University of Southern California Dissertations and Theses

Growth, characterization of gallium arsenide based nanowires and application in photovoltaic cells

(USC Thesis Other)

Growth, characterization of gallium arsenide based nanowires and application in photovoltaic cells

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Growth, characterization of gallium arsenide based nanowires and application in photovoltaic cells By Maoqing Yao A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment for the Degree of DOCTOR OF PHILOSOPHY (ELECTRICAL ENGINEERING - ELECTROPHYSIS) December 2014 Copyright 2014 MaoqingYao Dedication Dedicated to my dear wife, Linger Hong, and to my parents, Guohua Yao and Fengfeng Yao. 11 Acknowledgements Life is a journey, and completing my Ph.D. study is the most exciting adventure I have ever taken so far. To be honest, since I first started my study at USC in 2009, at many moments I thought that I might not be able to make it. However, I feel so blessed to meet so many people in the past five years and it is their inspiration, support, help, sharing and encouragement that helped me through this adventure and made my life in LA such a wonderful experience and invaluable memory of lifetime. Without them and their kindness, this dissertation would not be possible. Firstly, I would like to deliver my deepest appreciation to my two research advisors, Professor Chongwu Zhou and Professor Dan Dapkus for their support and guidance throughout my Ph.D. study. Their knowledge, experience, intelligence and vision inspired my research and made my study a wonderful learning experience. Not many people have the chance to work with two outstanding scholars at the same time but I am very fortunate to have this opportunity brought by the Center for Energy Nanoscience (CEN) at USC, a Department of Energy Energy Frontier Research Center. The exposure to the colleagues with different skillsets, various equipments, and distinct thinking philosophies made my research experience so rich and provided alternative solutions to technical problems that would otherwise be impossible. Prof. Zhou's vision and energetic team propelled me to always keep track of what is going on in the field and Prof. Dapkus' rich experience and old stories from the golden era of semiconductor research always remind me to be solid in knowledge basis and learn from the rigorous scientific attitude of old school researchers. I 111 would also like to thank Prof. Aiichiro Nakano, Prof. Michelle Povinelli, and Prof. Wei Wu for serving on my dissertation defense or qualify exam committee and extensive collaboration in CEN which contributed greatly to my dissertation work. I would like to thank other CEN faculty members including Prof. Stephen Cronin Prof. Grace Lu and Prof. Mark Thompson for discussion during CEN meetings and being supportive in collaborative projects. Given the opportunity to work on projects under CEN, I also had the chance to interact actively with colleagues in a multi-discipline team from different research groups, with complementary expertise and diversified cultural background. I would like to thank my CSL colleagues from Prof. Dapkus' group. Chun-Yung Chi and I worked on the same nanostructured solar cell project and had been sharing the same MOCVD reactor. Chun-Yung is a smart guy and easy to work with. He trained me on the system and took the lead in the maintenance of the machine. He is also always ready to share various information with you so I enjoyed working with him. Yen-Ting Lin and Yoshitake Nakajima are the other two guys working in the same MOCVD growth room with us. We discussed and chat a lot and worked synergistically to keep the lab running well. I would like to thank former CSL member Ting-Wei Yeh for his valuable guidance and help in growth, maintenance, TEM and other technical issues. I would like to thank my Nanolab colleagues from Prof. Zhou's group. I would especially appreciate Sen Cong for his hard work during the past year. He joined my direction after last summer but he is so diligent and determined that he picked-up things quickly and offered countless help to my research. The tandem cell would not have been achieved without his great contribution !V through so many midnights and weekends. I would like to thank my senior Anuj Madaria and Akshay Kumar who trained me on various experiments during the early stage of the program. I would like to thank other former and current Nanolab members including Dr. Furniaki Ishikawa, Dr. Po-Chiang Chen, Dr. Lewis Gomez, Dr. Hsiao-Kang Chang, Dr. Chuan Wang, Dr. Alexander Badrnaev, Dr. Yi Zhang, Dr. Jialu Zhang, Dr. Bilu Liu, Dr. Gang Liu, Dr. Yuchi Che, Haitian Chen, Jia Liu, Xiaoli Wang, Xue Lin, Zhen Li, Jing Qiu, Jing Xu, Kuan-Teh Li, Luyao Zhang, Noppadol Aroonyadet, Jiepeng Rong, Mingyuan Ge, Hui Gui, Pyojae Kirn, Younghyun Na, Rebecca Lee, Ning Yang, Pattararnon Vuttipittayarnongkol, Liang Chen, Xin Fang, Ahmad Abbas, Sen Cong, Yu Cao, Anyi Zhang, Yuqiang Ma, Fanqi Wu, Xuan Cao, Chenfei Shen and Yihang Liu. You made my Ph. D. study colorful and I cherish all the moments I spent with you. I would like to thank other researchers I had collaborated with including Chenxi Lin, Ningfeng Huang and Duke Anderson from Prof. Povinelli's group; Yunchu Li, Lawrence Stewart and Ashkan Seyedi from Prof. Dapkus' group; Chia-Chi Chang, Sherrnin Arab, Chun-Chung Chen, Rohan Dhall and Mohammed Arner from Prof. Cronin's group; Zaoshi Yuan and Chunyang Shen from Prof. Nakano's group; Liubing Huang and Mengyao Zhang from Prof. Lu's group; He Liu and Yuhan Yao from Prof. Wu's group; Shu Hu from Prof. Nathen Lewis' group at Caltech and Uguen Choi from Prof. Coleman's group at UIUC. I would like to express my appreciation to staff members at USC and other institutions. Their support allows me to focus on my research. Our USC cleanroorn manager Donghai is just amazing and so skillful that he keeps such a large cleanroorn running and responds quickly to any technical issues. Besides USC cleanroorn I also used equipments from v other institutions including UCLA nanolab, UCLA CNSI and UCSB Nanotech. I want to thank all the technicians there for making my experiments smooth and successful. I would like also to thank administrative staff members of EE-EP and VSOE, Jenny Lin, Eliza Aceves, Kirn Reid, Diane Dernetras, Jaime Zelada, Tracy Charles and Jennifer Gerson. I want to acknowledge the generous financial support from USC Provost's Ph. D. Fellowship and Center for Energy Nanoscience. Last but not least I am so indebted to my dear wife, Linger Hong, for her unconditional love, consolation, encouragement, and delicious dinner and lunch boxes every single day! I also owe a lot to my parents, Guohua Yao and Fengfeng Yao. They are always ready to offer whatever they have to support me moving forward. I sincerely dedicate my dissertation to my family. Vl Table of contents Dedication ................................................................................................................................. ii Acknowledgements .................................................................................................................. iii List of figures ........................................................................................................................... ix Abstract ................................................................................................................................... xv Chapter I Introduction ............................................................................................................... I I. I Global energy challenge .................................................................................................. I I. I. I Global energy consumption ...................................................................................... I 1.1.2 Demand of solar energy ............................................................................................ 