Efficient Spray-Coated Colloidal Quantum Dot Solar Cells

by

Gabriel Moreno-Bautista

A thesis submitted in conformity with the requirements for the degree of Master of Applied Science Edward S. Rogers Sr. Department of Electrical and Computer Engineering University of Toronto

© Copyright by Gabriel Moreno-Bautista 2015 Efficient Spray-Coated Colloidal Quantum Dot Solar Cells

Gabriel Moreno-Bautista

Master of Applied Science

Edward S. Rogers Sr. Department of Electrical and Computer Engineering University of Toronto

2015

Abstract

Colloidal quantum dots (CQDs) offer the promise of low-cost, high-performance solar cells due to their ability to be synthesized and deposited from solution, which makes it possible for this material to be adapted to production-scale manufacturing protocols such as roll-to-roll (R2R) processing. Here we describe the design and implementation of a spray- coating process for the fabrication of CQD solar cells. We find that spray-coated films are morphologically superior to films that were fabricated using the conventional spin-coating method. Spray coating is found to be effective at removing an electronic trap caused by an organic impurity, enhancing the diffusion length of the CQD film and leading to an average power conversion efficiency (PCE) of 6.5%, which is higher than the average PCE of spin- coated cells (5.2%). We also show that the spray process can be adapted to R2R methodologies and can be used to fabricate efficient solar cells with unconventional form factors, such as surfaces with multiple dimensions of curvature.

ii Acknowledgements

First of all, I would like to thank my supervisor, Prof. Edward H. Sargent, for his support and guidance throughout my degree. He has assembled a first-class group of researchers who endeavour to solve the hardest problems facing society and I am grateful for the chance to have worked in such a fast-paced, supportive environment. This work would not have been possible without the guidance of Dr. Illan Kramer. I would like to thank him for his mentorship and friendship throughout my time in Prof. Sargent’s group. I would also like to thank James Minor for helping me get up to speed during my first two weeks in the group. I extend my gratitude to Damir Kopilovic, who shared his vast technical expertise in this project whenever it was needed; and Elenita Palmiano, who made sure that we were always well supplied with quantum dots. Thank you to all the people who gave valuable contributions to this project: Dr. Pongsakorn Kanjanaboos, Dr. Susanna M. Thon, Graham H. Carey, Dr. David Zhitomirsky, Dr. Oleksandr Voznyy, Dr. Sjoerd Hoogland, Joel A. Tang, Dr. Kang Wei Chou, and Prof. Aram Amassian. I would like to thank Remigiusz Wolowiec, and Dr. Larissa Levina for keeping the lab running smoothly through every adversity, and Jeannie Ing for always being on top of our mountain of paperwork. I thank the other members of Sargent Group that have not been previously mentioned, especially the following people: I thank Dr. Jeffrey McDowell for enlightening scientific discussions as well as for keeping me company during long hours in the lab; I thank Lisa Rollny for her chemistry expertise and being a pleasant cubicle neighbour; I thank André Labelle for his contagious jolly demeanour and for always lending an ear during troubled times; I thank Alex Ip, Brandon Sutherland, Chris Wong, and Valerio Adinolfi for acting as big brothers during my time in this group. I would like to thank my friends who help me keep moving forward, especially Mina Labib, Derek Govier, Oliver Twardus, Sarah LeBlanc, and Duncan Strathearn. I thank my parents, Gabriel and Aurora, my siblings, David and Aurora E., and Cuzco, for their continued unconditional support. I finally thank Sam Yang, for her companionship, encouragement, and for always motivating me to be the best person I can be.

iii Contents

Abstract ...... ii

Acknowledgements ...... iii

Statement of personal contributions and collaborations ...... vi

Acronyms ...... vii

List of Figures ...... viii

1 Introduction ...... 1 1.1 Motivation ...... 1 1.2 Colloidal quantum dots ...... 2 1.2.1 Colloidal quantum dot photovoltaics ...... 3 1.3 Thesis objectives ...... 5

2 Background ...... 7 2.1 Solar cell fundamentals ...... 7 2.2 Batch-scale processing methods ...... 9 2.2.1 Standard CQD solar cell fabrication ...... 11 2.3 Roll-to-roll manufacturing ...... 12 2.4 Dynamics of Spray Coating ...... 15 2.5 Spray-coated photovoltaics in literature and their limitations ...... 16

3 Spray coating of CQD films ...... 20 3.1 Spray-coating setup ...... 20 3.1.1 Nozzle types ...... 20 3.1.2 Integration and automation ...... 22 3.2 Spray-coating procedure ...... 24 3.2.1 Layer-by-layer spray process ...... 24 3.2.2 Effect of nozzle type used for CQD deposition ...... 25 3.3 Conclusions ...... 28

4 Material characterization of sprayed CQD films ...... 29 4.1 Cross-sectional composition ...... 29

iv 4.2 Film morphology and physical properties ...... 32 4.3 Dot-to-dot spacing and arrangements ...... 35 4.4 Conclusions ...... 36

5 Spray-coated CQD photovoltaic devices ...... 38 5.1 Improving electronic properties of films via spray coating ...... 38 5.1.1 Identification and elimination of electronic defect ...... 38 5.1.2 Minority carrier diffusion length ...... 41 5.2 Characterization of photovoltaic devices ...... 43 5.3 Conclusions ...... 45

6 Applicability to large-scale manufacturing ...... 46 6.1 Roll-to-roll process simulation ...... 46 6.2 Flexible devices ...... 48 6.2.1 Importance of flexible solar cells ...... 48 6.2.2 Flat versus curved spraying configuration ...... 48 6.3 Spraying solar cells on unconventional form factor substrates ...... 50 6.3.1 Application of solar cells onto surfaces with multiple curvatures ...... 50 6.3.2 Sprayed solar cells on spherical lenses ...... 51 6.4 Conclusions ...... 52

7 Summary ...... 53 7.1 Thesis findings and conclusions ...... 53 7.2 Future work ...... 54

References ...... 55

Appendices ...... 58 A. Fabrication procedures ...... 58 B. Material characterization procedures ...... 60 C. Optoelectronic characterization procedures ...... 62

v Statement of personal contributions and collaborations

The idea for spray coating CQD devices was originally conceived by Dr. Illan Kramer, who also performed the initial optimization of the setup. Damir Kopilovic built all the custom parts of the setup and also put together the control box. James Minor and Dr. Kramer fabricated most of the unoptimized spray devices. I fabricated and tested most of the photovoltaic devices fabricated using the Ikeuchi nozzle, and made the CQD films that were used for the material characterization. The spin- coated devices that were used to compare against spray-coated devices were fabricated and tested by all members of Prof. Sargent’s group over a period of a year and a half, and I performed the Welch’s t-test for the population comparison with spray coating. I obtained the photoluminescence data that was used for the diffusion length measurements, and Dr. David Zhitomirsky helped me fit the data to his diffusion length model. I conceived the experiments regarding the flexible and spherical device fabrication, characterization, and optimization, while Dr. Kramer conceived and carried out the roll-to-roll simulation with the rotating drum. Cross-sectional SEM and TEM images, as well as EELS analysis, were obtained by staff at the Canadian Centre for Electron Microscopy, while Dr. Kramer and I analyzed the resulting data together. Dr. Pongsakorn Kanjanaboos performed the morphological and mechanical studies using AFM. Dr. Susanna M. Thon performed GISAXS measurements and analysis. Dr. Kramer performed the electroluminescence measurements. Lisa Rollny and Dr. Jeffrey McDowell obtained FTIR and NMR data and analyzed it with the help of Mr. Carey, Dr. Kramer, and myself. Dr. Oleksandr Voznyy performed DFT simulations to confirm our hypotheses originating from that data. Elenita Palmiano synthesized all the quantum dots used in these studies, and also cleaned all the FTO substrates used for devices.

vi Acronyms

AFM Atomic force microscopy MoO3 Molybdenum trioxide

AM1.5 Air Mass 1.5 G solar spectrum MPP Maximum power point

Ag Silver NMR Nuclear magnetic resonance

Au Gold OPV Organic photovoltaics

BHJ Bulk heterojunction PbS Lead sulphide

CQD Colloidal quantum dot PCB Printed circuit board

DFT Density functional theory PCE Power conversion efficiency

EELS Electron energy loss spectroscopy PET Polyethylene terephthalate

EQE External quantum efficiency PL Photoluminescence

FF Fill factor P-V Power-voltage

FIB Focused ion beam R2R Roll-to-roll

FTIR Fourier transform infrared RS Series resistance spectroscopy

FTO Fluorine-doped tin oxide RSh Shunt resistance

GISAXS Grazing-incidence small-angle X- SEM Scanning electron microscopy ray scattering

ITO Indium tin oxide TiCl4 Titanium tetrachloride

J-V Current-voltage TiO2 Titanium dioxide

JSC Short-circuit current density VOC Open-circuit voltage

LD Diffusion length TEM Transmission electron microscopy

vii List of Figures

Figure 1.1: Maximum photovoltaic efficiency of devices with multiple number of junctions. 3 Figure 1.2: Depleted heterojunction CQD solar cell architecture...... 4 Figure 2.1: Solar cell operation under illumination as measured by a current-voltage sweep. 8 Figure 2.2: Stages of the spin-coating process...... 10 Figure 2.3: Illustrations of coating techniques compatible with R2R processing ...... 13 Figure 2.4: Organic photovoltaic architectures...... 17 Figure 3.1: Paasche airbrush, VLS model...... 20 Figure 3.2: Ikeuchi nozzle, BIMV8002S model...... 22 Figure 3.3: Full procedure of layer-by-layer spray deposition...... 23 Figure 3.4: SEM characterization of films fabricated using Paasche and Ikeuchi nozzles. ... 27 Figure 4.1: High-resolution TEM of CQD film cross-sections ...... 30 Figure 4.2: EELS cross-sections of spin-coated films (top) and spray-coated films using dilute CQD solution (bottom)...... 31 Figure 4.3: Film thickness measured by Dektak profilometer versus the number of layers for sprayed films ...... 32 Figure 4.4: Top and angled views of spin-coated (top) and spray-coated (bottom) films obtained using AFM...... 33 Figure 4.5: Map of elastic modulus of spin-coated and spray-coated CQD films ...... 35 Figure 4.6: Inter-particle spacing information obtained using GISAXS...... 36 Figure 5.1: Electroluminescence measurements of spin-coated (top) and sprayed (bottom) CQD films ...... 39 Figure 5.2: Impurity characterization using NMR and FTIR...... 40 Figure 5.3: Identification of electronic defect...... 41 Figure 5.4: Photoluminescence of reporter layer on top of spin-coated and spray-coated CQD films to extract minority carrier diffusion length (LD)...... 42 Figure 5.5: Histograms of power conversion efficiency under AM1.5 illumination for spin and spray-coated photovoltaic devices with Gaussian fits overlaid...... 44 Figure 5.6: Photovoltaic performance of best sprayed CQD solar cell...... 45 Figure 6.1: Illustration of the R2R simulation setup...... 46 Figure 6.2: Photovoltaic characterization of devices that were fabricated while mounted to a rotating drum ...... 47

viii Figure 6.3: Illustration of a flexible PET substrate (blue crescent) with ITO and TiO2 layers wrapped around a 2 cm dowel (black circle) while being sprayed on using our spray-coating process...... 49 Figure 6.4: Characterization of flexible sprayed devices...... 50 Figure 6.5: Illustration of a spherical lens substrate (Thorlabs, LA1252, blue semicircle) with ITO and TiO2 layers being sprayed on using our spray-coating process...... 51 Figure 6.6: Characterization of sprayed spherical device...... 52

ix Chapter 1. Introduction 1

1 Introduction

1.1 Motivation

About 1.5×1022 J of energy from the sun reach the Earth’s surface every day. This is four orders of magnitude greater than the 1.3×1018 J that humans consume daily [1]. Solar energy could, in principle, make up a large portion of our electricity demand, decreasing our dependence on fossil fuels that pollute the environment and cause climate change. However, solar energy accounted for only a fraction of a percent of the electricity generated worldwide in 2012 [1]. Present-day methods of harnessing solar energy using photovoltaic devices often cost more than burning coal and natural gas. Conventional solar cells are made using crystalline silicon. This material has historically been expensive to produce in the purest, single-crystal form that is required to make efficient photovoltaic devices. The rigidity and weight of solar cells is also a limiting factor in their wide deployment because they are not easy to integrate onto a variety of surfaces. Overall, these factors have traditionally made solar energy too costly for wide adoption around the world. Solution-processed semiconductors have been proposed as a solution to the cost problem because they allow for the mass production of photovoltaic devices using large-area coating techniques in a roll-to-roll (R2R) system [2]. The low cost of the manufacturing technology required to build these solar cells, combined with the large rate at which devices can be fabricated, account for most of the cost savings that could come from solution- processed semiconductors. In addition, fabricating solar cells on flexible substrates, made possible by solution processing, reduces the balance of systems cost by removing many of the structural engineering requirements imposed by heavy and rigid modules. Large-area manufacturing is invariably one of the first points that are made in any published study about solution-processed solar cells [3]. However, most reports until now have limited their fabrication methods to batch-scale processes such as spin coating, drop

Chapter 1. Introduction 2 casting, and dip coating. In fact, these are not analogues to large-scale manufacturing protocols. In this work I take the view that solution-processed photovoltaic technologies should be evaluated using R2R-compatible fabrication methods in order to assess their viability for practical applications.

