SOLUTION PROCESSING FOR COPPER INDIUM SULFIDE SOLAR CELLS
A DISSERTATION SUBMITTED TO THE DEPARTMENT OF CHEMISTRY AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
Stephen Thacker Connor August 2011
© 2011 by Stephen Thacker Connor. All Rights Reserved. Re-distributed by Stanford University under license with the author.
This work is licensed under a Creative Commons Attribution- Noncommercial 3.0 United States License. http://creativecommons.org/licenses/by-nc/3.0/us/
This dissertation is online at: http://purl.stanford.edu/wr798sx5189
ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Yi Cui, Primary Adviser
I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Christopher Chidsey
I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.
T Stack
Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost Graduate Education
This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives.
iii Abstract
In recent years, the field of photovoltaics has become increasingly important due to rising energy demand and climate change. While most solar cells are currently composed of crystalline silicon, devices with thinner films of inorganic absorber materials might allow production at a greater scale due to their lower materials cost. In particular, thin films of CuInS 2 are promising solar absorber materials due to their high efficiencies and low required thicknesses. However, the fabrication of thin film solar cells currently requires expensive vacuum techniques. As an alternative, solution-based deposition techniques have been proposed as a route to low-cost and high-throughput electronic device fabrication. In this dissertation, I will describe two approaches developed to solution process CuInS 2 solar absorber layers from nanoparticle precursors. Many common solution techniques can quickly deposit thick layers of nanoparticulate material, but these layers must be subsequently converted into a final compact semiconductor thin film by annealing. The primary limitation in solution processed thin films tends to be the lack of grain growth during annealing, which degrades electrical properties. I have studied how film growth depends on precursor film quality, with the goal of producing large grained films of CuInS 2 through solution processing. In the first approach, we used solvothermal decomposition of organometallic precursors at moderate temperatures to produce nanoparticles of CuInS 2. In order to determine the growth mechanism of these nanoparticles, we analyzed the structure and phase of these nanoparticles by transmission electron microscopy and X-ray diffraction. Thin films of these nanoparticles were cast onto molybdenum coated glass and further processed to create CuInS 2 solar cells. We found that performance was dependent on film porosity, grain size, and stoichiometry of the nanoparticles. Films with grain sizes of ~200nm were attained, from which 1.3% efficient solar cells were made. In addition, we showed that this synthesis could be extended to produce CuInS 2 nanoparticles with partial substitution of Fe, Zn, and Ga. The substitution of In or Cu with other elements allows tuning of the band gap and is the first step in producing
iv more complicated structures such as layers with graded compositions for improved device performance. In the second approach, we synthesized an air-stable hybrid organometallic/nanoparticle ink at room temperature in ambient conditions through a vulcanization reaction. This ink could be coated onto substrates in smooth layers, and further reactive annealing formed large grained CuInS 2 films. This process was characterized, and a correlation between residual carbon and grain growth was found. Additionally, the chemical transformation between precursor layers and final sulfide thin film was analyzed, with an emphasis on the difference between sulfurization and selenization. We demonstrated that the sulfurization process was producing morphological defects due to its nucleation limited growth mechanism. However, it was modified to more closely resemble the diffusion limited selenization mechanism, thus producing flat films of CuInS 2 with grain sizes of ~500nm.
v Acknowledgements
This dissertation wouldn’t have been possible without the help and support of a whole gaggle of people. Foremost I’d like to thank my girlfriend Egle Cekanaviciute who edited almost every word in this document and also tolerated late night microscopy and other such tragedies. Second, I thank my adviser, Yi Cui, for his extensive guidance and keen scientific insights through the years. The entire Cui research group has been amazing and exciting to work within; in particular Ben Weil and Ching-Mei Hsu have been sounding boards for all of my terrible ideas, and thus deserve countless thanks. I’d like to thank my collaborators: Sumohan Misra and Mike Toney at the Stanford Linear Accelerator; Jung-Yong Lee and Peter Peuman in the Department of Electrical Engineering; and Mary Tang at the Stanford Nanofabrication Facility. In addition, the staff members of the Moore and McCullough buildings have been wonderful; in particular I offer heartfelt thanks to Christina Konjevich and Benita Givens for helping avert administrative doom many times and Mark Gibson for literally keeping the Moore building together. Finally, I’d like to thank my family and friends for the years of emotional support and good times. My mother and father, Cindy Floyd and Gordon Connor, have always stood by me, for which I’m very lucky and grateful. Lastly, I’d like to honor the memory of Jeanne Connor; I will always aspire to one day be as brilliant and good as she believed I was.
vi Table of Contents Abstract…………………….…………………………………………………………iv Acknowledgements…………..…………………………………………………….…vi Table of Contents…………………………………………………………………..…vii List of Tables……………………………………………………………….…………xi List of Figures…………………………………………………...……………………xii 1 Introduction and Background ...... 1
1.1 Solar Cell Design Principles ...... 2
1.1.1 Solar Spectrum ...... 2
1.1.2 Structure and Performance ...... 3
1.1.3 Important Parameters ...... 5
1.2 Thin Film Solar Cell Development ...... 6
1.3 Solution Processing ...... 8
1.4 Solution Deposition Techniques ...... 9
1.4.1 Spray Coating ...... 10
1.4.2 Spin Coating ...... 10
1.4.3 Paste Coating ...... 10
1.4.4 Bath Depositions ...... 11
1.5 Ink Compositions ...... 11
1.5.1 Solution-based Inks ...... 12
1.5.2 Nanoparticle-based Inks ...... 13
1.5.3 Hybrid Approaches ...... 14
2 Methods ...... 16
2.1 CuInS 2 Precursor Films from Nanoparticle Ink ...... 16
2.1.1 Oleate Precursor Synthesis ...... 16
2.1.2 CuInS 2 Nanoparticle Synthesis ...... 17
vii 2.1.3 Deposition of CuInS 2 nanoparticle film ...... 18
2.2 Cu-In-O Precursor Films from Hybrid Nanoparticle Ink ...... 18
2.2.1 Preparation of Hybrid Ink ...... 18
2.2.2 Deposition of Cu-In-O Film ...... 18
2.3 Sulfurization of Cu-In-O and CuInS 2 NP films ...... 19
2.4 Solar Cell Device Fabrication ...... 20
2.5 Analytical Techniques ...... 20
2.5.1 Transmission Electron Microscopy (TEM) ...... 20
2.5.2 Selected Area Electron Diffraction (SAED) ...... 20
2.5.3 Energy Dispersive X-Ray Spectroscopy (EDS) ...... 21
2.5.4 Scanning Electron Microscopy (SEM) ...... 21
2.5.5 Auger Electron Spectrometry (AES) ...... 21
2.5.6 UV/Visible Spectroscopy ...... 22
2.5.7 Powder X-Ray Diffraction (XRD) ...... 22
2.5.8 Anomalous Diffraction ...... 23
2.5.9 Electrical Measurements ...... 24
3 CuInS 2 Nanoparticle Growth Mechanism ...... 26
3.1 CuInS 2 Nanoparticle Synthesis ...... 26
3.2 CuInS 2 Nanoparticle Growth Mechanism ...... 26
3.2.1 Proposed Nanoparticle Growth Pathway ...... 27
3.2.2 Elemental Mapping ...... 28
3.2.3 UV/VIS Spectroscopy ...... 30
3.2.4 High Resolution Transmission Electron Microscopy (HRTEM) of Single Nanoparticles ...... 31
3.2.5 X-Ray Diffraction ...... 33
viii 3.2.6 Epitaxial Growth and Phase Conversion ...... 35
3.2.7 Growth Mechanism Modeling ...... 36
3.2.8 Summary of Growth Study ...... 39
3.3 Solar Cells from CuInS 2 Nanoparticles ...... 39
4 Band Gap Tuning of CuInS 2 Nanoparticles through Alloying with Zn, Ga, and Fe ...... 43
4.1 Alloying in CuIn(Se,S) 2 ...... 43
4.2 Alloying in CuIn(Se,S) 2 Nanoparticles ...... 44
4.3 Anomalous X-ray Diffraction ...... 45
4.4 Effects of Alloying on CuInS 2 Optical Band Gap ...... 45
4.5 Effects of Alloying on CuInS 2 Morphology and Phase ...... 46
4.5.1 Substitution of Fe ...... 46
4.5.2 Substitution of Zn ...... 49
4.5.3 Substitution of Ga ...... 51
4.6 Summary of Alloying Results ...... 53
5 Hybrid Nanoparticle Ink ...... 54
5.1 CuIn(S,Se) 2 Films from Oxide Precursors ...... 54
5.2 Solution Processing with Oxide Precursors ...... 55
5.3 Air-stable Ink Deposition ...... 55
5.4 Cu-In-O Precursor Film ...... 56
5.5 Electrical Measurements of CuInS 2 Solar Cells ...... 58
6 Mechanism of Sulfurization and Selenization of Cu-In-O Films ...... 61
6.1 Growth Mechanism of CuIn(S,Se) 2 ...... 61
6.2 Sulfurization of Solution Deposited Cu-In-O ...... 62
6.2.1 Sulfurization below 350°C: Partial Sulfurization ...... 62
ix 6.2.2 Sulfurization between 350°C-450°C: Intermediate Phases ...... 63
6.2.3 Sulfurization above 450°C: Formation of Pure CuInS 2 ...... 66
6.2.4 Development of Holes in CuInS 2 ...... 66
6.3 Selenization of Solution Deposited Cu-In-O Films ...... 67
6.4 Comparison of Sulfurization and Selenization Mechanisms ...... 72
6.5 Compensatory Growth Conditions for CuInS 2 ...... 73
6.6 Summary of Results ...... 74
List of References ...... 75
x List of Tables Table 1.1: Comparison of ink compositions ...... 15
Table 3.1: Device properties for 4 solar cells made from CuInS 2 nanoparticle films .. 41 Table 4.1: Optical band gaps of nanoparticles with substitution ...... 46
xi List of Figures Figure 1.1: Solar irradiance spectrum and optimal band gap for solar absorber ...... 3 Figure 1.2: Electrical properties of solar cells ...... 4
Figure 3.1: Schematic of proposed CuInS 2 growth mechanism ...... 28
Figure 3.2: TEM images and composition maps of CuInS 2 Nanoparticles ...... 29 Figure 3.3: UV/VIS spectra of nanoparticles ...... 31 Figure 3.4: HRTEM images of nanoparticles ...... 32
Figure 3.5: XRD of wurtzite CuInS 2 nanoparticles ...... 34 Figure 3.6: Crystal structures of relevant phases ...... 35 Figure 3.7: Time traces of nanoparticle phase conversion ...... 37 Figure 3.8: Nanoparticle films after 5 minute sintering in S vapor ...... 40 Figure 3.9: I-V curves of devices NP1-4 ...... 42
Figure 4.1: CuInS 2 NPs with Fe substitution...... 47
Figure 4.2: HRTEM image of a CuInS 2 nanoparticle with 10% Fe ...... 48
Figure 4.3: CuInS 2 NPs with Zn substitution ...... 50
Figure 4.4: CuInS 2 NPs with Ga substitution ...... 52
Figure 4.5: Fluorescence spectra of CuInS 2 samples with 0%, 5%, 10%, and 20% Ga substitution ...... 53 Figure 5.1: SEM cross-section of 3 layers of Cu-In-O film after air-baking ...... 57
Figure 5.2: Fabrication process and characeterization of CuInS 2 thin film...... 58
Figure 5.3: SEM and electrical performance of CuInS2 thin film ...... 59 Figure 6.1: SEM and XRD of film sulfurized at 335ºC ...... 63 Figure 6.2: SEM and AES of film sulfurized at 385ºC ...... 64 Figure 6.3: XRD patterns of samples sulfurized at 335ºC, 385ºC, 410ºC, and 585ºC . 65 Figure 6.4: SEM plan view and cross-sectional images of sulfurized films ...... 66 Figure 6.5: SEM and XRD of film selenized at 400ºC ...... 68 Figure 6.6: XRD patterns of samples sulfurized at 400ºC, 425ºC, 450ºC, 475ºC, and 500ºC ...... 69 Figure 6.7: SEM and AES of a sample selenized at 450ºC ...... 70 Figure 6.8: SEM plan view and cross-sectional images of selenized samples ...... 71
xii Figure 6.9: Growth mechanism models of sulfurization and selenization of oxide precursor films ...... 72
Figure 6.10: CuInS 2 films from two stage heating ...... 74
xiii
1 Introduction and Background
In recent years, the International Panel on Climate Change and other global organizations have synthesized massive amounts of data on global climate trends. Their 4 th annual review provides unequivocal evidence that the global climate is warming. Furthermore, the review indicates that the global temperature increase in the 20 th century is very likely due to fossil fuel use, and continuing with the current levels of carbon emissions will cause massive adverse global effects. Many countries have thus begun to fund large-scale energy projects in fuel cell, battery, and solar cells, which promise to dramatically reduce carbon emissions through the gradual replacement of fossil fuels. However, renewable energy technologies compete with traditional power generation and transmission, thus they must be able to be produced at a large scale and low cost. The scale of global power consumption is a daunting figure- 13TW in 2001, with demand predicted to rise to 27TW by 2050 1. Solar energy conversion is particularly well positioned in the renewable field; total incident solar power on earth can be estimated near 120,000 TW. Yet, the current percentage of global power coming from solar is meager by any measure, with all renewables accounting for less than 1%, and solar being a small fraction of that. Even with conservative growth estimates of 25% per year, solar energy would still not compose 1% of the total energy landscape in 25 years 2. Clearly, to become a significant portion of global energy production, breakthroughs in cost or efficiency are required. This thesis will focus on the solution processing of solar cells, which is a promising means of producing the globally required quantity of solar energy at a low cost and high rate. The operating principle in solar energy conversion is the photovoltaic effect, discovered by Edmond Becquerel in 1839 3. He observed that his silver battery electrodes produced a voltage in the electrolyte upon illumination. The next major development in photovoltaics came 38 years later, with the development of the first solid solar cell. Adams and Day demonstrated photovoltage generation in a selenium
1 bar with platinum contacts 4; this was quickly followed by a thin film analogue of the same device, with a 100 µm Se plate sandwiched between gold leaf and a brass plate, developed by Charles Fritts 5. While these devices were effectively solar cells, the advent of the semiconductor homojunction and heterojunction solar cell came in the mid-20 th century. This revolutionized the field because semiconductor junctions provide a highly effective means of collecting photogenerated electrons and holes.
1.1 Solar Cell Design Principles
1.1.1 Solar Spectrum All semiconductors and insulators have a band gap in their electronic structure and can thus excite electrons from valence to conduction bands, yet only a select group has band gaps appropriate to absorb solar photons. Figure 1.1(a) shows the AM1.5G terrestrial solar irradiation spectrum, a standard used to evaluate and compare solar cells, with the band gaps of several semiconductors marked to show absorption onsets. While a low energy (high wavelength) band gap enables the absorption of a broader portion of the spectrum, much of the energy of each photon is lost as the excited electron thermally relaxes to the conduction band edge. This trade- off between photon collection and energy has been analyzed extensively, and a range of maximally productive band gaps has been established by Shockley and Quiesser as being between 1.3 and 1.5eV 6. A plot of the maximum power conversion efficiency as a function of band gap is shown in figure 1.1(b); band gaps of technologically relevant semiconductors are labeled on the curve to illustrate their fit within the optimal range.
2
Figure 1.1: Solar irradiance spectrum and optimal band gap for solar absorber (a) Solar irradiance spectrum; band gaps of common semiconductors are labeled. (b) Maximum efficiency versus band gap, calculated with AM1.5G solar spectrum (solar zenith angle 48.19). Band gaps of Ge (0.7eV), Si (1.1eV), Cu 2S (1.2eV),
CuIn 0.7 Ga 0.3 Se 2 (1.3eV), GaAs (1.45eV), CdTe (1.5eV), CdS (2.4eV), CdSe (1.7) are marked by arrows. Spectrum of solar irradiance adapted from NREL reference data.
1.1.2 Structure and Performance There are several ways and means to collect photogenerated electrons from materials; however, this dissertation will be limited to semiconductor junction solar cells. Principally, n and p type semiconducting materials generate an internal electric field when put in contact, due to the difference between their Fermi energies. Electrons and holes are generated when light excites an electron across the band gap of a semiconductor, and the built-in electric field aids in their separation at the junction. However, since the electrons or holes generated in the semiconductor must diffuse close enough to the junction to be collected by the electric field, and the interface itself must not act as a site for carrier recombination, material quality is of paramount importance. Several critical characteristics in a solar cell are the light absorption profile, diffusion of carriers to the p-n junction, and extraction of carriers by the electrodes. The top and bottom electrodes are selected to provide an ohmic contact with the top and bottom semiconductors, respectively. The two semiconductor layers can be deposited by a number of techniques, but must be composed of films uniform in
3 thickness and have a well defined interface between the p and n materials. In most cases, the quality of both semiconductor materials and the interface determines the performance of a solar device. This representative solar cell can be approximately modeled by an electrical circuit which contains four components: a current generator, a diode, and two resistors. The corresponding circuit diagram can be seen in Figure 1.2, and a curve of current vs. voltage is shown for the device both in the dark and under illumination. In the dark, a solar cell acts purely as a diode, governed by equation 1.1, where the current and voltage are I and V, while I 0 and I L are the dark saturation current and light generated current, respectively. A and T are an ideality parameter and the temperature, and k and q are the Boltzmann’s constant and the charge of an electron. Many semiconductor properties dictate these values; in particular, the saturation current I0 is correlated to the rate of recombination in the system, and is thus a measure of impurities and other defects.