3 1.2 Photovoltaic solar energy conversion .............................................................................. 5 1.2.1 Physics of photovoltaic process ................................................................................ 5 1.2.2 Solar cell figures of merit ......................................................................................... 6 1.2.3 Shockley-Queisser limit and multijunction solar cells ............................................. 8 1.3 Semiconductor nanowires and its application in solar cells .......................................... 11 1.3.1 Economy of solar energy ........................................................................................ 11 1.3.2 Semiconductor nanowire solar cells ....................................................................... 12 1.4 Outline of the dissertation ............................................................................................. 14 References ........................................................................................................................... 17 Chapter 2 Selective area growth of nanowire arrays and scalable nanosphere lithography patterning ................................................................................................................................. 20 2.1 Introduction ................................................................................................................... 20 2.2 Nanowire growth techniques ......................................................................................... 23 2.2.1 Epitaxial growth using MOCVD ............................................................................ 23 2.2.2 Nanowire growth using VLS method ..................................................................... 25 2.3 Patterning techniques for selective area growth ............................................................ 34 2.3.1 Electron beam lithography ..................................................................................... 34 2.3.2 Nanoimprint lithography ........................................................................................ 36 2.3.3 Nanosphere lithography .......................................................................................... 38 2.4 Optical absorption study of nanowire array grown from pattern generated by nanosphere lithography ....................................................................................................... 44 Vll 2.5 Summary ....................................................................................................................... 51 References ........................................................................................................................... 53 Chapter 3 GaAs nanowire array solar cells with axial p-i-njunctions .................................... 58 3.1 Introduction ................................................................................................................... 58 3.2 Theoretical study of optical and electrical properties of GaAs nanowire solar cells .... 62 3.3 Fabrication and optimization ofGaAs nanowire array with axial p-i-njunction .......... 70 3.4 Summary ....................................................................................................................... 84 References ........................................................................................................................... 86 Chapter 4 Heteroepitaxial Growth of GaAs Nanowires on Si Substrates and Twin Formation Mechanism .............................................................................................................................. 89 4.1 Introduction ................................................................................................................... 89 4.2 GaAs nanowire grown on Si (111) substrate using SAG .............................................. 92 4.3 Twin formation mechanism in SAG and effect of growth condition .......................... 101 4.4 Thermodynamic model of twin probability ................................................................. 106 4.5 Summary ..................................................................................................................... 108 References ......................................................................................................................... 110 Chapter 5 tandem solar cells made by GaAs nanowires grown on Si ................................... 115 5.1Introduction ................................................................................................................. 115 5.2 Doping characteristics ofn-type GaAs nanowire doped by Si .................................... 118 5.3 GaAs nanowire on Si dual junction solar cell ............................................................. 132 5.4 Summary ..................................................................................................................... 137 References ......................................................................................................................... 139 Chapter 6 Conclusion and outlook ........................................................................................ 142 6.1 Conclusion ................................................................................................................... 142 6.2 Outlook ........................................................................................................................ 143 References ......................................................................................................................... 145 Vlll List of figures Figure 1.1 U.S. energy flow, 2013 (quadrillion Btu)1 ............................................................... 2 Figure 1.2 J-V characteristic ofa solar cell ............................................................................... 7 Figure 1.3 Schematic of a lattice matched triple junction solar cell with 41.6% efficiency made by Spectrolab 11 ........................................................................................................ 9 Figure 2. 1 Au-Si binary phase diagram .................................................................................. 26 Figure 2. 2 III-V nanowires grown from VLS method. GaAs nanowires grown on Si substrate using (a) :MBE 47 and (b) MOCVD 50 . Ga-self catalyzed nanowires grown on GaAs (11 l)B substrate using (c) :MBE 59 and (d) MOCVD by the author ................................. 28 Figure 2. 3 Schematic of SAG process. (a) Growth mask layer deposition, (b) opening generation by lithography, (c) nanowire SAG in MOCVD. ............................................ 30 Figure 2. 4 (a)-(c) 30° tilted SEM images of GaAs nanowires grown by SAG. Growth temperature is 730 ° C, pressure is 76 Torr, TMG partial pressure is 7.56x 10· 7 atm, AsH 3 partial pressure is (a) 2.14xl0" 5 atm, (b) 2.14xl0" 4 atm, and (c) l.43x10· 3 atm. (d) As trimer model of GaAs (11 l)B (2 x2) reconstruction: large open circles denote adsorbed As trimer atoms, small open circles denote first-layer As atoms, and small closed circles denote second-layer Ga atoms. (e) Detailed model of GaAs (lll)B ( J 19X J 19) reconstruction: large open circles denote top As atoms, medium closed circles denote second-layer Ga atoms, and small open circles denote third-layer threefold-coordinated As atoms .......................................................................................................................... 32 Figure 2. 