1.2 Colloidal quantum dots

Colloidal quantum dots (CQDs) are small crystals of semiconductor material that are dispersed in solution. The size of these crystals, which ranges from a few nanometers to tens of nanometers, is smaller than the dimension at which the quantum confinement effect becomes appreciable. When the crystal size becomes smaller than the Bohr radius, the bandgap energy of the particle increases. This effect allows for precise tuning of the bandgap via control over the size of the crystal. In addition to bandgap tuning, the quantum confinement effect allows the CQDs to have light emission with a high degree of spectral purity, which makes them attractive for many optoelectronic applications. The synthesis of CQDs involves a reaction in solution between precursors for each constituting element. In the case of lead sulphide (PbS) quantum dots, which is the material that is used for the studies in this thesis, the precursors are lead oxide and bis(trimethylsilyl)sulphide for lead and sulphur, respectively [4]. The precursors are reactive and produce bulk-like crystals if mixed. Particle nucleation needs instead to be carefully controlled in order to achieve precise size control over the crystals. This can be achieved by careful choice of reaction solvent as well as having a stabilizing ligand. Oleic acid serves well as a ligand in PbS quantum dot synthesis, and its concentration is crucial in determining the final particle size. The long carbon chain of the oleic acid prevents the aggregation of the quantum dots in the final dispersion. If the nanocrystals were to aggregate, this would compromise the quantum confinement effect and negate this useful property. Each quantum dot is one individual nanocrystal of the bulk semiconductor material, albeit one with only around 1000 atoms. As a result of its small size, the ratio of surface atoms to bulk atoms is very high in a CQD. Surfaces are usually not desirable for semiconductor materials because dangling bonds and other defects introduce electronic states

Chapter 1. Introduction 3 in the middle of the bandgap. These defects can be passivated using a variety of approaches, from introducing ligands with specific functional groups to attach to the surface of the dots, to growing an insulating shell around each dot during synthesis. The specific passivating approach depends on the quantum dot material and the final application. A dispersion of CQDs is effectively a semiconductor ink that can be applied onto a variety of substrates. Solution processing could enable facile mass-production of optoelectronic devices, as well as application to a variety of form factors that cannot be readily achieved using conventional crystalline semiconductors grown by epitaxy.

1.2.1 Colloidal quantum dot photovoltaics

Size tuning of the bandgap of CQDs means that the wavelength of solar radiation that can be absorbed by a photovoltaic device can be adjusted. PbS CQDs are the only solution- processed material whose bandgap can be tuned into the infrared. Considerable solar radiation lies in the infrared and is lost – not captured by single-junction organic or silicon solar cells (Figure 1.1). CQDs are ideally poised to become infrared sensitizers to existing photovoltaic technology at low cost.

Figure 1.1: Maximum photovoltaic efficiency of devices with multiple number of junctions. (a) Portion of solar spectrum covered by the given number of junctions at the ideal wavelengths for maximum power conversion efficiency (PCE). (b) Value of PCE for each number of junctions. Reprinted with permission from [2]. Copyright (2012) Nature Publishing Group.

Chapter 1. Introduction 4

In addition to the potential of PbS CQDs to sensitize other photovoltaic platforms to infrared radiation, their bandgap can be tuned down to the ideal value for a single-junction solar cell, which lies approximately between 950 nm and 1250 nm (Figure 1.1). If the bandgap wavelength of the solar cell were higher than this value, intraband relaxation loses would start to become significant for photons with energy in the area of the solar spectrum with highest intensity (from around 400 to 800 nm). On the other hand, if the bandgap wavelength were shorter than the ideal, less solar radiation would be absorbed, compromising photovoltaic efficiency. The efficiency of single-junction CQD solar cells has progressed rapidly since the first reports of the photovoltaic effect in CQD-sensitized devices [5]. One of the first major performance breakthroughs came when the architecture of CQD devices was improved to locate the charge-separating junction on the illumination side (Figure 1.2), which made the collection of photogenerated carriers more efficient [6]. The power conversion efficiency (PCE) of solar cells when illuminated using Air Mass 1.5 G solar spectrum (AM1.5) was improved from 3.6% to 5.1% using this depleted heterojunction architecture. The next advance came with increased understanding of how the surface chemistry of the quantum dots affects their electronic properties. A hybrid organic-inorganic passivation approach using a combination of thiol ligands and halide anions led to a certified PCE of 7%, with some cells reaching over 8% in the laboratory [3].

Figure 1.2: Depleted heterojunction CQD solar cell architecture. (a) Illustration of the device, which is illuminated from the fluorine-doped tin oxide (FTO) side. The charge-separating junction is formed between the n-type titanium dioxide (TiO2) layer and the p-type CQD layer. (b) The spatial band diagram of this architecture under illumination, showing that the photogenerated electrons flow into the TiO2 and the photogenerated holes flow towards the gold (Au) contact. Reprinted with permission from [6]. Copyright (2010) American Chemical Society.

Chapter 1. Introduction 5

While the studies outlined above have demonstrated that photovoltaic devices based on CQDs can achieve high performance, the methods upon which they have relied for the deposition of the active CQD layer were small-scale processing methods, such as spin coating and dip coating. CQDs offer the potential for high-performance solar cells manufactured using large-scale fabrication methods because, once synthesized, they do not require the implementation of complex nano- and microscale morphologies that other solution-processed technologies need for efficient photovoltaic function. The electron and hole transport layers in the CQD studies described above were deposited using thin-film sputtering and evaporation techniques. These processes have been demonstrated to be compatible with mass manufacturing by various large companies that fabricate organic electronics for displays. However, solution-processed versions of these materials have been developed by other groups and can be deposited using R2R deposition techniques [7]–[9]. In order to assess the viability of CQDs as a platform for low-cost, mass-producible solar cells, the active CQD layer needs to be deposited using R2R-compatible fabrication processes. These scalable deposition techniques also need to achieve photovoltaic devices with PCE values of equal or greater magnitude than devices fabricated using batch-scale processing.

1.3 Thesis objectives

The goal of the proposed project is to develop a spray-coating process, which is a technique compatible with R2R protocols, to deposit the active layer of CQD solar cells. Devices that are fabricated using this procedure are intended to perform as well as their spin-coated counterparts, or better. The specific objectives of this thesis are outlined as follows: 1. Design and build a spray-coating setup for the deposition of CQD films. Determine the appropriate equipment for each step in the fabrication process and integrate all the pieces into a system that can run automatically to simulate industrial fabrication and decrease sources of human error. 2. Investigate the effects of the spray-coating process on the material and electronic properties of CQD films. By learning how the spray-coating process

Chapter 1. Introduction 6

affects the morphology, quantum dot packing density, purity, and electronic trap density of the films, we can determine if the method is capable of delivering photovoltaic devices with high performance. 3. Realize sprayed CQD photovoltaic devices that are statistically equal to or better than the standard spin-coated devices. If solar cells of equal or better quality can be achieved using spray coating, it would mean that CQDs are well suited as a material for mass-produced solar cells. 4. Evaluate the transferability of our spray-coating procedure to a roll-to-roll process and to unconventional form factors. By simulating a roll-to-roll system we can further determine if the spray-coating process we develop can indeed be adapted for mass manufacture. In addition, exploring the fabrication of solar cells onto objects with non-planar and non-cylindrical surfaces will allow us to explore the suitability of CQDs as a photovoltaic material for applications that have not been considered previously. All of the listed thesis objectives are addressed through experimental analysis. The relevant background and theory is given in Chapter 2. In Chapter 3 we explore the design and early optimization of the spray-coating set up. In Chapter 4 we characterize the material properties of spray-coated films and compare them to spin-coated films. Planar and unconventional photovoltaic devices are described and characterized in Chapters 5 and 6. Chapter 7 provides a summary of the findings of this thesis and provides suggestions for future work.

Chapter 2. Background 7

2 Background

2.1 Solar cell fundamentals

The function of a solar cell is to turn electromagnetic energy from the sun into electricity.

Each photon has energy (Eph) according to the following equation: ℎ� � = ℎ� = (2.1) !! � where h is Planck’s constant, ν is the photon frequency, c is the speed of light in vacuum, and λ is the photon wavelength. When a photon interacts with a semiconductor material that has a bandgap energy that is smaller or equal to Eph, the photon excites an electron from the valence band to the conduction band of the semiconductor, leaving behind a positively charged hole. This process of photon absorption is governed by the Beer-Lambert Law: � = �!!" (2.2) where T is the optical transmission fraction, d is the thickness of the film, and α is the absorption coefficient. The inverse of the absorption coefficient is called the absorption length (LABS). Photogenerated carriers are then separated by an internal electric field, usually caused by a depletion region formed around a p-n junction, which drives the electrons and holes to opposite ends of the device. If the depletion region does not encompass the full thickness of the device, at least one of the charge carriers will need to diffuse across a quasi-neutral region, a volume in the device that is minimally affected by the internal electric field. The minority carrier diffusion length (LD) is the average distance that a minority carrier can travel by diffusion before recombining. This quantity impacts the efficiency of carrier extraction at the electrodes, which in turn affects the performance of the device. If the thickness required for near-complete absorption of light, LABS, is larger than LD, as in CQD films, then there will be a compromise between light absorption and carrier extraction: if the film is thin, it ensures efficient carrier extraction at the electrodes but there are not many carriers being generated; if

Chapter 2. Background 8 the film is thick, all of the incident light will generate carriers by being absorbed, but carrier collection at the electrodes will be poor. Solar cell operation under illumination is characterized by a current-voltage (J-V) sweep (Figure 2.1). The J-V curve of a solar cell in the dark is a typical diode curve, but under illumination it shifts towards the fourth quadrant of the plot, representing generated power. The convention is to invert the current axis. The current density that is generated by an illuminated solar cell at a bias of 0 V is referred to as short-circuit current density (JSC). This is the maximum current density that is generated in the solar cell by illumination, and it is affected by light absorption and carrier collection efficiency. The JSC is the main figure of merit that is affected by the absorption- extraction compromise.

VOC, the open circuit voltage, is generated between the terminals of an illuminated photovoltaic device when there is no outside connection between them. This quantity is related to the difference in the quasi-Fermi levels between the p-type and the n-type material under illumination. It is reduced by any source of recombination, such as electronic defects in the film.

Figure 2.1: Solar cell operation under illumination as measured by a current-voltage sweep. In this case, full current (I) is being measured, but current density (J) can be calculated by dividing by the area of the device. Reprinted with permission from [2]. Copyright (2012) Nature Publishing Group.

The maximum power point (MPP) of the device occurs at the point in the J-V curve in which the product of the current and the voltage (JMPP and VMPP, respectively) is at its maximum. The power conversion efficiency (PCE), which is defined as the ratio of the

Chapter 2. Background 9

maximum power density generated by the device (Pout) divided by the incident solar power density (Pin), is given by the following equation: � � � � � �� ��� = !"# = !"" !"" = !" !" (2.3) �!" �!" �!" where FF is the fill factor of the solar cell, given by the ratio of Pout to the product of VOC and

JSC.

Two factors that determine FF are series resistance (RS), which is the sum of the resistance within the film, the resistance between the film and the electrodes, and the resistance within the electrodes; and shunt resistance (RSh), which is a measure of how well the solar cell resists alternative paths to the photogenerated current. These values shape the area under the J-V curve (Figure 2.1), which determines how much it fills out of the VOC-JSC square, defining the FF. An ideal photovoltaic device would have a series resistance of zero, making the J-V curve go vertically from VOC towards the MPP; and a shunt resistance of infinity, which would make the J-V curve go horizontally from JSC to the MPP. This would lead to a FF of 100%.

2.2 Batch-scale processing methods

In research and development of coating materials it is useful to fabricate small-scale samples using simple processes that do not require very specialized equipment. Batch-processing methods have the advantage that they only require simple apparatus that can be easily operated by any researcher. The flexibility of material selection and processing parameters make these methods very attractive for optimizing thin films for a myriad of applications in a relatively short amount of time. High performances can usually be achieved using batch- scale processing methods because the small dimensions that are coated avoid many performance-limiting defects that are more likely to appear when processing is done more rapidly and on larger areas. As a result, performance metrics achieved using small-scale processing methods may give an optimistic picture compared to what could be achieved by using large-area coating methodologies.

Chapter 2. Background 10

Drop casting is one of the simplest small-scale deposition techniques that can be used to fabricate solution-processed solar cells [10]. To make a film using this method, the only thing that needs to be done is to apply a drop of solution on a substrate and allow it to dry. Spin coating is one of the most common batch-scale coating techniques used in research and development of solution-processed photovoltaics. This fabrication process, depicted in Figure 2.2, can be divided into four stages: deposition, spin-up, spin-off, and evaporation [11]. During the deposition stage, a few drops of the liquid of interest are placed at the centre of the rotation axis. The amount of material that is deposited far exceeds the amount that will be incorporated in the film, but this is a requirement for the uniform coating of a substrate. The substrate then begins to rotate during the spin-up stage. The centrifugal force that is generated by the rotating substrate causes the liquid to flow radially outward, displacing the gas that was originally in contact with the surface of the substrate. This stage lasts for the first few seconds of rotational speed increase, which can reach up to a few thousand revolutions per minute.

Figure 2.2: Stages of the spin-coating process. Reprinted with permission from [11]. Copyright (2011) Cambridge University Press.

Excess liquid that remains on the surface after spin-up flows outwardly during the spin-off stage. Droplets form at the edge of the sample and fly off. Once the film reaches a uniform thickness throughout the substrate, it remains uniform as it continues to get thinner. Spin-off and solvent evaporation happen concurrently, both of which contribute to the

Chapter 2. Background 11 thinning of the film. Fluid flow slows down as the film gets thinner due to the film providing greater resistance. The liquid flow eventually stops and evaporation becomes the only source of further reduction in film thickness until the solvent fully evaporates. The full spin-coating process is complete sometime between thirty seconds to a full minute of spinning, depending on the solvent. There is evidence that the final film thickness is usually unaffected by the deposition and spin-up stages, making spin-off the critical step. Spin coating is mainly used for flat, smooth substrates because surface inconsistencies result in films with poor morphology and coverage. However, there may be certain situations when there needs to be some texturing on the substrate or some deviation from the planar geometry. Dip coating is a batch-scale material deposition technique that can achieve conformal coatings on various form factors. The first step is to dip the substrate into the desired coating liquid. This step should be done at a slow speed to carefully displace the gas that was previously in contact with the surface of the substrate. After soaking in the liquid for some time, the substrate is slowly removed. Some of the liquid will cling to the surface, forming a very thin film. Conformal films can be achieved on a variety of non-planar features if the process parameters and materials are chosen correctly. Dip coating can also be automated to achieve greater consistency from batch to batch, but it is not compatible with scale-up manufacturing because it is inherently a very slow process.

