. The simulation results (Figure A - 6a) show that distortions towards the α-GeTe
structure could account for the disappearance of the A4 peak; but it is not necessary that
the entire PbS QD undertakes the α-GeTe structure. For instance, it is possible that the
presence of a few unit cells of α-GeTe structured PbS grew as the base of the QDs, and
slowly changed into NaCl-structured PbS towards the top of the QDs. In such a case, the
Pb-S bonds would be distorted towards the α-GeTe structure, but the variation in exact
bond length would prevent the long-range order in the QDs necessary for observing
peaks in the XRD or electron diffraction results.
V. Energy Gap Calculations
Density functional theory calculation on PbS QDs were carried out using the
package QUANTUM ESPRESSO.5 Perdew–Burke–Ernzerhof (PBE) exchange-
correlation funtional were used and the kinetic energy cutoff for wavefunctions was set to
35 Ry that is sufficient for system energy convergence. HOMO-LUMO energy gap of
QDs with rocksalt structure were benchmarked to experimental values6 as shown in
Figure A - 7. Bond angel distortion were introduced by shifting the displacement of
sulfur atoms in the (110) direction.
Katherine Roelofs Ph.D. Dissertation
232
Figure A - 7. Comparison of experimental fit of energy gap as a function of quantum dot
size with results from density functional theory calculation.
By using a combination of S and O K-edge XAS data, diffraction data, and
theoretical simulations, it was possible to obtain a detailed atomic level understanding of
the interfacial structure between PbS QDs and two types of TiO2 surfaces. We found that
the periodic bonds at the anatase TiO2 surface distorts the crystallinity in PbS QDs by
inducing asymmetric Pb—S bonds, whereas a non-periodic TiO2 bonding environment is
flexible, allowing the growth of rocksalt PbS. As illustrated in Figure A - 8, only an
asymmetric, non-crystalline PbS structure has the potential to grow on the (101) surface
of anatase TiO2 NPs, while structural variations and flexibility in amorphous TiO2 enable
the formation of crystalline PbS QDs. With these detailed structural information, a
quantum simulation was performed to investigate potential effects of the bonding
environment on device performance. Figure A - 9 illustrates the HOMO-LUMO energy
gap of PbS QDs as a function of distortion in the Pb-S-Pb bond angle. With increased
distortion, the energy gap of the QD increases, and this effect becomes more pronounced
with larger sized QDs. This is in agreement with the observation that the onset of light
absorption shifts to higher energy with AN samples where distortion is present. As
Katherine Roelofs Ph.D. Dissertation
233
discussed above, the lattice mismatch occurs at the PbS and TiO2 interface and the
distortion becomes less pronounced as it propagates into the PbS QDs. Accordingly, the
energy gap will shrink as a function of distance towards the interface where charge
injection occurs. The band diagram can be modeled similar to that of a graded band gap
solar cell, but with larger energy gap closer to the injection interface. This energy gap
gradient would pose an energy barrier for electron injection at the interface which could
diminish device performance.7-10
Figure A - 8. Schematics of the PbS/TiO2 interfaces. Atomic arrangement
representations of the interfaces created between (a) structurally-distorted PbS QDs
grown on a (101) anatase TiO2 surface and (b) rocksalt PbS QDs grown on an amorphous
TiO2 surface. These schematics imply that the anatase TiO2 surface imposes some of its
periodicity onto the bottom-up deposited PbS QDs, given the significant lattice mismatch
between the anatase (a = 3.78 Å , b = 9.52 Å), and rocksalt (a = 5.94 Å) structures.
Conversely, PbS QDs might be able to maintain a rocksalt structure on an amorphous
surface by adjusting the underlying TiO2 bonding structure.
Katherine Roelofs Ph.D. Dissertation
234
Figure A - 9. (a) HOMO-LUMO energy gap of different sized PbS QDs as a function of
Pb-S-Pb bonding angle shown by (b) and (c). Distortion from the rocksalt structure is
introduced as a shift in displacements of sulfur atoms in the unit cell that is used for
constructing QD in the density functional theory calculations.