5 (a)-(c) GaAs nanowires grown by SAG with different diameters and heights. (d) Average diameter and height of each sample with standard deviation. Inset is height vs l/Dimater 2 , which should show linear relationship if volume is constant. ..................... 34 Figure 2. 6 Schematic of selective area growth process .......................................................... 37 Figure 2. 7 (a) SEM image of a fabricated mold with a 10 nm diameter ................................ 38 Figure 2. 8 Schematic diagram of nanosphere lithography and nanowire growth. (a) Spin coating nanospheres on the substrate. (b) Oxygen plasma to reduce the size of the nanospheres, which will determine the diameter of the nanowires after growth. ( c) Metal (Al or Fe) deposition onto the nanospheres. (d) Dissolution of nanospheres in chloroform w ith the aid of sonication. ( e) Transfer of the pattern to the underlying substrate by dry etching and removal of the metal mask by HCl wet etching. (f) Nanowire growth by SA-MOCVD ................................................................................. 39 Figure 2. 9 Demonstration of scalable GaAs nanowire synthesis on a GaAs substrate using NSL technique. (a) Photograph of the spin-coated polystyrene nanospheres on a wafer scale. SEM images of nanosphere assembly at two random locations on the wafer, showing mono layer assembly of the nanospheres. (b) SEM image of nanospheres after size reduction by oxygen plasma. ( c) SEM image of the pattern after metal deposition, nanosphere lift-off, and dry etching, just before the nanowire growth. (d) Vertical GaAs nanowires grown on the NSL pattern showing a hexagonal cross section having six {-1 10} side fac ets ........................................................................................................... .41 l X Figure 2. 10 Structural quality comparison ofGaAs nanowires grown using (a) electron beam lithography and (b) nanosphere lithography, using HRTEM. Inset in each case shows the twin planes and diffraction pattern corresponding to Zinc-blende structure. Both images show similar distribution of twin planes and crystal structures. The arrow indicates the twin planes present in the nanowires ............................................................................... 43 Figure 2. 11 (a) Schematic of the GaAs nanowire array (b) Calculated weighted reflection loss as a function of nanowire diameter and lattice constant - diameter ratio for GaAs nanowires of 130 nm in height. A range of optimum configurations can be extracted from the graph and used to fabricate GaAs nano wires using NSL. ............................... .45 Figure 2. 12 Comparison between experimentally measured and simulated reflectance spectrum for the as-fabricated GaAs nanowire array sample (a~360nm, d=l95nm). (a) SEM image of the top view of the GaAs nanowires grown using NSL. Four regions extracted from the SEM image (1, 2, 3 and 4 shown in the inset) were selected to simulate the reflection profile across the visible spectrum. (b) Corresponding reflectance spectra using section 1, 2, 3 and 4. ( c) Plot of reflectance spectra for GaAs nanowires grown on a GaAs substrate using NSL (experiment), simulated periodic nano wire array with a~360 nm and d=l95 nm, and average of simulated nanowire arrays in (b). (d) Reflected power density under AMI. 5 spectrum showing experimental measurement of the GaAs nanowire array grown on NSL patterned substrate, simulation of the periodic nano wire array, and simulation using nano wire arrays with SEM images in (a). The total integrated reflected power density is also calculated ..................................................... .4 7 Figure 2. I3 Comparison between experimentally measured and simulated reflectance spectrum for a GaAs nanowire array sample (a ~ 890mn, d ~ 450nm) fabricated using NSL. (a) SEM image of the nanowires (top view). Three regions extracted from the SEM image (labelled 1, 2 and 3) are shown in the inset, each with dimensions of 5µm by 5 µm. (b) Corresponding reflectance spectra using section 1, 2 and 3. ( c) Plot of reflectance spectra for GaAs nanowires grown on a GaAs substrate using NSL (experiment), simulated periodic nanowire array with a~ 890 mn and d ~ 450 mn, and average of simulated nanowire arrays in Figure 6(b ). ( d) Reflected power density under AMI .5 spectrum showing experimental measurement of the GaAs nanowire array grown on NSL patterned substrate, simulation of the periodic nanowire array, and simulation using nanowire arrays with SEM images in Figure 6(a). The total integrated reflected power density is also calculated ...................................................................... .49 Figure 2. I4 Comparison of reflection spectra between simulation and experimental results of GaAs nanowires array grown on EEL patterned GaAs (111 )B substrate. (a) 360 nm pitch, grown for 20 min and 30 min. (b) 890 mn pitch, grown for 20 min and 30 min ............ 50 x Figure 3. I Multi-junction solar cells. (a) Multi-junction solar cells consist of materials with different band-gaps. Materials with larger band-gaps stack on top so different parts of the solar spectrum can be preferentially absorbed at different depths from the surface. (b) A solar cell made from IIl-V nanowire p-n junctions grown on Si substrate. (c) A three-junction solar cell made from heterogeneous III-V nanowires grown on Si solar cells ................................................................................................................................. 58 Figure 3. 2 Device structure and optical absorption simulation. (a) Schematic of a solar cell made from GaAs nanowires with axial junctions. Carrier flow direction in solar cell operation under sunlight is shown in the zoomed-in graph of the junction. (b) Maximum achievable J" (mA/cm 2 ) versus pitch a & diameter/pitch ratio d/a. Nanowires are 3 µm tall and grown on a GaAs substrate. (c) Maximum achievable Jsc versus nanowire diameter for a fixed pitch of 600nm and height of 3 µm. ( d) Absorption spectra of nanowire arrays with 600 mn pitch and 100 nm (black), 250 nm (red) and 350 nm (blue) diameter. (e) Carrier generation rate profile under AM l.5G solar spectrum for nanowires embedded in BCE and capped by ITO for diameters of 100 mn, 250 nm and 350 nm. ........................................................................................................................... 62 Figure 3. 3 Comparison between nanowire solar cells with axial junctions and radial junctions. (a) Schematic diagram of axial (upper) and radial (lower) junctions in nanowires. (b) J"' (c) V" and ( d) efficiency of nanowire solar cells with axial junctions (black) and radial junctions (red) versus doping concentration in n-type base. Surface recombination velocity is 3xl0 5 emfs. See text for details of junction geometry and doping concentration. (e) 3D mapping of hole concentration in the dark and in thermal equilibrium for the axial junction (left) and the radial junction (right). Band diagrams under AM 1.5 short circuit conditions along the center of each nanowire (A-A' and B-B ') are shown on the left together with quasi Fermi levels of electrons (blue dashed line) and holes (red dashed line). (f) Band diagram along C-C' line in the radial junction in (e) together with quasi Fermi levels of electrons and holes ................................................. 65 Figure 3. 4 Solar cell fabrication process and SEM images. (a) Fabrication steps of GaAs nanowire array solar cells with axial junction: (i) electron beam lithography to form hole array in silicon nitride mask, (ii) SAG of p-i-n GaAs nanowire using MOCVD, (iii) BCE infiltration, (iv) RIE to expose nanowire tips, and (v) ITO deposition. (b) 30° tilted SEM image of as grown vertical GaAs nanowire array on GaAs (11 l)B substrate. (c) SEM image after nanowires are embedded in BCE and etched by RIE to expose short tips. ( d) SEM image after coating of ITO film by sputtering. A conformal dome-like cap is formed on the tips of nanowires .................................................................................. 