Eqn. 1.1: = exp − 1 −
Figure 1.2: Electrical properties of solar cells (a) A simulated I-V curve for a solar cell; parameters used to define efficiency are marked. (b) Simplified circuit diagram for a solar cell device.
4
1.1.3 Important Parameters The primary characteristics of the I-V curve in Figure 1.2(a) are the short circuit current (I SC ), open circuit voltage (V OC ), and fill factor (FF). As implied, the I SC is the maximum current when the circuit in Figure 1.2(b) is electrically shorted and there is no voltage drop, and the V OC is the maximum voltage when the circuit is open and has no current flowing. The maximum power point is found on the I-V curve, and it would ideally have the voltage V OC and the current I SC . The maximum power is limited by both material properties and also by defects. It can be seen from equation
1.1 that a high saturation current I0 will cause a reduction in the intercept V OC and thus a lower attainable power. Also, the current I SC is dictated by the number of photons absorbed, which is inversely correlated to the band gap E g of a material. Additionally, other physical limits of materials, namely resistances in the bulk and at interfaces and the ideality of the diode, limit the efficiency . The fill factor (FF), indicates the ratio between the real maximum power point and the product of the V OC and I SC ; it can also be seen as the “squareness” of the I-V curve. From these observables and the incident power of light, P in , efficiency can be calculated: Eqn 1.2: = × × /
The resistances that control the fill factor are generally called the series and shunt resistances, abbreviated R S and R SH . The series resistance is simply the entire resistance of the device, and it acts as a parasitic loss of current at positive voltages. Common sources are the resistivity of the semiconductors themselves, which can be very high without sufficient doping, and the contact between the semiconductors and the metal contacts, which depends strongly on energy alignments. The shunt resistance originates from defects in the semiconductor layers that act as alternate conductive pathways. These shunts can be either morphological, such as holes or thinnesses, or can even be precipitations of conductive domains. Both series and shunt resistances
5 mainly affect FF, unless they are extreme, in which case they reduce V OC and I SC as well. When these resistances are included, the governing equation can be rewritten:
Eqn 1.3: ( ) = exp + −
1.2 Thin Film Solar Cell Development
Almost contemporaneously in the 1950s four materials systems were demonstrated to produce high solar conversion efficiencies: Si, Cu 2S, CdTe, and GaAs. The first to be shown was the Si solar cell, demonstrated at Bell Labs in 1954 by Chapin, Fuller and Pearson 7. A p-n homojunction was formed by depositing a thin layer of p-type Si on top of a large wafer of n-type Si and achieved a conversion efficiency of 6%. This inspired research into finding other suitable materials, and GaAs was another material from which homojunction devices could be made; Jenny et al developed 6% efficient cells by diffusing Cd into n-type GaAs wafers to form a homojunction 8. Another important photovoltaic device, composed of the heterojunction of p- type Cu 2S and n-type CdS, was nearly simultaneously shown to also have an initial efficiency of 6% 9. Finally, the first CdTe based devices were prepared slightly by diffusing Cu into an n-type CdTe film to form p-type Cu 2Te, also achieving efficiencies near 6% 10,11 . Later additions to the set of highly efficient solar absorber 12 materials include CuInSe 2, which was first fabricated into devices in 1974 , and 13 CuInS 2, first made by Kazmerski and Sanborn in 1977 . The CuInSe 2 and CuInS 2 systems are collectively known as the CuIn(S,Se) 2 system due the ability to form electrically useful alloys with widely varying compositions of S and Se with similar processing parameters. In the decades following the discovery of these promising photovoltaic material systems, as could be expected, market share and technological improvements have varied widely. Si technology has achieved market dominance due partly to synergy with transistor technology, enabling production scaling with adapted tools and
6 facilities. The band gap of Si (1.1eV) is appropriate for solar absorption, but at least 100 µm of material are required for sufficient light absorption because it has an indirect band gap. As a result, thick films of extremely high quality Si must be made by high temperature processing and energy intensive purification. Thin film absorbers of direct band gap semiconductors are a desirable alternative due to the thinner layers required for adequate light absorption(<2 µm), which allows quicker deposition times and lower materials usage. However, currently all thin film systems are binary, ternary, or even quaternary systems, which introduces an element of complexity as compared to the single element system of Si.
While the Cu 2S/CdS system seemed to be a perfect match of cheap materials and moderate efficiency, devices were found to suffer degradation over time due to massive Cu diffusion, and the technology was abandoned. Meanwhile, the
CuInSe 2/CdS and CdTe/CdS systems showed much greater stability, and have currently reached record device power conversion efficiencies of 19.9% 14 and 16.7% 15 , respectively. Nonetheless, although these two systems are leading in efficiency, many governments have expressed concern over the environmental dangers of Se and Te. As a result, efforts have gone into the development of a less toxic analogue, CuInS 2.
The CuIn(S,Se) 2 class of materials has myriad benefits, including large absorption coefficients and good stability under solar radiation 16 . Thin films of
CuIn xGa 1-xSe 2 (CIGS) with a band gap of 1.3 eV have been demonstrated to have a high power efficiency of ~20%, which is the highest among thin-film solar cell technologies. The analogous sulfide system, CuInS 2, has a direct band gap of 1.5 eV, which is also well matched with the solar spectrum and is more environmentally friendly without Se 17 . Theoretical solar conversion efficiencies of 28% have been 18 calculated with CuInS 2 as an absorber , which is among the highest of any thin film material, and CuInS 2 thin film solar cells of 12% efficiency have been successfully made 19 .
While early high efficiency CuInS 2 solar cells were made by vacuum co- 20 evaporation in an analogous manner to CuInSe 2 , simpler processes have since produced comparable devices. In particular, a two stage process 2122 , which separates
7 metal deposition and sulfurization, has produced the most efficient devices recently. While this process reduces the steps involving vacuum conditions and toxic gases, it still requires sputtering in a vacuum.
1.3 Solution Processing
As abovementioned, both CuInS 2 and CuInSe 2 solar cells utilize sputtering or coevaporation for solar absorber fabrication, which gives independent control over elemental fluxes and process conditions such as pressure and chemical atmosphere. However, these and other techniques involving high vacuums have several economic problems. The deposition systems are a large capital cost, and the use of vacuum deposition is highly energy intensive. In addition, materials utilization is not optimal in these techniques, reaching a maximum of 80% with even the most efficient sputtering systems 23 . As an alternative, solution-based deposition techniques are widely considered to be a route to low-cost and high-throughput electronic device fabrication. Design of the ink and the deposition technique is an important and challenging problem that must be overcome to produce high-quality photovoltaic devices. To make high quality active layers, the deposition must be scalable, and the ink must be tuned to make dense, micron thick layers with minimal residue. Solution processing requires less capital investment than vacuum deposition due to utilizing ambient atmosphere for all or part of the process and less expensive deposition tools. In addition, liquid processing is inherently amenable to flexible substrates, which enable high throughput and rapid scalability.
Currently the biggest companies involved in CuIn(S,Se) 2 use vacuum technologies, including Wurth Solar, Showa Shell, MiaSolé, HelioVolt, and Sulfurcell. However, many newer startup companies are reaching production scale with partially vacuum-free solution-based technologies, such as Institut de Recherche et Developpement sur l'Énergie Photovoltaique (IRDEP), Solopower, International Solar Electric Technology (ISET) and Nanosolar. While IRDEP and Solopower use
8 electrodeposition, note that ISET and Nanosolar both use nanoparticulate-based technologies.
1.4 Solution Deposition Techniques
Almost every solution-based solar fabrication includes both a process of precursor deposition and a subsequent annealing step to form the final layer. The primary variables are the nature of the precursor ink and the deposition technique utilized. Direct liquid coatings cover a large class of deposition techniques, including spin-coating, spray coating, and various techniques used in roll-to-roll processing. These techniques are desirable for their generally high rates of deposition as compared to most vacuum techniques. In direct liquid coatings, metal precursors are dissolved or dispersed in a solvent along with surfactants and other additives, and the ink is then coated onto a substrate by repeated applications, commonly with an intermediate drying step. The thick film is then thermally decomposed into a compact precursor film, which typically is annealed in chalcogen or hydrogen chalcogenide vapor (the process known as “reactive annealing”) to form the desired final film and remove residual contaminants such as C, O, and Cl. Solution processing for solar cells may have one or all of these steps; details vary widely due to the diverse temperatures and rates of ink deposition. Arguably, the most important variable in solution processed absorbers is the size of the grains produced in the final film. The production of small grains necessarily introduces more grain boundaries, which act to reduce carrier transport and thus lower
JSC , FF, and ultimately, efficiency. However, grain size is reduced by numerous factors; the most important seem to be residual contaminants, insufficient annealing, surface roughness, and precursor film porosity. Special emphasis will therefore be placed on the comparative grain sizes of different processes, as this correlates well with electrical performance.
9 Many thorough reviews have been published recently on solution processing for solar absorbers 24,25,26 , and thus only a small sampling of the field most relevant to the rest of the dissertation will be presented below.
1.4.1 Spray Coating Spray pyrolysis is one of the most successful and easily scaled of the direct liquid coating processes, due to its low additive requirements and precursor versatility. A solution is sprayed onto a substrate which is heated such that the solvent evaporates quickly, and the dissolved compounds react to form an intermediate chalcogenide or oxide film. This film is then reactively annealed as aforementioned, though attempts have been made to form the final film in one spray step 27 . Most spray pyrolyses tend to suffer from long deposition times required by the low solution concentrations to avoid clogging. Commonly, small grain sizes are seen, likely due to incomplete removal of impurities from the solvent and the precursor or uneven deposition.
1.4.2 Spin Coating Simpler direct liquid coating techniques are common in laboratory settings, particularly spin coating and paste coating. In spin coating, the substrate is firmly held and spun at high speeds. Solution is deposited either before or after spinning begins, but the goal is both cases is the removal of solvent and even deposition of solute or particulates. While the deposition has poor materials utilization and is slow, spin coating can be performed with a wide range of ink viscosities and precursor types; this versatility makes it a useful tool for rapid prototyping ink formulations before optimization with a more industrial technique such as spray coating.
1.4.3 Paste Coating Paste coating is a technique which can be done with high viscosity inks, wherein a thick layer of ink is deposited in a single pass of a knife or other spreading
10 tool. Paste coating is preferable to spray coating in many applications because coating speeds can be ~20 times faster while deleterious surface roughness and porosity are halved as compared to spraying 28 . However, the large quantity of required additives makes film purity a challenge. As a result, few research efforts have achieved grain sizes over 100nm from paste coating.
1.4.4 Bath Depositions One of the simplest and oldest 29 techniques to solution deposit semiconductors, bath deposition, does not truly involve an ink but should be mentioned due to its technological relevance. A substrate is immersed in a solution of the elemental precursors, and a layer is formed by inducing precipitation. This can be done by applying a voltage between the substrate and another electrode, or the precipitation can be induced by a chemical reaction in the bath. These techniques are commonly labeled electro-deposition and electroless-deposition (or chemical bath deposition), respectively. The CdS buffer layer of CIGS solar cells is industrially deposited by an electroless deposition due to the quality of the interface it creates 30 , and there is active research efforts to develop comparable electroless deposition of Cd-free buffer layers. However, the range of reduction potentials of the different elements in CIGS necessitates complex electrodeposition profiles, and deposition times can be hours long if large grained material is needed. Despite these complications, CIGS devices produced by electrodeposition have achieved efficiencies of 15.4% 31 , and some companies are currently attempting to scale this technique to factory levels.
1.5 Ink Compositions
Inks of incredibly diverse molecular compositions have been tried in the
CuIn(Se,S) 2 system. Approaches vary in complexity from simple aqueous solutions of the metal chlorides to complex organometallics with ligands to tune coating properties. A brief survey of ink formulations will be given, with emphasis on the two approaches used in Chapters III and V to fabricate CuInS 2 solar cells. Table 1.1 summarizes the
11 benefits and drawbacks of leading ink formulations, along with the highest reported solar conversion efficiency for each class.
1.5.1 Solution-based Inks Aqueous solutions of Cu and In salts were some of the first inks used to solution process CuIn(S,Se) 2 layers, though the low viscosity of water limited the deposition to spray coating. An early all-spray device of ZnO/CuInSe 2 achieved 2% 32 efficiency with CuCl 2, InCl 3, and dimethylselenourea in water . The best reported 33 CuIn(Se,S) 2 device created from solely dissolved salts had an efficiency of 5% ; impurities from the solvent and salts, and secondary phase precipitations were challenging to avoid, leading to degraded performance. A method which both avoids early secondary phase precipitation and improves coating speed and smoothness is to create a sol-gel instead of a low viscosity solution. Separate additives are mixed with any solvent to control film thickness, prevent cracking, and stabilize metal species in solution. Kaelin et al. fabricated a 6.7% efficient CIGS cell by an aqueous paste of ethylcellulose and metal nitrates and then heat treating in air at 250°C and Se vapor at 560°C34 . This film had impressively high performance considering that there was a 1 µm thick layer of amorphous carbon under the absorber layer. As seen from this example, there is a significant trade-off between the coating benefits of additives and the electrical detriments resulting from their residues. The most effective solution-based ink developed to date is a sol-gel wherein the solvent, hydrazine, also coordinates with dissolved chalcogenide species. Typically, the metal chalcogenides are slowly dissolved in neat hydrazine, along with extra S or Se, to form hydrazinium species such as (Cu 6S4)-(N 2H5)2 and (In 2Se 4)- 35 (N 2H5)2 . This viscous ink is spun-cast onto Mo substrates, and requires minimal annealing at 525°C to completely remove the hydrazine and form the final CIGS film. Devices fabricated in this manner have maximum efficiencies of 12% 36 due to their large grain sizes and elemental homogeneity. However, hydrazine is a highly toxic and
12 explosive compound, and this technique requires extremely rigorous air exclusion for safety, and thus a safer approach is desired.
1.5.2 Nanoparticle-based Inks Recently, colloidal nanocrystals (NC) of solar absorber materials have attracted much attention 37,38,39,40,41,42 . This new class of materials is exciting because nanocrystal syntheses can utilize lower processing temperatures than bulk synthesis methods, and nanocrystals can be dispersed in solvents and deposited by most of the discussed direct liquid coating techniques. On the other hand, while NC inks offer the advantage of dispersability in organic solvents suitable for many coating techniques, they often require encapsulating ligands that can leave behind residue that hurt device performance 43,44 . Furthermore, sulfide and selenide NC synthesis is often air-sensitive, requiring complicated schlenk-line techniques. The deposition of nanoparticles is nearly identical to that of solution liquid precursors, except in the effect of the reactive anneal. While the reactive anneal forms
CuInS 2 from solution precursors, in nanoparticles annealing can sinter the individual 45 CuInS 2 nanoparticles together into larger CuInS 2 grains . However, complications can arise during this anneal, including incomplete removal of organic molecules required to stabilize nanoparticles and insufficiently dense films due to inefficient nanoparticle packing. Both of these effects impede contact between nanoparticles and thus reduce the grain growth from sintering. Chalcogenide nanoparticles were originally synthesized to make so-called quantum dot solar cells, which involve little or no post-deposition anneal 41 . Wu et. al made Cu 2S/CdS heterojunction solar cells from nanoparticles and nanorods with no post-deposition treatment, obtaining a photovoltaic conversion efficiency of 1.6% 39 . Adding a post-deposition heating step in inert atmosphere increased grain size and film quality, and Li et al. demonstrated ~4% efficient nanoparticleline CuInS 2 solar cells using paste coating with an in-situ nanoparticle synthesis, followed by a sinter step in nitrogen 46 . However, this approach has proven less fruitful than post-annealing films in chalcogenide vapor, which encourages grain growth through chemical
13 transformation. Guo et al. showed that large grained CuInSe 2 films could be formed by selenizing films of sulfide nanoparticles spin coated onto a substrate, creating 5% efficient devices 47,48 . It is generally seen that only the selenization of sulfides has produced grain growth sufficient for device quality films, however nanoparticle sintering is a relatively unexplored technique. Due to the toxicity of Se, it is desirable to have an all-
S process, starting from CuInS 2 NPs and sintering them into an electrically active
CuInS 2 thin film. Chapter III will show our work on synthesizing CuInS 2 nanoparticles that can be reactively annealed in S vapor to form solar absorber layers. In addition to sulfides, oxide nanoparticles can also be used as a precursor for both CuInS 2 and CuInSe 2 absorber layers. The primary advantage of working with dispersions of oxide nanoparticles is that they are extremely air-stable and can be cheaply and easily formed. Oxide nanoparticulate precursors commonly require a more intense reactive anneal than sulfide precursors 49,50 ; however, the transformation from oxides enables grain growth when forming either sulfide or selenide films. Yet, most of the work has been limited to selenization, due to the ease of achieving larger grained films of CuInSe 2 than CuInS 2. By utilizing paste coating, Kapur et al. and Eberspacher et al. cast down thick layers of oxide nanoparticles. Both studies had a reactive annealing step, and devices fabricated had efficiencies of 12% 51 and 14% 52 , respectively. These devices had the highest efficiencies of any CuInSe 2 solar cells through solution processing, largely due to extensive grain growth during the transformation from oxide to selenide. However, it is unclear why similar results have not been achieved with sulfurization, and Chapter IV will deal with a study of this issue.