2.2.1 Standard CQD solar cell fabrication

CQD solar cells can be fabricated using a variety of batch-processing methods, but the one that has been more commonly used is spin coating. Below is a description of the standard spin-coating process that is used for the fabrication of CQD solar cells, which is the process that is used for comparison with spray coating throughout this work. A colloidal suspension of oleic-acid capped quantum dots is prepared in octane at a mass per volume concentration of 50 mg/mL. Two drops of this solution are dropped onto a substrate with a previously deposited TiO2 n-type layer. The substrate is then rotated at 2,500 rpm for ten seconds to form a solid thin film of CQDs. This film is then treated using a solution of 3-mercaptopropionic acid (MPA) in methanol at a volume-to-volume concentration of 1%. The film is completely soaked with this solution for three seconds, after

Chapter 2. Background 12 which the substrate is made to rotate at 2,500 rpm for 5 seconds to get rid of the excess solution. MPA is a short ligand with a thiol group on one end that has a very high affinity for the PbS quantum dot surface. This ligand displaces the long oleic acid molecule that originally prevented the quantum dots from getting too close to each other in solution. The exchange of oleic acid to the shorter MPA molecule produces a reduced dot-to-dot spacing and the film is rendered more conductive, which improves the efficiency of the device. However, the film also experiences a volume contraction that can lead to surface cracking, an issue that will be explored in Chapter 4. After the MPA treatment, the film is soaked with methanol and the substrate is rotated at 2,500 rpm for 10 seconds. This step is repeated twice to wash away the organic impurities that were leftover from the ligand exchange. These final methanol-washing steps mark the completion of one layer of CQDs. In order to build up the CQD layer to the final device thickness, all of the steps are repeated to build up the film in a layer-by-layer fashion. The standard device thickness is achieved by using 10 layers. The standard device architecture, which is the architecture used throughout this thesis for spin-coated and spray-coated devices, is the depleted heterojunction architecture. The electron transport layer in this system, TiO2, is sputtered to a thickness of 50 nm on glass that was previously covered with fluorine-doped tin oxide (FTO), a transparent conductive oxide.

The sputtered layer is then treated with titanium tetrachloride (TiCl4) and sintered at high temperature to render it more conductive. This is the substrate upon which the CQD layers are deposited. The hole transport materials, 40 nm of molybdenum trioxide (MoO3) and 50 nm of gold, are evaporated on top of the CQD layer through a shadow mask to mark 16 distinct pixels with an area of 6.7 mm2. To complete the device, a layer of 120 nm of silver is evaporated on top of the device stack through the same mask. Silver serves as a good metallic contact for probing the device during operation.

2.3 Roll-to-roll manufacturing

One of the biggest motivations behind solution-processed solar cells is the possibility for low-cost devices compared to photovoltaic devices based on conventional epitaxially-

Chapter 2. Background 13 processed materials. This reduction in cost can be achieved in two ways: by employing inexpensive materials, and by fabricating the devices using inexpensive, large-scale manufacturing methods. Roll-to-roll (R2R) manufacturing is a well-developed methodology that can be applied for the mass production of solution-processed solar cells to address the production point. This process consists of a substrate being moved by rollers to different coating stages. R2R manufacturing has been used for many years for printing newspapers, which is an industry that requires production of several hundreds of thousands of copies every day. The only caveat to this technique is that the substrate needs to have a certain amount of flexibility so that it may be processed on cylindrical rollers without breaking. Some of the same techniques used in the newspaper industry can be used to print certain features of solution-processed solar cells. We will focus on techniques that are being investigated for the deposition of the active material in photovoltaic devices (Figure 2.3).

Figure 2.3: Illustrations of coating techniques compatible with R2R processing: (a) knife-over-edge (b) meniscus (c) slot-die (d) gravure. The coating rollers and coating units are shown in grey shading. The web (substrate) is shown as a thin line and the coated material is shown as a dotted line. Reprinted with permission from [12]. Copyright (2009) Elsevier.

A simple coating method that is compatible with R2R manufacturing is knife-over- edge coating (Figure 2.3a). This technique employs a long edge, referred to as a “knife”, in conjunction with an ink bath immediately behind it relative to the rolling direction of the web (the name given to the substrate in R2R processing)[12]. As the web rolls forward it pushes ink under the knife, which flattens it into a uniform thin film. The distance from the knife to the web determines the thickness of the film. A similar technique is meniscus coating (Figure 2.3b), but instead of a knife it is a secondary roller placed over the support roller that flattens

Chapter 2. Background 14 the ink into a film. This additional roller can rotate in either the same direction or against the direction of motion of the web to add further control over the final thickness of the film. Knife-over-edge and meniscus coating minimize ink waste by confining the ink bath dimensions to the volume of ink required to make a film. In the case of continuous processing, the ink bath can be replenished automatically using a variety of feed systems [12]. Another coating technique used in R2R manufacturing is slot die coating (Figure 2.3c). This method allows for one-dimensional patterning of the film to make stripes, which can be a useful feature in multilayer solar cells. A complex coating head deposits the ink onto the moving web. The ink is fed to a small reservoir inside the coating head by a feed system. The coating head may have a mask inside that allows for the deposition of stripes of any desired widths onto the surface of the web as it rolls underneath the coater. The thickness of the film is determined by the rolling speed of the web as well as the ink feed rate [12]. Like the coating methods described above, slot die coating does not waste much ink. Gravure coating is a deposition technique that allows for two-dimensional patterning of the ink. This method consists of two rollers: a coating roller that contains a two- dimensional pattern, and a support roller under the web. The coating roller is dipped in ink and a knife skims off excess material before it comes in contact with the web (Figure 2.3d). The main disadvantage to this technique, other than its complexity, is the fact that the coating roller needs to be engraved to change a pattern, which can be an expensive process. However, this coating method can allow for processing low viscosity inks at relatively fast web speeds (1-10 m/s) [12]. Another relevant coating technique that is compatible with R2R manufacturing is spray coating. This process involves pushing ink through a nozzle using high-pressure gas, creating a collection of droplets that coat the substrate. The use of electrostatic charges to direct the droplets towards the substrate and minimize wasted ink might also be used for some applications. A downside of this technique is that the thickness and roughness of the film can be difficult to control [12]. However, spray coating can be a powerful technique if used appropriately, as it can allow for precise two-dimensional patterning down to the millimeter scale to create complex features for devices.

Chapter 2. Background 15

Looking beyond R2R manufacturing, spray coating can also be used to coat surfaces with multiple degrees of curvature, a property that could allow solar cells to be integrated onto a variety of surfaces such as car roofs, airplane wings, or the back of curved smartphones. This is a degree of freedom that other production coating techniques described above generally do not possess.

2.4 Dynamics of Spray Coating

It is important to examine aspects of the physics of spray atomization, such as the main mechanisms of droplet formation and breakup, to obtain good quality films from spray coating. There are two stages of atomization (the breakup of liquid into droplets) called primary and secondary atomization. In the first stage, droplets are formed from the bulk liquid. In the second stage, which occurs in flight as the droplets are propelled from the nozzle, the droplet size reduces further. Two forces act against each other on a droplet as it comes out of the nozzle: the aerodynamic force serves to break the droplet apart, while the surface tension force keeps the droplet together [13]. The ratio of aerodynamic to surface tension forces is referred to as the Weber number, which is proportional to the square of the initial ambient velocity relative to the drop and the initial diameter of the droplet. It is also inversely proportional to the surface tension of the drop. The magnitude of the Weber number determines the breakup mode of droplets during secondary atomization. When the Weber number is low, the droplet is subject to weak aerodynamic forces that deform the droplet into an oblate spheroid. At the same time, the surface tension restores the drop to a spherical shape. This leads to a vibrating droplet that may break up but only into very few large fragments. This regime is referred to as the vibrational mode. The next break up mode, bag breakup, occurs at relatively low Weber numbers, which means that only small amounts of energy are needed for the droplet to achieve this mode of secondary atomization. As the name “bag breakup” implies, the forces that act on the droplet turn it into the shape of a bag. As the bag stretches, the body of the droplet bursts and breaks up into smaller pieces, leaving the rim of the bag initially intact as a toroidal ring.

Chapter 2. Background 16

The ring eventually breaks up into a few larger fragments as it flies. The diameter of the smaller droplets is around 4% the diameter of the original droplet, while the droplets that are formed from the ring are around 30% the diameter of the original [13]. Sheet-thinning breakup occurs at higher ambient velocities than bag breakup (and thus at higher Weber numbers). The drop is deformed into the shape of a sheet, which breaks up into ligaments that break up into smaller fragments. It is hypothesized that ambient phase inertia causes the periphery of the initial droplet to be deflected into the direction of airflow, which gives the droplet the shape of a sheet that eventually breaks up into smaller pieces. The average droplet size from sheet-thinning breakup is smaller than the size from bag breakup. Multimode breakup happens at Weber numbers between bag breakup and sheet- thinning breakup, and is something of a combination of both modes. The final breakup mode happens at extremely high Weber numbers. Unstable waves form at the leading edge of the drop and they break up the droplet into many fine fragments. This break up mode, termed “catastrophic”, has only been observed in shock tube experiments where very high initial ambient velocities are possible. Thus, the breakup modes that are relevant in ordinary spray systems are bag, sheet thinning, and multimode.

2.5 Spray-coated photovoltaics in literature and their limitations

Since spray coating is a versatile coating technique that is compatible with R2R manufacturing, there are some groups that have already explored applying this technique to the fabrication of solution-processed solar cells. Some of the materials systems that have been explored are organic [14], [15] and ternary chalcogenide nanocrystals [16]. Organic materials used in organic photovoltaics (OPV) have a very short excitonic diffusion length, which means that the donor and acceptor that make up the active material of these solar cells need to be very thin to allow the exciton to reach the donor/acceptor interface that splits the exciton into an electron and a hole. Making a thin active layer decreases the amount of light that can be absorbed, which means that the amount of excitons that are generated in the first place is low. In order to overcome this constraint, researchers devised a structure for the active layer of OPV devices in which a donor and an acceptor are

Chapter 2. Background 17 intricately interpenetrating at a scale similar to that of the excitonic diffusion length, which is in the order of 10 nm. This allows the absorbing layer to be relatively thick while the maximum distance between an exciton and a donor/acceptor interface is short enough for the exciton to reach before recombining (Figure 2.4). This structure is called a bulk heterojunction (BHJ) [17], and is needed for efficient photovoltaic devices based on organic materials. Park et al. used spray coating to fabricate organic solar cells of a fullerene:polymer blend [14]. They achieved a PCE of 3.35% after various treatments on the sprayed BHJ. Even though they do not provide an equivalent spin-coated device for comparison, the PCE they achieved using spray coating is lower than the PCE of greater than 5% that has been attained by other groups with the same active materials by using spin coating [18]. Kumar et al. published a different study on spray-coated organic solar cells using a low-bandgap polymer BHJ with a different fullerene [15]. They achieved a maximum PCE of 6.11%, while spin-coated solar cells with similar active layers have surpassed 7% PCE [19].

Figure 2.4: Organic photovoltaic architectures. (a) Bilayer architecture in which the donor (white layer) and acceptor (grey layer) have a planar interface. (b) Bulk heterojunction structure. Unfilled circles represent holes, filled circles represent electrons, dotted-line circles represent excitons, and the wavy arrows represent incident light. Reprinted with permission from [17]. Copyright (2003) Nature Publishing Group.

Spray-coated OPV devices suffer from a performance reduction when compared to their batch-processed counterparts. The quality of a BHJ is not only affected by the nature of the donor and acceptor materials, but also by the fabrication conditions. Since the BHJ

Chapter 2. Background 18 structure forms as two immiscible materials separate, it is very important to control the deposition conditions precisely. Spray coating is inherently a chaotic process in which the forces on individual droplets cannot be precisely controlled. Thus, spray-coated BHJs cannot be made of the same quality as their spin-coated counterparts and sprayed OPV technology suffers from a trade-off between manufacturability and performance. Ternary chalcogenide nanocrystals are another solution-processed semiconductor material system that can be used to fabricate solar cells. Even though they are colloidal crystals, they are not CQDs since they are not quantum-tuned. Photovoltaic devices based on these materials can reach over 5% after drop casting and annealing at 800°C in the presence of a selenide material for simultaneous crystal sintering and surface passivation with selenide [10]. This high-temperature step defeats the purpose of solution processing because there are no flexible substrates that can withstand such high temperatures, and thus cannot be integrated into R2R processing. Akhavan et al. published a study in which ternary chalcogenide nanocrystals were spray-coated and only subjected to mild processing conditions [16]. They achieved efficiencies of 1.9% and 1.1% on glass and plastic substrates respectively, which are much lower than the PCE values that can be achieved using batch- scale methods. CQDs are a material system that does not suffer from the disadvantages that restrict the two materials systems described above. The best performing CQD solar cell architectures are not affected by processing method because they do not require intricate structuring that OPV devices need for efficient carrier collection. In addition, the crystal formation and passivation of CQDs is all done in the solution phase, unlike ternary nanocrystals. These factors suggest that CQDs lend themselves well to spray coating. However, the CQDs also present some potential complications when adapting them to a new processing method. First, the ligand exchange process is critical for photovoltaic function. If the original ligands are not properly removed from the film, it could lead to poor electrical properties that diminish device performance. Second, the packing of the quantum dots in the film needs to be carefully controlled in order to obtain a film with the desired optical and electrical properties. The next chapters will explore the development of a spray- coating procedure for the production of CQD films. The resulting films will be characterized

Chapter 2. Background 19 and compared to spin-coated films to investigate whether these issues can be addressed and whether spray coating can produce efficient CQD photovoltaic devices.

Chapter 3. Spray coating of CQD films 20

3 Spray coating of CQD films

3.1 Spray-coating setup

3.1.1 Nozzle types

In spray-coating, the spray nozzle affects the droplet size and the degree of atomization of the ink, which ultimately determines the quality of the coating. During the development of the spray coating setup we considered different types of nozzles. At the beginning of the project we started with easily accessible airbrushes from Paasche, a well-known airbrush company for art applications that is sold at many artist stores (Figure 3.1). We picked the model VLS because of its versatility. The spread of the spray pulse can be changed from 1/32 of an inch to 1.5 inches by using one of three different adjustment sets that consist of a spray head, an adjustment needle, and a spray tip. The packing in this airbrush is made out of polytetrafluoroethylene, which is resistant to methanol and octane, the solvents used in CQD solar cell fabrication.