70 Figure 3. 5 Comparison between performance of nanowires with 100 nm diameter and 250 mn diameter. (a) 30° tilted SEM images of nanowires with diameter of 100 nm (sample A) and 250 mn (sample B). (b)J-V curves of typical devices made from 100 nm and 250 nm thick nanowires in dark and under AM 1.5 solar spectrum. ( c) V" (black) and FF (blue) of six devices from each batch. ( d) JIV-V curve of the device with 100 nm nanowires shown in (b) which shows linear region marked by the blue line. Insets are schematics of a fully depleted thin wire and of a partially depleted thick wire that still has conductive channel in the center ............................................................................... 73 Xl Figure 3. 6 Junction depth dependency and J-V under varied light intensity. (a) J-V curves of devices with different junction depths between 100 nm and 600 mn under AM 1.5 solar spectrum. For the device with 100 nm junction depth, diameter is increased to 320 mn. (b) Dark and AM 1.5 J-V curves of the device with 100 nm junction depth and 320 nm diameter shown in (a) plotted in semi-logarithmic scale. Ideality factor is found to be 2.34. (c) J-V curves of the device with 150 nm junction depth and 220 nm diameter shown in (a) under different illuminating power between 50 µWand 10 mW from an 850 mn laser. v" of 0.716 vis observed under the highest light intensity. Insets shows the curve of V" v.s. ln(I") with extracted ideality factor to be 1.72. (d) Current versus pump power for the device shown in ( c) at -2 V and 0 V bias ........................................ 75 Figure 3. 7 Dark and light J-V curves of device G with 320 nm diameter and 300 mn junction depth ................................................................................................................................ 77 Figure 3. 8 dV/d(lnI) versus I curve of the device showing the highest efficiency Based on Cheung's method, R, is the slope and nkT/q is the y-axis intercept. The extracted values of ideality factor and series resistance are 2.2 and 41 i.1, respectively ............................ 79 Figure 3. 9 Spectrum response and Cathodoluminescence mapping. (a) Carrier generation rate profile under monochromatic light of different wavelengths between 350 mn and 800 nm. All the light intensities are 100 mW/cm 2 . Nanowires are 2.5 µm tall and 350 nm in diameter with 600 mn pitch. Nanowires are surrounded by BCE and covered by ITO on top. (b) External quantum efficiency (left y-axis) v.s. wavelength of the four devices shown in Fig 3.6a. Dash-dot curve is the photon intensity of AM l.5G solar spectrum (right y-axis). J" of each device are calculated by integrating the product of EQE and photon intensity over the range of 350 nm to 880 nm and are indicated in the plot. (c) SEM images and normalized cathodoluminescence mapping at wavelength of 865 mn for nanowires with 400 nm deep junction (left) and 100 nm deep junction (right). Red curves are relative CL intensity along the center of each nanowire ................................ 81 Xll Figure 4. 1 Top view SEM images of GaAs nanowire arrays grown on Si with different hydrogen annealing times. Other conditions are all the same as described in text. Lower annealing time reduces vertical nano wire yield ............................................................. 93 Figure 4. 2 (a) Temperature profile of GaAs nanowire growth on Si (111). Red and blue blocks indicate the time AsH 3 and TMG are supplied, respectively (b) 30°tilted and (c) top view SEM images of uniform GaAs nanowire arrays grown on Si. ( d) Schematic diagram of crystal lattice at GaAs/Si interface, viewed from (1-10) orientation. (e) HRTEM image taken at the GaAs/Si interface. Arrows indicate sites with misfit dislocation ....................................................................................................................... 94 Figure 4. 3 Top view SEM images of nanowire arrays grown with and without low temperature nucleation step prior to the nanowire growth. Nucleation is 4 minutes if there is ............................................................................................................................. 96 Figure 4. 4 TEM image at the nanowire/Si interface. The starting point of nanowire is deeper than the Si substrate surface indicated by red dashed line because of over-RIE ............. 97 Figure 4. 5 30° tilted SEM images at the holes on (a) GaAs (11 l)B and (b) Si (111) substrates after 2 minutes growth at 790 °C. ( c) Top view SEM images of nuclei after 2 minutes growth under different temperature and AsH 3 partial pressure. TMG partial pressure is kept constant at 7.56x10· 7 atm. (d) Schematic ofrelevant low-index planes viewed from (11 l)B direction. (e) Projection of relevant low-index planes on (1-10) plane in GaAs zincblende lattice. ____________________________________________________________________________________________________________ 98 Figure 4. 6 (a) 30° tilted SEM images ofnanowires grown at 850 °C, the initial part is grown at 760 °C to help nucleate. (b) SEM of images of nanowire tips with different morphologies. Images are organized in a sequence according to the proposed twin formation model in ( c ). ( c) Schematic of twin formation process during nanowire growth. ( d) SEM image of a nanowire tip. After even number of twins form the wire pinches off (c) SEM image of a nanowire tip with odd number of twins. (f) TEM image of a region consists of even number of twins. The crystals at two sides of transitional region share the same atomic registry. (g) TEM image of a region consists of odd number of twins. The crystal at two sides of transitional region can be considered as rotated by 180 ° compared to each other. ----------------------------------------------------------------------102 Figure 4. 7 (a)-(c) TEM images ofnanowire grown at different temperatures: (a) 760 °C, (b) 790 °C and ( c) 850 °C. The initial segment in ( c) is grown at 760 °C in order to nucleate. (d)-(f) TEM images of nanowires grown at 850 °C with different diameters: (d) 95 nm, (e) 180 nm and (f) 280 nm. (g) Histogram showing length distribution of twin free segments of nanowires grown at 850 °C with different diameters . ............................... 104 Figure 4. 8 Probability distribution of the twin-free segment length in GaAs nanowires of the diameter 95, 180 and 280 nm at temperature (a) 850 °C and (b) 760 °C. _____________________ 108 Xlll Figure 5. 1 Top view SEM images of n-type GaAs nanowires grown with different disilane flow rate: (a) 0.2, (b) 1 and (c) 2.5 seem. Scale bar is 1 µm. (d) PL spectra of GaAs nanowire grown with different disilane flow rate. (e) Peak intensity and width (at half maximum) of spectra in ( d) ........................................................................................... 122 Figure 5. 2 (a) Schematic energy band diagram for direct band to band transition with k conservation. (b) Schematic energy band diagram for indirect band to acceptor transition without k conservation. (c) Experimentally measured PL spectrum of non-doped nanowires and theoretical fitting based on direct band to band transition in Eq 5.1. (d) PL spectrum of single non-doped nanowire measured at 4K ........................................ 124 Figure 5. 3 (a) Normalized measured PL spectra of nanowires grown with different disilane flow rate and curve fitting based on indirect band to acceptor transitions .................... 