VI. References
1. Trejo, O.; Roelofs, K. E.; Xu, J.; Logar, M.; Sarangi, R.; Norlund, D.; Dadlani,
A.; Kravec, R.; Dasgupta, N. P.; Bent, S. F.; Prinz, F. B., Submitted to Nano Letters 2015.
2. Zagorac, D.; Doll, K.; Schön, J. C.; Jansen, M., Phys. Rev. B 2011, 84, 045206-
045206.
3. Rehr, J. J.; Kas, J. J.; Prange, M. P.; Sorini, A. P.; Takimoto, Y.; Vila, F.,
Comptes Rendus Physique 2009, 10, 548-559.
4. Rehr, J. J.; Kas, J. J.; Vila, F. D.; Prange, M. P.; Jorissen, K., Physical Chemistry
Chemical Physics 2010, 12, 5503-5513.
5. Giannozzi, P.; Baroni, S.; Bonini, N.; Calandra, M.; Car, R.; Cavazzoni, C.;
Ceresoli, D.; Chiarotti, G. L.; Cococcioni, M.; Dabo, I.; Dal Corso, A.; de Gironcoli, S.;
Katherine Roelofs Ph.D. Dissertation
235
Fabris, S.; Fratesi, G.; Gebauer, R.; Gerstmann, U.; Gougoussis, C.; Kokalj, A.; Lazzeri,
M.; Martin-Samos, L.; Marzari, N.; Mauri, F.; Mazzarello, R.; Paolini, S.; Pasquarello,
A.; Paulatto, L.; Sbraccia, C.; Scandolo, S.; Sclauzero, G.; Seitsonen, A. P.; Smogunov,
A.; Umari, P.; Wentzcovitch, R. M., J. Phys. Condens. Matter. 2009, 21, 395502.
6. Moreels, I.; Lambert, K.; Smeets, D.; De Muynck, D.; Nollet, T.; Martins, J.;
Vanhaecke, F.; Vantomme, A.; Delerue, C.; Allan, G.; Hens, Z., ACS Nano 2009, 3 (10),
3023-3030.
7. Mora-Ser , I. Gim nez, S. Fabregat-Santiago, F. G mez, R.; Shen, Q.; Toyoda,
T.; Bisquert, J., Accounts of Chemical Research 2009, 42 (11), 1848-1857.
8. Li, Z.-X.; Xie, Y.-L.; Xu, H.; Wang, T.-M.; Xu, Z.-G.; Zhang, H.-L., Journal of
Photochemistry and Photobiology A: Chemistry 2011, 224 (1), 25-30.
9. Wang, H.; Wang, T.; Wang, X.; Liu, R.; Wang, B.; Wang, H.; Xu, Y.; Zhang, J.;
Duan, J., Journal of Materials Chemistry 2012, 22 (25), 12532-12532.
10. Chen, Z.; Peng, W.; Zhang, K.; Zhang, J.; Yanagida, M.; Han, L., Nanoscale
2012, 4 (24), 7690-7.
Katherine Roelofs Ph.D. Dissertation
236
Appendix B. Details of NiOx ALD as a hole-transport layer in PSCs
This Appendix contains the full details of the optimization of ALD NiOx layers as
hole-transport layers in perovskite solar cells (PSCs).
I. NiOx ALD Saturation Curves
The ALD process using nickelocene as the precursor and ozone as the counter-
reactant was explored in detail. A potential application is the deposition of NiOx atop the
perovskite as an inorganic layer that acts as both a hole-transport material and a
passivating/encapsulating layer to protect the perovskite and improve device
performance. However the NiOx ALD process itself can degrade the perovskite film, due
to the necessary use of ozone as the counter-reactant for the nickelocene precursor.
Below are a series of studies of the NiOx ALD process, for the growth of NiOx on Si
wafers. For all these studies, the nickelocene bubbler was kept at 70 °C, and the
nickelocene delivery lines were kept at all points between 85 °C and 110 °C.
Thicknesses were measured by ellipsometry.
