Colloidal Engineering for Infrared-Bandgap Solution-Processed Quantum Dot Solar Cells

by

Amirreza Kiani

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy

Edward S. Rogers Department of Electrical & Computer Engineering University of Toronto

© Copyright by Amirreza Kiani, 2017

ii

Colloidal Engineering for Infrared-Bandgap Solution-Processed

Quantum Dot Solar Cells

Amirreza Kiani

Doctor of Philosophy

Edward S. Rogers Department of Electrical and Computer Engineering

University of Toronto

2017

Abstract

Ever-increasing global energy demand and a diminishing fossil fuel supply have prompted the

development of technologies for sustainable energy production. Solar photovoltaic (PV) devices

have huge potential for energy harvesting and production since the sun delivers more energy to

the earth in one hour than the global population consumes in one year.

The solar cell industry is now dominated by silicon PV devices. The cost of silicon

modules has decreased substantially over the past two decades and the number of installed

silicon PV devices has increased dramatically. There remains a need for emerging solar

technologies that can harvest the untapped portion of the solar spectrum and can be integrated on

flexible and curved surfaces.

This thesis focuses on colloidal quantum dot (CQD) PV devices. CQDs are nanoparticles

fabricated using a low-temperature and cost-effective solution technique. These materials suffer

from a high density of surface traps derived from the large surface-to-volume ratio of CQD

nanoparticles, combined with limited carrier mobility. These result in a short carrier diffusion

length, a main limiting factor in CQD solar cell performance.

iii

This thesis seeks to address the poor diffusion length in lead sulfide (PbS) CQD films and

pave the way for new applications for CQD PV devices in infrared solar harvesting and waste

heat recovery. A two-fold reduction in surface trap density is demonstrated using molecular

halide treatment. Iodine molecules introduced prior to the film formation replace the otherwise

unpassivated surface sulfur atoms. This results in a 35% increase in the diffusion length and

enables charge extraction over thicker active layer leading to the world’s most efficient CQD PV

devices from June 2015 to July 2016 with the certified power conversion efficiency of 9.9%.

This represents a 30% increase over the best-certified PCE (7.5%) prior to this thesis. The

colloidal engineering highlighted herein enables infrared (IR) solar harvesting for the first time.

Addition of short bromothiol ligands during the synthesis significantly reduces the

agglomeration of 1 eV bandgap CQDs and maintains efficient charge extraction into the

selective electrodes. The devices can augment the performance of the best silicon cells by 7

power points where 0.8 additive power points are demonstrated experimentally. A tailored

solution exchanged process developed for 1 eV bandgap CQDs results in air-stable IR PV

devices with improved manufacturability. The process utilizes a tailored combination of lead

iodide (PbI2) and ammonium acetate for the solution exchange and hexylamine + MEK as the

final solvent to yield solar thick films with the filtered (1100 nm and beyond) performance of

0.4%. This thesis pushes the limit of CQD device applications to waste heat recovery. I

demonstrate successful harvesting of low energy photons emitted from a hot object by designing

and developing the first solution-processed thermophotovoltaic devices. These devices are

comprised of 0.7 eV bandgap CQDs that successfully harvest photons emitted from an 800°C

heat source.

iv

Dedication

To Mom & Dad.

v

Acknowledgments

First and foremost, I would like to acknowledge my supervisor, Professor Ted Sargent. His scientific

instinct and management skills helped me finish a very efficient PhD. I learned a lot from him specially

how to think outside the box and be a more goal-oriented person. The support, guidance, and advice he

provided throughout my PhD were invaluable.

I would like to thank Dr. Larissa Levina, Damir Kopilovic, and Elenita Palmiano for all their support.

Without them, this thesis would have taken years longer to complete assuming Ted was OK with it.

I would like to thank Professor Nazir Kherani and Professor Wai Tung Ng for their inputs to make this

thesis better and more clear.

I would like to thank my co-workers, Sargent Lab people, who helped me a lot and I enjoyed amazing

team-works with. Thanks Alex Ip for always helping me and for your confidence and presence. Thanks

Brandon Sutherland for being so positive and having the only loud voice that I am OK with. Thanks

Xinzheng Lan for having the best spirit. Thanks Sasha Voznyy for knowing EVERYTHINGs and your

amazing calmness. Thanks Andre Labelle for laughing so loud but so good. Thanks Illan for being a goal

in number of publications that I will never meet. Thanks Grant Walters for not being a nice co-worker but

the best roommate. Thank Remi Wolowiec for always reminding me getting what you want is not as easy

as you think, i.e. a better desk space. I would like to thank others who I had the chance to work with: Dr.

Sjoerd Hoogland, Hamidreza Movahed, Graham Carey, David Zhitomirsky, Lisa Rollny, Valerio

Adinolfi, Pelayo, Pongsakorn, Mingjian (MJ), Fengjia, Michael Adachi, Riccardo Comin, Ning, Dan

Sellan, Gabriel, Mengxia Liu, Olivier (O2), James, Andrew Proppe, Xiwen Gong, Chris Wong, Randy

Sabatini, and all others who I unintentionally forget to mention here but helped me accomplish the PhD

goal.

To my friends, who I missed and lost quite a lot of them. To Peyman, for being a true friend for me. To

Sina, who is going to be the funniest Dr. in my life soon. To Emad, for being my “Agha Oveisi”. To

Ramin, for all our high-level discussions and this true question: “iani, ie PhD be ma nemikhore?”. To my

Sabi, for coping with my stupid fears, having the nicest soul and being the loveliest “Khabaloo”.

I want to thank and send my love to my amazing family. To my Baba, who left us a year ago for the better

place but his memory is always with me and his love of life, integrity, hope, ambition, and vision are the

lights of my future path. To my Mom, her love is the most valuable thing I am ever going to have in life.

To Sara, for being the best sister anyone can imagine and to my first niece that will be with us very soon.

To Samira, for being the other best sister anyone can imagine and for reminding me what the definition of

hard work is. To all other family members, Mamani, Mohammad, David, Tara, Fariba, Farnia, Nazi, Dais

and all others for their love and support.

vi

Table of Contents

Acknowledgments........................................................................................................................... v

Table of Contents ............................................................................................................................ v

List of Tables ............................................................................................................................... viii

List of Figures ................................................................................................................................ ix

List of Acronyms .......................................................................................................................... xii

Chapter 1 Introduction and Motivation ........................................................................................... 1

Chapter 2 Background .................................................................................................................... 8

2.1 Photovoltaic Theory .....................................................................................................................8

2.2 Colloidal Quantum Dots ............................................................................................................13

2.2.1 CQD synthesis .......................................................................................................................13

2.2.2 CQD Film Formation .............................................................................................................15

2.3 CQD Photovoltaic Devices ........................................................................................................17

2.4 Limitations of CQD PV devices ................................................................................................18

Chapter 3 Surface Passivation Using Molecular Halide Treatment ............................................. 20

3.1 Introduction ................................................................................................................................20

3.2 Molecular Halide Passivation ....................................................................................................21

3.3 Molecular Halide Passivated CQD-Based Solar Cells ...............................................................24

3.4 Mechanistic Study of Physical Origin of Improved Performance .............................................27

Chapter 4 Infrared Solar Harvesting via Coupling Enhancement and Agglomeration Suppression

32

4.1 Introduction ................................................................................................................................32

4.2 Selecting Proper CQD Bandgap ................................................................................................33

4.3 Depleted Heterojunction CQD device .......................................................................................35

4.4 Ligand Engineering ....................................................................................................................37

4.5 Device Performance ...................................................................................................................43

Chapter 5 Single Step Film Formation for Infrared Solar Harvesting .......................................... 47

5.1 Introduction ................................................................................................................................47

vii

5.2 Solution Exchange Process ........................................................................................................48

5.3 Solution-Exchanged CQD IR PVs .............................................................................................52

Chapter 6 Thermophotovoltaic Harvesting of Waste Heat via Gradient-Doped Quantum Dots . 57

6.1 Introduction ................................................................................................................................58

6.2 TPV Setup ..................................................................................................................................59

6.3 Gradient-Doped TPV Devices ...................................................................................................61

6.4 Thermal Stability........................................................................................................................65

Chapter 7 Conclusion and Perspective ......................................................................................... 68

7.1 Conclusion .................................................................................................................................68

7.2 Perspective .................................................................................................................................69

7.2.1 Power Conversion Efficiency ................................................................................................69

7.2.2 Long-Term Impact .................................................................................................................70

7.3 Final Conclusion ........................................................................................................................71

List of PhD Publications ............................................................................................................... 73

References ..................................................................................................................................... 74

Appendix A−Experimental Details ............................................................................................... 83