127 Figure 5. 4 Normalized PL spectra measured at 77K for nanowires with different disilane flow rate and laser excitation power .............................................................................. 129 Figure 5. 5 temperature dependence of PL spectra of nanowires grown with 1 seem disilane flow rate ........................................................................................................................ 130 Figure 5. 6 (a) Schematic of n+-GaAs/p+Si heteroepitaxial tunnel junction device. (b) 30° tilted SEM images of the device proposed in (a) at each step: (i) as grown n+-GaAs nanowires on p+-Si (111) substrate; (ii) after nanowire array being planarized and etched by RIE, wire tips are exposed; (iii) after deposition of AuGe/Ni/Au on top; (iv) after rapid thermal annealing at 380 °C with 5 °C/s ramping rate ......................................... 132 Figure 5. 7 (a)Boron impurity profile after implantation with ion energy of 40 keV and ion dose of 8xl0 15 cm- 2 . (b) Phosphorus impurity profile after implantation with ion energy of 85 keV and ion dose of 2.5xl0 16 cm- 2 . (c) Photography of a completed Si solar cell with SiN AR coating and two front bus lines. (d) Dark and lightJ-V curve of a typical Si W~C~.-----------------------------------------------------------------------------------------------------------------------1~ Figure 5. 8 (a) Schematic of dual junction solar cell with GaAs nanowire top cell epitaxially grown on top of Si planar bottom cell. (b) Dark and light J-V curve of device A mentioned in text, disilane flow rate in n+ GaAs is 1 seem. ( c) Dark and light J-V curves of device C to E mentioned text. Nanowire lengths are 900, 1800 and 2300 mn for C, D and E, respectively. Disilane flow rate in n+ GaAs is 1.5 seem. Performance specifications are tabulated. _________________________________________________________________________________________ .135 Figure 5. 9 (a) to (c) absorption in GaAs nanowire, Si substrate and all structure for nanowires with different indicated nanowire lengths. ( d) to (f) corresponding J-V curves, which shows nearly matched current between GaAs and Si for the first geometry and largely mis-matched currents for the taller nanowires. -------------------------------------------------137 X!V Abstract This dissertation covers the growth, characterization of GaAs nanowires and application in photovoltaic solar energy conversion. Nanowires, due to their unique electronic, optical and crystallographic properties, are promising candidate for the next-generation high-efficiency and low-cost solar cells. They could be grown on inexpensive substrates regardless of the lattice mismatch. Better bandgap combinations for multijunction solar cells become possible. The strong interaction between nanowire arrays and sunlight allows significantly reduced material usage. In this dissertation, we systematically demonstrated how we progressively approached the goal of III-V nanowire-on-Si tandem solar cells. Chapter 1 first presents a big picture of the global energy problem and justifies solar energy would be a promising renewable energy source followed by some discussion of how to surpass the Shockley Queisser limit using multijunction solar cells. Then I discuss the motivation of using nanowire as the building blocks of multijunction solar cells. Chapter 2 presents the technique we used to synthesis GaAs nanowires in a non-catalytic and highly uniform fashion which is selective area growth using MOCVD. Patterning techniques associated with selective area growth are discussed with some emphasis on nanosphere lithography that is suitable for high-throughput patterning. GaAs nanowire homojunction solar cell is demonstrated in chapter 3. The device shows champion efficiency of 7.58% after junction depth and diameter optimization. Chapter 4 is an extension of the growth technique reported in chapter 2. Here we transfer the growth to Si (111) substrate and solved the lattice xv mismatch issue for the heteroepitaxy. Along the way we also come up with a model to coherently interpret twin formation during the growth which agrees with experimental observations and atomistic simulations. In chapter 5, we first demonstrate heavily doped GaAs/Si heterojunction that shows ohmic transport behavior. The doping is characterized by photoluminescence measurement based on Burstein-Moss effect. Eventually we are able to integrate the GaAs nanowire top cell with the Si planar bottom cell and to deliver a working tandem cell with 11.4% efficiency. This is the first time the concept of epitaxial III-Von Si solar cell is materialized and sets a milestone in the progress of multijunction solar cells. XV! Chapter 1 Introduction 1.1 Global energy challenge 1.1.1 Global energy consumption Among all the technical issues the world is facing today, energy is the most important one. It is energy that really drives all the human activities and pushes our civilization forward. Without energy we can do nothing. The ever-increasing global demand of energy and environmental impact it brings along require us, as a whole human community, to find a way to manage the grand energy challenge. Dr. Richard Smalley testified in Congress that energy is indeed the most critical of current day issues. He stated that "it is the single most important challenge facing humanity today". Specifically this challenge is for the world to sustainably produce enough energy to meet the global 16 TW demand today. Out of that 16 TW, the United States consumes about 3.2 TW (equivalent to 97.53 quadrillion Btu (quads) per year) while supplying about 3.5 TW (109.33 quads including domestic production and imports) 1 . In 2013, 79.796 (81.8%) quads out of the total 97.534 quads consumption is represented by fossil fuels which mainly include coal, natural gas and crude oil. Renewable forms of energy represent 9.291 quads (9.5%) and nuclear electric power represents 8.268 quads (8.5%) 2 . Figure 1.1 U.S. energy flow, 2013 (quachillion Btu) 1 One can take the proven reserves of various fossil fuels and divide that number by the burn rate for each of those fuels to compute approximately how long each fuel will be available for energy production. The results show that we have 40-80 years of oil supplies, 60-176 years of naturnl gas and more than 200 yeaTs of coal. However "proven reserves" are the quantities of fossil energy that the U.S. Securities and Exchange Commission allows a company, with 90% confidence to book and to tell its stockholders it has in the ground. What really matters is not proven reserves but the amount of fossil fuels that is still available in the ground. United Nation Development Program (UNDP), with 50% confidence, estimates we have from 50 to 150 yearn of oil resources, with oil discovery going on continuously. We also have about 200 to 600 years of natural gas and almost 2000 years of coal in our resource base 3 . Furthermore we are now able to conve1t both natural gas and coal into liquid hydrocarbons. Thus, a limited global supply of oil can be compensated for the additional consumption of another fossil fuel. Today fossil fuel is still the cheapest energy resource. Given all of these factors, the abundance of fossil fuels 2 is not going to be the driving force of the transition into renewable energy era anytime sooner. The pure market forces determine the inexpensive fossil energy will be the favorite in the market for a long time. 1.1.2 Demand of solar energy So far we have seen the continuous supply of fossil fuel is not a series problem because they can still last for centuries, even if the predicted global energy consumption by 2050 is about 28 TW 4 . The true problem is associated with the carbon dioxide (C0 2 ) emitted when the fossil fuels are