Every attempt was made to minimize the necessary deposition temperature (as the
perovskite films are only stable up to 120 °C at most, depending on the processing of the
perovskite). Figure B - 1 shows the film thicknesses produced at different stage
temperatures for 75 ALD cycles of NiOx. The maximum concentration of ozone (~200
g/m3) was used, as measured by an in-line detector downstream of the ALD process.
This ALD process was originally optimized for temperatures of 225 °C, and it can be
seen at 200 °C the film growth can still reach the range of thickness expected, from 25-30
Å for 75 ALD cycles. However, decreasing the stage temperature below 125 °C
significantly reduces growth by about 2/3 of the optimized thickness. However, NiOx
films will still grow at these temperatures.
Katherine Roelofs Ph.D. Dissertation
237
Figure B - 1. Film thicknesses of 75 ALD NiOx cycles grown on Si wafers at varied
stage temperatures. The ozone concentration was kept at 200 g/m3 and a pulse-purge
sequence of 3 s/30 s for the nickelocene and 3 s/30 s for the ozone were used.
Further, we sought to limit the exposure of the perovskite to ozone. Under
oxidizing conditions, the perovskite layer will revert to PbI2 as the CH3NH3+ and I
- leave
the film in the form of various gas products. It can be seen in Figure B - 2 that lowering
the ozone concentration can have a significant stabilizing effect on the perovskite layer,
where the yellow colored film has been fully converted to PbI2, but the films that remain
dark brown still have a significant portion of the perovskite phase present. But, even
small amounts of PbI2 in a perovskite solar cell can significantly harm device
performance.
Katherine Roelofs Ph.D. Dissertation
238
Figure B - 2. Effect of ozone exposure on un-coated perovskite films. These are
perovskite films deposited on nanoporous TiO2 using the two-step deposition method.
Preliminary results suggest that the NiOx growth begins to dramatically decrease
for ozone concentrations below 20 g/m3. We lowered the ozone concentration to the
edge of this window, keeping it close to 20 g/m3, though occasionally as low as 13 and as
high as 30 g/m3, and performed a saturation test on the ozone pulse time. As shown in
Figure B - 3, we found that the ozone pulse time can be decreased to 1 s (roughly the
time-response limit of the ALD valves), with no significant dip in growth. We note that
for this test, the stage temperature was kept at 200 °C and the films were grown for 75
NiOx ALD cycles. Based on the result in Figure B - 3, the pulse/purge times for the
ozone counter-reactant were set at 1 s/30 s.
Figure B - 3. Film thickness of 75 ALD NiOx cycles grown on Si wafers at varied ozone
pulse times, with the ozone purge time kept at 30 C.
Katherine Roelofs Ph.D. Dissertation
239
With the pulse/purge time of ozone set at 1 s/30 s, the concentration of the ozone
was then varied carefully, keeping the stage at 90 °C, where the NiOx growth is much
slower, but the perovskite film is expected to be stable. The results are shown by the
black diamonds in Figure B - 4. It was found that below 20 g/m3 the NiOx film thickness
resulting from 75 ALD cycles on Si does indeed begin to plummet. However, some
preliminary work on using slightly higher stage temperatures at these ultra-low ozone
conditions suggest higher temperatures may be able to recover the growth. Our work
with perovskite films suggest they are more susceptible to damage from these ozone
levels than from the heat treatment, so we are willing to go a bit higher on temperature in
order to further limit ozone exposure.
Figure B - 4. Film thicknesses of 75 ALD NiOx cycles grown on Si wafers at varied
ozone concentrations. Ozone pulse/purge time is kept at 1 s/30 s, and the stage
temperature is kept to 90 °C (black diamonds), though slightly higher stage temperatures
(colored diamonds) were also tried in this preliminary work.