Appendix B−Copyright Acknowledgments .................................................................................. 95

viii

List of Tables

Table 3-1 Static figures of merit for control and I2-treated devices. ............................................ 26

Table 4-1 Detailed current-voltage characteristics of the unfiltered and filtered devices. ........... 44

Table 5-1 Performance table summary of a champion cell .......................................................... 54

Table 6-1 Details of thermo photovoltaic performance for different architectures ...................... 65

ix

List of Figures

Figure 1-1 Energy consumption is expected to go up for the coming decades. ............................. 2

Figure 1-2 Germany power supply by hour .................................................................................... 2

Figure 1-3 Historical trend for the cost of silicon solar panels and the 2015 cost of silicon PV

systems. ........................................................................................................................................... 3

Figure 1-4 AM1.5 global solar spectrum ........................................................................................ 5

Figure 2-1 Schematic of p-n junction ........................................................................................... 10

Figure 2-2 Representative current-voltage curve .......................................................................... 12

Figure 2-3 Schematic of the LaMer model ................................................................................... 14

Figure 2-4 Absorption spectra of CQDs at different sizes ............................................................ 15

Figure 3-1 Characterization of PbS CQDs following the I2 treatment ......................................... 22

Figure 3-2 TEM image of PbS QDs following different I2 treatment conditions ........................ 23

Figure 3-3 XPS measurements of the key elements at different stages of ligand exchange ........ 23

Figure 3-4 Device architecture and performance .......................................................................... 24

Figure 3-5 CQD devices thickness study and absorption profile ................................................. 25

Figure 3-6 Figures of merit for a certified molecular iodine CQD device. .................................. 26

Figure 3-7 Optoelectronic model of the origin of improved device performance ........................ 27

Figure 3-8 Measurement of mobility in the I2-treated and control films ..................................... 28

Figure 3-9 SCAPS modeling of control and molecular iodine treated CQD photovoltaic devices

....................................................................................................................................................... 29

Figure 3-10 Experimental determination of diffusion length and improved passivation ............. 30

file:///C:/Users/Amirreza%20K.%20Admin/Dropbox/work/Graduation/Thesis/Thesis.docx%23_Toc464667333

x

Figure 4-1 Detailed balance approach to select CQD bandgap .................................................... 34

Figure 4-2 Depleted heterojunction CQD device. ........................................................................ 35

Figure 4-3 Effectiveness of carrier injection from CQD film into TiO2 electron acceptor with

respect to the CQD electron affinity ............................................................................................. 36

Figure 4-4 Schematic of the effect of short thiol ligands on preventing agglomeration .............. 37

Figure 4-5 Transmission electron microscopy (TEM) images of the 1 eV-bandgap CQDs with

and without thiol (i.e. bromothiol) ligands ................................................................................... 38

Figure 4-6 Bromothiol treated CQD film characterization ........................................................... 40

Figure 4-7 Absorption spectra near the excitonic peak for CQD films without (blue circles) and

with (green squares) bromothiol ligands....................................................................................... 40

Figure 4-8 Transient PL decay spectroscopy and Film PL spectra .............................................. 41

Figure 4-9 Conductivity study and average carrier density for QCD films with and without

bromothiols. .................................................................................................................................. 42

Figure 4-10 The effect of combined ligand strategy on the PV device performance. .................. 43

Figure 4-11 External quantum efficiency of the champion cell made from combined passivated

CQD solids. ................................................................................................................................... 45

Figure 5-1 Solvent optimization for CQD inks............................................................................. 50

Figure 5-2 Spectroscopic properties of solution-exchanged IR PbS CQDs capped with PbI2. ... 51

Figure 5-3 Solar cells based on solution exchanged IR PbI2-capped PbS CQDs dispersed in

HXA+MEK as the final solvent .................................................................................................... 53

Figure 5-4 Stability of the solution-exchanged CQD-based solar cells ........................................ 54

Figure 5-5 Effect of ZnO on the carrier extraction for different bandgap CQDs ......................... 56

Figure 6-1 A CQD-based TPV system ......................................................................................... 60

file:///C:/Users/Amirreza%20K.%20Admin/Dropbox/work/Graduation/Thesis/Thesis.docx%23_Toc464667351file:///C:/Users/Amirreza%20K.%20Admin/Dropbox/work/Graduation/Thesis/Thesis.docx%23_Toc464667356file:///C:/Users/Amirreza%20K.%20Admin/Dropbox/work/Graduation/Thesis/Thesis.docx%23_Toc464667359file:///C:/Users/Amirreza%20K.%20Admin/Dropbox/work/Graduation/Thesis/Thesis.docx%23_Toc464667359file:///C:/Users/Amirreza%20K.%20Admin/Dropbox/work/Graduation/Thesis/Thesis.docx%23_Toc464667361file:///C:/Users/Amirreza%20K.%20Admin/Dropbox/work/Graduation/Thesis/Thesis.docx%23_Toc464667361file:///C:/Users/Amirreza%20K.%20Admin/Dropbox/work/Graduation/Thesis/Thesis.docx%23_Toc464667362file:///C:/Users/Amirreza%20K.%20Admin/Dropbox/work/Graduation/Thesis/Thesis.docx%23_Toc464667363

xi

Figure 6-2 TPV device architectures and their operation ............................................................. 62

Figure 6-3 Gradient-doped TPV device characterization ............................................................. 64

Figure 6-4 Evaluation of thermal stability .................................................................................... 66

xii

List of Acronyms

PV − photovoltaic

CQD − colloidal quantum dot

PbS − Lead sulfide

PCE − power conversion efficiency

JSC − short circuit current density

VOC − open circuit voltage

FF − fill factor

XPS − X-ray photoelectron spectroscopy

XRD – X-ray diffraction

PL − photoluminescence

PLQY − photoluminescence quantum yield

TEM − transmission electron microscopy

SEM − scanning electron microscopy

EQE − external quantum efficiency

FTIR − Fourier-transform infrared spectroscopy

IQE − internal quantum efficiency

DFT − density functional theory

MPA − mercaptopropionic acid

TBAI − tetrabutylammonium iodide

EDT − ethanedithiol

BTA − n-butylamine

HXA − hexylamine

MEK − methyl ethyl ketone

DMF − N,N-dimethylformamide

TMAOH − tetramethylammonium hydroxide

PbI2 − lead iodide

CdCl2 − cadmium chloride

C2H3O2NH4 − ammonium acetate

MAPbI3 − methylammonium lead iodide

ITO − indium-doped tin oxide

FTO − fluorine-doped tin oxide

ZnO − zinc oxide

TiO2 − titanium dioxide

MoO3 − molybdenum oxide

PEIE − polyethylenimine ethoxylated

OA − oleic acid

NP − nanoparticle

LWP − long wave pass

EA − electron affinity

TPV − thermophotovoltaic

QJ − quantum junction

1

Chapter 1 Introduction and Motivation

Today, the adverse environmental impact of continued economic growth is affecting all

humanity. Climate change is no longer an unproven scientific theory, but a real-world challenge

with detrimental consequences threatening the world’s stability and food supply. Economic

growth is vital and energy sources are the engine of this unstoppable train. Hitherto, fossil fuels,

such as oil, natural gas and coal have driven the world’s economy and it is projected that these

energy sources will continue to account for more than 75% of the world’s energy consumption

by 2040 (Figure 1-1).1 While fossil fuels are attractive due to their high energy density, their

combustion generates harmful byproducts. Carbon dioxide (CO2), a major byproduct of

combustion reactions, is a powerful greenhouse gas. CO2 contributes significantly to global

warming and its average atmospheric concentration exceeded 400 ppm.2 In December 2015, 195

countries adopted the first-ever universal binding global climate deal known as “Paris

Agreements”.3 The agreement seeks to limit global temperature rise to 2°C over the

preindustrial average, beyond which numerous climate catastrophes are predicted. Climate

scientists have predicted that to realize this temperature goal, the CO2 level must be limited to

300-350 ppm.4 Given the long life time of carbon dioxide, even if CO2 emission halted today, it

will take decades to reach a safe concentration.5 Apart from carbon dioxide, other fossil fuel

sources (in particular coal) emit additional pollutants that are immediately harmful to human

health, such as soot particles and sulfur/nitrogen oxides.6

The world’s energy consumption will continue to go up for the coming decades to ensure

economic growth as shown in Figure 1-1. However, this increase imposes great risks on the

environment and human health. Therefore, economic growth must be decoupled from the

production of greenhouse gases and environmentally hazardous wastes. Renewable energies,

such as wind, geothermal, and solar offer a powerful solution to this challenge and are projected

to supply 20% of the world’s energy demand by 2040. Even today, renewable energy sources

have proven their capability. Figure 1-2 shows that clean energy supplied Germany’s entire

power demand for the first time from 1 pm to 2 pm on May 15th 2016.7 This has been a milestone

for Germany and a confirmation of the feasibility of renewable energy utilization.