1.5.3 Hybrid Approaches A hybrid approach can also be taken by mixing nanoparticle and molecular precursors to obtain the benefits of both. The organic component acts as both surfactant and binder, allowing the solvent to wet the substrate and coat evenly, while the nanoparticulate component enables coatings of thicker layers 53 . This can also
14 mitigate pore filling and stabilization problems found when using solely nanoparticles.
Chapter V will review efforts towards using this hybrid approach to form CuInS 2 films.
Precursor Benefits Drawbacks Material Salts Simple to implement; variety Ink stability can be poor, 5%33 of available salts allows many additives required for highly tunable processing coating thick layers. Sulfide NPs Minimal additives needed for Pore filling can be low; 6% 47 thick films. Only sintering organic ligands may remain. required. Oxide NPs Easy removal of additives and Reduced device performance 14%54 other impurities; grain growth from residual oxygen
Hydrazine Very low foreign element Extremely air sensitive 12%36 impurities processing and high toxicity Table 1.1: Comparison of ink compositions
15 2 Methods
All work was performed in conjunction with Ben Weil at Stanford University, and all anomalous diffraction measurements were performed with Sumohan Misra at the Stanford Linear Accelerator, beam line 2-1. Additional electron microscopy of
CuInS 2 nanoparticles was also performed by Shaul Aloni at the Molecular Foundry at Lawrence Berkeley National Labs.
In this section, the primary synthetic and analytical techniques used throughout the rest of the dissertation will be presented. Brief theoretical background is also provided for analytical techniques. All chemicals were used as purchased from Sigma- Aldrich, and rigorous air exclusion techniques were employed by means of a Schlenk line with argon gas, as needed.
2.1 CuInS 2 Precursor Films from Nanoparticle Ink
Relatively large nanoparticles, circa 100nm, were synthesized so as to minimize the ratio of organic capping ligand to CuInS 2 material. In addition, a non- ionic ligand was chosen such that the particles could be formulated into an ink with an organic solvent base. A reaction between the metal oleates and a long chain alkanethiol in a high boiling point coordinating solvent 55 was found to be appropriate for this size regime, and the alkanethiol also acted as a stabilizing ligand for organic solvents.
2.1.1 Oleate Precursor Synthesis Indium and copper oleates were synthesized by the ion exchange of 20mmol sodium oleate with 10mmol copper chloride or 30mmol sodium oleate and 10mmol indium chloride in 35mL hexane, 15mL water, and 20mL ethanol. The reaction mixture was heated at 60°C and stirred vigorously for 4 hours, followed by repeated
16 cleanings with water and evaporation of residual hexane to isolate the waxy product. The synthesis of Ga, Fe, and Zn oleates followed the same procedure as that of Cu, and the accompanying nanoparticle syntheses replaced Cu or In oleates in proportions which retained charge balance, assuming Ga 3+ , Fe 3+ , and Zn 2+ .
Eqn 2.1: MCl x + xNa-OC(=O)C 17 H33 ‰ xNaCl + M-(OC(=O)C 17 H33 )x
In Chapter III, all nanoparticles were made by a modified version of a solution 55 phase synthesis of CuInS 2 nanoparticles wherein oleylamine and dodecanethiol are used as a mixed solvent system to sulfurize metal oleates at moderate temperatures.
Eqn 2.2: Cu-(OC(=O)C 17 H33 )2 + In-(OC(=O)C 17 H33 )3 + C 12 H25 SH ‰
CuInS 2 + C 12 H24
2.1.2 CuInS 2 Nanoparticle Synthesis In a typical synthesis, 0.16 mmol of Cu-oleate and 0.2 mmol of In-oleate were mixed with 5mL dodecanethiol and 5mL oleylamine in a three neck flask, and the solution was vacuumed at room temperature for 20 minutes and at 100°C for 20 minutes. The reaction mixture was then put under argon atmosphere and heated to a temperature between 215-300°C for 20 minutes; cooling to room temperature was allowed to occur gradually under argon atmosphere. Biphasic nanoparticles were obtained at temperatures below 250°C or at early times (<5 min) at higher temperatures. The product was dispersed into ethanol and centrifuged four times to wash out large amounts of residual solvent and precursors. The nanoparticles were then dispersed in toluene for analysis, where they remained suspended for several days, and were easily redispersed by sonication. Small amounts of bulk Cu 2S were found to be present in samples with excesses of Cu-oleate, but they could be filtered out with PTFE filters with 450nm pores.
17 2.1.3 Deposition of CuInS 2 nanoparticle film
Prior to coating, CuInS 2 nanoparticles were additionally cleaned by centrifugation, as aforementioned, in dichloromethane, toluene, and methanol, so as to remove residual reactants and excess ligand molecules. A dispersion was prepared with mass loading of ~10mg/mL in methanol, and 100uL was spin-casted 15 times at 500rpm onto glass which was sputter coated with 500nm of Mo.
2.2 Cu-In-O Precursor Films from Hybrid Nanoparticle Ink
Due to complications in making CuInS 2 films from CuInS 2 nanoparticles, primarily low layer density, another coating method was developed using a hybrid nanoparticle/organometallic mixture. A coating agent was desired which would contain primarily noncarbonaceous species, and polysulfide compounds were chosen for their good coating properties and ease of synthesis in ambient atmosphere. The synthesis involves the reaction between sulfur and In-acetylacetonate to form a viscous In polysulfide in which Cu-rich CuInS2 nanoparticles could be dispersed.
2.2.1 Preparation of Hybrid Ink Solutions of Cu-acetylacetonate (acac) and In-acac were prepared in pyridine, with concentrations of 0.4M and 0.8M, respectively. 500uL of the In-acac solution was mixed with 20mg S powder and sonicated until the solution turned light amber. For each coat, 50uL of the In-acac solution was mixed with 100uL of the Cu-acac solution. The color of the solution immediately darkened to a deep green-brown upon mixing.
2.2.2 Deposition of Cu-In-O Film A typical Cu-In-O precursor layer was created by solution deposition of the above ink onto a Molybdenum-coated glass substrate. Solution deposition was done in two manners, with a setup similar to a Mayer rod and with spin casting. A Mayer
18 rod is a metal cylinder with wire tightly wrapped around it; liquid can be evenly and quickly distributed with such a rod. The analogue of this used in our depositions was a roughened steel pipe which was passed over 50uL of ink. The intrinsic roughness of the bar allowed it to coat in a manner similar to a Mayer rod, leaving a uniform wet film. Spin casting was performed with the same conditions as used in section 2.1, though roughly 20 coats were required. The resulting wet film from both methods was baked in air on a hot plate at ~370 oC between coats. The resulting Cu-In-O films will be discussed in greater depth in Chapter V.
2.3 Sulfurization of Cu-In-O and CuInS 2 NP films
Both precursor films required annealing in S vapor to form the final active layer, and sufurizations were performed with either sealed steel pressure vessels or a rapid thermal processing (RTP) system. Samples were loaded into the stainless steel vessel with ~500mg S, under nitrogen, and sealed by copper gaskets. The entire vessel was placed in a box furnace and heated to the desired sulfurization temperature, generally 500-550°C, over the course of 30 minutes. The sample was held at the maximum temperature for an hour and allowed to cool naturally before the vessel was opened.
In the case of the CuInS 2 NP films and several Cu-In-O films, samples were sulfurized in a RTP with S powder. Samples were loaded into a quartz tube, with
~100mg S, and sealed by rubber o-rings at the tube ends. Atmospheric pressures of N 2 or 2% H 2 in N 2 (to remove residual O 2 and H 2O) were maintained throughout heating by low flow conditions (~20 cm 3/min) and an oil bubbler. Temperature was controlled by six parallel heating lamps, and ramp speeds between 1-10°C/sec were used to achieve temperatures between 500-550°C.
19 2.4 Solar Cell Device Fabrication
After sulfurization, films were etched with 0.5M KCN for 5 minutes to remove any residual copper sulfide. An n-type CdS buffer layer was then deposited by a standard chemical bath deposition; details can be found elsewhere 30 . The CdS deposition was performed by immersing substrates into a growth solution for 12 minutes, followed by rinsing with deionized water. Intrinsic ZnO was then deposited by DC magnetron sputtering. Devices were completed by deposition of an In xSn yO (ITO) top-contact. A 1mm x 1mm shadow mask was used to define the device area of the ITO layer.
2.5 Analytical Techniques
2.5.1 Transmission Electron Microscopy (TEM) TEM was used to analyze the morphology and phase of all nanoparticle samples. TEM utilizes the absorption and diffraction of an electron beam by an ultrathin sample to image both atomic contrast and lattice spacings. TEM was primarily performed on a FEI CM20 at the Stanford Nanocharacterization Lab (SNL) at Stanford University. Additional imaging and elemental mapping by EDS were performed on a JEOL 2100-F at the Molecular Foundry’s Imaging Facility at Lawrence Berkeley National Lab. TEM samples were prepared by drop casting dispersion onto nickel grids, followed by repeated washings in acetone, methanol, and ethanol.
2.5.2 Selected Area Electron Diffraction (SAED) While the TEM is primarily an imaging tool, the nature of the electron imaging allows direct viewing of diffraction space as well. Thus, the beam can be limited to a small area, wherein an electron diffraction pattern is gathered. Lattice spacings are
20 extracted from the electron diffraction pattern by equation 2.3, which relates the spacing of a lattice plane, d, with the electron wavelength, , relative space distance, r, and camera length, L. Lattice spacings obtained in this manner were used in conjunction with angles between diffraction spots to identify phase in nanoparticles. SAED was performed on the FEI CM20 at the SNL during TEM imaging. Eqn 2.3: × = ×
2.5.3 Energy Dispersive X-Ray Spectroscopy (EDS) Another auxiliary function commonly used in conjunction with TEM imaging is Energy Dispersive X-Ray Spectroscopy. On the FEI CM20, signal was collected from areas of nanoparticle samples by use of limiting apertures, with a minimum size of 1 µm. EDS was performed with a JEOL 2100-F, in STEM (scanning TEM) mode, to determine the local composition of both the monophasic and biphasic nanoparticles.
2.5.4 Scanning Electron Microscopy (SEM)
SEM was used to analyze the structure of all films of CuInS 2 to determine film roughness, porosity, and grain sizes. SEM imaging is performed by collecting secondary electrons ejected from a sample surface as an electron beam is rastered across it. An FEI XL30 Sirion SEM at the SNL was used for all analysis. Samples were prepared by either mechanical cleavage of the substrate and mounting on a 45degreee holder or, in the case of nanoparticles, by dropcasting a dispersion onto a Si chip.
2.5.5 Auger Electron Spectrometry (AES) Elemental compositions and distributions of thin film surfaces and profiles were measured with AES. AES images by the same principle of an SEM; however, additional detectors measure the kinetic energy of secondary electrons. The kinetic energies of some of the collected electrons correlate to Auger transitions in the impacted atoms, which allows elemental identification. The Phi 700 Scanning
21 Electron Probe was used at the SNL for all analysis; sample preparation was identical to that used for SEM.
2.5.6 UV/Visible Spectroscopy UV/Vis spectroscopy was used to identify the band gaps of both nanoparticles and thin films of CuInS 2. The incident beam energy is scanned by means of a monochromator, and the absorption of light in a semiconductor primarily occurs when the energy of a photon can excite electrons across the band gap. UV/Vis spectra were collected on a Shimadzu UV-1700 and Carie 3000 UV- visible spectrometers. Samples were diluted into toluene and spectra were taken in a two beam transmission mode with pure toluene as the reference.
Band gaps were determined by assuming that CuInS 2 is a direct band gap material and Cu 2S is an indirect band gap material and thus follow equations 2.4 and 2.5.
1 A ~ (hν − E ) 2 Eqn 2.4 (direct): g 2 ()ν − A ~ h Eg
A ~ (hν − E )2 Eqn 2.5 (indirect): g 2/1 ()ν − A ~ h Eg
All band gaps of CuInS 2 and its alloys were extrapolated from plots of squared absorption and wavelength in the region between 1.6eV and 1.3eV. The band gap of an alloy usually follows Vegard’s law due to the dependence of band gap on lattice constants 56 . The nonlinear bowing parameter, b, was assumed to be zero for the purpose of predicting band gaps as a function of composition. Eqn 2.6: E = E × %x + E × %y − b × %x × %y g g x g y
2.5.7 Powder X-Ray Diffraction (XRD)
22 Powder XRD was the primary means of phase determination for all synthesized nanoparticles and films. XRD generates patterns of reflections when an X- ray beam incident on a crystalline material satisfies the Bragg condition (Equation 2.7). Eqn 2.7: = 2 sin
XRD was performed at the SNL on a PANalytical X'Pert with Cu K α radiation at 45kV and 40mA, and crystal modeling and XRD simulation were done with Materials Studio (Accelrys Inc.) at the SNL. Scans were acquired by rotating the incident beam and detector symmetrically around the sample, with 2 being the subtended angle. The acquired patterns were compared against appropriate reference spectra from the Joint Committee on Powder Diffraction Standards (JCPDS). XRD samples of nanoparticles were drop cast out of toluene, and all films showed no changes due to oxidation over several weeks in atmosphere.
2.5.8 Anomalous Diffraction
Synchrotron based XRD of the CuInS 2 nanoparticles substituted with Fe, Zn and Ga was performed on beamline 2-1 at Stanford Synchrotron Radiation Lightsource (SSRL). The nanoparticle samples were measured in θ–2θ geometry with 1 mm slits to define the diffracted beam acceptance and a Vortex Detector was used to collect the diffracted X-rays. The diffracting intensity was recorded as a function of the scattering vector, , where 2 θ is the angle between the = (4 ∙ sin( ))/ ) incident and diffracted X-rays and λ is the wavelength of the X-rays. The XRD data were first collected between Q of 1.7 and 3.6 Å −1 at 7112 eV for Fe samples; between Q of 1.75 and 4.0 Å −1 at 10367.1 eV for Ga samples. For the anomalous X-ray −1 diffraction, profiles were measured between Q of 1.7–3.5 Å around the (100) h, (112), (013), (020)/(004), (220), (024) peaks at 25 energies between 6950 and 7160 eV, across the X-ray absorption edge of Fe at 7112 eV. Similar profiles were measured for both Zn and Ga samples. Data was collected around the (100), (002) and (101) peaks at 29 energies between 9500 and 9800 eV, across the X-ray absorption edge of Zn at
23 9658.6 eV and 31 energies between 10210 and 10500 eV, across the X-ray absorption edge of Ga at 10367.1 eV. The peaks in the XRD profile was fitted to either one or two peaks with Pseudo-Voigt function using Origin 8.0 (Origin Lab Corporation). The integrated area (A int( hkl )) of each individual fitted peak were obtained at each X-ray 2 energy. A int( hkl ) is directly proportional to the structure factor |F| (hkl ) of the sample at that X-ray energy. The experimental data was corrected for beamline attenuation resulting from air path-length, ion-chambers and Be-windows (in the detector) and also attenuation due to sample thickness. The molar fractions of Fe, Zn and Ga substituted CuInS 2 nanomaterials are obtained by fitting the experimental data to the 2 simulated | F| (hkl ), using the relationship between the structure factor and the X-ray energy : f = f (Q) + f ' (E) + if '' (E) Eqn 2.8: n 0
atoms π + + = )(2( hxi n ky lz nn ) F lkh ),,( ∑ fn (E) e Eqn 2.9: n=1
Here, f0(Q) is the atomic scattering factor, f′(E) is the real part of the anomalous sc attering factor, f″(E) is the imaginary part of the anomalous scattering factor and xn, yn & zn are the atomic co-ordinates. The f0(Q), f′(E) and f″(E) are tabulated only for elements in their zero oxidation state. Hence, in order that the 3+ 2+ 3+ simulated F(h,k,l) included absorption edge shifts due to cations (Fe , Zn and Ga ) and fine structure features near the absorption edges, the following Kramers-Kronig relationship is used to obtain f′(E) from measured f″(E) (fluorescence data) values near the corresp onding absorption edges: 2 ∞ E f ' (E ) = f '' (E) dE 0 ∫ 2 2 π 0 E − E Eqn 2.10: 0
2.5.9 Electrical Measurements An Agilent 1500B parameter analyzer was used to perform all I-V measurements on solar cell devices. Samples were scribed to remove all material except the Mo back contact, and probe tips were touched down onto the exposed Mo
24 and the ITO pads. Voltage was scanned from -1V to 1V, with a voltage compliance of 1A. I-V curves were generated by the parameter analyzer, and subsequent analysis was done with Microsoft Excel and Origin Labs OriginPro 8. Current densities (J) were used in all cases for clarity of comparisons with standard solar cells. RSH and R S were approximated by taking inverse slopes of the I-V in the dark at -0.1V and 0.9V, respectively.