However, no matter the ALD conditions, uncoated perovskite films were not able
to withstand NiOx deposition using the nickelocene and ozone, and produce functioning
solar cells. X-ray diffraction (XRD) analysis of the films revealed some PbI2
components, which are likely the reason for the failed devices. Another possibility, or
perhaps an additional problem, is that the as-deposited ALD NiOx films were not that
conductive. For inverted perovskite solar cells, we grew ALD NiOx directly on ITO
Katherine Roelofs Ph.D. Dissertation
240
substrates, using the following conditions: 180 °C stage temperature, 70 °C nickelocene
bubbler, pulse/purge times of 1 s/30 s for nickelocene and 3 s/30 s for ozone, and ozone
concentrations of ~200 g/m3. We observed that the resulting films needed to be annealed
in air for 1 hr at 250 °C in order to produce working solar cells (see Chapter 7, section 3),
presumably in order to improve their conductivity.
II. Perovskite Stability to Oxidizing Conditions & Pre-
Coating by Metal Oxide Layers Prior to NiOx ALD
The strategy of pre-coating perovskite films prior to NiOx was explored as well.
ALD TiO2 and ALD Al2O3 processes optimized in previous work by Han Bo-Ram Lee
and Fatemeh Hashemi use water as a counter-reactant. For TiO2, titanium isopropoxide
(TTIPs) was used as the precursor, with a bubbler temperature of 65 °C. For Al2O3,
trimethylaluminum (TMA) is used as the precursor, with a bubbler temperature of 30 °C.
The normal growth rate of Al2O3 in the optimized temperature window is 1.1 Å/cyc,
while for TiO2 it is 0.3 Å/cyc. The Al2O3 and TiO2 ALD processes were run at 50 °C to
protect the perovskite from temperature degradation, or temperature-enhanced
degradation by the H2O pulse. For these preliminary studies, it was found that the Al2O3
and TiO2 growth rates at 50 °C were very close to the growth rates under optimized ALD
conditions.
The results of this work shed some light on the potential stabilizing effects of
inorganic layer coatings to oxidizing conditions – in this case, the ozone used for the
NiOx deposition. Before these tests, it was not known how stable the perovskite layers
would be to the water vapor introduced as the counter-reactant for the TiO2 and Al2O3
processes. However, for TiO2 or Al2O3 coatings of up to 10 nm and deposition times of
up to 3 hours, the perovskite films retain the dark color, and the perovskite structure as
measured by XRD. Figure B - 5shows camera images of perovskite films after ozone
exposure. The films were split in half, and the top half exposed for 25 s to 25 g/m3
ozone, while the bottom half was exposed for 25 s to 5 g/m3 ozone. The films range from
2 cyc Al2O3 to 80 cyc (~9 nm) Al2O3, and 2 cyc TiO2 to 20 cyc (~6 nm) TiO2. It is
Katherine Roelofs Ph.D. Dissertation
241
unclear why the uncoated film in Figure B - 5a appears more table compared to an ultra-
thin Al2O3 coating of 2 cycles shown in Figure B - 5b. From the XRD results, the Al2O3
process does not damage the perovskite. It could be that this reveals the impact of
processing variability of the perovskite film can have on stability. Overall, the thicker the
Al2O3 or TiO2 coating, the more resistive the perovskite is to oxidation. As can be seen
in Figure B - 5h, coated perovskite films appear similar visually before and after ozone
exposure.
Figure B - 5. Camera images of perovskite films, deposited on ~350 nm thick
nanoporous TiO2 films by the two-step deposition process. The films were split and the
top half exposed to 25 s of ~25 g/m3 ozone while the bottom half was exposed to 25 s of
~5 g/m3 ozone. The films had a variety of ALD coatings prior to ozone exposure: (a)
uncoated film, (b) 2 cyc Al2O3, (c) 5 cyc Al2O3, (d) 20 cyc Al2O3, (e) 80 cyc Al2O3, (f) 2
cyc TiO2, (g) 20 cyc TiO2. (h) Shows light shining through duplicate films coated with
20 cyc TiO2, one un-exposed to ozone (h), and one exposed to 25 s of ~5 g/m3 ozone.