2

Figure 1-2 Germany power supply by hour. On May 15th 2016, Germany’s renewable energy

sources delivered equal power to the country’s entire demand. Reprinted from Bloomberg

News.7

Figure 1-1 Energy consumption is expected to go up for the coming decades. While oil,

gas and coal continue to be the main sources for energy production, renewable energies are

expected to supply 20% of the world’s energy consumption by 2040. Adapted from

International Energy Outlook 2016.1

3

Solar energy is the most abundant source of energy available on Earth. Sunlight provides

vastly more energy daily than the global population consumes and therefore is a compelling

option to power the electricity-centric future. Today, the photovoltaic (PV) industry is a robust

$100 billion industry, which is mostly dominated by silicon photovoltaic devices (approximately

90%).8,9 Global cumulative installed PV capacity in 2016 is 270 GWP and it is projected to

increase to 760 GWP in 2025.10

Figure 1-3 Historical trend for the cost of silicon solar panels and the 2015 cost of silicon

PV systems. (a) The cost for silicon panels has declined significantly to $0.30/WP in 2015.

Reprinted from Cleantechnica.com.10 (b) The cost of silicon PV systems is still well beyond the

$1/WP threshold, mainly due to expensive installation requirements, labor, and overhead costs.

Reprinted from NREL 2016 cost report.14

a

b

4

While these numbers are promising, in order to compete with traditional energy sources,

in particular cheap and abundant natural gas, installed solar PV modules must cost less than $1

per watt-peak. This means an energy cost of $0.05/kWh over a system lifetime.11 While the cost

of silicon modules has gone below the $1/WP threshold (Figure 1-3a) mostly due to economies of

scale, the cost of silicon PV systems is well above the competitive edge.12,13 This is mainly due

to the significant hardware costs required for silicon PV installations and soft costs, such as

installation labor, customer acquisition and overhead cost. Therefore, the 2016 cost for

residential silicon PV systems is close to $3/WP (Figure 1-3b).14 While residential solar

electricity has remained expensive, the cost for large-scale solar electricity generation has

decreased more rapidly and is currently below $1.5/WP for a 100 MW power plant. On June 27th

2016 Bloomberg news agency reported that a 800 MW solar power plant is expected to be built

in Dubai by 2020.15 The power plant will generate electricity at a projected price of 2.99

cents/kWh. This projected price of electricity is comparable to that of natural gas.

Third-generation photovoltaic systems, including dye-sensitized, organic, perovskite and

colloidal quantum dot (CQD), offer a low-cost, light-weight, and potentially high efficiency solar

energy capture and conversion. While most of these technologies currently operate at lower

efficiencies than what is commercially feasible, they have the potential to reach efficiencies well

beyond the commercial barrier. Furthermore, they do not possess some of the silicon’s inherited

limitations. Silicon panels are heavy, bulky, inflexible, and require costly mounting racks to

ensure long-term stability. Silicon cells are mainly based on highly pure and crystalline silicon

that requires high-temperature processing, clean-room facilities, and rigorous process control.

Apart from cost-related issues, silicon cells are transparent to a significant portion of the solar

spectrum: silicon cannot absorb any photons beyond 1100 nm in wavelength due to its

unchangeable bandgap (highlighted portion in Figure 1-4). The ideal solar material is cheap,

light weight and highly manufacturable. It has the capability to harvest a broader portion of the

solar spectrum than what silicon absorbs.

5

Figure 1-4 AM1.5 global solar spectrum. The silicon cells are transparent to the highlighted

portion of the solar spectrum. Generated based on the spectral irradiance raw data available

online.

In this thesis, I highlight the potential of CQDs to fill this niche. CQDs are semiconductor

nanoparticles that are stabilized in solution by long-chain organic molecules attached to their

surface. These nanoparticles are synthesized using a low-temperature and reproducible method

and therefore are cheap. These nanoparticles can be employed as an ink and are, therefore

compatible with large-scale fabrication processes. The CQD-based solar technology is also light

weight. Their large absorption coefficient enables full absorption in a 1μm-thick film.

Furthermore, their bandgap can be simply tuned by changing the size of the nanoparticles, and

therefore enables a more precisely-positioned absorption edge than silicon in a single-junction

architecture or even the capability to absorb the entire solar spectrum in multi-junction

architectures. While the CQDs are an interesting material system, solar PV devices made based

on these nanoparticles have yet to deliver a meaningful output capacity to be commercially

feasible. Carrier diffusion length, the average distance a carrier diffuses before recombining in an

active PV film, is three orders of magnitude lower for colloidal quantum dot films than for

silicon (~100 nm for CQDs compared to hundreds of micrometer the for silicon).16,17 As a result,

CQD film thicknesses for efficient charge extraction are significantly lower than the ones

6

required for complete absorption. Carrier diffusion length is therefore the main limitation

holding back progress in the CQD PV field.

This thesis highlights strategies to improve the limited diffusion length in colloidal

quantum dot films. It also, through colloidal engineering, offers for the first time, a new

application for CQD PV devices to augment the performance of highly efficient PV systems by

harvesting the untapped portion of the solar spectrum in the near-infrared region. The Sargent

group’s initial calculations confirm that given the low cost of CQD technology, the addition of 4

power points to a 20% silicon cell (20% additional power), results in a 40% increase in gross

profit margin of the silicon module manufacturer with a break-even point at 2-2.5 additive power

points. Given the theoretical capability of the CQD technology to add 7 power points to the best

silicon cells (highlighted in this thesis), this technology is attractive for commercial applications.

Another perspective is that this technology lowers the $/kWh for existing solar systems, making

them more competitive and more financially appealing for the end costumer. This results in

larger integration of the solar systems and more revenue for the module manufacturers. This

thesis further pushes the limit of CQD PV applications to waste heat recovery by utilizing the

smallest-bandgap colloidal quantum dots ever used in a PV system.

A brief outline of the thesis is as follows: Chapter 2 highlights the existing research in the

CQD field and studies the CQD material system, its limitations/benefits and applications in a

greater detail. Chapters 3 through 6 focus on original work produced for this thesis. Chapter 3

highlights an efficient halide-based surface treatment to reduce the CQDs’ trap density and

shows world-record efficiency for CQD PV devices. Chapters 4 and 5 outline colloidal

engineering strategies to develop 1 eV bandgap CQDs for infrared solar harvesting. Guided by

the findings in Chapter 3, Chapter 4 identifies the optimal bandgap for CQDs to harvest infrared

photons (1100 nm in wavelength and beyond) and employs a strategy for surface passivation and

agglomeration suppression. Chapter 5 builds upon previous chapters’ achievements and

highlights a ligand exchange strategy that improves manufacturability and air-stability of IR PV

devices. These chapters demonstrate efficient and air stable IR PV devices with improved

manufacturability for the first time. Chapter 6 pushes the CQD PV limit further by designing a

material system/device architecture that successfully harvests photons emitted from an 800°C

heat source by employing the smallest-bandgap CQDs ever employed in a PV device. Chapter 7

7

draws together the overall conclusion and findings generated in this thesis and will discuss

relevant future research directions.

8

Chapter 2 Background

This thesis showcases the capability of colloidal quantum dot (CQD) photovoltaic (PV) devices

for broad solar energy harvesting employing CQDs with different sizes and hence different

bandgaps. The capability that makes CQDs powerful candidates for capturing the infrared

portion of the solar spectrum to augment the performance of high efficiency PV systems and

waste heat recovery has been highlighted in this thesis for the first time. CQD PVs have

progressed significantly since their initial demonstration in 2005.18 In more than a decade, their

power conversion efficiency has improved from sub-1% to more than 11%.19 This chapter

provides background on the science of CQD PVs through examining the research that has

informed and led to the work presented in this thesis. First, the basis of photovoltaic theory will

be covered briefly. Second, colloidal quantum dots will be discussed in greater detail laying out

their synthesis and CQD film formation. Finally, CQD PV device architecture and the key

factors that limit the efficiency of these devices will be outlined.