burned. Earth's temperature is a result of an equilibrium established between the incoming radiation from sun and the energy radiated into space by the earth. The presence and composition of the earth's atmosphere have significant impact on the outgoing radiation emitted by the Earth. Water vapor and C0 2 are the two major gases that absorb the outgoing radiation and keep the Earth warm. Ice core data from over 600,000 years shows the C0 2 level has been in a narrow band of between 200 and 300 ppm but not higher 5 • 6 . Over this same time period, the atmospheric C0 2 concentration has been highly correlated with temperature swings that have repeatedly caused ice ages on the planet. The C0 2 concentrations in the past 50 years have been rising because of anthropogenic C0 2 emissions from fossil fuel consumption and they are now in excess of 380 ppm, a new level never seen before in measured history. Without being too aggressive in prediction, a more than double of the pre-anthropogenic C0 2 is expected within the century if the current trend of emission continues 7 • 8 . Some climate 3 change models predict senous sea level mcrease, ocean acidification, changes of hydrological cycle, melt of permafrost and other effects to occur for atmospheric C0 2 at or in excess of 550 ppm 9 . To hold the C0 2 concentration at 550 ppm and yet meet the demand in 2050 we need more than 10 TW power produced by carbon-free energy resources assuming 0.45 kg of carbon emitted per year per watt of power produced under business-as-usually scenario (right now this value is around 0.5 kgC/Wyr)7. This requires a much greater portion of the total energy consumption to come from carbon-free power than it does right now. The only two practical resources for these 10 TW carbon-free powers are nuclear fission and renewable energy under the existing technology. A typical nuclear fission reactor is designed at 1 GW electricity output level for the safe handle of heat and radiation flux consideration. So we need 10,000 such reactors to be built in the next 35 years. However the proven reserves and the resource base of all terrestrial uranium combined would be enough to provide 10 years of operation for this 10,000 nuclear power plants. Another option is to use carbon-free or carbon-neutral renewable energy which is defined as energy that comes from resources which are naturally replenished on a human timescale. According to U.S. energy information administration, renewable energy includes seven sources: hydroelectric, wood, biofuels, wind, waste, solar/PY and geothermal. Among all these renewables, solar energy is the only practically viable terrestrial resource to meet the 10-20 TW carbon-energy demand by the middle of this century. The sunlight is free, abundant and widely distributed, available to every country and person in the world. The 4 energy flux of 165,000 TW from sunlight to the Earth means more energy from the sun hits the earth in one hour than all of the energy consumed on our planet in an entire year. To put that another way, the area in the box is the amount of land that would be required for a solar energy farm operating at 10% efficiency to supply the 3 TW power consumption of the entire country. This box represents 1.7% of the U.S. land, comparable to the area devoted to the nation's highways 4 . 1.2 Photovoltaic solar energy conversion 1.2.1 Physics of photovoltaic process Photovoltaic effect was first discovered m 1954, when scientists at Bell Telephone discovered that silicon created an electric charge when exposed to sunlight. Soon solar cells were being used to power space satellites and smaller items like calculators and watches. In a photovoltaic energy conversion process, electricity is produced in the form of voltage and current upon the irradiation of electromagnetic waves. A device that utilizes photovoltaic effect to convert solar energy into electricity is called a solar cell. Four basic steps are involved to complete one cycle of such energy conversion: 1. A light absorption step in which a photon is captured by a material and its energy is transferred to a charge carrier so that the carrier is stimulated from a ground state to an excited state, 2. An exciton dissociation step in which a pair of free negatively charged and positively charged carriers is created, 5 3. A transportation step in which negatively-charged carriers move in one direction to a contact called cathode and positively charged carriers move in another direction to a contact called anode, and 4. After an energetic negatively charged carrier reaches the cathode it dissipating the energy by traveling through an external path and doing useful work at an electrical load. Then it returns to the anode and recombine with a positively charged carrier so that the absorbing material restores to the ground state. In some materials, such as most semiconductors, the first two steps combines because of the relatively small exciton binding energy. In some other materials such as organic molecules and quantum dots, these two steps are distinct. 1.2.2 Solar cell figures of merit In order to quantify the performance of a solar cell certain standard conditions and parameters are established. Firstly, the sunlight is distributed, so the exact pattern of the solar spectrum depends on time, latitude and atmospheric condition. Several standardized solar spectra based on average energy intensity are developed to characterize solar cells unambiguously and to compare devices fairly. These spectra define the number of photons (or sometimes power) per unit area per unit time per unit bandwidth in the sunlight. The Air Mass (AM)O spectrum ,based on ASTM standard E-490, is the spectrum outside the atmosphere and is used for space power applications such as satellites. AM l.5G spectrum, based on ASTM standard G-173, is another important spectrum for 6 terrestrial applications which includes both direct and diffuse light. The integrated power density is 1000 W/m 2 . The AM 1.5D spectrum, on the other hand, only takes the direct light into account so its integrated power density is 888 W/m 2 . Solar simulators duplicate these standard spectra so that devices can be characterized in a laborat01y . ..., 1: ·c;; c Q) 0 c: ~ !:5 0 Maximum Power Point With J = Jmp & V = V mp Voltage Developed V Figure 1.2 J-V characteristic of a solar cell Current density J is defined as current divided by the cell area which together with corresponding voltage V, completely describe the performance of a sola!' cell. Fig 1.2 represents a typical J-V characteristic of a solal' cell. Short-circuit cmrent density Jsc is the current density when no voltage is applied between the anode and cathode while open-circuit voltage V 0 c represents the voltage when no current flows between the anode and cathode. Areal output power Pout at a given voltage is given by the product of J and V. The current density and voltage at maximum Pout is labeled as Jmp and Vmr Therefore the best thermodynamic efficiency 1/ of the solar cell is: where Pin is the total power impinging on the solar cell and is 1000 W/m 2 if the light 7 source is AM1.5G spectrum. Another important figure of merit called fill factor FF is invented to describe how closely a given curve is to conforming to the ideal rectangular J-V characteristic. Thus the fill factor is given by: J xVmp FF=-mp~-~ Jsc X V:c Obviously, FF is always less than one. Other parameters can also be extracted from the J-V characteristic, such as series resistance, shunt resistance and ideality factor, which help to understand the physical nature of a device and will be discussed in detail in following chapters. 