25 3 CuInS 2 Nanoparticle Growth Mechanism
3.1 CuInS 2 Nanoparticle Synthesis
Nanoparticle syntheses can utilize lower processing temperatures than bulk synthesis methods, and device fabrication with nanoparticles can benefit from high throughput solution phase processing 57 . In addition, chemically synthesized nanostructures can provide well-defined domains for understanding complex phase behaviors such as vacancy ordering and Cu(I) ion diffusion 58,59 , which are particularly important in I-III-VI 2 semiconductors.
CuInS 2 nanoparticles have previously been made by standard synthetic methods such as solvothermal 60,61 , hydrothermal 62 , and single-source decomposition 63,64 . Nanoparticles have been mostly synthesized in the chalcopyrite phase (JCPDS #85-1575) of which all solar absorber layers of CuInS 2 are composed. 60 However, Lu has shown that a new wurtzite phase of CuInS 2 is accessible by a hot- injection synthesis method. The wurtzite phase allows flexibility of stoichiometry, due to the copper and indium sharing a lattice site, and thus might avoid the deleterious vacancies found in Cu-poor films of CuInS 2. However, hot-injection syntheses necessarily have small yields and require careful processing, so a more robust solvothermal synthesis of CuInS 2 would be preferred. In addition, the transformation from wurtzite to chalcopyrite CuInS 2 at high temperatures and sulfur pressures might assist in grain growth through a recrystallization process. In this chapter, I will describe the details of the synthesis of wurtzite CuInS 2 nanoparticles, and then show their transformation into functional chalcopyrite CuInS 2 solar cells.
3.2 CuInS 2 Nanoparticle Growth Mechanism
A recent solvothermal synthetic procedure developed by Hyeon 55 and Tang 65 utilized metal-oleate complexes in oleylamine and dodecanethiol to produce
26 heterostructured nanoparticles. We chose this method due to its simplicity, low cost of precursors, and scalability of the synthesis. However, nanoparticles thus made were identified as Cu 2S-In 2S3 heterostructures without CuInS 2 in the original work.
Briefly, we discovered that the growth process starts with nucleation of Cu 2S nanodiscs, followed by epitaxial overgrowth of CuInS 2 NPs onto only one side of
Cu 2S nanodiscs, resulting in biphasic Cu 2S-CuInS 2 heterostructured NPs. We also tracked the phase transformation of biphasic Cu 2S-CuInS 2 heterostructured NPs into monophasic CuInS 2 over synthesis time. We studied this complex phase behavior using a variety of characterization techniques including Selected Area Electron Diffraction (SAED), Transmission Electron Microscopy (TEM), X-Ray Diffraction (XRD), UV-Visible Spectroscopy, and Energy Dispersive X-Ray Spectroscopy (EDS).
3.2.1 Proposed Nanoparticle Growth Pathway We studied the evolution of nanoparticle shape and composition during the reaction of copper-oleate, indium-oleate, dodecanethiol, and oleylamine, at 260°C, by taking aliquots over the course of the reaction. Figure 3.1 shows a schematic representation summarizing growth steps along with corresponding TEM micrographs.
Nanoparticles have been observed to first nucleate as Cu 2S nanodiscs (diameters of 22±5nm and widths of 9±2nm) (Figure 3.1(a)), followed by anisotropic overgrowth of wurtzite CuInS 2 onto only one side of Cu 2S nanodiscs. This forms biphasic Cu 2S-
CuInS 2 heterostructured NPs (Figure 3.1(b)). As growth continues, the Cu 2S segment is chemically transformed into wurtzite CuInS 2, with an intermediate state seen in
Figure 3.1(c). After 20 minutes, the final monophasic CuInS 2 NPs are formed (Figure 3.1(d)). The NPs are tapered in shape and the maximum diameter is found to be constant throughout the growth.
27
Figure 3.1: Schematic of proposed CuInS 2 growth mechanism
(a) Formation of Cu 2S nanodiscs at earliest times. (b) Epitaxial growth of CuInS 2 on
Cu 2S nanodiscs leads to biphasic NPs. (c) Continued growth of CuInS 2 occurs simultaneously with conversion of Cu 2S into CuInS 2. (d) Solely monophasic CuInS 2 present upon completion.
The maximum diameter is found at the interface between the Cu 2S and CuInS 2 phases in the biphasic NPs, and at the least tapered region in the monophasic NPs. All following analyses were performed on an aliquot taken at 5 minutes and the final product, which were chosen as representative samples of biphasic and monophasic nanoparticles, respectively.
3.2.2 Elemental Mapping
EDS mapping in Figure 3.2(a-d) shows that the monophasic CuInS 2 NPs grown for 20 minutes are elementally homogeneous along the whole NP within instrumental resolution. Figure 3.2(e-h) are EDS mapping data of Cu 2S-CuInS 2 biphasic NPs grown for 5 minutes. Figure 3.2(f) shows that Cu is present throughout the NPs, but is in much higher concentrations in the caps of the NPs. However, the In
28 signal in Figure 3.2(g) is located solely in the main body of the NPs, not at the cap, while the S signal is constant throughout the whole NP (Figure 3.2(h)).
Figure 3.2: TEM images and composition maps of CuInS 2 Nanoparticles
(a) monophasic CuInS 2 and (e) biphasic CuInS 2-Cu 2S nanoparticles. EDS mapping of
(b-d) monophasic CuInS 2 and (f-h) biphasic CuInS 2-Cu 2S. A cap of a representative biphasic nanoparticle is highlighted with a white box in (f-h) for clarity. The intensity gradient in all biphasic nanoparticles is due to a corresponding thickness gradient of the cone shaped bodies.
Large area EDS gives the ratio of Cu:In:S in the monophasic CuInS 2 NPs as
23:27:50, which is close to the 1:1:2 ratio of CuInS 2. The elemental ratio in the whole biphasic Cu 2S-CuInS 2 NPs is typically Cu-rich and In-poor by an amount corresponding to the cap volume, with a usual Cu:In:S ratio between 32:18:50 and 42:12:46. Elemental ratios of Cu:S in isolated copper sulfide nanodiscs were found to be approximately 64:36, indicating that Cu 2S is present instead of CuS. It can be
29 inferred from EDS mapping results that the chemical transformation involves the anisotropic interdiffusion of Cu and In between the Cu 2S and CuInS 2 phases, coupled with In absorption from the growth solution and diffusion through the NP.
3.2.3 UV/VIS Spectroscopy
Spectroscopy was used to track the progression of Cu 2S-CuInS 2 NP growth via optical bandgap. We noted that the nanoparticle size is much larger than the Bohr radius of these materials, therefore quantum confinement is not relevant 66 . In figure 3.3(a), the spectrum of a sample taken from a reaction at the first sign of color change, shows particles having a bandgap of ~1.38 eV (~ 900 nm), which is consistent with 67 the bandgap of Cu 2S in the chalcocite phase . The spectrum of an aliquot at an intermediate stage of the reaction, in Figure 3.3(b), shows the emergence of a shoulder at ~1.63 eV (~760 nm), indicating CuInS 2 nanoparticle formation. This shoulder is possibly from defect-based absorption 63 , and vacancies likely form during the low temperature conversion process. Here, the shoulder is used as an indication of the progress of the growth of CuInS 2 and the biphasic to monophasic transition. The final product’s spectrum in Figure 3.3(c) gives the bandgap of CuInS 2 as ~1.5 eV (~830nm). The stronger shoulder suggests the completion of the conversion to monophasic
CuInS 2 nanoparticles.
30
Figure 3.3: UV/VIS spectra of nanoparticles
(a) Cu 2S with small CuInS2 body, (b) CuInS 2 with small Cu 2S cap, and (c) pure
CuInS 2 nanoparticles in dilute toluene dispersions. Bulk band gap energies are 1.38 eV for Cu 2S and 1.5 eV for CuInS 2.
3.2.4 High Resolution Transmission Electron Microscopy (HRTEM) of Single Nanoparticles
HRTEM and SAED on monophasic CuInS 2 (Figure 3.4(a,c)) and biphasic
Cu 2S-CuInS 2 (Figure 3.4(b,d)) nanoparticles give crystallographic information about the growth direction and relative orientation of the two phases. HRTEM and SAED of monophasic NPs are consistent with hexagonal wurtzite CuInS 2 with the NP long axis along the <001> direction. The SAED patterns of the biphasic nanoparticles in Figure
3.4(d) can be indexed as a hexagonal chalcocite Cu 2S phase (JCPDS #26-1116) and a hexagonal CuInS 2 phase. HRTEM shows the epitaxial interface between Cu 2S and
CuInS 2. Both Cu 2S and CuInS 2 have the same <001> crystallographic direction pointing along the long axis of NPs. We note that there is only 4% and 1.5% difference in the c and a lattice parameters, respectively, between Cu 2S and CuInS 2, so the epitaxial interface is expected. In a similar fashion to other ternary chalcogenide
31 nanostructures 58 , both monophasic and biphasic nanoparticles show evidence of two or four-fold superlattice formation in the <100> direction, as seen in Figure 3.4(c,d).
Figure 3.4: HRTEM images of nanoparticles
(a) monophasic CuInS 2 and (b) Cu 2S-CuInS 2 nanoparticles with phase boundary in inset of (b). (c) SAED of monophasic CuInS 2. (d) SAED of Cu 2S-CuInS 2 nanoparticles.
Superlattice formation can be indicative of extensive defect formation in ternary chalcogenides68 . Defect formation would allow anisotropic vacancy-based diffusion 69 , which can contribute to the generation of one dimensional structures, along with parallel mechanisms such as differential surfactant absorption between crystal faces, as seen in CdSe nanoparticle growth 70 .
32 3.2.5 X-Ray Diffraction Our TEM and SAED analysis is supported by XRD in Figure 3.5(a), which also shows that the CuInS 2 NPs can be indexed as the wurtzite phase. Figure 3.5(b) shows a simulated powder diffraction pattern generated from the proposed CuInS 2 structure of Lu 60 , which agrees in intensity and position to the experimental pattern.
While a combination of the tetragonal In 2S3 and hexagonal Cu 2S reference patterns in Figure 3.5(c,d) can approximate some of the peaks, the match is weak and has several extraneous reflections.
33
Figure 3.5: XRD of wurtzite CuInS 2 nanoparticles (a) Experimental XRD pattern of CuInS2 nanoparticles. (b) Simulated powder diffraction of proposed hexagonal CuInS2 structure. Reference patterns of (c)
34 tetragonal In 2S3 and (d) hexagonal Cu 2S. Guidelines overlayed to demonstrate matching of (a) with (b) and mismatching of (a) with (c,d).
3.2.6 Epitaxial Growth and Phase Conversion
The most interesting elements of the growth process are the nucleation of Cu 2S nanodiscs, epitaxial overgrowth of CuInS 2, and phase conversion of Cu 2S into CuInS 2 so that biphasic Cu 2S-CuInS 2 NPs can turn into monophasic CuInS 2 NPs. There are several important properties of the conversion mechanism. First, CuInS 2 and Cu 2S both have a hexagonal structure and share nearly identical packing of the S sub-lattices.
Figure 3.6(a,b) show the crystal structures of Cu 2S and CuInS 2, respectively, projected along the <001> direction. Figure 3.6 (c) and (d) show the crystal structure from an isometric point of view roughly along the <110> for Cu 2S and CuInS 2.
Figure 3.6: Crystal structures of relevant phases
(a) hexagonal chalcocite Cu 2S and (b) hexagonal CuInS 2 with zone axis of <001>.
Crystal structure of (c) hexagonal chalcocite Cu 2S and (d) hexagonal CuInS 2 viewed from another angle.
For Cu 2S, the Cu(I) ions are color-coded to indicate different S coordination: tetrahedral (blue), trigonal (red), and linear (green). Second, at the growth temperature
35 71 of 250 ºC, the chalcocite Cu 2S phase is in the superionic conducting state . Cu(I) ions are randomly distributed in the interstitial sites formed by the S sub-lattice. Cu(I) ions have very high mobility, as if they are in a fluidic state, which speeds up exchange with In(III) ions and chemical transformation. Third, the conversion from Cu 2S to
CuInS 2 requires little lattice distortion and thus can be done with a low energy barrier. As visualized in Figure 3.6, the S sub-lattices do not need to move while half of the Cu(I) ions (trigonally and linearly bonded Cu(I) ions) are required to move out of the 72 Cu 2S phase . Half of the remaining Cu(I) ions (tetrahedrally bonded) are replaced by In (III) ions. Finally, both Cu(I) and In(III) ions randomly distribute among tetrahedral sites, resulting in the CuInS 2 phase.
3.2.7 Growth Mechanism Modeling Understanding the growth mechanism can allow precise control over the size and shape of resultant nanoparticles. This reaction can be modeled as three separate steps: the nucleation of Cu 2S nanodiscs, the growth of the CuInS 2 phase on only one side of Cu 2S nanodiscs through absorption of reactants, and the conversion from biphasic to monophasic nanoparticles by interdiffusion of Cu and In between the
CuInS 2 and Cu 2S phases. The progressions of growth in CuInS 2 NP length and Cu 2S nanodisc thickness and phase transformation with respect to time are shown in Figure
3.7. The seemingly linear increase in CuInS 2 NP length, seen in the blue curve of Figure 3.7(a), can be described as diffusive growth in one dimension with a constant concentration of precursors. However, the detailed time trace also shows a slight exponential at times between 0 and 9 minutes (red curve of Figure 3.7(a)), caused by the limiting step of forming Cu 2S seeds. The depletion of Cu and In reactants also causes the length to reach a saturation limit at later times between 22 and 30 minutes (green curve of Figure 3.7(a)).
36
Figure 3.7: Time traces of nanoparticle phase conversion
(a) CuInS 2 length La (nm), (b) Cu 2S thickness Lb (nm), and (c) conversion progress (% monophasic/total nanoparticles). Fittings are (a) exponential, linear, and logistic; (b) powe r law; (c) sigmoidal Ln 100(− conversion %) = kt n with n=4.
This shape is commonly seen for fast nucleation steps followed by diffusive growth 73 , but the middle linear growth stage is prolonged due to the high stability of the metal-oleate precursors. Their stability at high temperatures leads to a low and
37 constant concentration of active metal species in solution, which in turn leads to a constant rate of NP elongation as seen in the middle linear region of Figure 3.7(a).
The disappearance of the Cu 2S phase also occurs at later times, and is modeled here as a solid-state phase change. Progress is tracked, in Figure 3.7(c), as the fraction of the number of monophasic to total nanoparticles versus the same reaction time as in
Figure 3.7(a,b). This transformation occurs concurrently with the saturation of CuInS 2 growth at around 20 minutes, likely due to a reduction in the concentration of Cu- oleate in solution. The only source of Cu at subsequent times is the Cu 2S phase, though it does not seem that Cu is gradually leeched, but is instead suddenly removed.
The surprising abruptness of this transition is reflected in the constant Cu 2S nanodisc thickness prior to transformation, as seen in Figure 3.7(b), which coincided with the absence of partially transformed Cu 2S nanodiscs. We think that the abruptness of the transition is due to the thinness (9nm) of the Cu 2S phase and the high Cu ion mobility in the superionic state. The curve in Figure 3.7(c) can then be fit well to the Avrami- Erofe’ev growth model 74,75 , which simulates a solid-solid seed-based phase transformation, with an estimated order of ~4 by least squares fitting. This indicates that the transition is likely three dimensional despite the linearity of the NPs and involves a high rate of nuclei generation at the CuInS 2-Cu 2S interface, which would corroborate the sudden disappearance of the Cu 2S phase. The original causes of one dimensional growth are likely the inhibition of radial diffusion by the surfactant oleylamine and enhancement of growth rates at the 76 high energy (001) faces of wurtzite chalcogenides . The fact that CuInS 2 only nucleates onto one side of Cu 2S nanodiscs suggest that the (001) face terminated with S anions has different binding affinity from the face terminated with Cu and In cations.
In this study, it is not yet determined which face favors the CuInS 2 overgrowth. Our observation of preferential growth of one side of a nanoparticle has previously been seen in the case of wurtzite CdSe NPs 77,70 , which showed preferential growth on the cationic (001) face, leading to a “tadpole” shape. Overall, our system’s novel combination of stable precursors, well separated nucleation and growth steps, and
38 asymmetric crystal structure all provide an easily analyzed and controlled growth of one-dimensional CuInS 2 nanoparticles.
3.2.8 Summary of Growth Study In conclusion, we have synthesized a range of Cu-In-S structures, with fine control over both phase and size, by a low temperature solution phase synthesis. We found that the growth process starts with Cu 2S nanodiscs followed by epitaxial overgrowth of CuInS 2 on only one side of nanodiscs, resulting in biphasic Cu 2S-
CuInS 2 heterostructures consisting of hexagonal chalcocite Cu 2S and wurtzite CuInS 2. We showed that the biphasic NPs can be easily converted to monophasic NPs at temperatures above 250°C due to their similar crystal structures and fast diffusion of Cu(I) ions in the superionic state. These compositional, structural, and mechanistic findings provide insight into the controlled solution growth of ternary chalcogenide nanoparticles.