2.1 Photovoltaic Theory

Energy harvesting through the photovoltaic effect is the conversion of light energy into electrical

energy. In the simplest form, any photovoltaic device operation comprises of three key steps: 1.

absorption of light (i.e. photons) and generation of electron-hole pair, 2. separation of the photo-

generated charges and transporting them to asymmetric contacts, and 3. extraction of the charges

into the electrodes. While some PV systems rely on separate materials to absorb and transport

photo-generated carriers (i.e. step 1 and 2) in silicon PVs and most of the quantum dot solar cells

(including all solar cells in this thesis) the light absorber and charge separating material are a

single component. In a typical solar cell, the front electrode is transparent to allow light to reach

the absorber material, while the back electrode is a highly reflective metal to reflect back the

unabsorbed light to the absorber for a second pass. Photonic enhancement strategies, such as

roughening the back electrode, can allow for multiple passes and improve absorption for a given

thickness of the absorber material. The absorber material is a semiconductor with a bandgap

optimized for optimal absorption. Upon absorption of a photon, an electron will be promoted

from the valance band (Ev) to the conduction band (Ec) and hence a tightly-bound electron-hole

pair will be generated (i.e. an exciton). Therefore, only photons with energies equal to or greater

than the semiconductor bandgap can be absorbed. The exciton has equivalent energy to the

9

absorbed photons. If the energy is larger than the bandgap energy of the semiconductor absorber,

the exciton relaxes to the energy level of the semiconductor absorber over a short period of time

via thermalization.20 In order to generate a useful output, the exciton should be dissociated into

its constituent carriers which then should be transported and extracted by the asymmetric

electrodes. Depending on the absorbing/transporting material(s), various trap states and

recombination states may hinder the full extraction of the photo-generated carriers, which limits

the output current.

In order to extract charges, an electric field should be present. The required field can

either be generated through an external bias or by a built-in electric field from a semiconductor-

metal or semiconductor-semiconductor junction. Most solar cells, including the solar cell devices

highlighted in this thesis, are based on a junction formed at the interface between p-doped and n-

doped semiconductors (i.e. p-n junction). When these materials are put in contact, the carriers

flow under the driving force of diffusion until the chemical potential on the two sides

equilibrates, which means that the position of the Fermi level is at the same level at both p and n

sides. A concentration gradient causes electron to diffuse from the n-type semiconductor to the

p-type semiconductor and holes to diffuse in the opposite direction. This diffusion establishes a

quasi-neutral region in each semiconductor and sets up a built-in electric field in the depleted

region at the interface, Figure 2-1. The width of the depletion region in a given junction depends

on the acceptor and donor concentration in a p- and n-type semiconductor, respectively. The loss

of charge neutrality at the junction can be also expressed in terms of the potential, which is

referred to as built-in voltage. Built-in voltage (Vbi) is simply the difference of the Fermi levels

in p- and n-type semiconductors before they were joined. Electrons going from the n-type to the

p-type semiconductor and holes going in the opposite direction are called minority carriers. The

Vbi is a barrier for minority carrier injection and eventually hinders the minority carrier injection

at the equilibrium condition, Figure 2-1.

In a solar cell device, thick enough to absorb a significant portion of the incident light,

both depleted and quasi-neutral regions are typically present. While in the depletion region the

dominant carrier transport mechanism is drift (determined by carrier mobility and the magnitude

of the electric field), in the quasi-neutral region carrier diffusion is the dominant mechanism

(determined by carrier mobility and recombination lifetime). Efficient transport mechanism (both

diffusion and drift) is required for optimal solar cell performance.

10

Figure 2-1 Schematic of p-n junction. (a) When a p-type and an n-type semiconductor are put

into physical contact, carriers diffuse due to a concentration gradient at the interface to form a

depletion region. A built-in electric field will form at the depleted region that eventually blocks

the minority carrier diffusion when equilibrium is established. (b) Band diagram of a p-n

junction at equilibrium condition. Built-in voltage (Vbi) is simply the difference of the Fermi

levels in p- and n-type semiconductors before they were joined. The Vbi is a barrier for minority

carrier injection (np and nn are minority and majority electrons, respectively, with the same

applying for the holes). Carrier transport in the central depletion region is electric-field driven

(i.e. drift) and in the far left/right quasi-neutral region it is based on diffusion. Adapted from

MIT’s OpenCourseWare lecture 18 on the p-n junction basics.

Within the depletion region there is a built-in potential, or an electric field, which

immediately sweeps the photogenerated holes and electrons into opposite directions. This means

that the carriers generated within the depletion region or those that have diffused into the

depletion region will be pushed by the field into the bulk of the material (holes into the p-side

and electrons into the n-side). When a p-n junction is under illumination, electron-hole pairs are

generated. The electrons and holes will diffuse upon dissociation and if they reach the depletion

region (i.e. the electric field) they will be swept into the n- and p-type semiconductor,

respectively. If we connect p-side to n-side (i.e. make a short circuit), then these carriers will

flow at zero applied voltage (just under built-in voltage). This means we will observe short-

b

a

11

circuit current called photocurrent (Iph). If we isolate the contacts the carriers accumulating on p

and n sides would eventually lead to a potential build-up (effectively lowering built-in voltage),

which would increase the dark current through the diode cancelling the photocurrent. The

potential at which the net current across the junction is zero, is called open circuit voltage (Voc).

Solar cell performance is characterized under one sun simulated illumination, typically using the

American Society for Testing and Materials AM 1.5 G standard spectrum with an input power of

100 mW cm-2.21 To evaluate the solar cell performance under AM 1.5, current density-voltage

sweep measurements (JV) are required, Figure 2-2. The blue curve is a typical JV curve of a

solar cell under AM 1.5 illumination. Voc is the voltage at zero current and Jsc (short-circuit

current density) is the current density obtained at zero bias (as discussed above). The point at

which current density-voltage product is maximized is called the maximum power point (MPP)

which has related current density (JMPP) and voltage (VMPP). The slope of the JV curve at Jsc and

Voc is called shunt (Rsh) and series resistance (Rs), respectively. Smaller series resistance and

larger shunt resistance, bring JMPP and VMPP closer to JSC and VOC, respectively and increase the

squareness of the JV curve. Fill factor (FF) is a measure of JV curve squareness (maximum value

is therefore 1) and can be calculated using Equation 2-1:

𝑭𝑭 = 𝑱𝑴𝑷𝑷×𝑽𝑴𝑷𝑷

𝑱𝑺𝑪×𝑽𝑶𝑪 2-1

The overall power conversion efficiency of a solar cell (𝜂) can be calculated using Equation 2-2.

It depends on fill factor JSC and VOC and input power (100 mW cm-2 for AM 1.5 illumination).

𝜂 =𝐽𝑆𝐶×𝑉𝑂𝐶×𝐹𝐹

𝑃𝑖𝑛 2-2

External quantum efficiency (EQE) which measures the probability that a given incident

photon (with a particular wavelength λ) creates an electron that is successfully delivered to the

external circuit is another important figure of merit for solar cells. Non-optimal (

12

probability of charge extraction by an absorbed photon at a given wavelength and is called

internal quantum efficiency (IQE).

Figure 2-2 Representative current-voltage curve. Characteristic points of interest are short

circuit current density (Jsc), open circuit voltage (Voc), and maximum power point (MPP) and its

related current density and voltage. Slope of the JV curve at the JSC and VOC is called shunt (Rsh)

and series resistance (Rs), respectively. The area of the highlighted square to the area of the

square defined by Jsc and Voc is called fill factor. Reprinted by permission from Macmillan

Publishers Ltd: Nature, copyright 2012.22

The maximum achievable power conversion efficiency (PCE) for a single junction solar

cell is ~ 31% using an absorber material with 1.37 eV bandgap.23,24 One approach to determine

the maximum achievable PCE is solely employing fundamental thermodynamic considerations

when analyzing the performance of a solar cell.23 This approach relies on the bandgap of the

semiconductor absorber and temperature of the sun, i.e. the shape of the solar flux which is

approximated by a blackbody source radiation at 6000oC. This approach was demonstrated in a

paper by Shockley and Queisser in 1960. The other approach was developed by Henry in 1980.24

This approach looks at specific mechanisms that limit the solar cell performance. Upon light

exposure, photons with the energy smaller than the bandgap of the semiconductor absorber

cannot be absorbed. The photons that have energies larger than the bandgap of the active

13

material lose their extra energy as they quickly relax down to the energy of the active material

bandgap. The extra energy of the absorbed photon is lost in the form of heat. Furthermore, the

work done by an absorbed photon is less than the energy of that photon. In other words, the

energy at which the maximum power can be extracted from a solar cell is less than the bandgap

energy of the active material of the solar cell. However, by employing multiple active materials

having different bandgaps (i.e. multi-junction solar cell), the maximum achievable PCE will

increase. In a hypothetical case of an infinite number of junctions, the limit for the PCE would be

more than 68%.

2.2 Colloidal Quantum Dots

Quantum dots are semiconducting nanoparticles that are small enough to experience quantum

confinement effects. Quantum confinement occurs when the dimensions of the nanostructures

are of the same magnitude as the electrons wavelength. At this state, the energy levels of

electrons become discrete; therefore, the band gap can be varied by changing the sizes of

nanostructures. This enables tuning the optoelectronic properties of QDs via size control.