1.2.3 Shockley-Queisser limit and multijunction solar cells For a solar cell using a single p-n junction, there exists a theoretical maximum efficiency. This value is called Shockley-Queisser limit or detailed balance limit, first calculated by William Shockley and Hans Queisser in 1961 rn The limit places maximum solar conversion efficiency around 33.7% assuming a single p-n junction with a band gap of 1.34 eV under AM 1.5 solar spectrum. Several fundamental loss mechanisms are pointed out in the limit which is inherent to any solar cell design. The first are the losses due to blackbody radiation, a loss mechanism that affects any material object above absolute zero. The second mechanism is radiative recombination, meaning not all the all the electron-hole pairs created by photovoltaic effect can be collected but some of them will recombine and emit photons. The third and most dominant mechanism is spectrum losses. Only photons with energy larger than the bandgap can be absorbed and produce a 8 photoelectron. For example silicon has a bandgap of 1.1 eV, so it cannot absorb far infrared light, microwaves and radio waves which compose about 19% of the total power in AM 1.5 solar spectrnm. Conversely, although a photon with energy more than the bandgap can be absorbed, the extra energy is immediately lost through collision with crystal lattice known as "thermalisation". The lost energy turns into heat in the cell. This limit however is delived based on several fundamental assumptions: the cell contains a single p-n junction; any extra energy of a photon compared to the bandgap is lost; only one electron-hole pair is created per incoming photon; and the solar cell is illuminated by unconcentrated sunlight. None of these assumptions is necessarily tme and valious approaches have been used to surpass the Shockley-Queisser limit including multi junction architectme, hot electron capture, intrnduction of inte1mediate-band, photon upconversion. multiple exciton generation, light concentrntion. etc. GalnP top cell { i---------4' p-AlGalnP BSF Wide-bandgap tunnel junction l==::l·~ ··.~;:::::;::=:11' I n-GalnP window n-Ga(ln)As emitter .,. .. n-•-TJ n.Ga(ln)As buffer Tunnel junction """'"'';'" l { l==~ n- !ii)j e em ~:;; er==V. Ge bottom cell p-Ge base and substrate t::::::~c:::::=V Figure 1.3 Schematic of a lattice matched triple junction solar cell with 41.6% efficiency made by Spectrolab 11 . 9 Among all these approaches, multijunction solar cell is the most widely explored and with most significant research and commercial success. A multijunction solar cell consists of multiple materials with different bandgaps. By stacking them in series with higher bandgap material on top each layer is tuned to response to a particular band of the solar. Materials with higher bandgaps can utilize higher energy photon more efficiently by alleviating the thermalisation loss mentioned above while the materials with lower bandgaps can capture the lower energy photons that will be otherwise wasted in a single junction solar cell. Yet to produce such a tandem cell is not easy. Monolithically integrated multijunction solar cells are usually made from III-V compound semiconductor given the vast choice of bandgap and lattice constant available and maturity in growth and process technology. Fig 1.3 represents a typical triple junction solar cell made by Spectrolab"- Ga(In)As and GainP subcells with matched lattice to Ge bottom cell are grown epitaxially. This device can deliver 41.6% efficiency at 364x concentration, way beyond the Shockley. However the bandgap of Ge is lower than the optimum for a triple junction solar cell, and a subcell with around 1 e V bandgap can further improve the efficiency. By using a combination of GainNAsSb/GaAs/InGaP with bandgap of (1/1.4/1.9 e V), Solar Junction successively demonstrated a lattice matched triple junction solar cell with 44% efficiencyn Recently Sharp hits a Fraunhofer-Institute-confirmed, world-record 44.4% for its triple junction solar cell made by inverted metamorphic technique"- The top InGaP (1.9 e V) and GaAs (1.4 e V) cells are lattice matched to Ge or GaAs substrates while the bottom GainAs cell is grown through a transparent buffer layer to accommodate the lattice difference between GaAs and GalnAs 14 . 10 1.3 Semiconductor nanowires and its application in solar cells 1.3.1 Economy of solar energy Given so many advantage of solar energy, an apparent question arises: why does solar energy not compose a larger fraction of the total energy production? The answer lies in the cost. The use of energy is decided not only by policy but also to a large degree shaped by the economic factors. Electricity produced from a coal power plant is 1-4 cents per kilowatt-hour (kWh) which is much cheaper than the price per kWh from any forms of renewable energies. In 2011 U.S. Department of Energy (DOE) launched the SunShot Initiative to drive research, manufacturing, and market solutions to make the abundant solar energy resources in the United States more affordable and accessible for Arnericans 15 . In order to make solar energy cost effective and competitive to fossil fuel, the SunShot Initiative aims to reduce the total installed cost of solar energy system to 6 cents per kWh by 2020. Today, SunShot is 60% of its way toward achieving the program's goal and the average price per kWh of a utility-scale photovoltaic project has dropped from about 21 cents to 11 cents. According to the latest Greentech Media Research and Solar Energy Industries Association's (SEIA) Solar Market Insight Year in Review 2013 report, photovoltaic installations continued their impressive growth, increasing 41 % over 2012 to 4,751 megawatts (MW) of installed power in the United States, totaling 12,000 MW of PV. In 2013, China installed 12,000 MW and increased total PV capacity dramatically to 20,300 MW. 14,000 MW is planned for 2014 to further cut the carbon emission and meet the goal of sustainable society. After the nuclear disaster in Fukushima Japan also actively seeks expansion of PV capacity to replace the 11 nuclear fission reactors shut down. Total installed PV capacity by the end of 2013 is 14,000 MW with 7,000 MW installed in 2013. In the first quarter of 2014 the U.S. installed another 1,330 MW to 14.8 GW installed capacity. 74% of all new electricity generation in the U.S. came from solar in 2014 QI and 100% of new electricity generation in Arizona, Illinois, Massachusetts, New Jersey, Missouri, Vermont and the District of Columbia came from solar energy 16 . Further reduction of the cost of photovoltaic system would partly come of the continuous technology development. Higher conversion efficiency and lower manufacturing cost would both play an important role in this regard. Industry-leading Chinese crystalline silicon PV module manufacturers anticipate the production costs will fall from 50 cents per watt in the fourth quarter of 2012 to 36 cents per watt by the end of 2017. The majority of these cost declines will derive from technology innovations that would roughly contribute to 7 cents per wattn 1.3.2 Semiconductor nanowire solar cells We have seen so far that multijunction solar cell is the most promising approach to surpass the Shockley-Queisser limit. However it is difficult to fabricate monolithically integrated multijunction solar cells in a traditional planar fashion. High crystal quality requires lattice matched materials which is a big constrain of material choice and add complication in the synthesis process. Solar Junction used complicated Galn.NAsSb as lattice-matched I e V subcell. Sb concentration is restrictly controlled within the range of 0.1-3% as surfactant and constituent to minimize N-related defects, not higher or lower. 