3.3 Solar Cells from CuInS 2 Nanoparticles
Nanoparticles were cast as described in chapter II, forming films of ~1 µm after 10 coats. Four representative devices were made, labeled NP1-4, from four batches of
CuInS 2 nanoparticles with different stoichiometries and sizes; their respective properties can be seen in table 3.1. It has been seen in the literature that a high Cu:In 78 ratio, sometimes as much as 1.5:1, can enable greater grain growth and higher V OC ; therefore we tested ratios from 0.95:1 to 1.2:1. Also, it was assumed that larger NPs, with their lower surface:volume ratio, would have lesser organic contamination. So, our nanoparticles, with their sizes of 50-100nm, were well suited for this experiment.
Upon annealing CuInS 2 NPs in S vapor, the films’ grain structure grew to a greater or lesser extent depending on several factors (Figure 3.8), and no films showed noticeable residual carbon deposits. Grain sizes were estimated by taking average grain sizes across multiple samples.
39
Figure 3.8: Nanoparticle films after 5 minute sintering in S vapor Starting from precursor films (a) NP1 (b) NP2 (c) NP3 (d) NP4. All scale bars are 500nm
The largest factor in grain growth was the size of the original nanoparticles, with 50nm nanoparticles transforming into films with grains of ~200nm, compared to ~100nm with 100nm nanoparticles. It can be reasoned that the more anisotropic shape of the larger nanoparticles, which led to poorer packing and thus lesser interparticle contact, was the main cause. The second factor that was found to matter was the Cu:In ratio of the original nanoparticles. Figure 3.8 shows comparisons of grain growth between Cu-rich and Cu-poor CuInS 2 films sintered in S vapor. Previous studies with sputtered films suggest that excess Cu forms Cu xS, which enables grain growth by acting as a fluxing agent. While these grains sizes are smaller than the ~500nm to 1 µm grains in vacuum-deposited CuInS 2 solar cells, grain sizes between 100-200nm are typical in many solution processed films of chalcogenides.
40 Sample NP size Grains Cu:In VOC JSC Eff. RS RSH (nm) (nm) (V) (mA/ (%) (ohm* (kohm* cm 2) cm 2) cm 2) NP1 200x50 90 1.05 0.13 0.2 0.007 50 30 NP2 100x50 200 1.05 0.37 2.5 0.32 15 38 NP3 50x30 70 .95 0.23 0.4 0.03 12 32 NP4 50x30 240 1.2 0.35 9.8 1.27 2.5 10.6
Table 3.1: Device properties for 4 solar cells made from CuInS 2 nanoparticle films
Device performance was found to be highly dependent on both grain size and the Cu:In ratio. Figure 3.9 shows comparisons of the devices made from the four aforementioned films, with device properties compared in Table 3.1. The efficiency differences between samples can be explained by the different amounts of collected current and series resistances. The series resistance is strongly correlated to the number of resistive interfaces, such as grain boundaries, and so the films with larger grains showed lower series resistance. While the shunt resistances of all samples were reasonably large in the dark, indicating the absence of morphological defects, large light-activated parasitic conduction can be seen in all samples under illumination. The sample with the largest grains and thus longest possible minority lifetimes correspondingly had the highest efficiency and J SC . However, all samples suffered from low V OC , possibly due to the very high roughness of the junction between CuInS 2 and CdS. Such recombination at the p-n interface is commonly the limiting factor in 79 CuInS 2 solar cells , though the n-type contact is believed to also be sub-optimal and thus also reducing V OC .
41
Figure 3.9: I-V curves of devices NP1-4 (a) NP1 is close to stoichiometric large NPs, (b) NP2 is close to stoichiometric small NPs, (c) NP3 is Cu-poor small NPs, (d) NP4 is Cu-rich small NPs.
42 4 Band Gap Tuning of CuInS 2 Nanoparticles through Alloying with Zn, Ga, and Fe
4.1 Alloying in CuIn(Se,S) 2
While extensive work has shown the benefits of alloying in CuInSe 2, the effects of similar processes are not as well understood in CuInS 2. The most well known alloy of CuInSe 2 is CuIn 1-xGa xSe 2, with x~0.3, which both raises the band gap from 1.05eV to 1.3eV and reduces the effect of Cu:In stoichiometry on device 80 81 performance . The analogous substitution in CuInS 2 also increases V oc and increases tolerance to off-stoichiometry 82 , though segregation of the Ga to the back 83 contact during traditional processing limits its benefit . It has been demonstrated that interdiffusion during the deposition of buffer layers of CdS onto CIGS films may create a buried homojunction 84,85 . This type of junction typically reduces interfacial recombination and thus might be a contributing factor to the higher performance of CuInSe 2-based solar cells as compared to CuInS 2.
Therefore, a suitable means of spatial dopant inclusion in CuInS 2 may enable the creation of a similar buried homojunction. In addition, spatial control of composition could allow the creation of a concentration grading, increasing minority carrier diffusion and efficiency 86 , similar to what is seen in CIGS solar cells. Inclusion of alloying elements which both raise and lower the band gap could also be used to create new materials with a wide range of band gaps. One example is the case of substituting In with Sn and Zn in Cu 2ZnSnS 4, a similar material composed of more Earth-abundant elements 87 , to achieve a similar band gap with more abundant materials.
While most CuInS 2 films are made by cheaper sulfurization techniques, more expensive co-evaporation and other vacuum techniques can control elemental profiles precisely 14 . However, spatial control of composition is possible by solution processing multiple layers of material, at a lower cost than traditional vacuum techniques.
43
4.2 Alloying in CuIn(Se,S) 2 Nanoparticles
Solution processing of alloy nanoparticles by the means described in Chapter III requires modification of the previous synthesis to accommodate desired elements. In addition, multiple means of verification must be implemented to confirm elemental inclusion at the desired quantities. Several studies have described the synthesis and 88 89 90,91 analysis of alloyed NPs of CuIn 1-xGa xS2 , CuIn 1-xGa xSe 2 , and CuIn 1-xZn xS2 . However, all of these studies have involved hot-injection synthesis, which follow different mechanisms than the aforementioned solvothermal route. A critical difference which arises is that most syntheses of CuInS 2 alloys have focused on the chalcopyrite phase (jcpds#047-1372); while the synthesis of interest creates nanoparticles in the related wurtzite phase. Nanoparticles formed in the wurtzite phase were shown to grow controllably in the <002> direction, with sizes ranging from 5nm to 200nm, notably larger than nanoparticles formed in the chalcopyrite phase. As mentioned in chapter III on the mechanism of CuInS 2 NP formation, the CuInS 2 attains its anisotropic shape and wurtzite structure from the epitaxial overgrowth off an initial hexagonal Cu 2S NP. Herein, we will describe the substitution of Fe, Zn, and Ga into wurtzite
CuInS 2 nanoparticles in order to understand compositional changes and optical effects pertinent to solar absorbers. These elements are chosen to allow either a lowering or raising of the band gap, and alternately to replace the In, as has been demonstrated with Cu 2ZnSnS 4. Traditional laboratory X-ray diffraction experiments cannot easily distinguish between neighboring elements in the periodic table, such as Cu and Zn. However, a synchrotron radiation source provides high incident X-ray intensity and allows variation of the X-ray energy, which enables compositionally sensitive analysis of phases by anomalous X-ray diffraction (AXRD). This technique combines both structural (diffraction) and chemical (absorption spectroscopy) analysis techniques, thereby being sensitive to specific crystallographic phases and unique crystal sites
44 within a particular phase. The above technique was used to determine the distribution of Fe, Zn, and Ga in CuInS 2.
4.3 Anomalous X-ray Diffraction
To determine whether the distribution of Fe, Zn, and Ga in CuInS 2 is substitutional, interstitial, or segregated, we have used AXRD measurements below, through, and above the absorption edges of the above cations. Using this technique, we perform diffraction scans at a variety of energies for a particular Bragg peak corresponding to an unique crystallographic phase (chalcopyrite vs . wurtzite) – we see a decrease in the diffracted intensity as we traverse through the absorption edge if the cation is present in the said crystallographic phase, while if the cation is absent then we do not see such a change. To quantify the amount of substituent present in a particular crystallographic phase, the experimental data was fitted to the simulated 2 3+ 2+ 3+ |F| (hkl ), containing both absorption edge shifts (due to cations – Fe , Zn and Ga ) and fine structures near the corresponding absorption edges.
4.4 Effects of Alloying on CuInS 2 Optical Band Gap
The primary expected effect of elemental substitution in CuInS 2 is a shift in the band gap from its original value of 1.5eV. Values of observed band gaps from UV- visible spectroscopy are shown in Table 4.1, along with projected values from simple alloying with CuFeS 2, ZnS, and CuGaS 2. Representative absorption spectra, along with optical band gap fitting parameters, can be seen in Supplementary Information. Assuming alloying with ZnS (Eg=3.6eV), Zn substitution of Cu and In should lead to a rise in the band gap, and the actual rise was close and proportional to the expected amount. Fe showed a similar correlation between the experimental and the expected drop from alloying with CuFeS 2 (Eg=0.6eV). Ga, however, showed no observable change in band gap as would be expected from alloying with CuGaS 2 (Eg=2.2eV). To
45 determine the origin of these spectroscopic phenomena, structural information was gathered on all samples by XRD, TEM, and AXRD.
Eg(eV)
Element 5% 10% 20%
Ga 1.49 1.41 1.49 (1.50) (1.55) (1.66) Fe 1.43 1.38 1.35 (1.45) (1.36) (1.28) Zn 1.47 1.51 1.56 (1.50) (1.56) (1.67) Table 4.1: Optical band gaps of nanoparticles with substitution Expected values from alloying are shown in parantheses.
Elemental additions will be discussed separately in the following sections, so that the effect of each element on the structure of CuInS 2 will be clearer.
4.5 Effects of Alloying on CuInS 2 Morphology and Phase
4.5.1 Substitution of Fe
Fe substitution in CuInS 2 can be represented as alloying with chalcopyrite
CuFeS 2, which has Eg=0.6eV. Fe inclusion in CuInS 2 in extremely low dopant levels has been previously studied 92 , though it has not been thoroughly explored as a substitutional element. UV-visible spectroscopy showed optical band gaps which were within 5% of the projected alloy values, thus it was expected that Fe would be observed in the nanoparticles.
The substitution of Fe into the CuInS 2 nanoparticles produced radical morphological changes, and EDS also confirmed the presence of Fe in samples. At the maximum tested substitution of 20% Fe, NPs with a size of 11±2nm are observed
46 (figure 4.1(a)), and HRTEM shows only lattices with 3.2Å and 2Å spacings. The absence of the wurtzite (100) reflection and the prominence of the chalcopyrite (112) reflection in XRD (figure 4.1(b)) confirms that the sample is entirely in the chalcopyrite phase. Scherrer broadening of the (112) peak gives an estimate of ~17nm for average nanoparticle size.
Figure 4.1: CuInS 2 NPs with Fe substitution
(a) TEM image of CuInS 2 nanoparticles with 20% Fe substituted. (b) XRD patterns of
CuInS 2 samples with 5%, 10%, and 20% Fe substituted; chalcopyrite and wurtzite peaks are indicated for clarity. (c) AXRD spectra of the (112) and (100) reflections in samples with 10% Fe substitution.
At lower Fe substitutions of 5% and 10%, the synthesis yields NPs with diameters of 17nm±4nm and 32±5nm, respectively. XRD patterns obtained on 5% and 10% samples (figure 4.1(b)) show an apparent mixture of chalcopyrite and wurtzite phases. High intensity at the (112) reflections indicates a majority of the chalcopyrite
47 phase, while intensity in the (100) peak indicates the presence of some wurtzite phase material. Anomalous XRD of a 10% Fe sample is shown in Figure 4.1(c), which compares the Fe signal of the minority wurtzite versus majority chalcopyrite reflections. Similar dips in both a peak that is unique to wurtzite, the (100), and the prominent chalocpyrite peak, (112), indicate that Fe is present at similar levels in both phases, instead of favoring the chalcopyrite phase. To distinguish if phase segregation was between or within nanoparticles, HRTEM and SAED were performed on many individual nanoparticles from a 10% Fe sample. A representative nanoparticle is shown in Figure 4.2, with two grains of indeterminate phase. Both sides have lattice spacings of 3.2A and 2.0A, corresponding to the either the (211) and (024) chalcopyrite reflections or the (002) and (110) wurtzite reflections. However, the right grain can be positively identified as wurtzite due to the (102) and (001) planes, with spacings of 2.3A and 6.4A. This indicates that each nanoparticle is likely biphasic, with an epitaxial stack of wurtzite-CuIn 1-xFe xS2 and chalcopyrite-CuIn 1-xFe xS2.
Figure 4.2: HRTEM image of a CuInS 2 nanoparticle with 10% Fe Lattice spacings are marked on both sides of the grain boundary for phase determination.
48
Combining the results of TEM, XRD and AXRD, it would seem that the chalcopyrite phase of CuIn 1-xFe xS2 grows off of the small seeds of the wurtzite CuIn 1- xFe xS2, and at some critical concentration converts the whole nanoparticle into the chalcopyrite phase. The preferential formation of chalcopyrite CuIn 1-xFe xS2 is not surprising given the instability at room temperature of the wurtzite phase of CuFeS 2 in the bulk. The low temperature conversion of this alloy to the chalcopyrite phase might be advantageous for the sintering of nanoparticle films, though further studies are required. In addition, the addition of Fe is shown to be an effective means of lowering the band gap of CuInS 2.
4.5.2 Substitution of Zn UV-visible spectroscopy shows that there is a strong increase in optical band gap with increasing Zn substitution. Assuming that the sample is the alloyed phase of wurtzite CuInS 2 (Eg=1.5eV) and wurtzite ZnS (Eg=3.6eV), the measured and expected values of optical band gap compare favorably, as seen in Table 4.1. The general raise in band gap indicates likely Zn inclusion in the wurtzite CuInS 2 lattice, but the values are lower than estimates by 4-8%, possibly due to incomplete substitution of Zn. TEM (figure 4.3(a)) shows that Zn substitution results in little change to the overall morphology of the CuInS 2 nanoparticles. The average size and shape matches well with that of normal CuInS 2 nanoparticles, implying that Zn substitution does not change the growth mechanism substantially.
49
Figure 4.3: CuInS 2 NPs with Zn substitution
(a) TEM image of CuInS 2 nanoparticles with 20% Zn substituted. (b) XRD pattern of
CuInS 2 samples with 20% Zn substitution, with pure CuInS 2 inset for comparison. (c) AXRD spectra of the (101) and (100) reflections in samples with 5%, 10%, and 20% Zn substitution.
Substitution of Zn is challenging to confirm with XRD alone; the pattern of even the highest substitution amounts (Figure 4.3(b)) show little variation from the original wurtzite CuInS 2 NPs’ pattern. ZnS has a lattice matched (2% difference in a, 3% difference in c) wurtzite phase (jcpds#010-0434), and can thus easily form solid solutions with wurtzite CuInS 2 with insignificant changes in lattice constants. To further confirm the alloying of CuInS 2 and ZnS, Anomalous XRD measurements were performed on a representative sample. In the AXRD scan, the (101) or (100) peak’s intensity was scanned, and Zn substitution in either Cu or In sites of the corresponding plane is expected to produce the corresponding anomalous scattering factor’s lineshape at the absorbtion edge. Figure 4.3(c) shows the peak intensity as a function of incident X-ray energy for the three substitution amounts. The characteristic dip at
50 9.658keV is present in all samples, with increasing intensity with higher concentrations of Zn. This finding indicates that Zn is a feasible alloying material to controllably increase the band gap of CuInS 2.
4.5.3 Substitution of Ga
Ga inclusion in CuInSe 2 solar cells has been instrumental in reaching high efficiencies. The most important effect of Ga substitution is an increase in the band gap, which allows a higher voltage in the solar cell.
Interestingly, the substitution of Ga in the CuInS 2 nanoparticles resulted in no changes in the optical band gap, XRD pattern, or morphology (figure 4.4(a,b)).
Contrary to what is observed in all thin film solar cells of Cu(In,Ga)Se 2, this implies that Ga did not go into the crystal structure. While there were no measurable changes in the position of the X-ray reflections which initially resulted from the wurtzite
CuInS 2, Ga was still detected by EDS inside the TEM. The contradictory finding that Ga does not seem to change the optical band gap yet is still present is likely related to the wurtzite phase of the CuInS 2 NPs and was explored in greater depth with AXRD.
51
Figure 4.4: CuInS 2 NPs with Ga substitution
(a) TEM image of CuInS 2 nanoparticles with 20% Ga substituted. (b) XRD pattern of
CuInS 2 samples with 20% Ga substitution, with pure CuInS 2 inset for comparison. (c) AXRD spectra of the (101) and (100) reflections in samples with 5%, 10%, and 20% Ga substitution.