Colloidal quantum dots (CQDs) are quantum dots that are synthesized in solution. The ability to

manipulate the semiconducting properties of the CQDs, along with their relatively rapid and

cost-effective synthesis make them promising candidates for applications in photovoltaics.

Below, I will discuss in detail the synthesis of CQDs, followed by a discussion on the

process of formation of a solid network of CQDs for incorporation into photovoltaic devices.

2.2.1 CQD synthesis

Similar to conventional solution-based synthesis, the CQD synthesis process is driven by

temperature control and concentration changes. The temperature should be low enough to

promote formation of nanocrystals from solution and high enough to allow for growth of initial

nuclei and subsequent rearrangement of atoms.

CQD precursors are dissolved in a solvent at levels higher than the threshold for

nucleation. Therefore, plenty of small nanocrystals are formed initially. Consequently, the

concentration of precursors drops under the threshold of nucleation and the remaining reagents

contribute only to the growth of existing nuclei, maintaining a constant density of nanocrystals.

As the solution gets depleted from precursor species, the nanocrystals undergo Ostwald ripening.

14

During this process, CQDs of different sizes exchange materials until becoming monodispersed.

The CQDs’ nucleation and growth process is explained by La Mer,25,26 and is depicted in Figure

2-3.27 The final CQD sizes can be determined by controlling the temperature of synthesis,

concentration and reactivity of precursors, and duration of the ripening stage.28

Figure 2-3 Schematic of the LaMer model. The model describes the nanocrystals nucleation

and growth over time. Adapted with permission from The Royal Society of Chemistry, 2015.27

As discussed above, by controlling the CQDs’ dimensions, thus varying their band gap

energy, their optoelectronic properties can be tuned.28 According to the Brus equation (Equation

2-3),29 by reducing the CQDs’ diameter, their band gap energy increases, which results in a

blueshift in the absorbance/emission peaks of the CQD, Figure2-4.28

𝑬 = 𝑬𝒈 + 𝒉𝟐

𝟖𝒓𝟐 (

𝟏

𝒎𝒆∗ +

𝟏

𝒎𝒉∗ ) 2-3

15

Figure 2-4 Absorption spectra of CQDs at different sizes, ranging from 3 to 10 nm in

diameter. As the size of the CQDs increases, the bandgap energy decreases and

absorption/emission spectra redshift. Reprinted with permission from American Chemical

Society, 2011.28

Lead-based CQDs (i.e. lead sulfide) is the focus of this thesis. Lead sulfide’s large Bohr

radius and small bulk bandgap (

16

A common method of film formation, which is also employed in this thesis, is spin-

coating: after the deposition step, the substrate is spun at speeds of 1000-5000 rpm. This process

is repeated until the desired film thickness is achieved.31–33 Compared to the other film formation

techniques, such as drop-casting and dip-coating, spin-coating results in higher-quality films.

This is because of the consistent film thicknesses achieved from a constant spinning speed and

facilitated solvent evaporation during substrates revolutions.

The CQDs spin-coating is often coupled with a ligand-exchange process to enhance the

electronic transport properties of the CQD film. As-prepared, CQDs are normally capped with

long organic ligands to ensure their dispersion in the solution (e.g. oleic acid). However, the

presence of these ligands in CQD films prevents formation of a highly packed CQD network.

Larger CQD spacing deteriorates carrier transport, as the electrons’ wavefunction overlap to a

lesser extent between the dots. Therefore, to decrease the dot-to-dot spacing, after the formation

of a layer of a CQD network, a solution of shorter ligands is deposited onto the substrate. The

solution of shorter ligands is highly concentrated to avoid re-dispersion of CQDs; and the shorter

ligands have higher binding affinity to CQDs compared to longer ligands to be able to replace

them effectively (e.g. mercaptopropionic acid, tetrabutylammonium iodide, or ethanedithiol).

The process is called layer-by-layer deposition (LBL). LBL deposition of CQD solutions

followed by deposition of the short ligand solution is repeated until the desired film thickness is

achieved.

In the LBL process, the ligand exchange step is performed after the formation of the film,

i.e. solid-state ligand exchange. However, alternatively this process can be performed while the

CQDs are still dispersed in solution. This approach (solution exchange) has shown to be very

effective in yielding high-quality CQD films in-part by improving mono-dispersity.19,34,35 During

this process a ligand-containing polar solvent will be added to the oleic acid-capped CQDs in a

non-polar solvent. CQDs will undergo a ligand exchange and transfer into the polar solvent

subsequently. The process enables deposition of thick CQD films (solar-relevant thicknesses of

hundreds of nanometers) in a single step and therefore improves manufacturability

.

17

2.3 CQD Photovoltaic Devices

Power conversion efficiency of CQD PV devices has progressed rapidly from sub-1% to more

than 11%, in a decade of research.19,30 This progress is, in part, due to advances in the CQD PV

device architecture. Schottky architecture was among the first developed.36,37 It comprises of a p-

type semiconductor that has been sandwiched between a transparent conductive oxide such as

indium-tin oxide (ITO) and a shallow-work function metal to form a rectifying junction. The

rectifying junction formed at the semiconductor-metal interface, results in a depletion region and

a built-in electric field that separates and drives out the carriers. This architecture, particularly

suffers when the thickness of the active layer increases beyond the depletion region width to

maximize the absorption. In this case, most of the absorption and charge generation happen far

from the closer-to-the-back-contact depletion region, i.e. at the quasi-neutral region. Given the

minority carrier diffusion length in today’s QD solids, the charge extraction won’t be effective

anymore. Therefore, the power conversion efficiency for the Schottky devices has been limited

to sub-6% values.38

To solve the limitations of Schottky devices and shifting the charge separation closer to

the illumination side of the device, the depleted-heterojunction architecture was designed.39 This

architecture flips the rectifying-junction side to the front transparent conductive oxide, in which

the CQD layer is deposited atop of a shallow work function electron acceptor such as titanium

dioxide (TiO2) or zinc oxide (ZnO). The rectifying back contact is replaced by a deep work

function metal (e.g. gold or silver) or metal-oxide (e.g. molybdenum oxide (MoO3)) to form an

ohmic contact. Higher PCEs (>9%) have been reported for devices with this architecture

compared to the Schottky architecture.40 This architecture is employed in the research presented

in the Chapter 4 of this thesis. To further improve charge collection in an increased device

thickness, bulk-heterojunction branched-out of the depleted-heterojunction architecture.41–43 In

this architecture, instead of using planar electron acceptor, nanostructured electron acceptor, such

as nanowire arrays, was employed. Although bulk-heterojunction architecture inherently holds a

significant theoretical advantage over the depleted-heterojunction and has demonstrated

improved current density experimentally, it never surpassed the performance achieved using the

depleted-heterojunction architecture. This, in part, is due to the non-uniformity of the

nanostructured electron acceptor over a large area, which results in lower FF and Voc.42

18

Better control over the doping concentration of the QD layer enables replacing the

typically n-doped oxide electron acceptor with an n-doped QD layer in the quantum-junction

architecture. In this case, the junction will form between two appropriately-doped p- and n-type

QD layers.40,44–46 This architecture becomes particularly interesting as the size of the CQD

increases for infrared solar harvesting. For the dots large-enough to harvest light at 1100 nm

wavelength and beyond, in the depleted-heterojunction architecture, the injection into the

electron acceptor layer, such as TiO2 and ZnO deteriorates and therefore limits the overall

performance of the solar cell.47 This limitation can be avoided in the quantum-junction

architecture. This architecture is the focus of the Chapter 6 of this thesis.

A blend of the depleted-heterojunction and quantum-junction architectures has also been

reported (currently the most efficient);19 creating a p-i-n structure by sandwiching a thick

intrinsic (or mildly n-doped) QD layer between a thin p-doped layer and an n-doped transparent

conductive oxide.19,40,44,48 This architecture has been utilized in the research presented in

Chapters 3 and 5 of this thesis.

2.4 Limitations of CQD PV devices

In order to achieve highly-efficient CQD PV devices, advances in device architecture should be

combined with research on better-quality CQD films. CQDs are nanoparticles with extremely

high surface-to-volume ratios. This will significantly increase the probability of surface trap

formation. These surface states, if left unpassivated, will have detrimental effects on solar cell

performance. The as-synthesized nanocrystals tend to have an excess of metal atoms on the

surface.28,49 The non-stoichiometric charge imbalance on the surface, acts as a trap and should be

passivated with anionic ligands. The choice of the ligand has a significant effect on the band

alignment of the CQD films and can even introduce mid-gap states if chosen inappropriately.