Figure 3.1: Paasche airbrush, VLS model. (a) Picture of fully assembled airbrush. (b) Schematic of the working parts of the airbrush. The numbered parts are: 1 – needle protecting cap, 2 – spray head, 3 – spray tip, 4 – ink inlet, 5 – finger lever, 6 – gas inlet, 7 – paint control mechanism, 8 – adjustment needles. Product picture and schematic were obtained from the supplier website: http://www.paascheairbrush.com

This airbrush is simple in its function. The CQDs are fed in by connecting the ink inlet (part #4 in Figure 3.1b) to a reservoir of CQDs in octane by a plastic tube. Pressured gas

Chapter 3. Spray coating of CQD films 21 is fed into the brush via the gas inlet (part #6 in Figure 3.1b). The gas is allowed to pass over the ink inlet through the inside of the brush when the finger lever (part #5 in Figure 3.1b) is pressed down, which causes suction force that drives ink up into the body of the brush and then out through the tip. The tip of the adjustment needle reduces the size of the opening for the liquid, which causes the ink to atomize into droplets. The head of the brush controls the spread of the spray. When automating a system that uses this kind of airbrush it is important to secure the finger lever in the ‘down’ position so that the gas inlet is continuously open without the need for a user to activate the brush manually. We also looked for a more sophisticated nozzle that could potentially give a better spray. We turned to Ikeuchi, a company that specializes in the manufacture of spray nozzles and systems for industrial applications. Their nozzles in the BIMV-S series are described as “fine fog nozzles” with flat sprays. These have the smallest available droplet size, which ranges from 10 to 50 µm, depending on the gas pressure applied. A higher air pressure leads to a smaller droplet size, in good agreement with the theory presented in Section 2.4. We purchased a nozzle with a USP type adaptor (model BIMV8002S), which makes the spray direction somewhat adjustable by being able to rotate the spray tip by 15° in all directions. This nozzle is hereby referred to as “Ikeuchi nozzle”. The operation of this nozzle lends itself well to integration into an automated system. This nozzle has two separate controls for liquid flow and airflow, which allow them to be activated independently of each other for better control over the atomization process. The liquid is fed by gravity through an inlet at the top of the body (Figure 3.2c), which can be opened by applying gas pressure to the pilot air inlet (Figure 3.2b). This action only opens the liquid inlet to let the ink into the chamber by the gravity pressure. If no gas were input through the airflow inlet, then the ink would simply drip out of the tip of the nozzle. In order to create a spray, compressed gas needs to be fed in through the separate airflow inlet (Figure 3.2c). This gas mixes with the liquid in the chamber and expels it out of the spray tip in the form of a fine mist. The spray shape is a long and narrow oval, with each edge making an angle of 80° with the spray tip at its vertex (Figure 3.2a). The Paasche and the Ikeuchi airbrushes were the two types of spray nozzles that were explored for this work. The next few sections of this chapter will describe how they were

Chapter 3. Spray coating of CQD films 22 integrated into an automated spray system and what fabrication steps they were used for in the production of CQD solar cells.

Figure 3.2: Ikeuchi nozzle, BIMV8002S model. (a) Spray tip in operation, showing the spray angle of 80°. (b) Top view of full nozzle apparatus, indicating the location of the pilot air inlet. The spray tip is to the left of the diagram. (c) Side view full nozzle apparatus, indicating the location of the liquid and compressed air inlets. Product schematic was obtained from supplier website: http://www.ikeuchiusa.com/catalogs/bimv.pdf

3.1.2 Integration and automation

To create a process that approximates industrial-level manufacturing, it is important to integrate all components of the spraying apparatus into an automated system (Figure 3.3). We achieved control over the spray-coating process by passing the pressurized gas lines through solenoid valves that were controlled by electromechanical relays on a printed circuit board (PCB). The number of valves corresponded to the number of gas lines that were needed at the spray setup. This PCB was connected to a computer via serial port connection to allow for automatic control of the gas outflow towards the spray setup. There were two main options for the pressurized gas for the spray-coating setup: inert nitrogen gas and compressed air. The source of nitrogen gas was a cryogenic storage dewar with a gas outlet attached to a regulator. The gas line was connected to the control box where it was split into multiple lines with their own valves to allow the individual control of gas outflow to each of the multiple parts in the setup that needed nitrogen. The gas connections in the control box allowed for the inclusion of secondary regulators to individually control the pressure of each of the outlets.

Chapter 3. Spray coating of CQD films 23

The source of the compressed air was a gas line that is supplied to the tenants of the building that housed our equipment. The advantage of this line is that it comes at higher pressures than what can be achieved using the nitrogen dewar, and thus can be used as the gas source for any steps that require very high gas pressure. This gas line was also connected to a valve in the control box before going on to the deposition setup.

Figure 3.3: Full procedure of layer-by-layer spray deposition. Stage 1 involves the spraying of CQDs, which is done by an Ikeuchi nozzle in this schematic. Stages 2 and 3 use commercial airbrushes to spray MPA diluted in methanol and pure methanol, respectively. In stage 4, an air blade applies a curtain of high-pressure compressed dry air (CDA) to aid in solvent drying. All solenoid valves are controlled by a computer through a control-printed circuit board. The looping of the sample through the four stages has been implemented as a loop in time with all the nozzles pointed at the same position in space. Adapted with permission from [20]. Copyright (2014) John Wiley and Sons.

The spray nozzles need a reservoir of material in order to keep operating without constant human supervision. The ink feed tubes for the Paasche airbrushes were connected to jars with enough solution to last at least one full round of spray coating. The ink for the Ikeuchi nozzle was fed by gravity from above using a tubular reservoir connected to the ink inlet in the nozzle by Teflon tubing. The gas valves on the relay board were controlled by a computer using MATLAB scripts. The scripts allowed for the opening and closing of the valves for predetermined amount of time using ‘pause’ commands, which allowed for well-defined spray pulses and

Chapter 3. Spray coating of CQD films 24 wait times between steps. To activate the spray of a Paasche airbrush only one gas line needs to be opened due to the simple operation of this device. The operation of the Ikeuchi nozzle is more complicated since the ink feed into the nozzle chamber is controlled by a pneumatic valve and the gas that drives the ink out of the nozzle is fed in through a separate inlet. These two functions can be activated separately or at the same time. We found that the smallest droplets were obtained when the pilot gas was activated a fraction of a second after the carrier gas, possibly because primary atomization happened as soon as the ink dripped into the nozzle chamber with carrier gas already flowing through it. Initial results suggested that the pilot gas needed to be turned off very quickly to end the spray cycle more effectively. By adding an outlet to atmosphere to the valve that controls the pilot air in the control box, the gas line to the pilot gas can be depressurized very quickly, thus shutting off the spray function quickly. The level of automation that we achieved with this system allowed us to eliminate human error from the process so that we could isolate individual steps in the procedure for easier control over optimization. In the next section, we describe the process of adapting this automated setup to fabricate CQD films using previously understood ligand exchange procedures.

3.2 Spray-coating procedure

3.2.1 Layer-by-layer spray process

The standard CQD film fabrication procedure is a layer-by-layer process using spin coating, as described in Subsection 2.2.1. Spray coating lends itself well to adapting this fabrication procedure because each step in the process can be assigned to a separate nozzle. First, a pulse of a solution of oleic-acid capped CQDs in octane is sprayed onto the substrate using either a Paasche airbrush or an Ikeuchi nozzle. The effect of the nozzle type used for the CQD spraying step is explored in Subsection 3.2.2. This is the most critical step of the spray-coating procedure since it determines the morphology and coverage of the film, which cannot be improved by subsequent steps in the ligand exchange procedure. The

Chapter 3. Spray coating of CQD films 25 pressurized gas used for this step was nitrogen at 45 psi. The thickness of each layer can be controlled by the concentration of CQDs in the feed solution and the duration of the pulse. The ligand exchange step is carried out by spraying a solution of the MPA ligand in methanol using a Paasche airbrush and nitrogen gas. We found that using 45 psi for this step resulted in films with visible white residues and poor photovoltaic devices, so we included a secondary regulator for this line inside the control box and set it to 35 psi. This caused a gentler MPA spray that resulted in visibly better quality films with good solar cell performance. After the MPA spray pulse, the film was washed by methanol using a strong stream from a Paasche airbrush. The problem with this step is that it moistens the film, which does not dry very quickly on its own. In order to speed up the drying process, we added an air- drying step after the methanol wash. The equipment for this step was made in-house and consisted of a hollow metal block with a thin outlet slit to provide a curtain of pressurized gas. We found that this step required higher pressure than the 45 psi that could be provided by the nitrogen dewar, so instead we used the compressed dry air that is supplied by the Galbraith building. We found that a pressure of 85 psi worked well for drying the films. The steps of our spray-coating process for CQD fabrication are summarized in Figure 3.3. Step 1 refers to the CQD deposition step, which is performed by an Ikeuchi nozzle in the figure. We found that a pulse duration of 0.4 s worked best for this step, followed by a 3 s pause. Step 2 is the spraying of the MPA solution in methanol for 1 s, after which the substrate is rinsed with methanol in Step 3 for 4 s. After the wash is finished, the air blade dries the substrate for 40 s. Upon completion of all four steps, only one thin layer of CQDs is deposited. This process needs to be repeated several times in order to fabricate a CQD film of device thickness in a layer-by-layer fashion.

3.2.2 Effect of nozzle type used for CQD deposition

One of the most important factors that affect the morphology of sprayed films is the nozzle that is employed for the spraying step. This apparatus affects the initial droplet size, the behaviour of the droplet in the air, and the spread of the spray. All of these parameters affect the morphology and coverage of the final film, which in turn have a direct effect on device performance.

Chapter 3. Spray coating of CQD films 26

We first turned to a Paasche airbrush (described in Subsection 3.1.1), which was the most accessible instrument at the beginning of this project. This airbrush has excellent spray density at short distances because it is meant as an artist’s tool for painting lines, an application in which incomplete coverage is aesthetically unacceptable. The shorter distance that was needed to provide dense coverage of the substrate also meant that the initial droplets only had a small amount of time to further break up in flight (Section 2.4). This caused the droplets to be relatively large upon hitting the substrate, and they were also impacting the surface at a higher velocity than they would from a longer distance. After we optimized the distance from the nozzle to the substrate as well as the CQD concentration, the fabrication of a film of device thickness that provided full coverage of the substrate was achieved by doing 15 spray cycles with a CQD concentration of 16.7 mg/mL and an MPA volume-to-volume concentration of 0.16%. Depositing the CQD layer using the Paasche airbrush resulted in a very rough surface at the micron scale, as evidenced by a scanning electron microscope (SEM) image of the surface of the film (Figure 3.4a). Imaging the cross-section of a full device reveals that the device thickness varies wildly between around 300 nm to above 600 nm (Figure 3.4c). Such variations in active layer thickness lead to inconsistencies in light absorption and carrier collection throughout the device area, negatively impacting device performance. The inset of Figure 3.4c also reveals large cracks that extend through the thickness of the device. These defects can provide a path for the top contact to touch the bottom contact, creating a short circuit that destroys the photovoltaic performance by turning the device into a simple conductor. The yield of functioning pixels from devices fabricated using the Paasche airbrush can be expected to be very low given the poor morphological quality of the CQD film, and the performance of the pixels that function as photovoltaic devices is likely to be relatively low since the cracks can also act as shunting paths. After this result, we turned to the Ikeuchi nozzle that is described in Subsection 3.1.1. This nozzle is described as a “fine fog” nozzle, which means that the droplet size is very small (in the range of 10 to 50 µm). The spray coming out of this nozzle is also not very forceful since the ink is atomized to a high degree and the spread of the spray is very wide (Figure 3.2a). This allows the small droplets to gentle settle onto the substrate surface.

Chapter 3. Spray coating of CQD films 27

The effect that these two factors have on the morphology of the film is quite striking. The film in Figure 3.4b and d was fabricated using the same concentration of CQDs and the same number of spray cycles as the one that was made with the Paasche airbrush. The surface of a CQD film deposited by the Ikeuchi nozzle, shown in Figure 3.4b, is much smoother than the film deposited by the Paasche airbrush. This SEM image also shows ring patterns formed by the droplets carrying CQD material in the last spray pulse, indicating that the droplets are around 30 µm or smaller in diameter. A cross-sectional micrograph of a full device shows that the CQD film is quite uniform in thickness, which is around 300 nm in this case (Figure 3.4d). This image also reveals that the film is free of cracks.

Figure 3.4: SEM characterization of films fabricated using Paasche and Ikeuchi nozzles. (a, c) Films from Paasche airbrush. (b, d) Films from Ikeuchi nozzle. (Top) Images of the surface of each CQD film. (Bottom) Cross-sectional images of complete devices. The jagged layer at the bottom is FTO, the dark layer immediately on top is 50 nm of sputtered TiO2. The CQD layer is immediately on top of the TiO2. The layers on top of the CQDs are MoO3 (too thin to be resolved), gold, and silver, which make up the top contact of the device. The inset of (c) shows a characteristic crack of films fabricated using the Paasche brush, and the scale bar is 200 nm.

Chapter 3. Spray coating of CQD films 28

These factors demonstrate that CQD films of good morphological quality can be fabricated using the Ikeuchi nozzle, which is an indication that this nozzle can be used to produce functional photovoltaic devices. The Paasche airbrush does not yield films with sufficient morphological consistency to make high-performing devices. Unless otherwise indicated, the spray equipment used for the CQD deposition step in the rest of this thesis is the Ikeuchi nozzle.

3.3 Conclusions

In this chapter, we reviewed the CQD spray-coating procedure that we developed and how the automation was achieved. We also described the two types of spray nozzles that we used in the setup, both of which are commercially available. After analyzing micrographs of films that were fabricated using each nozzle, it was determined that the high-performance Ikeuchi nozzle should be used for the CQD deposition step. This nozzle is designed for industrial applications, and could easily be integrated into a large-scale system. The Paasche art airbrush is sufficient for the ligand exchange and washing steps of the procedure, which are not very sensitive to droplet size or spray uniformity.

Chapter 4. Material characterization of sprayed CQD films 29

4 Material characterization of sprayed CQD films

4.1 Cross-sectional composition

The purity of the active layer of a solar cell is of utmost importance because it affects device performance. If an impurity is electronically active, it can form a state in the middle of the bandgap of the CQD layer, trapping electronic carriers and aiding in their recombination before they can be collected at the electrodes. It is important to remove any foreign substances from the active layer to minimize unwanted results. An effective way to characterize the purity of a CQD layer is to make a very thin and clean cross-section of the layer using focused ion beam (FIB) milling and obtain high resolution images using transmission electron microscopy (TEM). Not only is this microscopy technique useful for seeing visual abnormalities in the film, but it also enables elemental mapping using electron energy loss spectroscopy (EELS), which can be useful to observe if any elements are concentrated in particular regions in the sample. We first sought to characterize a CQD film fabricated using the standard layer-by- layer spin-coating procedure that was described in Subsection 2.2.1. The FIB TEM image of a typical spin-coated film is show in Figure 4.1a. Each separate layer that was deposited can be individually distinguished due to the presence of a visible dark impurity between CQD layers. We hypothesized that this impurity consisted of leftover organic material from the ligand exchange process. To test this hypothesis, we obtained the EELS spectra of elements that would allow us to distinguish between organic impurities and the quantum dot layer. We picked cadmium (which should only be found on the surface of the quantum dots), sulphur (which is a component of the lead sulphide quantum dots as well the MPA ligand, one of the potential impurities), and carbon (the main component of organic molecules). The EELS spectra of the spin-coated CQD film are shown on the top half of Figure 4.2. It is clear from the intensity image of carbon that it is more heavily concentrated along the dark stripes that are seen between CQD layers in the dark field TEM image. The relative element variation graph on the right end of Figure 4.2 shows that there is a relative increase

Chapter 4. Material characterization of sprayed CQD films 30 of carbon coupled with a relative decrease of cadmium and vice versa at regular intervals along the thickness of the film. Since the presence of cadmium is a signature of CQD content, the alternate concentrations of carbon and cadmium suggest the presence of organic impurities between well-defined CQD layers. Further investigation on the nature of this impurity is provided in Subsection 5.1.1.