Anomalous X-ray diffraction was performed to identify the bonding environment of Ga in the CuInS 2 lattice. Similar to the Zn substitution case, AXRD of a peak containing an In site should show anomalous scattering at the Ga band edge. As can be seen in figure 4.4(c), no dip is observed when scanning the X-ray energy through the Ga edge. This indicates that the Ga is not incorporated in the wurtzite crystal lattice, however it does not exclude inclusion on the surface or in the body of the NPs. HRTEM in figure 4.4(a) shows no visible amorphous shell, which would need to be >3nm to contain the observed Ga content. To check for Ga presence in the nanoparticles, fluorescence spectroscopy was done on Ga-containing samples. Figure 4.5 shows an inflection point near the Ga-edge, approximately proportionate to substituted Ga amount. This shows that Ga is present in the matrix of the nanoparticles, yet is not in the wurtzite crystal phase.
52
Figure 4.5: Fluorescence spectra of CuInS2 samples with 0%, 5%, 10%, and 20% Ga substitution
4.6 Summary of Alloying Results
In conclusion, substitution of Fe, Ga, and Zn in CuInS 2 have been analyzed by AXRD, and the nature of their incorporations were determined. Band gap could be effectively tuned over a wide range through the use of Zn and Fe substitution. It is believed that phase and morphology are determined largely by the relative stability of the chalcopyrite and wurtzite phases of the alloyed materials. It was determined that the activity of the Ga precursor, in this reaction, were insufficient to form a stable substituted wurtzite phase. Also, the substitution of Fe into CuInS 2 results in formation of a secondary chalcopyrite CuInS 2 phase. While the substitution of Ga contradicted expected band gap changes, it was only possible to observe the unusual distribution of Ga in the structure with AXRD due to its dual sensing of chemical and structural properties. Further studies on Ga in these structures could be performed with other X- Ray techniques, such as X-Ray Absorption Near Edge Structure, to further probe its bonding environment. Understanding and controlling the elemental distribution in complicated solar cell structures is a necessary step in optimizing performance.
53 5 Hybrid Nanoparticle Ink
The fabrication of solar cells from CuInS 2 nanoparticles in chapter III raised the issues of maximum attainable grain sizes and layer densities in sintered nanoparticle films. Therefore, we decided to simultaneously pursue an alternate route that would compare the grain growth obtained by sintering NPs with that possible from reactively annealing a non-sulfide precursor film. From the options discussed in chapter I, oxides were chosen to be an appropriate precursor material due to their massive recrystallization during reactive annealing, which should lead to large grain structures if controlled carefully. A hybrid solution/particulate formulation was developed, which allowed easy deposition of the oxide film through an organometallic/chalcogenide intermediate ink. The following chapters will discuss our efforts to produce a contaminant-free oxide precursor film and understand and control its transformation into CuInS 2.
5.1 CuIn(S,Se) 2 Films from Oxide Precursors
Over the last two decades, the transformation of Cu and In oxides into sulfides and selenides has been explored as an alternative pathway towards inexpensive solar cell absorber layers. Previous works have employed diverse techniques, such as spray pyrolysis 93,49 , printing 34 , spin-coating 94 , pulsed laser deposition 95 , and sputtering 96 to deposit oxide precursors. After precursor deposition, the chalcogenide transformation is typically 49,34,97 98 performed at ambient pressure, with either S/Se vapor or H 2S/H 2Se gas in Ar. The resulting films from this class of reactions vary, with some reporting incomplete transformation 50 , and a wide variety of microstructures have been seen. Solar cells of
CuInSe 2 have reached efficiencies of 14% using a two step oxide reduction and selenization process, and have been commercially produced 54 . Sulfurization temperatures between 500ºC and 550ºC have been reported, though large grain sizes
54 97,99 have been more challenging to produce in CuInS 2 than in CuInSe 2 . While previous reports demonstrated a 7.5% efficient CuInS 2 solar cell from reactively annealed oxides 96 , understanding of the sulfurization process is lacking.
5.2 Solution Processing with Oxide Precursors
The key advantage of fabricating sulfide and selenide solar absorbers from oxides is the ease of solution processing the oxide precursors. While it is possible to solution process both chalcogenides and metals, oxides are a uniquely advantageous material for solution processing. Oxides are air-stable, and can be milled or synthesized to have nanoscale dimensions. In addition, surfactants and other additives, which enable solution processing, can be burned out of oxides safely. While trace quantities of oxygen can be expected to remain in a solar cell fabricated in this manner, previous studies have shown that oxygen incorporation is tunable and can even be beneficial 100 .
5.3 Air-stable Ink Deposition
In this section, we present a methodology and process to create an easily decomposable vulcanized ink from commercially available precursors. This ink will then be utilized to make precursor films of Cu-In-O which can be transformed into
CuInS 2 solar cells with an absorber layer that is flat, contaminant-free, and large grained. As mentioned in Section 2.6, the synthesis of the ink involves the reaction between In-acetylacetonate (acac) and S in the solvent pyridine. We speculate that this reaction produces an easily decomposable polymer, through acetylacetonate vulcanization 101 , which allows the ink to wet the substrate well, removing the necessity to add other additives. The vulcanized ink created in this process allows deposition of a flat, contaminant-free precursor film. We chose pyridine as a solvent,
55 because it dissolves elemental sulfur and metal acetylacetonates (acac). We noticed, that upon mixing of In(acac) 3 and sulfur, the ink quickly undergoes a color transition from clear to a deep yellow-orange. Furthermore, while many metal salts are commercially available for a CuInS 2 ink, most have counter-ions, such as nitrates or chlorides, that require high temperatures to remove93,102 . By contrast, acetylacetonates are better metal precursors due to their commercial availability, high solubility in pyridine (~0.5M), moderate decomposition temperatures (~220 oC), and gaseous decomposition products 103 . Finally, in case we desire to alter the chemical composition, the commercial availability of many metal acacs allows for the addition of nearly any metal to our ink. In summary, we designed an easily decomposable vulcanized ink using Cu and In acetylacetonate salts, elemental sulfur, and pyridine to avoid contamination and obtain dense, uniform precursor layers of Cu-In-O.
5.4 Cu-In-O Precursor Film
Efficient CuInS 2 devices must be dense films and have minimal contamination, thus the precursor Cu-In-O film must be both densely packed and free of foreign elements. The SEM and depth profile of our air annealed precursor film using Auger Electron Spectroscopy in figure 5.1 and figure 5.2(a) reveals that the ink decomposes into an oxide bi-layer.
56
Figure 5.1: SEM cross-section of 3 layers of Cu-In-O film after air-baking
The oxide is important because previous attempts to achieve grain growth at high temperature in sulfur atmospheres when starting from nanoparticleline CuInS 2 have generally not demonstrated sufficiently large-sized grains 47 . Furthermore, an ideal absorber layer would have a columnar grain structure to aid in carrier collection.
Therefore, it is important that our process succeeded in forming large-grained CuInS 2 films after sulfurization of the oxide (figure 5.3(a) ). An AES depth profile of our film in figure 5.2(e) indicates the conversion of the oxide film to uniform CuInS 2.
57
Figure 5.2: Fabrication process and characeterization of CuInS 2 thin film
(a-c) Schematic of the fabrication of CuInS 2 films from the vulcanized ink, with processing temperatures and atmospheres. (d) AES depth profile the Cu-In-O precursor film and (e) the final CuInS 2 film.
5.5 Electrical Measurements of CuInS 2 Solar Cells
The current-voltage measurement for a CuInS 2 solar cell is shown in figure 2 5.3(b). The short-circuit current, J SC , is 18.5 mA/cm , which is comparable to 104105106 commercial CuInS 2 solar cells . However, the fill factor (FF=0.37) and open- circuit voltage (V OC = 320 mV) result in a photovoltaic conversion efficiency, η, of 2.2%. The shunt resistance, estimated at 100ohm*cm2 from the I-V, is likely the cause of the reduced FF and V OC . We believe that the low shunt resistance is the result of stresses created during sulfurization. Chapter VI details our investigation into the mechanical properties of our film before and after sulfurization in order to increase the
58 shunt resistance, which should significantly increase V OC , FF, and the photovoltaic conversion efficiency.
Figure 5.3: SEM and electrical performance of CuInS2 thin film
(a) SEM x-section and (inset) plan view of a CuInS 2 film sulfurized at 550°C. The buffer layers, CdS and ZnO, are shown as well. (b) I-V curve with inset device parameters of a CuInS 2/CdS heterojunction solar cell.
In summary, we have demonstrated an air-stable, easily decomposable, vulcanized ink deposition process for producing flat, contaminant-free, and large grain
CuInS 2 absorber layers from simple, commercially available precursors. As compared to the sintered nanoparticle films, grain sizes are larger, ranging from ~200nm to 1µm, sometimes spanning the entire thickness of the film. While these increased grain sizes
59 dramatically reduced the series resistance in devices, leading to almost double the 2 current, with a JSC of 18.5 mA/cm , morphological defects are believed to be limiting device performance.
60 6 Mechanism of Sulfurization and Selenization of Cu-In-O Films
The previous chapter described the development of a simple solution-phase process involving oxides, in what was labeled Air-stable Ink Deposition 107 . In our experiments, large grained films of CuInS 2 were achievable by the AID process, but efficiencies of solar cells only reached up to ~2%, which is significantly lower than that of commercialized CuInS 2 devices, despite large grain sizes of ~700nm. The shunt resistance was low, and it was observed that film defects such as holes were causing shorts through the absorber. Since we observed that holes did not form during selenization of Cu-In-O films, we analyzed the mechanism of selenization for comparison. The current chapter focuses on the mechanism of sulfurization and selenization of oxide films with specific analysis on controlling the morphology of grain growth to avoid hole formation.
6.1 Growth Mechanism of CuIn(S,Se) 2
The growth mechanisms of various CuIn(S,Se) 2 film synthesis methods have been analyzed in depth, with the goal of optimizing device fabrication and improving performance. The techniques used generally fall into two categories: in-situ and ex- situ methods. In-situ X-Ray Diffraction (XRD) measurements are commonly used as a means of gathering kinetic information during film growth, so as to understand phenomena such as secondary phase precipitation. Stacked metal films are commonly analyzed as they are sulfurized 108,109 or selenized 110,111 , as this is the easiest setup to transpose into an XRD chamber. Although in-situ methods are valuable for studying the kinetics of phase transformations, ex-situ methods can provide more spacial analysis, with techniques such as Transmission Electron Microscopy 112 (TEM) or Auger Electron Spectroscopy 113 (AES).
61 6.2 Sulfurization of Solution Deposited Cu-In-O
Similar to metal sulfurization processes, the sulfurization of our Cu-In-O films occurs via intermediate Cu-In-S phases. However, several steps of sulfuization of metals are slowed or nonexistant with oxides due to their lower reactivity. Experimental methods are explained in section 2.3; the main variable we have tested is the temperature at which Cu-In-O films are reactively annealed in a S atmosphere.
6.2.1 Sulfurization below 350°C: Partial Sulfurization
At the lowest temperatures of 285-335ºC, CuInS 2 does not form, but the Cu and In in the Cu-In-O film begin to separate due to Cu diffusion towards the surface to form CuS. SEM of these samples shows a layer of material under hexagonal ~1 µm plates (figure 6.1(a)). XRD of these low temperature samples shows only peaks corresponding to the (101), (102), (103), and (006) reflections of hexagonal CuS (figure 6.1(b)), with a (006) texture. It is likely that this phase is a room temperature polymorph of the high temperature Cu xS hexagonal phases which are seen in in-situ studies 108 . The bottom material is an oxide primarily composed of In as determined by AES depth profiling of the film.
62
Figure 6.1: SEM and XRD of film sulfurized at 335ºC (a) SEM image of Cu-In-O sample sulfurized at 335ºC with inset above to show protruding orientation of CuS. (b) XRD pattern of same, with CuS phase indexed.
6.2.2 Sulfurization between 350°C-450°C: Intermediate Phases
108 109 Common growth temperatures for CuInS 2 formation are 250 -320ºC with metals, and 400 114 -450ºC 97 from oxides. Accordingly, dominant peaks corresponding to phases of Cu-In-S appear at 385ºC, coincident with the peaks from CuS. Hexagonal CuS plates are observed, as in the lower temperature samples, but a finer grained layer is seen under the plates (figure 6.2(a)). To further analyze this sample, the CuS was removed by KCN solution etching, which exposed the underlayer for analysis. An AES map of the exposed surface shows that the CuS plates were resting on a film of Cu-In-S (figure 6.2(c)) with a small amount of residual oxygen still present as Cu-O. AES depth profiling of the KCN-cleaned film confirms that while the front of the sample contains primarily Cu-In-S, oxide remains closer to the back interface with the Mo-glass substrate (fig 2(b)). Surface Cu-O indicates that Cu diffuses to the surface of the film, then reacts with S, though some Cu-O remains unreacted. This diffusion process was also observed while heating of the film in air at temperatures between 300-400ºC 107 . XRD of the original film in figure 6.3 indicates that the sample contains CuS and a Cu-In-S phase, with no crystalline oxides. (figure 6.3)
63
Figure 6.2: SEM and AES of film sulfurized at 385ºC (a) SEM image of sample sulfurized at 385ºC. (b) AES depth profile of back of film after KCN etching. (c) AES map of film surface after KCN etching.
64
Figure 6.3: XRD patterns of samples sulfurized at 335ºC, 385ºC, 410ºC, and 585ºC
Phase-pure CuInS 2 and CuS are indexed in the 585ºC and 335ºC patterns, respectively.
In XRD patterns taken of samples from 335-585ºC, the phase transition from
CuS/In-O to CuInS 2 can be tracked. Figure 6.4 shows that, at 385ºC, CuS is only coincident with a Cu-poor phase of CuInS 2 best matched as CuIn 5S8. At temperatures even marginally higher, such as 410ºC, only CuInS 2 is present. Rodriguez-Alvarez et 115 al. also have reported a formation pathway involving CuIn 5S8 and Cu xS during the
65 sulfurization process with stacked metal films; however, this reaction was only dominant at high S pressures.
6.2.3 Sulfurization above 450°C: Formation of Pure CuInS 2
The sphalerite (cubic) and chalcopyrite (tetragonal) phases of CuInS 2 can both be formed, but they are distinguishable by splitting of the (004)/(020) and (204)/(220) peaks in XRD patterns, which results from the tetragonal distortion in the S sublattice.
The reflections of CuInS 2 sharpen between 410ºC and 585ºC, indicating grain growth, and the splitting of the (004)/(020) peaks is first observed at temperatures over 500ºC, corresponding to the chalcopyrite phase.
Figure 6.4: SEM plan view and cross-sectional images of sulfurized films (a,d) 385ºC (b,e) 435ºC (c,f) 535ºC
6.2.4 Development of Holes in CuInS 2
The evolution of large holes in the CuInS 2 films can be seen by SEM as the sulfurization temperature was raised. Figure 6.4(a-c) shows wide plan views of the
66 films sulfurized at 385ºC, 435ºC, and 535ºC. CuS plates are not observed on the top surface at temperatures higher than 385ºC, and the films produced at higher temperatures primarily consist of homogeneous and smooth CuInS 2. However, at 435ºC, small fissures appear scattered randomly throughout the film, and these fissures become 10-20 µm holes by 535ºC. However, cross-sectional SEM correlated with XRD (figure 6.4(d-f)) to show that the grain size is increasing from a few nm at 435ºC to ~500nm at 535ºC. The sulfurization of Cu-In-O films creates voids that appear in the film concurrently with grain growth. As we can infer from the above observation, nucleation occurs only at 385ºC, after which the entire volume quickly and almost homogeneously undergoes the oxide to sulfide transformation. This indicates that the transformation is likely limited by nucleation of Cu-In-S in the oxide films. We propose the following mechanism of hole formation: nucleation occurs at all points in the film where Cu, In, and S are present, and after Cu xS formation, there is still residual Cu in the oxide layer. The multitude of jutting CuS plates provides many nucleation points, which allows nucleation of CuIn 5S8 at numerous sites on the surface. Nucleation is followed by rapid diffusion of Cu from both the Cu xS plates and residual Cu-O. The diffusion of Cu towards the CuIn5S8 and simultaneous growth of many grains generates voids, which coalesce as the grains of CuInS 2 grow. This leads to the cracks and holes which are observed in fully transformed films.
6.3 Selenization of Solution Deposited Cu-In-O Films
We selenized Cu-In-O films, under similar conditions as sulfurization, to compare the growth mechanisms and understand why sulfurization tends to produce smaller grained films than selenization. The different reactivities between oxides and metals lead to a mechanism with a Cu xSe-InO x reaction interface. This mechanism is closely related to the Cu 2Se-In 2Se 3 interfacial reaction pathway seen with stacked metal films.
67 Similar to sulfurization, CuSe plate structures were also formed at low temperatures, such as 400ºC, yet they lay flat on the substrate as seen in Figure 6.5(a), unlike CuS, which protrudes out of the film. XRD showed peaks at (132), (006), (008), and (00 10) indicating that the CuSe was in the hexagonal phase, with a strong (006) texture. The underlayer in the 400ºC sample was determined to be In-O from AES.
Figure 6.5: SEM and XRD of film selenized at 400ºC (a) SEM image of Cu-In-O sample selenized at 400ºC with inset above to show planar orientation of CuSe. (b) XRD pattern of same, with CuSe phase indexed.