Complete ligand coverage is also challenging. Furthermore, during the ligand exchange process

(i.e. exchanging the non-conductive, as-synthesized oleic acids with much shorter and

conductive ligands) the nanoparticles are prone to fusion and agglomeration.19,34 The

agglomerates can act as low energy trap states for the carriers, limit the overall device

performance. While advances have been made to passivate the excess metal sites, it remains

highly challenging to passivate the dangling bonds associated with the sulfur sites.48 In Chapter 3

of this thesis, I discuss an effective strategy to replace the uncoordinated surface sulfur sites and

19

hence improve the overall passivation. Undesired non-radiative recombination arises from these

unpassivated surface sites and can be eliminated by adding a shell of a larger bandgap

semiconductor.50 While shelling is effective in reducing the non-radiative recombination, if

implemented improperly, it can hinder the charge transport and inter-particle coupling.

Studies suggest improving the mobility of the CQD solids will not have an appreciable

effect on the solar cell performance unless the trap density is reduced significantly.51 In the case

of a high trap-density film, higher mobility means carriers can reach trap sites faster, hence

recombine faster, reducing the chance of escaping from the shallow traps. Therefore, improving

the diffusion length in the CQD films, the key limiting factor in CQD PV performance, only

happens when the trap density in CQD solids is significantly reduced.

In the next chapter, I highlight a surface treatment to reduce the surface trap density in CQD

solids. The treatment results in a 35% increase in diffusion length and a 15% increase in the

PCE. The PCE (certified 9.9%) was the world’s record at the time of publication. The work

presented in Chapter 3, showcases the significance of the discussed limitations and highlights the

promise this direction holds for the future of CQD PV devices.

20

Chapter 3 Surface Passivation Using Molecular Halide Treatment

Surface trap density is considered to be the dominant limiting factor in the CQD-based solar cell

performance. This chapter focuses on CQDs with 1.4 eV bandgap and describes an experimental

procedure to replace the otherwise unpassivated sulfur sites using molecular halide treatment.

The thicker active layer enabled by better passivated CQDs resulted in a certified 9.9% power

conversion efficiency, world record from June 2015 to July 2016.

Section 3.1 to 3.4 contain material from the equally contributed (co-first authored): X.

Lan, A. Kiani, O. Voznyy, F. Pelayo García de Arquer, A. Saud Abbas, Gi-Hwan Kim, M. Liu,

Z. Yang, G. Walters, J. Xu, M. Yuan, Z. Ning, F. Fan, P. Kanjanaboos, I. Kramer, D.

Zhitomirsky, P. Lee, A. Perelgut, S. Hoogland, E. H. Sargent, Passivation Using Molecular

Halides Increases Quantum Dot Solar Cell Performance, Advanced Materials 28, 299-304, 2016.

Figures are reprinted with permission, copyright 2016 WILEY-VCH Verlag GmbH & Co.

KGaA, Weinheim. I was co-first author and contributed to all experimental design, halide

surface treatment, device fabrication and characterization, data interpretation and writing.

SCAPS simulation and XPS were performed by O. Voznny. PCE Certification was done with the

help of X. Lan. Diffusion length and trap density measurements were performed with the help of

F. Pelayo García de Arquer. Experimental details for this chapter can be found in Appendix A,

page 83-87.

3.1 Introduction

Colloidal quantum dot (CQD) solar cells have advanced significantly in performance over the

past decade.18,44,52,53 Recent progress in CQD solar cell power conversion efficiency (PCE) has

been accompanied by improvements in device stability,40,44,54 which further increases their

promise.

Improved light absorption and charge collection are the two major requirements for

increasing solar cell efficiency.55,56 Photon management techniques, such as nanoscale plasmonic

inclusions57,58 and hierarchically structured electrodes,52,59 have demonstrated enhanced light

absorption in CQD solar cells. Significant improvements in charge extraction have been

21

achieved as the device architecture progressed from Schottky-junction to depleted

heterojunction,36,37,39 bulk heterojunction41,42,60–62 and quantum junction devices.40,44,45,63

While notable advances in performance have been accomplished through the

aforementioned efforts, further progress still relies on addressing the quality of the light-

absorbing film itself.64–68 In CQD solids, the large surface-to-volume ratio results in added

opportunities for electronic defect formation. Increased recombination losses risk compromising

charge collection if not suitably addressed. Ligand-assisted surface passivation is a proven

strategy to combat these losses. A range of short-chain-organic (e.g. mercaptopropionic acid,

ethanedithiol, thiocyanate) and inorganic ligands (e.g. Cl-, Br- and I-) have been explored for this

purpose.37,69–73 These surface passivation schemes have been carried out based on solid-state

ligand exchange processes driven by the difference in CQD-ligand binding strengths.74 Ideally,

the ligand exchange approach will preserve the passivation state existing prior to the exchange

process (i.e. the passivation state of long-chain organic-ligand-capped CQDs). However, it is

unlikely to repair any trap states already present in the as-synthesized CQDs. New passivation

strategies, applied before the solid-state ligand exchange, could thus potentially further improve

CQD characteristics.

Here I report a solution-based passivation scheme that features a redox reaction (PbS + I2

= PbI2 + S) occurring at the PbS CQDs surface in the presence of iodine molecules (I2) dissolved

in a nonpolar solvent. Solution treatment allows for increased control over halogen reactivity,

helping to avoid overtreatment and improving overall passivation of CQDs in solution.75,76

Better-passivated CQDs resulting from the optimized reactivity show a longer carrier diffusion

length in film. Hence, we were able to fabricate photovoltaic devices with thicker active region

without compromising the charge collection. This optimization ultimately led to CQD solar cells

with certified power conversion efficiency (PCE) of 9.9%. This is the highest certified PCE for

CQD solar cells.

3.2 Molecular Halide Passivation

X-ray photoelectron spectroscopy (XPS) confirms the adsorption of iodine on the surface of PbS

CQDs following I2 treatment (Figure 3-1a). The absorption spectra of the PbS CQDs before and

after I2 treatment (Figure 3-1b) show no appreciable change in the exciton peak position,

suggesting that CQD size is not affected by the process. At the same time, the

22

photoluminescence spectra of the dots show that the full-width-at-half-maximum is preserved,

proving that ensemble monodispersity remained intact. The photoluminescence quantum yield

(PLQY) of the dots with and without I2 treatment are 19% and 15%, respectively, an indication

of improved surface passivation.

Figure 3-1 Characterization of PbS CQDs following the I2 treatment. (a) I 3d peak

showcasing the successful incorporation of iodine into the dots. (b) Absorption and

photoluminescence spectra of untreated and I2-treated PbS CQDs in the solution phase. No

changes in monodispersity are observed. The increase of photoluminescence intensity after I2

treatment is consistent with improved surface passivation.

The amount of I2 applied to the CQDs was systematically optimized in order to avoid

overtreatment. In developing the I2-treatment process, we have found that the quantity of applied

I2 plays a crucial role. Under the optimized I2 treatment conditions, the CQDs maintain good

monodispersity (Figure 3-2a). Heavier than optimized I2 treatment leads to several undesirable

effects, such as stripping of an undesirably large number of ligands from the QD surface, causing

QDs to fuse (Figure 3-2b). Furthermore, it can contribute to excessive n-type doping due to

changes in stoichiometry and leading to self-compensation via surface reconstruction and trap

formation.77,78 The optimal I:Pb ratio in XPS is found to be 5% which corresponds to ~15-25%

of available surface S sites.64

23

Figure 3-2 TEM image of PbS QDs following different I2 treatment conditions. (a)

optimized I2 concentration and (b) heavier I2 treatment showing the fusing of QDs.

XPS confirms that I2 molecules break up and bind to the PbS CQDs’ surface (Figure 3-

3a) during the process without affecting surface lead significantly (Figure 3-3b). The narrower

XPS peaks for sulfur after I2 treatment (Figure 3-3c) indicates a more homogeneous local

environment for sulfur atoms. This may arise from the elimination of surface sulfur sites with

reduced coordination by replacing them with iodine, considering that the eliminated sites

possessed a distinct XPS signature compared to fully-coordinated S in the bulk of PbS.

Figure 3-3 XPS measurements of the key elements at different stages of ligand exchange.

(a) Iodine signal before and after I2 and TBAI exchanges showing that I2 molecules break-up and

bind to the surface, (b) Pb signal showing little change upon iodine treatments along with

decomposition into bulk and surface components, (c) reduction in sulfur signal strength upon I2

overtreatment (40% iodine incorporation compared to 5% in CQDs optimized for device

performance) showing that molecular iodine displaces surface sulfur (the signal intensities are

normalized to Pb).