Figure 4.1: High-resolution TEM of CQD film cross-sections prepared using FIB milling. The films were fabricated using (a) spin coating, (b) spray coating using concentrated CQD solution, (c) spray coating using dilute CQD solution. The white scale bars represent 500 nm.

The first CQD films of device quality that were fabricated using the Ikeuchi nozzle were made with 15 spray cycles using a solution with a CQD concentration of 16.7 mg/mL. When we characterized these films using high-resolution TEM we observed similar dark stripes of impurities between CQD layers throughout the thickness of the film (Figure 4.1b). However, these dark stripes seemed to be much thinner and less continuous than the ones in the spin-coated film. We hypothesized that if the CQD layer were made very thin, ideally approaching a thickness of a single quantum dot, the methanol rinsing step after ligand exchange would have access to the surface of every quantum dot that was just exchanged, thereby increasing the effectiveness of the removal of the left-over ligand. Having a purer active layer would potentially increase the device performance of a CQD solar cell.

Chapter 4. Material characterization of sprayed CQD films 31

Figure 4.2: EELS cross-sections of spin-coated films (top) and spray-coated films using dilute CQD solution (bottom). Left to right are the dark field, sulfur, cadmium, and carbon signals, along with the relative variations of each element along the thickness of the film (right graph). Adapted with permission from [20]. Copyright (2014) John Wiley and Sons.

In order to achieve thin layers per CQD spray, we reduced the CQD concentration of the ink to 3.33 mg/mL and increased the number of spray cycles to 75 to get a similar final film thickness as before. Figure 4.3 shows the relationship between the thickness of the final film and the number of sprayed layers using the new procedure. The slope of the line of best fit indicates that the average deposited thickness per sprayed layer is around 4.7 nm, which translates to roughly 1.5 quantum dot monolayers per spray pulse given a dot-to-dot spacing of 3.0 nm (calculated later in this chapter). The high-resolution TEM image of the full CQD film fabricated using this method shows that the impurity lines that were present in the spin-coated and thick spray-coated cases were essentially removed (Figure 4.1c). It is evident from the EELS cross-sections of sprayed films (bottom half of Figure 4.2) that the carbon content in the film is no longer concentrated in thick horizontal stripes like the carbon in the spin-coated film. The reduction in impurity content was successfully achieved by using extremely thin CQD layers with vigorous rinsing after ligand exchange. Organic impurities are not expected to be conductive,

Chapter 4. Material characterization of sprayed CQD films 32 so the removal of a periodic cushion of impurities is expected to increase the conductivity of the film along the direction of carrier collection. The relationship between the removal of this impurity and electrical properties that affect photovoltaic performance of devices is explored in Chapter 5. Spraying ultrathin layers was shown to produce purer films than the spray-coating process that deposited thicker layers per pulse. As a result, we decided to continue characterizing films produced from spraying dilute CQD solution. Unless otherwise indicated, this is the spray-coating process mentioned in the rest of this work.

Figure 4.3: Film thickness measured by Dektak profilometer versus the number of layers for sprayed films and a line of best fit. The offset of 78 nm appears as an artefact due to the simultaneous scratching of the CQD film and the underlying FTO film to generate the requisite trench for profilometry. This is confirmed through the zero-layer measurements constituting the substrate alone. Reprinted with permission from [20]. Copyright (2014) John Wiley and Sons.

4.2 Film morphology and physical properties

Film morphology is a significant factor in thin-film solar cell technology because it can affect various aspects of solar cell performance. If the roughness of the film is too high, it can lead to poor or incomplete coverage of the substrate, creating short-circuits between the top and the bottom electrodes that negate the operation of the device as a solar cell. Another aspect that can be affected by film roughness is the deposition of the top contact. At the research scale, the metallic top contacts of a solar cell are deposited using physical vapour deposition methods, which do not provide conformal coatings. As a result, a rough film will create

Chapter 4. Material characterization of sprayed CQD films 33 regions where the top contact is either very thin or missing completely, as shown in Figure 3.4c for CQD films that were spray-coated using the Paasche airbrush. Other morphological defects such as cracking can cause reduction in device performance. In order to characterize the morphology of our spin-coated and spray-coated films we turned to atomic force microscopy (AFM), which is a very sensitive technique that measures surface elevation variations with resolution in the order of fractions of a nanometer. We first applied this technique to measure the morphology of CQD films that were fabricated using our standard spin-coating fabrication procedure (Subsection 2.2.1). An AFM image of the surface of this film can be seen in the top half of Figure 4.4, shown as a top view on the left and as an angled topographical view on the right. The bottom half of Figure 4.4 shows similar AFM images of a film that was fabricated using our spray-coating process. It is immediately evident from the AFM characterization that the spin-coated film contains cracks on the surface, while the surface of the sprayed film does not have such defects. The cracking can be attributed to the ligand exchange process that is performed during the fabrication of the quantum dot films.

Figure 4.4: Top and angled views of spin-coated (top) and spray-coated (bottom) films obtained using AFM. The white scale bars represent 500 nm. Surface roughness (as the standard deviation of surface height) of spun and sprayed films is 2.0 and 1.8 nm, respectively. Adapted with permission from [20]. Copyright (2014) John Wiley and Sons.

When the long oleic-acid ligands that cap the CQDs in solution are exchanged for the shorter MPA molecule in the solid state, the dot-to-dot spacing is reduced in the layer. As a

Chapter 4. Material characterization of sprayed CQD films 34 result, a volume reduction occurs that places tensile forces throughout the film. If strong enough, these forces can cause the film to crack. This effect is evident in spin-coated films, where each layer in the layer-by-layer process is around 30-40 nm thick (Figure 4.1a). On the other hand, our spray-coating process deposits extremely thin layers that average out to be around one quantum dot monolayer per spray pulse. The volume of a sprayed layer is much smaller than that of a spin-coated layer, and thus the volume contraction that is experienced upon ligand exchange is not as dramatic. In addition to being extremely thin, there are 75 spray-coated layers in a CQD film of device thickness. The large number of layers mitigates deleterious effects from volume contraction from any individual layer across the thickness of the film. The cracks that are observed on the spin-coated film do not penetrate the full thickness of the film, as evidenced by the cross-sectional TEM in Figure 4.1a. As a result, their effect would not be as catastrophic as the cracking observed in films that were spray- coated using the Paasche airbrush (Figure 3.4c, inset). However, these cracks could still provide non-ideal shunting paths between individual CQD layers that would adversely affect photovoltaic performance by increasing shunt resistance. The surface roughness of each film was calculated from the standard deviation of surface height in the AFM scan, which gave 2.0 and 1.8 nm for spin-coated and spray-coated films, respectively. However, the number for the spin-coated films takes into account the height deviation caused by the surface cracks. It is evident from the top view of Figure 4.4 that the roughness of the spin-coated film between the surface cracks is lower than the roughness of the spray-coated film, but not significantly so. Even though spin coating can lead to slightly smoother films than our spray-coating process, it also leads to unwanted surface cracking. The AFM apparatus can also be used to map the nanomechanical properties of films, such as the elastic modulus, by using the tip to make an indentation on the surface of the film with a given force and measuring the film’s restorative forces [21]. Using this method, we determined that the elastic modulus of spray-coated CQD films was more than one order of magnitude larger than those of spin-coated films (Figure 4.5). This result means that spray- coated films are harder than their spin-coated counterparts, which can be attributed to the

Chapter 4. Material characterization of sprayed CQD films 35 removal of the thick layers of organic impurities that were evident in spin-coated films (Figure 4.1a).

Figure 4.5: Map of elastic modulus of spin-coated and spray-coated CQD films measured using AFM. Adapted with permission from [20]. Copyright (2014) John Wiley and Sons.

4.3 Dot-to-dot spacing and arrangements

Electrical transport in quantum dot films is viewed as a “hopping” mechanism in which electrical carriers jump from dot to dot [22]. The likelihood of a hopping event to happen is increased when the distance between the nanocrystals is reduced. The goal of the ligand- exchange process described in Subsection 2.2.1 is to decrease this distance, which facilitates hopping events between dots and in turn makes the film more conductive. A higher conductivity in a CQD film increases the diffusion length of the minority carriers, which would allow for better charge extraction at the electrodes and thus better photovoltaic performance. Grazing-incidence small-angle X-ray scattering (GISAXS) using high-energy synchrotron radiation is a technique that can be used to probe the inner structure of films that were fabricated with colloidal nanocrystals [23]. We applied this technique to measure the spatial arrangement of quantum dots in spin-coated and spray-coated films (Figure 4.6a). The spray-coated films in this particular study were fabricated using the Ikeuchi nozzle but with thick layers, as shown in Figure 4.1b.

Chapter 4. Material characterization of sprayed CQD films 36

The inter-particle spacing information that was extracted from the GISAXS data is plotted in Figure 4.6b. The distance from the centre of one quantum dot to another was determined to be 3.1 ± 0.2 nm for spin-cast films and 3.0 ± 0.2 nm for spray-cast films. These values fall within the uncertainty error of each other, but nevertheless indicate closer packing for spray-coated films. The packing of films fabricated using dilute CQD solution is expected to be even closer because of the effectiveness of this process at removing organic impurities.

Figure 4.6: Inter-particle spacing information obtained using GISAXS. (a) GISAXS plots of spin and unoptimized spray. (b) Inter-particle spacing given by azimuthal integration and (c) isotropicity in spacing given by radial integration of the intensity plots from (a). Adapted with permission from [20]. Copyright (2014) John Wiley and Sons.

In addition to differences in dot-to-dot distance, the spatial arrangement of the quantum dots also differs. The out of plane ordering of the quantum dots in the spray-coated case is better than in the spin-coated case (Figure 4.6c). This means that spray coating leads to improved particle packing in the direction of charge transport and extraction, which should lead to better photovoltaic performance than in the spin-coated films. The better packing in this direction is attributed to the better removal of organic impurities between CQD layers using spray coating, which was explored earlier in this chapter.

4.4 Conclusions

In this chapter we found that by using spray coating to deposit ultra-thin CQD layers we are able to remove thick layers of organic impurities that are commonly found in spin-coated films. This approach resulted in crack-free films that are one order of magnitude harder than their spin-coated counterparts. In addition, spray-coated films show better quantum dot

Chapter 4. Material characterization of sprayed CQD films 37 packing than spin-cast films, especially in the direction of charge collection. All of these improvements in material properties and composition are indications that fabricating films using spray-coating may, in fact, lead to better performing photovoltaic devices than the standard batch-scale spin-coating process.

Chapter 5. Spray-coated CQD photovoltaic devices 38

5 Spray-coated CQD photovoltaic devices

5.1 Improving electronic properties of films via spray coating

In Chapter 4 we found that our spray-coating method of CQD films fabrication resulted in better material properties than the standard spin-coating process. One of the most significant achievements was the removal of an impurity that appeared as dark stripes in high-resolution TEM images of spin-coated films (Figure 4.1). It was hypothesized that the removal of this impurity would lead to better electronic properties in the film. In this chapter, we will characterize the electronic defects that can be observed in CQD films and will explore how the spray-coating process helps their elimination, creating films with longer minority carrier diffusion lengths.

5.1.1 Identification and elimination of electronic defect

One way in which electronic traps in photovoltaic devices can be investigated is by applying a forward bias to the device and capturing any light emission from the pixel into a spectrometer. When we characterized the device as a light-emitting diode, any mid-gap states in the CQD film that facilitate charge recombination will manifest themselves as an emission peak in a longer wavelength than the wavelength that corresponds to the bandgap energy. The results of applying forward current to CQD solar cells that were fabricated via spin coating and spray coating are shown in Figure 5.1. All solar cells had the standard

FTO/TiO2 bottom contact and MoO3/Au/Ag top contact. The spectra for both fabrication processes display a large emission peak at around 1100 nm, which corresponds to the bandgap energy of the CQD film. The spectrum corresponding to the spin-coated device has a weak peak at around 1600 nm, which is lower in energy than the main bandgap emission peak. This is evidence of a mid-gap trap state that serves as a recombination pathway for carriers in the CQD film. The spectrum generated by the spray-coated device does not have the lower-energy emission, indicating that the spray-coating process is effective at removing the source of the trap state.

Chapter 5. Spray-coated CQD photovoltaic devices 39

Figure 5.1: Electroluminescence measurements of spin-coated (top) and sprayed (bottom) CQD films, illustrating the presence of an electronic defect in the spin-coated film as manifested by a peak at around 1600 nm. Adapted with permission from [20]. Copyright (2014) John Wiley and Sons.

We then turned to solid-state nuclear magnetic resonance (NMR) – a sensitive technique that helps identify chemical groups in a sample – to gain insight into the nature of the chemical species that causes the trap state. We found that the main difference between the two deposition methods is that the spray-coated samples displayed a peak that corresponds to protonated carboxylic acid group (COOH) (Figure 5.2a). The carboxylic acid functional group is the free end of the MPA ligand, and if it is protonated it means that it is not part of a cross-linked complex that could potentially cause an electronic trap. Fourier transform infrared spectroscopy (FTIR) is a different technique that measures bond energies in chemical species of samples. This technique also shows a marked increase in COOH in the spray-coated film and a higher concentration of deprotonated carboxylate (COO-) in the spin- coated film (Figure 5.2b). The deprotonated group can bind itself to other species because of its negative charge, potentially forming a complex that causes the mid-gap trap state. In order to identify the impurity, we turned to density functional theory (DFT) to simulate various molecules that would likely be around the vicinity of our quantum dots, which would help to determine which one would cause an electronic trap state in the middle of the bandgap of the quantum dot. Figure 5.3a shows the ideal case in which MPA is the only organic species on the quantum dot surface. The DFT density of states calculation shows a clean quantum dot bandgap between around -4 and -5 eV. We also simulated the

Chapter 5. Spray-coated CQD photovoltaic devices 40 case of a molecule containing deprotonated carboxylic acid groups on the surface of the quantum dot, such as a deprotonated MPA molecule, and the mid-gap trap was not created. We came to the conclusion that the trap state is most likely caused by a complex in which the COO- group is bound to another charged species.