68
Figure 6.6: XRD patterns of samples sulfurized at 400ºC, 425ºC, 450ºC, 475ºC, and 500ºC
Phase-pure CuInSe 2 and CuSe are indexed in the 525ºC and 400ºC patterns, respectively. Insets added for clarity in 400ºC and 425ºC patterns.
69
Figure 6.7: SEM and AES of a sample selenized at 450ºC (a) before KCN etching and (b) after KCN ectching. (c) AES map of the KCN-etched sample surface.
The growth mechanism of selenization involves simpler intermediate states than sulfurization of oxide films, and unlike sulfurization, presents a discernible growth front at several growth temperatures. At temperature ranging from 400-525ºC (figure 6.6), XRD show a gradual decline in the CuSe (006) peak, along with a similar rise in the CuInSe 2 (112) peak. The (112) peak is sharp in all patterns, even of the transition state at 450ºC, indicating that CuInSe 2 forms large grains at all points in which it is observable.
SEM of a sample selenized at 450ºC shows large grains of CuInSe 2 clustered around the CuSe plates (figure 6.7(a)). After KCN etching of the same sample, to remove CuSe, SEM shows only the remaining CuInSe 2 crystals, clustered around the depressions where the CuSe plates were etched away (figure 6.7(b)). AES mapping of
70 the same etched sample shows that the shallow depressions left by the CuSe plates are primarily In-O, and confirm that the large-grain material is CuInSe 2 (figure 6.7(c)). The remaining surface is composed of In-O-Se. At temperatures above 450ºC, CuSe disappeared concurrently with the growth of the CuInSe 2, according to XRD (figure
6.6). From these observations, it can be assumed that the CuInSe 2 is growing at the interface of Cu xSe, In-O, and Se vapor.
Figure 6.8: SEM plan view and cross-sectional images of selenized samples (a,d) 450ºC (b,e) 475ºC (c,f) and 500ºC.
Another major difference from the sulfurization case was the total absence of film cracking and hole formation in all CuInSe 2 samples. Figure 6.8 shows plan view
SEM of CuInSe 2 films, along with corresponding cross-sectional SEM to show the evolution of grain sizes as growth temperature is increased. We observed that grain sizes at 500C were as large as 1 µm, which is significantly larger than the grains of comparably sulfurized films.
71 6.4 Comparison of Sulfurization and Selenization Mechanisms
Figure 6.9 shows the proposed steps of the nucleation-limited versus diffusion- limited mechanisms for CuInS 2 and CuInSe 2, respectively. CuInSe 2 growth occurs exclusively at the top interface of the In-O film with Cu xSe, over a wide temperature range. The planar growth of Cu xSe on the surface, along with the relatively slow diffusion of Cu in Cu xSe, limits nucleation to the edges of the Cu xSe plates which are exposed to both Se vapor and In-O. Thus, nuclei form in limited numbers at these interfaces and slowly grow as the In-O and CuSe are consumed, pressing all void space unidirectionally down and out of the film. This process is similar to that found in selenization of stacked metal films, where the formation of CuInSe 2 occurs at the 111 interface of Cu xSe and InSe . In both cases, large grained films are commonly produced.
Figure 6.9: Growth mechanism models of sulfurization and selenization of oxide precursor films
The sulfurization pathway’s critical difference from selenization is the high number of nuclei due to the jutting Cu xS plates, which provides ample nucleation points on the surface. The ready availability of residual Cu in the oxide, the angular
72 116 orientation of the Cu xS plates, and the high diffusivity of Cu in Cu xS , allows simultaneous nucleation of CuIn 5S8 at many points on the film. These nuclei rapidly grow into the film, trapping void space and causing the cracks and holes that lead to poor solar cell performance, such as the low shunt resistances seen in experiments in chapter V. This model points out the critical distinction between the sulfurization and selenization of oxides. Further comparing these models with the sulfurization of metal films provides some insight into film growth design. With metal films, an alloy forms 108 at low temperatures, and then the alloy undergoes a direct transformation to CuInS 2 . However, the sulfurization of oxide films undergoes a segregation of the Cu in the form of Cu xS before the formation of CuInS 2 occurs. Finally, metal film selenization usually produces the binary selenides CuSe and In 2Se 3, which then react to form
CuInSe 2. By controlling the segregation of Cu into CuS, the sulfurization of oxides could produce films with morphologies similar to CuInSe 2 films in a way that metal film sulfurization cannot. One way this could be achieved would be to deposit In 2O3 and CuO sequentially, which would form their binaries separately upon sulfurization and sequentially react them together. An alternative method, which could be performed in the framework of our deposition technique, is to segregate the Cu and In prior to sulfurization at high temperature. It is believed that this greater separation would lead to large grained CuInS 2 through a CuS templating type reaction, as has 117 been seen from annealing stacked CuS/Cu xInS 2 with x<1 .
6.5 Compensatory Growth Conditions for CuInS 2
We have concluded that the sulfurization process could be made to follow a pathway more like that of selenization. Therefore, we have developed a simple method to compensate for cracking and hole formation during sulfurization of oxide films. Specifically, the heating profile was designed with a hold at a lower temperature of 335ºC for 30 minutes before heating to the final sulfurization
73 temperature for 30 minutes. This allows the complete diffusion of the Cu into Cu xS crystals on the surface of the film, inhibiting nucleation of CuIn 5S8 in most of the film at intermediate temperatures. The Cu would then diffuse back into the film and react with the In-O at a more controlled interface, in a manner resembling the selenization case. The results in figure 6.10 show that a hole-free film with ~1 µm grains could be produced in this manner.
Figure 6.10: CuInS 2 films from two stage heating (a) plan view and (b) cross-sectional SEM images of a sample sulfurized in two stages, at 335ºC and 585ºC.
6.6 Summary of Results
Understanding the mechanism of CuIn(S,Se) 2 formation is an important step in improving production and device efficiencies. While most films of CuIn(S,Se) 2 are currently made with vacuum techniques, solution deposition holds promise for future cost reductions. We have described a route to large grained films of CuInS 2 and
CuInSe 2 through oxide precursors, along with an analysis of void formation during sulfurization. We hope that these findings will enable future works in solution processing of oxides for solar cell fabrication.
74
List of References (1) Lewis, N. S.; Nocera, D. G. Powering the planet: Chemical challenges in solar energy utilization. Proceedings of the National Academy of Sciences 2006 , 103 , 15729-15735. (2) Morton, O. Solar energy: A new day dawning?: Silicon Valley sunrise. Nature 2006 , 443, 19-22. (3) Becquerel, E. Action de la Radiation sur les Lames Metalliques. Compt. Rend. 1839 , 9, 561. (4) Day, R. E..; Adams, W. G. The Action of Light on Selenium. Proceedings of the Royal Society, London 1877 , A25 , 113. (5) Fritts, C. E. On a New Form of Selenium Photocell. American J. of Science 1883 , 26 , 465. (6) Shockley, W.; Quassar, H. Detailed Balance Limit of Efficiency of p-n Junction Solar Cells. Journal of Applied Physics 1961 , 32 , 510-519. (7) Chapin, D. M. Fuller, C. S.; Pearson, G. L. A New Silicon p-n Junction Photocell for Converting Solar Radiation into Electrical Power. Journal of Applied Physics 1954 , 25 , 676-677. (8) Jenny, D. A. Loferski, J. J.; Rappaport, P. Photovoltaic Effect in GaAs p-n Junctions and Solar Energy Conversion. Physical Review 1956 , 101 , 1208. (9) Reynolds, D. C. Leies, G. Antes, L. L.; Marburger, R. E. Photovoltaic Effect in Cadmium Sulfide. Physical Review 1954 , 96 , 533. (10) Cusano, D. A. CdTe solar cells and photovoltaic heterojunctions in II-VI compounds. Solid-State Electronics 6, 217-218. (11) Goldstein, B. Properties of Photovoltaic Films of CdTe. Physical Review 1958 , 109 , 601. (12) Wagner, S. Shay, J. L. Migliora.P; Kasper, H. M. Culnse2-Cds Heterojunction Photovoltaic Detectors. Applied Physics Letters 1974 , 25 , 434-435. (13) Kazmerski, L. L.; Sanborn, G. A. CuInS2 thin-film homojunction solar cells. J. Appl. Phys. 1977 , 48 , 3178.
75 (14) Repins, I. Contreras, M. A. Egaas, B. DeHart, C. Scharf, J. Perkins, C. L. To, B.; Noufi, R. 19·9%-efficient ZnO/CdS/CuInGaSe 2 solar cell with 81·2% fill factor. Prog. Photovolt: Res. Appl. 2008 , 16 , 235-239. (15) Wu, X. High-efficiency polycrystalline CdTe thin-film solar cells. Solar Energy 2004 , 77 , 803-814. (16) Lewerenz, H. J. Development of copperindiumdlsulfide into a solar material. Solar Energy Materials and Solar Cells 2004 , 83 , 395-407. (17) Yoshino, K. Ikari, T. Shirakata, S. Miyake, H.; Hiramatsu, K. Sharp band edge photoluminescence of high-purity CuInS2 single crystals. Applied Physics Letters 2001 , 78 , 742-744. (18) Meese, J. M. Manthuruthil, J. C.; Locker, D. R. Cuins2 Diodes for Solar- Energy Conversion. Bulletin of the American Physical Society 1975 , 20 , 696- 697. (19) Braunger, D. Hariskos, D. Walter, T.; Schock, H. W. An 11.4% efficient polycrystalline thin film solar cell based on CuInS2 with a Cd-free buffer layer. Solar Energy Materials and Solar Cells 1996 , 40 , 97-102. (20) Scheer, R. Klenk, R. Klaer, J.; Luck, I. CuInS2 based thin film photovoltaics. Solar Energy 2004 , 77 , 777-784. (21) Nakabayashi, T. Miyazawa, T. Hashimoto, Y.; Ito, K. Over 10% efficient CuInS2 solar cell by sulfurization. Solar Energy Materials and Solar Cells 1997 , 49 , 375-381. (22) Klenk, R. Blieske, U. Dieterle, V. Ellmer, K. Fiechter, S. Hengel, I. Jäger- Waldau, A. Kampschulte, T. Kaufmann, C. Klaer, J. Lux-Steiner, M. C. Braunger, D. Hariskos, D. Ruckh, M.; Schock, H. W. Properties of CuInS2 thin films grown by a two-step process without H2S. Solar Energy Materials and Solar Cells 1997 , 49 , 349-356. (23) Wilkommen, U.; Dimer, M. Thin film photovoltaic production technology. In; Ispra, Italy, 2007.
76 (24) Habas, S. E. Platt, H. A. S. van Hest, M. F. A. M.; Ginley, D. S. Low-Cost Inorganic Solar Cells: From Ink To Printed Device. Chem. Rev. 2010 , 110 , 6571-6594. (25) Hibberd, C. J. Chassaing, E. Liu, W. Mitzi, D. B. Lincot, D.; Tiwari, A. N. Non-vacuum methods for formation of Cu(In, Ga)(Se, S)2 thin film photovoltaic absorbers. Prog. Photovolt: Res. Appl. 2010 , 18 , 434-452. (26) Todorov, T.; Mitzi, D. B. Direct Liquid Coating of Chalcopyrite Light- Absorbing Layers for Photovoltaic Devices. Eur. J. Inorg. Chem. 2009 , NA- NA. (27) Mere, A. Electrical properties of sprayed CuInS2 films for solar cells. Journal of Physics and Chemistry of Solids 2003 , 64 , 2025-2029. (28) Eberspacher, C. Pauls, K.; Serra, J. Non-vacuum processing of CIGS solar cells. Conference Record of the Twenty-Ninth Ieee Photovoltaic Specialists Conference 2002 2002 , 684-687. (29) Bhattacharya, R. N. Solution Growth and Electrodeposited CuInSe[sub 2]Thin Films. J. Electrochem. Soc. 1983 , 130 , 2040. (30) Contreras, M. A. Romero, M. J. To, B. Hasoon, F. Noufi, R. Ward, S.; Ramanathan, K. Optimization of CBD CdS process in high-efficiency Cu(In,Ga)Se2-based solar cells. Thin Solid Films 2002 , 403-404 , 204-211. (31) Bhattacharya, R. N. Batchelor, W. Hiltner, J. F.; Sites, J. R. Thin-film CuIn[sub 1−x]Ga[sub x]Se[sub 2] photovoltaic cells from solution-based precursor layers. Appl. Phys. Lett. 1999 , 75 , 1431. (32) Tomar, M.; Garcia, F. A ZnO/p-CuInSe2 thin film solar cell prepared entirely by spray pyrolysis ☆. Thin Solid Films 1982 , 90 , 419-423. (33) Duchemin, S. Bougnot, J. Ghzizal, A. E.; Belghit, K. Studies on the improvement of sprayed CdS–SuInSe2 solar cells. .), Proceedings of the 9th EPVSEC, Freiburg, Germany 1989 , 476-479. (34) Kaelin, M. Rudmann, D. Kurdesau, F. Zogg, H. Meyer, T.; Tiwari, A. N. Low- cost CIGS solar cells by paste coating and selenization. Thin Solid Films 2005 , 480 , 486-490.
77 (35) Chung, C.-H. Li, S.-H. Lei, B. Yang, W. Hou, W. W. Bob, B.; Yang, Y. Identification of the Molecular Precursors for Hydrazine Solution Processed
CuIn(Se,S) 2 Films and Their Interactions. Chem. Mater. 2011 , 23 , 964-969. (36) Liu, W. Mitzi, D. B. Yuan, M. Kellock, A. J. Chey, S. J.; Gunawan, O. 12%
Efficiency CuIn(Se,S) 2 Photovoltaic Device Prepared Using a Hydrazine Solution Process †. Chem. Mater. 2010 , 22 , 1010-1014. (37) Kim, S. Fisher, B. Eisler, H. J.; Bawendi, M. Type-II Quantum Dots: CdTe/CdSe(Core/Shell) and CdSe/ZnTe(Core/Shell) Heterostructures. J. Am. Chem. Soc. 2003 , 125 , 11466-11467. (38) Wang, P. Abrusci, A. Wong, H. M. P. Svensson, M. Andersson, M. R.; Greenham, N. C. Photoinduced Charge Transfer and Efficient Solar Energy Conversion in a Blend of a Red Polyfluorene Copolymer with CdSe Nanoparticles. Nano Lett. 2006 , 6, 1789-1793. (39) Wu, Y. Wadia, C. Ma, W. Sadtler, B.; Alivisatos, A. P. Synthesis and Photovoltaic Application of Copper(I) Sulfide Nanocrystals. Nano Letters 2008 , 8, 2551-2555. (40) Robel, I. Subramanian, V. Kuno, M.; Kamat, P. V. Quantum dot solar cells. Harvesting light energy with CdSe nanocrystals molecularly linked to mesoscopic TiO2 films. J. Am. Chem. Soc. 2006 , 128 , 2385-2393. (41) Gur, I. Fromer, N. A. Geier, M. L.; Alivisatos, A. P. Air-Stable All-Inorganic Nanocrystal Solar Cells Processed from Solution. Science 2005 , 310 , 462-465. (42) Johnston, K. W. Pattantyus-Abraham, A. G. Clifford, J. P. Myrskog, S. H. MacNeil, D. D. Levina, L.; Sargent, E. H. Schottky-quantum dot photovoltaics for efficient infrared power conversion. Applied Physics Letters 2008 , 92 , 151115. (43) Huang, D. Liao, F. Molesa, S. Redinger, D.; Subramanian, V. Plastic- Compatible Low Resistance Printable Gold Nanoparticle Conductors for Flexible Electronics. Journal of The Electrochemical Society 2003 , 150 , G412- G417.
78 (44) Panthani, M. G. Akhavan, V. Goodfellow, B. Schmidtke, J. P. Dunn, L. Dodabalapur, A. Barbara, P. F.; Korgel, B. A. Synthesis of CuInS2, CuInSe2, and Cu(InxGa1-x)Se2 (CIGS) Nanocrystal “Inks⠀ for Printable Photovoltaics. Journal of the American Chemical Society 2008 , 130 , 16770- 16777. (45) Goldstein, A. N. Echer, C. M.; Alivisatos, A. P. Melting in Semiconductor Nanocrystals. Science 1992 , 256 , 1425-1427. (46) Li, L. Coates, N.; Moses, D. Solution-Processed Inorganic Solar Cell Based on in Situ Synthesis and Film Deposition of CuInS2 Nanocrystals. Journal of the American Chemical Society 2009 , 132 , 22-23. (47) Guo, Q. Ford, G. M. Hillhouse, H. W.; Agrawal, R. Sulfide Nanocrystal Inks for Dense Cu(In1−xGax)(S1−ySey)2 Absorber Films and Their Photovoltaic Performance. Nano Letters 2009 , 9, 3060-3065. (48) Guo, Q. Hillhouse, H. W.; Agrawal, R. Synthesis of Cu2ZnSnS4 Nanocrystal Ink and Its Use for Solar Cells. Journal of the American Chemical Society 2009 , 131 , 11672-11673. (49) Beck, M. E.; Cocivera, M. Thin-film copper indium diselenide prepared by selenization of copper indium oxide formed by spray pyrolysis. Thin Solid Films 1996 , 272 , 71-82. (50) Kaelin, M. Rudmann, D. Kurdesau, F. Meyer, T. Zogg, H.; Tiwari, A. N. CIS and CIGS layers from selenized nanoparticle precursors. Thin Solid Films 2003 , 431 , 58-62. (51) Eberspacher, C. Fredric, C. Pauls, K.; Serra, J. Thin-film CIS alloy PV materials fabricated using non-vacuum, particles-based techniques. Thin Solid Films 2001 , 387 , 18-22. (52) Kapur, V. K. Bansal, A. Le, P.; Asensio, O. I. Non-vacuum processing of CuIn1-xGaxSe2 solar cells on rigid and flexible substrates using nanoparticle precursor inks. Thin Solid Films 2003 , 431 , 53-57.