50 nm 100 nm

24

3.3 Molecular Halide Passivated CQD-Based Solar Cells

In the next step, I utilized the improved passivation by incorporating CQDs into solar cells with

an improved quantum junction architecture (Figure 3-4a).40,45 Tetrabutylammonium iodide

(TBAI) exchanged PbS CQD film (7-8 layers), serving as the n-type active region, was deposited

on ZnO-nanoparticle-coated indium-doped tin oxide (ITO) glass substrates. This is followed by 2

layers of p-type CQD film prepared via ethanedithiol (EDT) exchange.44 It should be noted that

the bandgap of the CQDs used in this study was blue-shifted compared to previous reports in

order to improve charge injection from the PbS into the ZnO electrode and increase the

VOC.40,44,79

Figure 3-4 Device architecture and performance. (a) Schematic of the device structure. A

TBAI-exchanged CQD film is deposited on top of a ZnO electrode, followed by 2 layers of

EDT-exchanged CQDs and gold as the top contact. (b, c) Cross-sectional SEM images of a 200

nm CQD thick control device and a 220 nm thick I2 treated PbS CQD device. (d) EQE of the

control and I2-treated CQD devices. Fabry-Perot effects visibly change the shape of the EQE

compared to single-pass absorption spectra. (e) J-V characteristics under simulated AM1.5G

illumination for the control and I2-treated devices. A PCE = 10.18% is obtained for molecular-

iodine treated dots as a result of increased Jsc and FF.

25

We systematically optimized the thickness of both I2- and non-I2-treated CQD devices

(Figure 3-5a). By using the I2-treated CQDs throughout all device active region, we were able to

increase the device thickness by one extra CQD layer without losing the fill factor (Figure 3-

4b,c), which is an indication of improved passivation that yields longer carrier diffusion length.

External quantum efficiency (EQE) spectra shown in Figure 3-4d confirm that the extra

photocarriers generated in a thicker device can still be successfully extracted (double-pass

absorption of the devices can be seen in Figure 3-5b).

Figure 3-5 CQD devices thickness study and absorption profile. (a) Thickness-dependent

device performance for both I2- and non-I2-treated CQD devices. (b) Double-pass absorption of

PbS CQD devices for control and molecular-iodine treated CQDs, collected with an integrating

sphere. The thickness of the PbS photoactive region is respectively 200 nm and 220 nm for

control and I2-treated CQDs. The internal quantum efficiency at 820 nm was estimated in both

cases by dividing the EQE by the double pass absorption at this wavelength.

Current-voltage characteristics of the representative devices under simulated AM1.5G

illumination are shown in Figure 3-4e, with the relevant figures of merit and statistical analysis

summarized in Table 3-1. For I2 treated devices, a PCE of 10.2% was measured in our lab with a

certified PCE of 9.9% (Figure 3-6). The 10.2% is significantly higher than the PCE of 9.2%

obtained for non-I2-treated control samples (see Table 3-1). The extra PCE point is the result of

improved Jsc and FF.

a b

26

Table 3-1 Static figures of merit for control and I2-treated devices. Statistics is based on 49

different devices.

Figure 3-6 Figures of merit for a certified molecular iodine CQD device. It was measured by

an accredited PV calibration laboratory (Newport Technology and Application Center – PV

Lab).

Voc (V) Jsc (mA cm-2) FF (%) PCE (%)

Control devices 0.632 ± 0.004 21.40 ± 0.31 69.47 ± 0.98 9.24 ± 0.13

Molecular iodine devices 0.639 ± 0.004 22.28 ± 0.27 72.37 ± 0.94 10.18 ± 0.20

27

3.4 Mechanistic Study of Physical Origin of Improved Performance

In order to gain insight into the physical origins of the improved performance, we built an

optoelectronic device model (Figure 3-7) that takes into account the electron affinity of CQDs

with different ligand treatments, absorption profiles, thicknesses, and carrier diffusion lengths

starting from previously reported values for TBAI-treated films.40,61,80 The improvement of the

EQE in the longer wavelength region can be fully attributed to the increased device thickness

(Figure 3-7c). However, for a constant photocarrier diffusion length, the increased device

thickness can eventually result in a deterioration of the charge collection. Holes photoexcited

near the front electrode have the longest distance to travel and therefore experience this

limitation first, resulting in an EQE drop in the blue region.

Figure 3-7 Optoelectronic model of the origin of improved device performance. (a) Band

diagram of the graded device at open-circuit voltage conditions. Ec, Fn, Fp and Ev describe

respectively the energy level of the conduction band, quasi-Fermi level of electrons and holes,

and valence band across the length of the device (x). (b) J-V curves and (c) Simulated EQEs

demonstrating the effect of the device thickness and carrier diffusion length, LD, on the device

performance and EQE respectively.

28

Figure 3-7b,c show this phenomenon for the case when the diffusion length is chosen to

differ by ~50% (to visually amplify the difference in EQE). Therefore, the improvement in the

EQE over all wavelengths observed experimentally for I2 treated dots can only be justified by

both thicker active region and longer effective diffusion length, given that mobility has not been

affected (Figure 3-8).

Figure 3-8 Measurement of mobility in the I2-treated and control films. The mobility was

measured on TBAI ligand exchanged films using a photoluminescence-based method that

utilizes smaller-bandgap inclusions as carrier transport reporters.[46] Mobility is directly

proportional to the line slope and is the same in both types of samples.

Experimentally observed differences are subtler than simulated 50% in Figure 3-7. The

increase in fill-factor and short-circuit current density, obtained experimentally, can be

reproduced with 80 nm and 110 nm carrier diffusion lengths for control and I2-treated films,

respectively (Figure 3-9). Such an increase in diffusion length allows for thicker devices while

maintaining efficient carrier collection, leading to a full power point improvement in device

performance.

29

Figure 3-9 SCAPS modeling of control and molecular iodine treated CQD photovoltaic

devices. (a) Current-voltage characteristics under AM 1.5G illumination. (b) EQE of control and

I2 treated CQD devices. In the modelling CQD film thicknesses obtained from SEM cross-

sectional images were used (200 nm and 220 for control and I2 films, respectively).

To further validate the predictions of the theoretical model, we measured experimentally

the carrier collection efficiency (𝜂) of the I2-treated and control devices (Figure 3-10a). We used

an analytical model also employed in prior reports to fit the experimental data,81

𝜼(𝑽) = 𝑰𝑸𝑬(𝝀)𝑱𝑯(𝑽)−𝑱𝑳(𝑽)

𝑱𝒔𝒄,𝑯−𝑱𝒔𝒄,𝑳 3-1

where IQE(λ) is the internal quantum efficiency at short-circuit condition under

monochromatic illumination (λ=820 nm), and JH and JL are the currents at two different powers

for the same illumination (see experimental section for more details). By fitting 𝜂 one can

estimate the diffusion length of the CQD films. The experiment confirms that the molecular

iodine treatment increases diffusion length substantially, from 85 to 115 nm, which is in a good

agreement with the diffusion lengths obtained from optoelectronic device simulations. Given the

similar depletion widths of the control and I2 treated devices (~85 nm at maximum power point,

measured using capacitance-voltage spectroscopy), I can pinpoint diffusion length to be

responsible for the improved Jsc and FF.

Since the carrier diffusion length is primarily determined by the electronic trap state

density in the CQD solid, we sought to verify how the trap density is affected by molecular

iodine treatment by using transient photovoltage measurements.64 Based on the photovoltage

studies, I observed a two-fold reduction in the trap density for I2-treated dots at maximum power

30

point condition (Figure 3-10b). From 𝐿𝐷 = √𝐷𝜏, where the lifetime 𝜏 is inversely proportional to

trap density,17,51 this predicts a ~√2 factor of improvement (i.e. ~40% increase) in the diffusion

length, which is in a good agreement with the 35% increase obtained experimentally. The

reduced trap density is also in agreement with the increase of VOC in I2-treated devices.

Figure 3-10 Experimental determination of diffusion length and improved passivation. (a)

Collection efficiency and diffusion length for control and molecular I2-treated dots. A 35%

increase is observed for I2 treated devices, (b) density of trap states extracted from VOC decay

measurements for control and I2-treated films. The better passivation of molecular iodine leads to

a two-fold reduction in the trap density at maximum power point conditions, which is in

agreement with the observed increase in the diffusion length.

In conclusion, I demonstrated that molecular iodine improves the passivation of traps in

PbS CQDs; the traps that haven’t been addressed using prior solid-state ligand exchange

treatments. A two-fold decrease in trap density in the molecular iodine-treated CQD film leads to

improved diffusion length and a full power point increase in device performance. This

improvement resulted in a certified PCE of 9.9%, a new record for colloidal quantum dot solar

cells.

PV industry is dominated by the silicon solar cells. Currently, silicon cells possess better

efficiency and stability compared to solution-processed PV devices (e.g. CQD PVs). However,

their efficiency has only seen a marginal 10% increase over the past two decades. Silicon,

because of its bandgap, is transparent to the photons at 1100 nm wavelength and beyond.