Figure 5.2: Impurity characterization using NMR and FTIR. (a) Proton solid-state NMR measurements of spin-coated films (top) and of spray-coated films (bottom) (solid lines). Peak fits indicating phosphonic acid (~12 ppm), the MPA carbon chain (~0 to 2 ppm), and residual methanol and water (~3 ppm and ~5 ppm, respectively) are represented by dashed lines. The presence of a protonated carboxylic acid is observed in the sprayed case at ~8 ppm (highlighted in green). (b) FTIR spectroscopy of the same film types as (a). The acid peak shifts toward a deprotonated carboxylate (shoulder at ~1500-1550 cm-1) in the spin case, whereas the increase at ~1600-1700 cm-1 for spray indicates the enhanced presence of carboxylic acid. Adapted with permission from [24]. Copyright (2014) American Chemical Society.

Upon further examination of our quantum dot system, we realized that a molecule that could potentially remain in the final film is lead oleate, which is one of the by-products that are formed during the chemical synthesis of CQDs when lead oxide and oleic acid are reacted together [4]. Lead oleate can bind to deprotonated MPA to form a complex that displays a mid-gap trap state when simulated in the vicinity of a quantum dot (Figure 5.3b). To test this theory we added excess lead oleate to the CQD solution and spray-coated a full solar cell device. When forward current was applied to this device, the light emission spectrum has a prominent peak at a wavelength of around 1600 nm (Figure 5.3c), which matches well with the trap state emission observed for spin-coated devices (Figure 5.1, top). This study identified the nature of the source of the deleterious mid-gap trap state. It is evident that our spray-coating process is effective at eliminating the trace amounts of the MPA-lead oleate complex that causes this electronic flaw. By being able to spray ultrathin layers and having access to the surface of almost every quantum dot, the methanol-rinsing

Chapter 5. Spray-coated CQD photovoltaic devices 41 step between layers is able to wash away the complex at the concentrations that are normally found in the CQD solution. The effect that the elimination of this defect has on the electronic properties of the film and the photovoltaic device efficiency is explored in the following sections of this chapter.

Figure 5.3: Identification of electronic defect. (a,b) Atomic model of organic species (left) and density of states calculations when each respective molecule is at the surface of a PbS quantum dot (right). (a) The ideal ligand exchange case where MPA is the only organic molecule bound on the quantum dot surface. (b) A case in which a complex of lead oleate and MPA is near the quantum dot surface, showing a mid-gap trap state at around -4.5 eV. (c) Electroluminescence of spray-coated film with excess lead oleate added to the CQD solution, showing the emergence of the trap state emission peak at ~1600 nm. Adapted with permission from [24]. Copyright (2014) American Chemical Society.

5.1.2 Minority carrier diffusion length

The distance that an excited carrier can diffuse in a semiconductor before recombining is referred to as the minority carrier diffusion length (LD) and it is crucial for photovoltaic device operation because it affects how many of the photogenerated carriers will be collected at the electrodes and contribute to the generated current. This quantity is influenced by electrical transport in the film and recombination. As discussed in Chapter 4, reduced inter- particle spacing in the film should aid transport, thereby increasing LD. A higher mid-gap trap density in a film leads to a shorter minority carrier diffusion length because the likelihood of a diffusing carrier encountering a recombination centre is increased, which decreases the average distance that is travelled by photogenerated charges in the quasi-neutral region [25]. Hence, it is important to reduce the mid-gap trap density and the dot-to-dot spacing for efficient photovoltaic device operation.

Chapter 5. Spray-coated CQD photovoltaic devices 42

In Subsection 5.1.1 we demonstrated that our spray-coating process was more effective than the standard spin-coating process at removing a recombination centre caused by an MPA-lead oleate organic complex. We sought to characterize the effect of the trap removal by measuring the diffusion lengths of spray-coated and spin-coated films using the method of Zhitomirsky et al. [25]. In short, this method consists of the deposition of a thin film of small-bandgap CQDs (referred to as a reporter layer) on top of photovoltaic-quality CQD films of different thicknesses. The large-bandgap film is illuminated by a laser to generate charge carriers while the photoluminescence (PL) spectrum is measured. The PL intensity of the reporter layer, manifested by a longer-wavelength peak, will depend on the thickness and the diffusion length of the CQD layer of interest. By varying the film thickness and plotting the long-wavelength PL intensity, a mathematical fit can be applied to extract LD of the larger-bandgap CQD film [25]. The results of performing this procedure to measure the diffusion length of spin-coated and spray-coated films are shown in Figure 5.4.

Figure 5.4: Photoluminescence of reporter layer on top of spin-coated and spray-coated CQD films to extract minority carrier diffusion length (LD). The films exhibit LD of approximately 80 and 100 nm, respectively, according to the method of Zhitomirsky et al [25]. Adapted with permission from [20]. Copyright (2014) John Wiley and Sons.

The curve that corresponds to spray-coated films in Figure 5.4 has a longer right tail than the curve representing the spin-coated films, which means that at any given thickness, the photoluminescence of the reporter layer is stronger for spray-cast films. This can be

Chapter 5. Spray-coated CQD photovoltaic devices 43 attributed to more photogenerated carriers reaching the reporter layer after diffusing through the thickness of the film, which implies a longer diffusion length. After applying the mathematical fit, the extracted diffusion lengths were 100 nm and 80 nm for spray-coated and spin-coated films, respectively. The effective removal of the trap-causing impurity and better dot-to-dot packing (Section 4.3) achieved by spray coating leads to a diffusion length that is 25% longer than the standard batch-scale method of spin coating. This is achieved by being able to deposit layers that approach one CQD layer in thickness, and thus having more access to the CQD surface for washing away the MPA-lead oleate complex. The impact that the longer diffusion length has on device performance will be discussed in the following section.

5.2 Characterization of photovoltaic devices

The end goal of making CQD films via a spray-coating process is to allow for the mass- production of solar cells. We have found in previous sections that the spray-coating process we developed produced films with better morphology, fewer electronic impurities, and longer minority carrier diffusion lengths than the standard spin-coating process. These systemic improvements can contribute to more efficient photovoltaic devices in addition to the inherent applicability to large-scale manufacturing. All photovoltaic devices were characterized under AM1.5 illumination, which was limited to 4.9 mm2 on each pixel by a circular aperture. The maximum power point (MPP) was determined from current-voltage (J-V) and power-voltage (P-V) curves and the reported power conversion efficiency (PCE) was determined by measuring the generated power when the bias was held at the MPP for 7 seconds. In Figure 5.5, we show a histogram of the PCE of the best pixel of every device that was spray-coated using our best spray-coating process alongside a histogram of every spin- coated device fabricated by our group that used the standard architecture and quantum dot type. The data are normalized to the area under the Gaussian fits. It is clear from this plot that our spray-coating procedure produces devices with a higher average PCE than the standard spin-coating procedure (6.5% and 5.2%, respectively). In addition, the distribution of the performance of sprayed devices is narrower than that of the performance of spin-coated

Chapter 5. Spray-coated CQD photovoltaic devices 44 devices. This achievement can be attributed to the automated nature of our spray-coating process, which does not suffer from the user-to-user variability of the manual spin-coating procedure.

Figure 5.5: Histograms of power conversion efficiency under AM1.5 illumination for spin and spray- coated photovoltaic devices with Gaussian fits overlaid. Data series were normalized to the Gaussian area to accentuate the higher mean value and narrower distribution of sprayed samples as compared with spin-coated ones. Adapted with permission from [20]. Copyright (2014) John Wiley and Sons.

Since the population of the spin-coated devices was much higher than the spray- coated device number, we performed a Welch’s t-test to determine if the higher mean performance of the spray-coating devices was an artifice caused by the smaller sample population. A p value of 0.0001 was obtained, proving, with 99.99% confidence, that our spray-coated device data are part of a separate population from the spin-coated devices and that spray-coated devices have statistically better performance. These results corroborate our hypothesis that the more thorough removal of the trap-inducing organic impurity, in conjunction with better film morphology, would lead to higher photovoltaic performance from spray coating than spin coating. The photovoltaic performance of the best spray-coated device is shown in Figure 5.6a. The measured static PCE was 8.1%, which was, at the time, a performance record for a solar cell employing a spray-coated active layer deposited under ambient conditions. The calculated short-circuit current density (JSC) from the external 2 quantum efficiency (EQE) curve matches the measured JSC of approximately 23 mA/cm , confirming the appropriate use of conversion factor in matching the lamp spectrum to the solar spectrum [26].

Chapter 5. Spray-coated CQD photovoltaic devices 45

Figure 5.6: Photovoltaic performance of best sprayed CQD solar cell. (a) J-V and P-V curves under AM1.5 conditions with static PCE at 8.1% illustrated with a round marker. (b) EQE curve resulting in 2 a predicted JSC of around 23 mA/cm matching the measured JSC under AM1.5 illumination. Adapted with permission from [20]. Copyright (2014) John Wiley and Sons.

5.3 Conclusions

In this chapter, we demonstrated that the minority carrier diffusion length in CQD films was enhanced from 80 nm to 100 nm when using the spray-coating procedure due to a smaller dot-to-dot distance and the efficient removal of an MPA-lead oleate complex that caused an electronic trap in spin-coated films. The improved diffusion length of spray-coated films resulted in better average photovoltaic performance than spin-coated devices (6.5% versus 5.2% PCE). In addition, the best spray-coated device achieved a PCE of 8.1%, which was a performance record for solar cells with a spray-coated active layer deposited under ambient conditions. These results demonstrate that there does not need to be a trade-off between performance and manufacturability when using spray coating as a fabrication process for CQD photovoltaics, which is a compromise that limits the wide deployment of other solution-processed solar cell technologies.

Chapter 6. Applicability to large-scale manufacturing 46

6 Applicability to large-scale manufacturing

6.1 Roll-to-roll process simulation

Spray coating is a technique that is compatible with roll-to-roll (R2R) manufacturing, a scale- up production process during which a flexible roll passes quickly through various deposition methods with the help of cylindrical rollers. We sought to simulate the R2R process to test the adaptability of our spray-coating protocol to large-scale manufacturing, which was the initial motivation for the development of this processing technique. In order to simulate R2R processing, we secured six substrates to each side of a hexagonal drum. The drum was then made to rotate at 200 rpm while the spray-coating process was being performed on its surface (Figure 6.1). Since all the nozzles were pointed at the same point in space, every step of the spray-coating procedure started after a time period that was a multiple of 0.3 s (the inverse of the rotational speed) after the start of the previous step. This way, each step would start on the same substrate and the treatment would be the same for all six samples. Since each side of the hexagonal drum was 5 cm in length, this test was analogous to a roll of substrate passing through the spray coating process at a speed of 1 meter per second.

Figure 6.1: Illustration of the R2R simulation setup. Six planar glass substrates (red rectangles) with FTO and TiO2 layers were affixed to the sides of a hexagonal drum that rotated at 200 rpm while our spray-coating procedure occured. Adapted with permission from [27]. Copyright (2014) AIP Publishing LLP.

Chapter 6. Applicability to large-scale manufacturing 47

Each of the substrates was a planar square of FTO-covered glass with a TiO2 n-type layer on top, the same kind of substrate that was used for the solar cell studies described in Chapter 5. After a final-device thickness was reached in the CQD deposition step on the hexagonal drum, the samples had the standard top contacts of MoO3/Au/Ag evaporated through a shadow mask to mark 16 distinct pixels. The photovoltaic performance under AM1.5 illumination conditions of the best pixel of each of the six substrates is shown in Figure 6.2. The J-V plots show that the six samples were remarkably consistent, with an average PCE of 6.7% and a small standard deviation of

0.4 percentage units (Figure 6.2, left). The static PCE, VOC, and JSC plots highlight the stability of the device performance over a period of seven seconds (Figure 6.2, right). These performance metrics fall within one standard deviation of the distribution of spray-coated solar cells that were fabricated while the samples were stationary (Figure 5.5), demonstrating that the effectiveness of our spray-coating process is not diminished when depositing a film onto a moving substrate. These findings support the claim that our spray-coating procedure can be adapted to R2R processing to fabricate consistently efficient CQD solar cells at high throughputs.

Figure 6.2: Photovoltaic characterization of devices that were fabricated while mounted to a rotating drum (each sample is given a different color). (left) J-V curves (right) Figures of merit over time highlighting the small variation over the six samples. Static measurements illustrate the absence of transient or hysteretic behaviour. Data points are color coded to the J-V curves on the left. Adapted with permission from [27]. Copyright (2014) AIP Publishing LLC.

Chapter 6. Applicability to large-scale manufacturing 48

6.2 Flexible devices

6.2.1 Importance of flexible solar cells

One of the main selling points of solution-processed semiconductors is their ability to produce flexible devices, something that is not possible with conventional crystalline semiconductors. Physical flexibility opens up many opportunities for the deployment of solar cells. First of all, flexible devices are widely portable because they can be folded or rolled to get around space constraints in suitcases and bags. They are also not as heavy as crystalline silicon solar cells because the substrate is thin plastic. These characteristics would make it more likely for an end user to take the solar cell wherever they needed it, such as going camping or on a trip. In addition to portability, R2R processing requires a flexible substrate so that it can be wound and moved to and from different processing steps. In order for us to continue to evaluate the viability of our spray-coating process for scale-up manufacturing, we need to demonstrate the successful fabrication of efficient solar cells on flexible substrates.