Katherine Roelofs Ph.D. Dissertation
242
Figure B - 6. Perovskite films grown on ~350 nm nanoporous TiO2 film, using the two-
step deposition process. The films were then coated with either 15 ALD cycles of Al2O3
(~1.7 nm) or 15 ALD cycles of TiO2 (~0.5 nm). The films were then exposed to ozone
for 25 s at ~ 5 g/m3
(or ~1 vol% ozone). XRD measurements show that after exposure,
the PbI2 phase in the film increases significantly, and that 0.5 nm TiO2 provides superior
protection than 1.7 nm Al2O3.
Katherine Roelofs Ph.D. Dissertation
243
Figure B - 7. XRD measurements of a perovskite film deposited on ~350 nm
nanoporous TiO2 using the two-step deposition process. The black trace shows the as-
deposited perovskite film, the blue trace is after 80 Al2O3 ALD cycles (~9 nm), and the
red trace shows the Al2O3-capped film after ozone exposure for 25 s at ~5 g/m3.
Perovskite peaks are indicated by black circles, the main PbI2 peak is indicated on the left
with a diamond.
Figure B - 7 shows that by using a fairly thick ALD film coating, in this case ~9
nm of Al2O3, the perovskite can be fairly successfully protected from ozone exposure.
Comparing the PbI2 peak growth with thinner ALD coatings in Figure B - 6, we conclude
that ALD films of at least 5 nm thick are required to protect the perovskite. However,
our previous studies of Al2O3 ALD layer coating interfaces in DSSCs and QDSSCs
suggest that Al2O3 ALD layers thicker than ~1 nm are insulating enough to block all
charge flow through the device. Future studies could try more closely studying the
stability with TiO2 ALD capping layers, as this seems more promising than the Al2O3.
TiO2 is also more conductive, so it could be that thinner TiO2 layers can successfully coat
the perovskite without blocking charge flow.
Katherine Roelofs Ph.D. Dissertation
244
The ozone exposure levels studied here (~25 s at ~5 to ~25 g/m3) were based on
the levels studied for an ozone pulse for NiOx ALD growth. We had hoped that using
water-based ALD systems (TiO2 or Al2O3), we would be able to pre-coat the perovskite
layer, and then could successfully deposit the NiOx layer atop the perovskite to act as the
hole-transport layer in the place of spiro-OMeTAD. Our preliminary tests of this strategy
indicate that the ALD pre-coating layers have to be too thick to adequately protect the
perovskite film from NiOx deposition.
Figure B - 8 shows the J-V curve results for various ALD coatings prior to NiOx.
The perovskite solar cell structure is shown in Figure B - 8, these devices were planar
perovskite films deposited by chemical vapor deposition at 120 °C by collaborators at the
Okinawa Institute of Science and Technology (OIST). The perovskite films were first
pre-coated with various thicknesses of TiO2 or Al2O3, followed in some cases by a thin
film of NiOx. Finally, the devices were coated with spiro-OMeTAD layers prior to
deposition of the top Au contact. The goal of this study was to determine if the
combination of the ALD layers would harm device performance in and of themselves –
the top spiro-OMeTAD layer is to ensure that the devices don’t fail out of a too-thin
coatings intermediary between the perovskite and the Au. Unfortunately, the ALD
coatings proved too thick for the device. The highest JSC curve indeed is seen for the
thinnest coating of the most conductive ALD layer – 1.3 nm of TiO2. The poor
performance of the control cell was attributed to the fact that all the devices rested for 1
week in the glovebox in the un-coated stage, during which we believe they were exposed
to low levels of dimethylformamide (which the PbI2 components of the perovskite film
are soluble in) due to other depositions conducted in the glovebox over that time.
Katherine Roelofs Ph.D. Dissertation
245
Figure B - 8. Performance of planar perovskite solar cells where the perovskite was
deposited by chemical vapor deposition (CVD) by collaborators at OIST. The basic
structure is shown in the bottom right, while variants on this structure are indicated in the
legend. For instance, 2.0 nm TiO2/ 0.2 nm NiOx /spiro indicates that the perovskite layer
was coated with 2 nm of ALD TiO2, then 0.2 nm of NiOx, followed by a standard coating
of spiro-OMeTAD.