Considerable portion of the solar energy resides in that unabsorbed infrared region. In the next

31

chapter, the effectiveness of the CQDs to harvest infrared photons (1100 nm in wavelength and

beyond) has been demonstrated. Chapter 4 discusses a ligand exchange strategy that enables 1

eV bandgap CQDs to maintain efficient charge injection into the TiO2 electron acceptor. The

following chapter highlights the importance of a combined strategy to achieve working devices

employing a smaller bandgap CQDs. This involves improving surface passivation and hence the

diffusion length on one hand, and reducing agglomeration and hence low-energy trap states on

the other.

32

Chapter 4 Infrared Solar Harvesting via Coupling Enhancement and Agglomeration Suppression

Research discussed in the previous chapter showcased the importance of reducing surface trap

density in order to improve the diffusion length in quantum dot solids. Having longer diffusion

length results in efficient charge extraction from thick CQD films and is the path toward

realization of highly efficient single-junction CQD PV devices. The present chapter highlights a

new strategy for the CQDs to be used for infrared solar light harvesting. It discusses a combined

ligand exchange strategy that provides surface passivation for 1 eV bandgap CQDs and

significantly reduces agglomeration. Both aspects of the combined ligand exchange scheme, are

discussed in this chapter and demonstrated to be critical in achieving working devices based on

the 1 eV bandgap CQDs. Silicon solar cells and other highly efficient PV systems (e.g.

perovskite solar cells) are transparent to the infrared portion of the solar spectrum due to their

bandgap. This limitation opens a potential niche for the CQD solids to harvest the otherwise

wasted photons’ energy as the CQDs’ bandgap can be tuned via quantum size effect. The CQD

technology is shown to offer up to 7 additional power points to the best silicon cells. Given the

marginal increase of 2% (in absolute PCE) for the silicon cells over the past 20 years and

ubiquity of the silicon solar cells, this technology has a substantial potential to revolutionize the

energy industry.

Section 4.1 to 4.5 contain material from the equally contributed (co-first authored): A. H.

Ip, A. Kiani, I. J. Kramer, O. Voznyy, H. F. Movahed, L. Levina, M. M. Adachi, S. Hoogland,

E. H. Sargent, Infrared Colloidal Quantum Dot Photovoltaics via Coupling Enhancement and

Agglomeration Suppression, ACS Nano 9, 8833-8842, 2015. Figures are reprinted with

permission, copyright 2015, American Chemical Society. I was co-first author and contributed to

all experimental design, device fabrication and characterization, data interpretation and writing.

Density functional theory simulation was performed by O. Voznny. Detailed balance modeling

was performed by I. J. Kramer. XRD and PL measurements were performed with the help of A.

Ip. Experimental details for this chapter can be found in Appendix A, page 87-90.

4.1 Introduction

Silicon solar cells dominate solar photovoltaic devices (PVs),82 with laboratory power

conversion efficiencies (PCEs) that reach 25%.83 Recently, solution-processed perovskite solar

33

cells have gained significant attention as their certified efficiencies have rapidly increased to

over 20%.84–87

These solar cells cannot harvest large portion of the available solar spectrum. There is a

large opportunity to improve spectral utilization as half of available photons reside in the infrared

(IR)88. High-efficiency III-V multi-junction cells absorb the IR region of solar spectrum89 but are

costly and require complex concentrator and tracking systems.

Colloidal quantum dots (CQDs) are a solution-processed material that can potentially

overcome these limitations. CQDs’ bandgap can be tuned via the quantum size effect. The

absorption onset can be simply and precisely tuned across the entire solar spectrum.55 Solution

processability of CQDs means that they are amenable to rapid, large-area fabrication on myriad

of substrates (e.g. lightweight and flexible).90,91 Performance advances in CQD PV have focused

on the passivation of trap states40,51,64,70 as well as development of new device structures for

efficient charge extraction and optical absorption enhancement.41,52,57 The efficiency of CQD

cells has thereby been advanced to certified efficiencies of more than 10%.44,48

4.2 Selecting Proper CQD Bandgap

We used a detailed balance approach to select a CQD bandgap that will combine well with a

variety of systems. For this study, we considered crystalline silicon and methylammonium lead

halide perovskite front absorbers because of their commercial ubiquity and rapid rise in

efficiency, respectively. In Figure 4-1a bromide perovskite refers to CH3NH3PbBr3, while iodide

perovskite refers to CH3NH3PbI3. Figure 4-1b shows the available short circuit photocurrent

density (Jsc) for CQD cells as a function of CQD bandgap for different front cells. Notably, there

is sufficient light available to current-match to perovskites in a series-connected tandem, while

the same is not true of the smaller-bandgap silicon cell. The theoretical open-circuit voltages

(Voc) are shown in Figure 4-1c, and the theoretically achievable PCE is shown in Figure 4-1d.

For the bromide perovskite maximal additive PCE can be achieved with CQDs with bandgap in

the range of 900 to 1300 nm, with a peak of potential 22 additive power points at 1100 nm. The

iodide perovskite peaks with approximately 12 power points for 1300 nm. For the silicon cell the

peak PCE is achieved at 1770 nm, although it also shows a local maximum at about 1300 nm (6

additive power point). Decreasing the bandgap in the broad-maxima regions results in potential

Voc loss with insufficient benefit to Jsc, due to the dips in the AM1.5G spectrum caused by

34

Figure 4-1 Detailed balance approach to select CQD bandgap. (a) Solar spectrum

(black curve) with portions accessible to labeled material: blue is a lead-bromide

perovskite, blue + pink is a lead-iodide perovskite and blue + pink + grey is crystalline

silicon. Everything to the right of each absorption edge represents the available solar

spectrum for any subsequent cells. (b-d) Detailed balance simulations for solar cells

accessing the solar spectrum transparent to the front cells identified in (a) according to the

same color scheme including (b) short-circuit current density (JSC), (c) open-circuit voltage

(VOC) and (d) power conversion efficiency (PCE).

atmospheric water absorption of sunlight. In light of these findings, we focused on developing

the 1.3 µm bandgap CQD solid with the goal to augment the performance of silicon and

perovskite PVs.

35

4.3 Depleted Heterojunction CQD device

We built our devices based on the fully-developed depleted heterojunction structure.39 Figure 4-

2a shows a cross-sectional SEM image of the device and Figure 2b shows 2D and its

corresponding 3D device schematics. At first, a very thin layer of the polyethylenimine

ethoxylated (PEIE) was spin-coated over the pre-etched indium-tin oxide (ITO) coated glass

substrate, after which I applied a 50 nm-thick layer of TiO2 nanoparticles (NPs).92 The

PEIE/TiO2 stack forms a low-temperature electron acceptor layer. I have found that PEIE not

only makes the charge collection at the electrode more efficient through lowering the barrier

between ITO and TiO2 NPs,93 but also improves the adhesion of the TiO2 and CQD layers

substantially and hence yields a better quality film. A CQD film with a thickness of over 500 nm

was used as the absorber. The thickness was optimized to maximize the absorption while

allowing efficient charge collection.

Figure 4-2 Depleted heterojunction CQD device. (a) Cross-sectional SEM image and (b)

schematic of the depleted heterojunction CQD device.

b a

36

Mercaptopropionic acid (MPA) was used to replace the poorly conductive, as-

synthesized oleic acid (OA) ligands in a solid-state ligand exchange process. A thin layer of

molybdenum oxide (MoO3) was used as the hole collector. Gold and silver (140 nm combined)

were used to form the top electrode. Figure 4-3 shows schematically the conduction band edge of

the PbS CQD as a function of its absorption onset and compares it to the electron affinity of

TiO294 for three different bandgaps. For the optimal CQD single junction bandgap (i.e. 1.3 eV to

1.4 eV bandgap), a favorable offset allows electron injection into the TiO2 even for a broad

distribution of bandedge states, while for 1770 nm CQDs (i.e. 0.7 eV bandgap) a large barrier

between the CQD conduction band and that of TiO2 prevents electron transfer. The conduction

bandedge of optimal ~1300 nm PbS CQDs, studies in this chapter, lies at the threshold of

efficient electron injection: carriers in tail states would not be extracted.

Figure 4-3 Effectiveness of carrier injection from CQD film into TiO2 electron acceptor

with respect to the CQD electron affinity. Comparing TiO2 and PbS CQD conduction band

with respect to CQD absorption onset (or electron affinity) at bandgaps of interest. Positive

values for the offset (towards yellow) indicate a favorable injection of electrons from CQD

conduction band into the TiO2 acceptor. Negative values for the offset (towards blue) means

there is a barrier for the electrons injecting into the TiO2.

37

4.4 Ligand Engineering

Choosing 1300 nm CQDs (~1 eV bandgap), we were particularly attentive to possible

agglomeration