6.2.2 Flat versus curved spraying configuration

We first sought to fabricate a flexible solar cell on a polyethylene terephthalate (PET) substrate with pre-deposited indium tin oxide (ITO), a transparent conductive oxide. We cut this material into a square piece with one-inch sides, the standard substrate size in our lab, for easier processing. 50 nm of TiO2 were sputtered onto the substrate, but it was not treated with

TiCl4 solution or sintered because this treatment would damage the PET substrate, rendering the device unusable. We then applied our standard CQD spray-coating process while the substrate was lying flat on the sample holder and later evaporated our standard MoO3/Au/Ag top contact. We characterized the device under AM1.5 conditions to test its photovoltaic performance. We found that that the VOC of the flexible device was around 10% lower than the typical sprayed device on a rigid substrate (0.56 V compared to 0.61 V), and the fill factor was much lower as well (42% compared to 52%). We hypothesized that while handling the flexible device between fabrication and testing, any bending would create some

Chapter 6. Applicability to large-scale manufacturing 49

strain in the active layers and create recombination and shunt paths, reducing VOC and decreasing shunt resistance, which would in turn reduce the fill factor. To test this, we measured the dark J-V characteristics of the device and found poor rectification (Figure 6.4a, red curve). We further hypothesized that if all the active layers of the device were deposited while the substrate was flexed with the surface of interest on the outside, then when unflexed as the final device, the planar films would be under compressive strain that would mitigate any tensile forces that would arise from the device being flexed during manipulation. To test this hypothesis, we wrapped a piece of the ITO-covered PET substrate around a 2 cm diameter dowel. We put this object into the sputtering chamber to deposit the TiO2 layer, and then applied our spraying process on the still-flexed substrate (Figure 6.3). We then unwrapped the substrate, cut it to standard size, and deposited our standard top contacts. Figure 6.4c shows a picture of the final device, demonstrating its flexibility.

Figure 6.3: Illustration of a flexible PET substrate (blue crescent) with ITO and TiO2 layers wrapped around a 2 cm dowel (black circle) while being sprayed on using our spray-coating process. Adapted with permission from [27]. Copyright (2014) AIP Publishing LLC.

We measured the dark J-V characteristics of this device and the rectification ratio was about two orders of magnitude better than the ratio corresponding to the flexible device that was sprayed while flat (Figure 6.4a), which means that reverse current was more efficiently rejected. The fill factor of the device that was sprayed while flexed was 54% (Figure 6.4b), an increase of 12 percentage units over the sprayed-flat device, indicating that shunting paths were minimized by processing the active layer while the substrate was flexed. The VOC and PCE were 0.59 V and 7.2%, respectively, in good agreement with our typical performance on rigid glass substrates (Figure 5.5). Therefore, performance does not suffer when applying our spray-coating method onto substrates that are commonly used in the R2R process, particularly when they are flexed during the deposition process.

Chapter 6. Applicability to large-scale manufacturing 50

Figure 6.4: Characterization of flexible sprayed devices. (a) Dark I-V curves of devices sprayed on unflexed (red) and flexed (green) ITO-coated PET sheets. (b) Illuminated J-V curve of the best flexible device (green curve from (a)). (c) A photograph of a finished device that was sprayed on a flexible PET substrate while being wrapped around a 2 cm diameter dowel. The device was unwrapped and deposited with sixteen 6.7 mm2 devices (illuminated using 4.9 mm2 of light) on the same substrate. Reprinted with permission from [27]. Copyright (2014) AIP Publishing LLC.

6.3 Spraying solar cells on unconventional form factor substrates

6.3.1 Application of solar cells onto surfaces with multiple curvatures

Wide deployment of solar harvesting devices might be achieved if photovoltaic devices could be integrated onto surfaces with a variety of form factors. For example, solar cells on car roofs could provide the energy to power the electronics inside, and could aid in extending the driving range of electric cars that are currently proliferating in the market. Solar cells on airplane roofs and wings would provide similar applications. An interesting concept would be if we could coat balloons and kites with solar cells and send them high in the sky to collect sunlight and send the electricity to the ground by the supporting wires. Any solar cell that would be used for these applications would need to be lightweight and, perhaps harder to achieve, would need to conform to surfaces with multiple dimensions of curvature. Conventional crystalline solar cells fail on both of these requirements. Solution- processed semiconductors offer an attractive option to coat these surfaces with solar cells. However, most solution deposition techniques appropriate for mass production, such as slot- die or knife-over-edge coating, can only accommodate one dimension of curvature. This is

Chapter 6. Applicability to large-scale manufacturing 51 why these techniques work well when coating a flexible substrate rolling on a cylindrical drum in R2R processing. On the other hand, spray coating allows for the deposition of material onto surfaces with any form factor as long as the line of sight is uninterrupted from the nozzle to all the points of the surface of interest. We have demonstrated the ability of our spray-coating process to produce efficient CQD solar cells, which could potentially be applied onto many unconventional surfaces for the wider deployment of photovoltaic devices.

6.3.2 Sprayed solar cells on spherical lenses

In order to test the ability of our spray-coating process to create device-quality films on surfaces with more than one dimension of curvature, we sprayed films onto spherical lenses with a radius of curvature of 13.1 mm (Thorlabs, LA1252, Figure 6.5). The lenses were made into working solar cell substrates by depositing 250 nm of ITO via heated sputtering, after which the standard 50 nm of TiO2 were sputtered. The substrates were then treated with

TiCl4 solution and sintered to make a more conductive TiO2 layer.

Figure 6.5: Illustration of a spherical lens substrate (Thorlabs, LA1252, blue semicircle) with ITO and TiO2 layers being sprayed on using our spray-coating process. Adapted with permission from [27]. Copyright (2014) AIP Publishing LLC.

The top and middle pictures of Figure 6.6a show images of the spherical substrate after the deposition of a CQD film using our standard spray-coating process. The coating is smooth and even throughout the area around the apex of the dome. We then evaporated our standard top contacts on the films through special masks that allowed for the definition of 4 distinct circular pixels of 6.7 mm2 (Figure 6.6a, bottom). The completed device achieved a PCE of 5.0%, demonstrating the ability of our spray-coating process to produce efficient CQD photovoltaic devices on unconventional form factors.

Chapter 6. Applicability to large-scale manufacturing 52

Figure 6.6: Characterization of sprayed spherical device. (a) Top and side views of a spray-coated spherical lens illustrating coverage along multiple axes of curvature (top and middle) and a contacted spray-coated spherical lens solar cell device (bottom). (b) J-V curve of the best spherical device. Adapted with permission from [27]. Copyright (2014) AIP Publishing LLC.

6.4 Conclusions

In this chapter we demonstrated that the spray-coating procedure that we developed is indeed compatible with mass-manufacturing methodologies. We simulated the integration of our spray-coating process into R2R manufacturing by spraying onto substrates that were attached to a rapidly rotating drum, yielding devices that performed just as well as our stationary- fabricated devices described in Chapter 5. We also sprayed onto a flexible substrate and produced an efficient CQD device. It is conceivable to extrapolate these results to a spraying onto a rapidly moving flexible web in a R2R manufacturing setup. In addition to the adaptability of spray coating to large-area manufacturing, we showcased the unique ability of this process to coat surfaces with form factors other than planar or cylindrical, which would enable a wider deployment of solar cell technology. We fabricated a working CQD device on a spherical lens, demonstrating the versatility of the process we developed.

Chapter 7. Summary 53

7 Summary

7.1 Thesis findings and conclusions

The efficiency of colloidal quantum dot photovoltaics has rapidly increased in recent years, surpassing 8% power conversion efficiency [28]. Though they offer the promise of low-cost photovoltaics via large-scale manufacture, most studies limit the fabrication of the active layer of CQD solar cells to batch-scale processing methods such as spin coating. These fail to represent the device performance that could be achieved by mass-fabrication procedures. This thesis presents major contributions to the field as follows: 1. We developed an automatic spray-coating procedure for CQD deposition and ligand exchange steps, building a thick film in layer-by-layer fashion. We show that by choosing a fine-mist spray nozzle for the CQD deposition step, films of sufficient morphological quality (roughness of 2.0 nm and no cracks) could be produced. 2. It was found herein that by spraying ultrathin layers, approaching a CQD monolayer in thickness, we were able to fabricate dense, crack-free films with lower levels of organic impurities than spin-coated films. These improvements resulted in CQD films with longer minority carrier diffusion length, a key determinant of photovoltaic efficiency. 3. We demonstrated that we could fabricate devices with statistically better photovoltaic performance using our spray-coating process than the standard spin- coating procedure. This is a remarkable achievement because it indicates that there does not need to be a compromise between manufacturability and device performance for CQD solar cells. A hero PCE of 8.1% was achieved, which was a record for a device with a spray-coated active layer fabricated under ambient conditions. 4. In this work we simulated a R2R process by spray coating onto substrates affixed to a rapidly rotating drum, which resulted in devices with performance that fell under the statistical distribution of the devices that were fabricated while being stationary. We also demonstrated the versatility of our process by spraying onto curved

Chapter 7. Summary 54

surfaces, resulting in an efficient flexible device and a working device on a spherical lens. This was evidence that by using spray coating we could apply solar cells into a variety of previously unexplored surfaces for the wide deployment of CQD photovoltaics.

7.2 Future work

Here we present some suggestions for future work: 1. We used the relatively small illumination aperture size of 4.9 mm2 (0.049 cm2) throughout the device studies because bigger pixel sizes increase the likelihood of shorted devices when the CQD film is spin-coated. A pixel that allows for illumination of at least 1 cm2 is needed for the industry to start considering a particular photovoltaic platform. We showed in this thesis that our spray-coating method results in CQD films with no surface cracking compared to spin-coated films. Larger pixel sizes should be a possibility with spray-coated films, and should be explored in order to increase the relevance of CQD photovoltaic technology to the renewable energy industry. 2. This work used the best device architecture available at the time of the project. However, recent architecture breakthroughs have pushed CQD photovoltaic efficiencies to above 9% while still using spin coating [29]. A spray-coating process should be developed for this architecture to test its adaptability to mass production. If similar improvements on film quality are attained using spray coating of this architecture, the device performance could surpass 11%. 3. R2R processing was simulated in this work to demonstrate that spray coating of CQD films is compatible with this mass production protocol. The next step in realizing large-scale production of CQD solar cells is to adapt spray coating to a true R2R setup, such as the one described by Krebs et al. in their review of processing methods for solution-processed electronics [12]. The film quality, device performance, and production cost should be evaluated in order to determine the feasibility of mass production of CQD solar cells using spray coating.

References 55

References

[1] “International Energy Statistics.” [Online]. Available: http://www.eia.gov/cfapps/ipdbproject/IEDIndex3.cfm. [Accessed: 10-Feb-2015]. [2] M. Graetzel, R. A. J. Janssen, D. B. Mitzi, and E. H. Sargent, “Materials interface engineering for solution-processed photovoltaics,” Nature, vol. 488, no. 7411, pp. 304– 312, Aug. 2012. [3] A. H. Ip, S. M. Thon, S. Hoogland, O. Voznyy, D. Zhitomirsky, R. Debnath, L. Levina, L. R. Rollny, G. H. Carey, A. Fischer, K. W. Kemp, I. J. Kramer, Z. Ning, A. J. Labelle, K. W. Chou, A. Amassian, and E. H. Sargent, “Hybrid passivated colloidal quantum dot solids,” Nat. Nanotechnol., vol. 7, no. 9, pp. 577–582, Jul. 2012. [4] M. A. Hines and G. D. Scholes, “Colloidal PbS Nanocrystals with Size-Tunable Near- Infrared Emission: Observation of Post-Synthesis Self-Narrowing of the Particle Size Distribution,” Adv. Mater., vol. 15, no. 21, pp. 1844–1849, Nov. 2003. [5] S. A. McDonald, G. Konstantatos, S. Zhang, P. W. Cyr, E. J. D. Klem, L. Levina, and E. H. Sargent, “Solution-processed PbS quantum dot infrared photodetectors and photovoltaics,” Nat. Mater., vol. 4, no. 2, pp. 138–142, Feb. 2005. [6] A. G. Pattantyus-Abraham, I. J. Kramer, A. R. Barkhouse, X. Wang, G. Konstantatos, R. Debnath, L. Levina, I. Raabe, M. K. Nazeeruddin, M. Graetzel, and E. H. Sargent, “Depleted-Heterojunction Colloidal Quantum Dot Solar Cells,” ACS Nano, vol. 4, no. 6, pp. 3374–3380, Jun. 2010. [7] A. R. Madaria, A. Kumar, and C. Zhou, “Large scale, highly conductive and patterned transparent films of silver nanowires on arbitrary substrates and their application in touch screens,” Nanotechnology, vol. 22, no. 24, p. 245201, Jun. 2011. [8] H. Liu, J. Tang, I. J. Kramer, R. Debnath, G. I. Koleilat, X. Wang, A. Fisher, R. Li, L. Brzozowski, L. Levina, and E. H. Sargent, “Electron Acceptor Materials Engineering in Colloidal Quantum Dot Solar Cells,” Adv. Mater., pp. 3832–3837, Jul. 2011. [9] P. Liu, Y. Wu, Y. Li, B. S. Ong, and S. Zhu, “Enabling Gate Dielectric Design for All Solution-Processed, High-Performance, Flexible Organic Thin-Film Transistors,” J. Am. Chem. Soc., vol. 128, no. 14, pp. 4554–4555, Apr. 2006. [10] Q. Guo, G. M. Ford, H. W. Hillhouse, and R. Agrawal, “Sulfide Nanocrystal Inks for Dense Cu(In1−xGax)(S1−ySey)2 Absorber Films and Their Photovoltaic Performance,” Nano Lett., vol. 9, no. 8, pp. 3060–3065, Aug. 2009. [11] L. E. Scriven, “Physics and applications of dip coating and spin coating,” in MRS Proceedings, 1988, vol. 121, p. 717. [12] F. C. Krebs, “Fabrication and processing of polymer solar cells: A review of printing and coating techniques,” Sol. Energy Mater. Sol. Cells, vol. 93, no. 4, pp. 394–412, Apr. 2009. [13] D. R. Guildenbecher, C. López-Rivera, and P. E. Sojka, “Droplet Deformation and Breakup,” in Handbook of Atomization and Sprays, N. Ashgriz, Ed. Boston, MA: Springer US, 2011, pp. 145–156. [14] H.-Y. Park, K. Kim, D. Y. Kim, S.-K. Choi, S. M. Jo, and S.-Y. Jang, “Facile external treatment for efficient nanoscale morphology control of polymer solar cells using a gas- assisted spray method,” J. Mater. Chem., vol. 21, no. 12, p. 4457, 2011.