79 (53) Todorov, T. K. Reuter, K. B.; Mitzi, D. B. High-Efficiency Solar Cell with Earth-Abundant Liquid-Processed Absorber. Adv. Mater. 2010 , 22 , E156- E159. (54) Kapur, V. K. Fisher, M.; Roe, R. Fabrication of light weight flexible CIGS solar cells for space power applications. II-VI Compound Semiconductor Photovoltaic Materials. Symposium (Materials Research Society Symposium Proceedings Vol.668) 2001 , H3.5.1-6|xv+570. (55) Choi, S. H. Kim, E. G.; Hyeon, T. One-pot synthesis of copper-indium sulfide nanocrystal heterostructures with acorn, bottle, and larva shapes. Journal of the American Chemical Society 2006 , 128 , 2520-2521. (56) Denton, A.; Ashcroft, N. Vegard’s law. Phys. Rev. A 1991 , 43 , 3161-3164. (57) Redinger, D. Molesa, S. Shong, Y. Farschi, R.; Subramanian, V. An ink-jet- deposited passive component process for RFID. Electron Devices, IEEE Transactions on 2004 , 51 , 1978-1983. (58) Peng, H. L. Xie, C. Schoen, D. T. McIlwrath, K. Zhang, X. F.; Cui, Y. Ordered vacancy compounds and nanotube formation in CulnSe(2)-CdS core- shell nanowires. Nano Letters 2007 , 7, 3734-3738. (59) Peng, H. Schoen, D. T. Meister, S. Zhang, X. F.; Cui, Y. Synthesis and Phase Transformation of In2Se3 and CuInSe2 Nanowires. J. Am. Chem. Soc. 2007 , 129 , 34-35. (60) Pan, D. C. An, L. J. Sun, Z. M. Hou, W. Yang, Y. Yang, Z. Z.; Lu, Y. F. Synthesis of Cu-In-S ternary nanocrystals with tunable structure and composition. Journal of the American Chemical Society 2008 , 130 , 5620-+. (61) Xiao, J. P. Xie, Y. Xiong, Y. J. Tang, R.; Qian, Y. T. A mild solvothermal route to chalcopyrite quaternary semiconductor CuIn(SexS1-x)(2) nanocrystallites. Journal of Materials Chemistry 2001 , 11 , 1417-1420. (62) Xiao, J. P. Xie, Y. Tang, R.; Qian, Y. T. Synthesis and characterization of ternary CuInS2 nanorods via a hydrothermal route. Journal of Solid State Chemistry 2001 , 161 , 179-183.
80 (63) Castro, S. L. Bailey, S. G. Raffaelle, R. P. Banger, K. K.; Hepp, A. F. Synthesis and characterization of colloidal CuInS2 nanoparticles from a molecular single-source precursor. Journal of Physical Chemistry B 2004 , 108 , 12429-12435. (64) Nairn, J. J. Shapiro, P. J. Twamley, B. Pounds, T. vonWandruszka, R. Fletcher, T. R. Williams, M. Wang, C.; Norton, M. G. Preparation of Ultrafine Chalcopyrite Nanoparticles via the Photochemical Decomposition of Molecular Single-Source Precursors. Nano Lett. 2006 , 6, 1218-1223. (65) Han, W. Yi, L. Zhao, N. Tang, A. Gao, M.; Tang, Z. Synthesis and Shape- Tailoring of Copper Sulfide/Indium Sulfide-Based Nanocrystals. J. Am. Chem. Soc. 2008 . (66) Czekelius, C. Hilgendorff, M. Spanhel, L. Bedja, I. Lerch, M. Müller, G. Bloeck, U. Su, D.-S.; Giersig, M. A Simple Colloidal Route to Nanocrystalline ZnO/CuInS2 Bilayers. Advanced Materials 1999 , 11 , 643-646. (67) Kuzuya, T. Itoh, K. Ichidate, M. Wakamatsu, T. Fukunaka, Y.; Sumiyama, K. Facile synthesis of nearly monodispersed copper sulfide nanocrystals. Electrochimica Acta 2007 , 53 , 213-217. (68) Tham, A.-T. Su, D. S. Neumann, W. Schubert-Bischoff, P. Beilharz, C.; Benz, K. W. Transmission Electron Microscopical Studies of the Layered Structure of the Ternary Semiconductor CuIn5Se8. Crystal Research and Technology 2001 , 36 , 303-308. (69) Konstantin, G. Leonid, C. Vera, L. David, C. Vladimir, D. Vitaliy, K. Roald, M. Elena, S.; Valentina, U. Direct evidence for diffusion and electromigration of Cu in CuInSe2. Journal of Applied Physics 1997 , 82 , 4282-4285. (70) Manna, L. Scher, E. C.; Alivisatos, A. P. Synthesis of Soluble and Processable Rod-, Arrow-, Teardrop-, and Tetrapod-Shaped CdSe Nanocrystals. Journal of the American Chemical Society 2000 , 122 , 12700-12706. (71) Wakamura, K.; Tsubota, I. Small band gap and high ionic conduction in Cu2S. Solid State Ionics 2000 , 130 , 305-312.
81 (72) Buerger, M. J.; Wuensch, B. J. Distribution of Atoms in High Chalcocite, Cu2S. Science 1963 , 141 , 276-277. (73) Robb, D. T.; Privman, V. Model of Nanocrystal Formation in Solution by Burst Nucleation and Diffusional Growth. Langmuir 2008 , 24 , 26-35. (74) Francis, R. J. O’Brien, S. Fogg, A. M. Halasyamani, P. S. O’Hare, D. Loiseau, T.; Ferey, G. Time-Resolved In-Situ Energy and Angular Dispersive X-ray Diffraction Studies of the Formation of the Microporous Gallophosphate ULM-5 under Hydrothermal Conditions. Journal of the American Chemical Society 1999 , 121 , 1002-1015. (75) Avrami, M. Kinetics of Phase Change. I General Theory. The Journal of Chemical Physics 1939 , 7, 1103-1112. (76) Zhao, L. Lu, T. Yosef, M. Steinhart, M. Zacharias, M. Gosele, U.; Schlecht, S. Single-Crystalline CdSe Nanostructures: from Primary Grains to Oriented Nanowires. Chemistry of Materials 2006 , 18 , 6094-6096. (77) Manna, L. Wang, L. W. Cingolani, R.; Alivisatos, A. P. First-Principles Modeling of Unpassivated and Surfactant-Passivated Bulk Facets of Wurtzite CdSe: A Model System for Studying the Anisotropic Growth of CdSe Nanocrystals. J. Phys. Chem. B 2005 , 109 , 6183-6192. (78) Klenk, R. Klaer, J. Scheer, R. Luxsteiner, M. Luck, I. Meyer, N.; Ruhle, U. Solar cells based on CuInS?an overview. Thin Solid Films 2005 , 480-481 , 509-514. (79) Turcu, M.; Rau, U. Recombination Mechanisms in Cu(In,Ga)(Se,S)2 Solar Cells. In Wide-Gap Chalcopyrites ; Siebentritt, S.; Rau, U., Eds. Springer- Verlag: Berlin/Heidelberg, 2006; Vol. 86, pp. 91-111. (80) Wei, S.-H. Zhang, S. B.; Zunger, A. Effects of Ga addition to CuInSe[sub 2] on its electronic, structural, and defect properties. Appl. Phys. Lett. 1998 , 72 , 3199. (81) Neisser, A. Hengel, I. Klenk, R. Matthes, T. W. Álvarez-García, J. Pérez- Rodríguez, A. Romano-Rodríguez, A.; Lux-Steiner, M. C. Effect of Ga
82 incorporation in sequentially prepared CuInS2 thin film absorbers. Solar Energy Materials and Solar Cells 2001 , 67 , 97-104. (82) Watanabe, T.; Matsui, M. Improved Efficiency of CuInS2-Based Solar Cells without Potassium Cyanide Process. Jpn. J. Appl. Phys. 1999 , 38 , L1379- L1381. (83) Hengel, I. Current transport in CuInS2:Ga/Cds/Zno – solar cells. Thin Solid Films 2000 , 361-362 , 458-462. (84) Persson, C.; Zunger, A. Anomalous grain boundary physics in polycrystalline CuInSe2: The existence of a hole barrier. Physical Review Letters 2003 , 91 . (85) Siebentritt, S. Sadewasser, S. Wimmer, M. Leendertz, C. Eisenbarth, T.; Lux- Steiner, M. C. Evidence for a neutral grain-boundary barrier in chalcopyrites. Phys Rev Lett 2006 , 97 , 146601. (86) Brown, G. Faifer, V. Pudov, A. Anikeev, S. Bykov, E. Contreras, M.; Wu, J. Determination of the minority carrier diffusion length in compositionally graded Cu(In,Ga)Se[sub 2] solar cells using electron beam induced current. Appl. Phys. Lett. 2010 , 96 , 022104. (87) Katagiri, H. Sasaguchi, N. Hando, S. Hoshino, S. Ohashi, J.; Yokota, T. Preparation and evaluation of Cu2ZnSnS4 thin films by sulfurization of E B evaporated precursors. Solar Energy Materials and Solar Cells 1997 , 49 , 407- 414.
(88) Zhang, W.; Zhong, X. Facile Synthesis of ZnS−CuInS 2 -Alloyed Nanocrystals for a Color-Tunable Fluorchrome and Photocatalyst. Inorg. Chem. 2011 , 50 , 4065-4072. (89) Tang, J. Hinds, S. Kelley, S. O.; Sargent, E. H. Synthesis of Colloidal CuGaSe
2 , CuInSe 2 , and Cu(InGa)Se 2 Nanoparticles. Chem. Mater. 2008 , 20 , 6906-6910. (90) Nakamura, H. Kato, W. Uehara, M. Nose, K. Omata, T. Otsuka-Yao-Matsuo, S. Miyazaki, M.; Maeda, H. Tunable Photoluminescence Wavelength of Chalcopyrite CuInS2-Based Semiconductor Nanocrystals Synthesized in a Colloidal System. Chemistry of Materials 2006 , 18 , 3330-3335.
83 (91) Wang, X. Pan, D. Weng, D. Low, C.-Y. Rice, L. Han, J.; Lu, Y. A General Synthesis of Cu−In−S Based Multicomponent Solid-Solution Nanocrystals with Tunable Band Gap, Size, and Structure. J. Phys. Chem. C 2010 , 114 , 17293-17297. (92) Sato, K. Isawa, M. Takahashi, N.; Tsunoda, H. Optical Absorption Spectra in CuInS2 Doped with Fe, Mn and Cr. Japanese Journal of Applied Physics 1988 , 27 , 1359-1360. (93) Weng, S.; Cocivera, M. Preparation of copper indium diselenide by selenization of copper indium oxide. Journal of Applied Physics 1993 , 74 , 2046-2052. (94) Luo, P. F. Zuo, R. Z.; Chen, L. T. The preparation of CuInSe2 films by combustion method and non-vacuum spin-coating process. Solar Energy Materials and Solar Cells 2010 , 94 , 1146-1151. (95) Wada, T. Negami, T.; Nishitani, M. Preparation of Chalcopyrite-Type Semiconductor-Films by Sulfuration of Oxide-Films. Japanese Journal of Applied Physics Part 1-Regular Papers Short Notes & Review Papers 1993 , 32 , 41-42. (96) Negami, T. Hashimoto, Y. Nishitani, M.; Wada, T. CuInS2 thin-films solar cells fabricated by sulfurization of oxide precursors. Solar Energy Materials and Solar Cells 1997 , 49 , 343-348. (97) Todorov, T. Cordoncillo, E. Sanchez-Royo, J. F. Carda, J.; Escribano, P. CuInS2 films for photovoltaic applications deposited by a low-cost method. Chemistry of Materials 2006 , 18 , 3145-3150. (98) Terauchi, M. Negami, T. Nishitani, M. Ikeda, M. Wada, H.; Wada, T. Preparation of Cuinse2 Thin-Films by Selenization of Cu-in-O Precursors. Solar Energy Materials and Solar Cells 1994 , 35 , 121-127. (99) Oliveira, L. Todorov, T. Chassaing, E. LinCot, D. Carda, J.; Escribano, P. CIGSS films prepared by sol-gel route. Thin Solid Films 2009 , 517 , 2272-2276.
84 (100) Noufi, R. Powell, R. C.; Matson, R. J. On the effect of stoichiometry and oxygen on the properties of CuInSe2 thin films and devices. Solar Cells 1986 , 21 , 55-63. (101) Vratislav, D. ccaron; ek; Antonín, K. Pavel, P. rbreve; ibyl Efficiency of metal activators of accelerated sulfur vulcanization. Journal of Applied Polymer Science 1993 , 47 , 743-746. (102) Zouaghi, M. C. Nasrallah, T. B. Marsillac, S. Bernède, J. C.; Belgacem, S. Physico-chemical characterization of spray-deposited CuInS2 thin films. Thin Solid Films 2001 , 382 , 39-46. (103) Hoene, J. V. Charles, R. G.; Hickam, W. M. Thermal Decomposition of Metal Acetylacetonates: Mass Spectrometer Studies. The Journal of Physical Chemistry 1958 , 62 , 1098-1101. (104) Siemer, K. Klaer, J. Luck, I. Bruns, J. Klenk, R.; Braunig, D. Efficient CuInS2 solar cells from a rapid thermal process (RTP). Solar Energy Materials and Solar Cells 2001 , 67 , 159-166. (105) Ellmer, K. Hinze, J.; Klaer, J. Copper indium disulfide solar cell absorbers prepared in a one-step process by reactive magnetron sputtering from copper and indium targets. Thin Solid Films 2002 , 413 , 92-97. (106) Scheer, R. Walter, T. Schock, H. W. Fearheiley, M. L.; Lewerenz, H. J. CuInS2 based thin film solar cell with 10.2% efficiency. Appl. Phys. Lett. 1993 , 63 , 3294. (107) Weil, B. D. Connor, S. T.; Cui, Y. CuInS2 Solar Cells by Air-Stable Ink Rolling. Journal of the American Chemical Society 2010 , 132 , 6642-6643. (108) Scheer, R. von Klopmann, C. Djordjevic, J.; Rudigier, E. Real-time studies of phase transformations in Cu-In-Se-S thin films 2. Sulfurization of Cu-In precursors. Journal of Crystal Growth 2006 , 289 , 121-33. (109) Rodriguez-Alvarez, H. Koetschau, I. M. Genzel, C.; Schock, H. W. Growth paths for the sulfurization of Cu-rich Cu/In thin films. Thin Solid Films 2009 , 517 , 2140-2144.
85 (110) Scheer, R. Djordjevic, J.; Rudigier, E. Real-time studies of phase transformations in Cu-In-Se-S thin films-3: Selenization of Cu-In precursors. Journal of Crystal Growth 2006 , 294 , 218-30. (111) Brummer, A. Honkimaki, V. Berwian, P. Probst, V. Palm, J.; Hock, R. Formation of CuInSe2 by the annealing of stacked elemental layers - analysis by in situ high-energy powder diffraction. Thin Solid Films 2003 , 437 , 297- 307. (112) Sandino, J. Romero, E. Oyola, J. S. Gordillo, G.; Lichte, H. Study of the Mo/CuInS2/ZnS system by TEM. Solar Energy Materials and Solar Cells 2009, In Press, Corrected Proof . (113) Calvo-Barrio, L. Perez-Rodriguez, A. Romano-Rodriguez, A. Barcones, B. Morante, J. R. Siemer, K. Luck, I.; Klenk, R. Combined in-depth scanning Auger microscopy and Raman scattering characterisation of CuInS2 polycrystalline films. Vacuum 2001 , 63 , 315-321. (114) Wada, T. Negami, T.; Nishitani, M. Preparation of Cuins2 Films by Sulfurization of Cu-in-O Films. Applied Physics Letters 1993 , 62 , 1943-1945. (115) Rodriguez-Alvarez, H. Kotschau, I. M.; Schock, H. W. Pressure-dependent real-time investigations on the rapid thermal sulfurization of Cu-In thin films. Journal of Crystal Growth 2008 , 310 , 3638-3644. (116) Bartkowicz, I.; Stokłosa, A. Kinetics of copper sulfidation at temperatures 570–1123 K. Oxidation of Metals 1986 , 25 , 305-318. (117) Rodriguez-Alvarez, H. Mainz, R. Marsen, B. Abou-Ras, D.; Schock, H. W. Recrystallization of Cu-In-S thin films studied in situ by energy-dispersive X- ray diffraction. Journal of Applied Crystallography 2010 , 43 , 1053-1061.
86