References 56

[15] P. Kumar, S. Kannappan, S. Ochiai, and P.-K. Shin, “High-performance organic solar cells based on a low-bandgap poly-thienothiophene-benzodithiophene polymer and fullerene composite prepared by using the airbrush spray-coating technique,” J. Korean Phys. Soc., vol. 62, no. 8, pp. 1169–1175, Apr. 2013. [16] V. A. Akhavan, B. W. Goodfellow, M. G. Panthani, D. K. Reid, D. J. Hellebusch, T. Adachi, and B. A. Korgel, “Spray-deposited CuInSe2 nanocrystal photovoltaics,” Energy Environ. Sci., vol. 3, no. 10, p. 1600, 2010. [17] P. Peumans, S. Uchida, and S. R. Forrest, “Efficient bulk heterojunction photovoltaic cells using small-molecular-weight organic thin films,” Nature, vol. 425, no. 6954, pp. 158–162, Sep. 2003. [18] G. Dennler, M. C. Scharber, and C. J. Brabec, “Polymer-Fullerene Bulk- Heterojunction Solar Cells,” Adv. Mater., vol. 21, no. 13, pp. 1323–1338, Apr. 2009. [19] Y. Liang, Z. Xu, J. Xia, S.-T. Tsai, Y. Wu, G. Li, C. Ray, and L. Yu, “For the Bright Future-Bulk Heterojunction Polymer Solar Cells with Power Conversion Efficiency of 7.4%,” Adv. Mater., vol. 22, no. 20, pp. E135–E138, May 2010. [20] I. J. Kramer, J. C. Minor, G. Moreno-Bautista, L. Rollny, P. Kanjanaboos, D. Kopilovic, S. M. Thon, G. H. Carey, K. W. Chou, D. Zhitomirsky, A. Amassian, and E. H. Sargent, “Efficient Spray-Coated Colloidal Quantum Dot Solar Cells,” Adv. Mater., vol. 27, no. 1, pp. 116–121, Jan. 2015. [21] K. E. Mueggenburg, X.-M. Lin, R. H. Goldsmith, and H. M. Jaeger, “Elastic membranes of close-packed nanoparticle arrays,” Nat. Mater., vol. 6, no. 9, pp. 656–660, Sep. 2007. [22] P. Guyot-Sionnest, “Electrical Transport in Colloidal Quantum Dot Films,” J. Phys. Chem. Lett., vol. 3, no. 9, pp. 1169–1175, May 2012. [23] A. E. Saunders, A. Ghezelbash, D.-M. Smilgies, M. B. Sigman, and B. A. Korgel, “Columnar Self-Assembly of Colloidal Nanodisks,” Nano Lett., vol. 6, no. 12, pp. 2959– 2963, Dec. 2006. [24] G. H. Carey, I. J. Kramer, P. Kanjanaboos, G. Moreno-Bautista, O. Voznyy, L. Rollny, J. A. Tang, S. Hoogland, and E. H. Sargent, “Electronically Active Impurities in Colloidal Quantum Dot Solids,” ACS Nano, vol. 8, no. 11, pp. 11763–11769, Nov. 2014. [25] D. Zhitomirsky, O. Voznyy, S. Hoogland, and E. H. Sargent, “Measuring Charge Carrier Diffusion in Coupled Colloidal Quantum Dot Solids,” ACS Nano, vol. 7, no. 6, pp. 5282–5290, Jun. 2013. [26] H. J. Snaith, “The perils of solar cell efficiency measurements,” Nat. Photonics, vol. 6, no. 6, pp. 337–340, May 2012. [27] I. J. Kramer, G. Moreno-Bautista, J. C. Minor, D. Kopilovic, and E. H. Sargent, “Colloidal quantum dot solar cells on curved and flexible substrates,” Appl. Phys. Lett., vol. 105, no. 16, p. 163902, Oct. 2014. [28] P. Maraghechi, A. J. Labelle, A. R. Kirmani, X. Lan, M. M. Adachi, S. M. Thon, S. Hoogland, A. Lee, Z. Ning, A. Fischer, A. Amassian, and E. H. Sargent, “The Donor– Supply Electrode Enhances Performance in Colloidal Quantum Dot Solar Cells,” ACS Nano, vol. 7, no. 7, pp. 6111–6116, Jul. 2013. [29] “NREL Best Research-Cell Efficiencies.” [Online]. Available: http://www.nrel.gov/ncpv/images/efficiency_chart.jpg. [Accessed: 10-Feb-2015].

References 57

[30] J. M. Kroon, M. M. Wienk, W. J. H. Verhees, and J. C. Hummelen, “Accurate efficiency determination and stability studies of conjugated polymer/fullerene solar cells,” Thin Solid Films, vol. 403–404, pp. 223–228, Feb. 2002.

Appendices 58

Appendices

A. Fabrication procedures

Synthesis of PbS CQDs and metal halide treatment PbS quantum dots were synthesized according to a previously published method [4]. A solution-phase metal halide treatment (CdCl2) was then carried out following a previously published method [3]. Specifically, the metal halide precursor (1 mL of CdCl2) and tetradecylphosphonic acid (TDPA) were dissolved in oleylamine with 13.6:1 Cd:TDPA molar ratio. This mixture was introduced into the CQD reaction flask after the sulfur source injection during the slow cooling process. A 6:1 Pb:Cd molar ratio was adopted during the synthesis. At 30–35°C, the nanocrystals were isolated by the addition of 60 mL of acetone and then subjected to centrifugation. The nanocrystals were then purified by dispersion in toluene and reprecipitation with a mixture of acetone/methanol (1:1 volume ratio), and then redissolved in anhydrous toluene. The solution was further washed with methanol two more times before finally dispersing it in octane at a concentration of 50 mg/mL.

Substrate preparation Planar glass substrates: Cleaned glass substrates coated with fluorine- doped tin oxide (TEC

15; Pilkington) were employed in this study. Two equivalent TiO2 electron-accepting layers were found to be equal in performance and used interchangeably in the process. The first used a sol–gel TiO2 mixture that was prepared, deposited, and annealed according to a previously published method [8]. The second used a sputtered 50 nm layer of TiO2 (Kurt Lesker) using an argon pressure of 7.5 mTorr in an Angstrom Engineering Å mod deposition system in an Innovative Technology glovebox and a deposition rate of 0.08 Å/s. In either -3 case, the substrates were then treated with a 120⋅10 M TiCl4 solution at 70°C for 30 min followed by a rinse with deionized water and annealing step on a hot plate at 520°C for 45 min in air ambient. The samples were then stored in a nitrogen-filled glovebox until just before device fabrication.

Appendices 59

Flexible PET substrates: We mounted an ITO-coated flexible polyethylene terephthalate sheet into our sputtering chamber and deposited 50 nm of TiO2 as in the case of FTO-coated glass substrates mentioned above. In the case of the flexed substrate, the PET sheet was wrapped around a 2 cm diameter dowel prior to TiO2 deposition. No TiCl4 treatment was done on these substrates, as the PET would not support the high temperature anneal. Spherical lens substrate: 250 nm of ITO were deposited on spherical lenses (Thorlabs, LA1252) via heated sputtering at 350°C. The substrates were then sputtered with 50 nm of

TiO2 and treated with TiCl4 solution as above.

CQD spray deposition The stock 50 mg/mL CQD in octane solution was diluted to 3.33 mg/mL immediately prior to use. The total solution volume required for one device was 18.75 mL, yielding a mass of oleic acid capped CQDs of 62 mg. This solution was placed in a reservoir connected to the solution gravity-fed inlet of an Ikeuchi fine mist nozzle (BIMV8002S). The nozzle was pressurized to 45 psi using a nitrogen gas line. Another 45 psi nitrogen gas line provides activated piston control for the nozzle. MPA was diluted in methanol (MeOH) to 0.16% (v:v) and placed in a reservoir for a Paasche VL airbrush pressurized with a 35 psi nitrogen gas line. A third 45 psi nitrogen gas line pressurized an additional Paasche VL airbrush loaded with MeOH. Finally, a custom-made air blade was connected to an 85 psi compressed dry air gas line. Fabrication consisted of between 65 and 85 layers of a sprayed layer-by-layer procedure where each layer included: 1. 0.4 s actuated CQD nozzle followed by a 3 s pause; 2. 1 s actuated MPA nozzle; 3. 4 s MeOH rinse for airbrush; 4. 40 s air blade drying.

Deposition of top contacts The top contacts were deposited using an Angstrom Engineering Å mod deposition system in an Innovative Technology glovebox and consisted of 40 nm thermally evaporated MoO3 deposited at a rate of 1.0 Å/s, followed by e-beam deposition of 50 nm of Au deposited at 1.5 Å/s, and finally 120 nm of thermally evaporated Ag deposited at 2.0 Å/s.

Appendices 60

B. Material characterization procedures

Nanomechanical properties characterization by AFM The AFM measurements were performed using PeakForce Quantitative Nanomechanical Property Mapping by Bruker®. Fast force curves were performed as the AFM scanned the samples’ surfaces. The PeakForce QNM provides modulus data in addition to surface topology. Prior to measurement, the cantilever tip’s radius and reflection sensitivity were measured via rough surface imaging and peak force measurement on quartz. In addition, the spring constant was measured via thermal vibration measurement. The surface indents for our samples were less than 1 nm using an indentation force of 5 nN. Only one cantilever was used and the samples were tested back to back to ensure comparability.

GISAXS measurements GISAXS measurements were performed on Beamline 06ID-1 (HXMA) of the Canadian Light Source. Monochromatic light was used with energy of 7 keV. The marCCD SX-165 detector with a pixel size of 80⋅80 µm2 and a total of 2048⋅2048 pixels was used to record the scattering patterns. The images were dark-current-corrected, distortion-corrected, and flat- field-corrected by the acquisition software. Using a silver behenate powder standard, the sample-to-detector distance was determined to be 679 mm. The angle of incidence of the X- ray beam was varied between 0.08° and 0.12°, and an exposure time of 30 s was used. All films show primarily ring-like GISAXS patterns. We plotted azimuthally integrated intensity profiles and used Gaussian fitting and an exponential background to determine the location of the scattering rings at q ≈ 0.2 Å–1. Conversion to real-space coordinates gave the average center-to-center nanocrystal spacings.

Solid-state nuclear magnetic resonance All solid-state NMR experiments were performed on an Agilent DD2 700 MHz spectrometer with a 1.6 mm T2 NB HX Balun probe. Magic angle spinning experiments were conducted while spinning the sample at 25 kHz to ensure that there were no spinning sidebands in the region of interest. 1H chemical shifts were referenced to trimethyl silane (δiso = 0 ppm) using adamantane as a secondary reference (1H:1.85 ppm for the high-frequency resonance). For

Appendices 61

NMR analysis, the curves were fit by a superposition of Gaussians with peak centers corresponding to constituent ligands (MPA ≈ 0-2 ppm), residual methanol (~3 ppm) and water (~5 ppm) and by a residual synthesis precursor, tetradecyl phosphonic acid (~12 ppm). The additional ~8 ppm peak was assigned to a protonated carboxylic acid.

Fourier transform infrared spectroscopy FTIR was performed on a Bruker Tensor spectrometer in transmission mode. Analysis of CQD films was performed by fabricating films as outlined above on glass substrates, manually scraping material off the substrate and mixing the resultant powder with KBr powder in roughly a 1:100 w/w ratio. The mixture was compressed into a thin pellet using a PIKE Technologies pellet press. Reference measurements for pyruvic and oleic acids were carried out by depositing and spreading a small volume of each compound on a KBr disk and immediately measuring. A background subtraction against air was carried out for the entire data set.

Density functional theory calculations Calculations were carried out using the Quickstep module of the CP2K program suite utilizing a dual basis of localized Gaussians and plane waves. The plane wave cutoff was 300 Ry, appropriate for the Goedecker-Teter-Hutter pseudopotentials that we employed, and the localized basis set of double-ζ plus polarization (DZVP) quality optimized to reduce the basis set superposition errors. Calculations were performed using the Perdew-Burke-Ernzerhof (PBE) exchange correlation functional. Simulations were performed with nonperiodic boundary conditions in a 50×50×50 Å unit cell for 2.5 nm quantum dot sizes. The quantum dot was carved out of bulk PbS. All singly bonded atoms were discarded, resulting in a faceted cuboctahedron shape. A mixture of Cl and thiol ligands was used to passivate all dangling bonds on (111) and (110) facets, with the (100) facets left unpassivated. Care was taken to select the stoichiometry that preserves the charge neutrality of the dot, necessary to position the Fermi level in the mid-gap. For all calculations, a single CQD was modeled with appropriate additions (MPA, deprotonated MPA, lead oleate as well as the monomer and dimer of the MPA-lead oleate complex). The lead oleate and complexes were not bound to the surface of the CQD but placed near the surface (approximately 3-5 Å). The distance was

Appendices 62 sufficient to observe no overlap between the electronic wave function of the states on the complex and those on the CQD itself.

C. Optoelectronic characterization procedures

AM1.5 photovoltaic performance characterization Current-voltage data were measured using a Keithley 2400 source meter. The solar spectrum at AM 1.5G was simulated within class A specifications (less than 25% spectral mismatch) with a xenon lamp and filters (ScienceTech; measured intensity of 100 mW/cm2). The source intensity was measured with a Melles-Griot broadband power meter through a circular 0.049 cm2 aperture. We used an aperture slightly smaller than the top electrode (0.067 cm2) to avoid overestimating the photocurrent: the entire photon fluence passing through the aperture was counted as incident on the device for all analyses of JSC and EQE [26]. The spectral mismatch of the system was characterized using a calibrated reference solar cell (Newport). The total AM1.5 spectral mismatch—taking into account the simulator spectrum and the spectral responsivities of the test cell, reference cell, and broadband power meter— was remeasured periodically and found to be ∼5%. This multiplicative factor, M = 0.95, was applied to the current density values of the J-V curve to most closely resemble true AM1.5 performance [30]. The uncertainty of the current–voltage measurements was estimated to be ±3.3%.

EQE measurements External quantum efficiency measurements were obtained by applying chopped (220 Hz) monochromatic illumination (450 W xenon lamp through a monochromator with order- sorting filters) collimated and cofocused with a 0.7 Sun intensity white light source on the device of interest. The power was measured with calibrated Newport 818-UV and Newport 818-IR power meters. The response from the chopped signal was measured using a Stanford Research system current preamplifier feeding into a Stanford Research system lock-in amplifier set to voltage mode. The uncertainty in the EQE measurements was estimated to be 2.9%.

Appendices 63

Luminescence measurements Electroluminescence measurements were carried out by connecting a Keithley 2410 source meter to our devices and applying a range of forward bias voltages while reading the resultant current. The luminescence was collected through a set of lenses focused on an optical fiber and connected to an Ocean Optics NIR-512 spectrophotometer. Photoluminescence measurements cofocused the input to the same signal collection optics with a 630 nm wavelength continuous-wave laser.