12 Furthermore the quality ofGainNAsSb is sensitive to carbon and hydrogen contamination so Molecular Beam Epitaxy (MBE) is used instead of Metalorganic Chemical Vapor Deposition (MOCVD). In inverted metamorphic (IMM) devices, although the lattice-match constrain can be relieved to certain degree, the complexity in growth of buffer layer and in the wafer peeling-off process is by no means easy. All these factors combined with the expensive Ge and GaAs substrates and limited supply of In significantly increase the manufacturing cost so that the efficiency advantage of multijunction solar cells is outweighed by the low cost of Si solar cells. These years, the advancement in the field of nanotechnology provides us innovative methods to overcome the aforementioned difficulties. Over the past decade, solar cells based on semiconductor nanowires have been a topic of intense research and development for next-generation photovoltaics 18 - 38 . Many unique characteristics become available in nanowires due to their small feature size. It has been shown that lattice mismatches can be accommodated during the growth of nanowire heterostructures through strain relaxation at nanowire sidewall without generating dislocations 39 - 44 _ This capability allows us to choose optimum bandgap combination for multijunction solar cells. Unique interaction between light and nanowires can lead to efficient absorption otherwise can only be achieved by much larger amount of material in conventional fihns. Thirdly, nanowires with core/shell radial junction greatly reduce the charge diffusion distances thus allowing using materials with relatively shorter carrier diffusion length. Last but not least, it is possible to use Si or even flexible and/or transparent material to replace the expansive 13 GaAs or Ge substrates making multijunction solar cells cost-effective for terrestrial implementation. Isoefficiency curves of two junction solar cells can be plotted as a function of the top cell and bottom cell bandgap 45 . Curves plotted by Olson et al. indicate that a two junction cell with maximum efficiency of 38.8% (I sun, AM l.5G) consists of a top bandgap of 1.75 eV and a bottom bandgap of I.I eV. Geisz et al. estimated a limiting efficiency of 37.4% for a top bandgap of I. 7 e V and a bottom bandgap of I. I e V 46 . Tobias et al. determined a limiting efficiency of 40.7% for bandgaps of 1.65 eV and 0.97 eV 47 . Finally, from thermodynamic considerations, Green et al. determined a limiting efficiency of 42.5% for a bottom bandgap of 0.97 e V and top bandgap of I. 7 e V 48 . Similarly under concentration of 500 suns, the limiting efficiency has been estimated as 43.8% and 44.2% for the optimum two-junction and lattice-matched triple junction cells, respectively 46 . These calculations indicate a 1.7 eV bandgap material on Si solar cell should have optimum efficiency comparable to that of the commercial lattice matched (to Ge) triple junction cells. Obviously, the use of Si as substrate significantly reduces the cost of the former and III-V nanowires grown on Si provides us a promising path toward this Holy Grail. 1.4 Outline of the dissertation Chapter 2 starts with general discussion of nanowire growth using Vapor-Liquid-Solid (VLS) and Selective Area Growth (SAG) methods with the emphasis on the growth 14 behavior of SAG method and its application in growmg GaAs nanow!fes. Different patterning techniques for making SAG template are briefly introduced and compared. Among these techniques, nanosphere lithography (NSL) is discussed in detail. Theoretical and experimental study of light absorption in NSL patterned nanowire is presented and proves slight disordering does not decrease light absorption. Chapter 3 focuses on GaAs nanowires' application in solar cells. 7.58% efficiency is demonstrated from solar cells based on GaAs nanowires with axial p-i-n junction. We present detailed optical and electrical simulation which reveals advantages of axial junction over radial junction. Device optimization is performed experimentally, showing significance of junction depth and diameter in determining the overall performance. With the success m nanow!fe growth and solar cell device fabrication discussed in Chapter 2 and 3, the next goal will be making GaAs nanowire-on-Si. Chapter 4 discusses the heteroepitaxial growth of GaAs nanowires on Si (l ll) substrates using SAG. A growth model is proposed to explain the twin formation during growth. Twin density is studied as a function of growth temperature and wire diameter. Theoretical thermodynamic analysis of the nucleation process is presented at the end, which well explains the twin energetics under the various experiment conditions. Finally in Chapter 5, we report the developments in the GaAs nanowire-on-Si dual junction solar cells. We choose n+-GaAs/p+-Si heterojunction as the interconnecting junction. In order to obtain heavily doped n-GaAs nanow!fe, we carried out 15 photoluminescence (PL) measurement to investigate the doping characteristics. The PL spectra can be well explained by non-vertical transition with band tails. Fermi-energy, tail depth and carrier concentration are extracted through curve fitting. Si self-compensation is observed and results in deep acceptor levels. Nearly ohmic behavior is observed between heavily doped GaAs nanowire and heavily doped Si substrate. This is used as tunnel junction to fabricate the dual junction solar cells. Best device with matched current shows voltage add-up and efficiency as high as 11.4%. Chapter 6 will conclude this dissertation by providing an overview of main findings in this work and a brief discussion of several interesting further research directions. 16 References 1. http://www.eia.gov/totalenergy/data/monthly/pdf/flow/total energy.pdf 2. http://www.eia. gov/totalenergy/data/monthly/pdf /mer.pdf 3. Goldemberg, J.; Programme, U. N. D.; Economic, U. N. D. o.; Affairs, S.; Council, W. E., World energy assessment: energy and the challenge of sustainability. United Nations Development Programme: 2000. 4. Lewis, N. S.; Nocera, D. G. Proc. Natl. Acad. of Sci. 2006, 103, (43), 15729-15735. 5. Petit, J. R.; Jouzel, J.; Raynaud, D.; Barkov, N. I.; Bamola, J. M.; Basile, I.; Bender, M.; Chappellaz, J.; Davis, M.; Delaygue, G.; Delmotte, M.; Kotlyakov, V. M.; Legrand, M.; Lipenkov, V. Y.; Lorius, C.; Pepin, L.; Ritz, C.; Saltzman, E.; Stievenard, M. Nature 1999, 399, (6735), 429-436. 6. Siegenthaler, U.; Stocker, T. F.; Monnin, E.; Luthi, D.; Schwander, J.; Stauffer, B.; Raynaud, D.; Bamola, J.-M.; Fischer, H.; Masson-Delmotte, V.; Jouzel, J. Science 2005, 310, (5752), 1313-1317. 7. Hoffert, M. I.; Caldeira, K.; Jain, A. K.; Haites, E. F.; Harvey, L. D. D.; Potter, S. D.; Schlesinger, M. E.; Schneider, S. H.; Watts, R. G.; Wigley, T. M. L.; Wuebbles, D. J. Nature 1998, 395, (6705), 881-884. 8. Wigley, T. M. L.; Richels, R.; Edmonds, J. A. Nature 1996, 379, (6562), 240-243. 9. Watson, R. T.; Albritton, D. L., Climate change 2001: Synthesis report: Third assessment report of the Intergovernmental Panel on Climate Change. Cambridge University Press: Washington, DC, 2001. 10. Shockley, W.; Queisser, H.J. J. Appl. Phys. 1961, 32, (3), 510-519. 11. King, R.; Boca, A.; Hong, W.; Liu, X .; Bhusari, D.; Larrabee, D. ; Edmondson, K.; Law, D.; Fetzer, C.; M esropian, S. In Band-gap-engineered architectures for high-efficiency multi.function concentrator solar cells, 24th European Photovoltaic Solar Energy Conference and Exhibition, Hamburg, Germany, 2009. 12. http://www.sj-solar.com 13. Green, M. A.; Emery, K. ; Hishikawa, Y.; Warta, W.; Dunlop, E. D. Progress in Photovoltaics: Research and Applications 2014, 22, (7), 701-710. 14. http://sharp-world.com/corporate/news/120531.html 15. http://energy.gov/eere/sunshot/sunshot-initiative 16. http://www.seia.org/research-resources/solar-market-insight-report-201 4-g 1 17. http ://www. greentechm edia. com/articles/read/s olar-pv-module-cos ts-to-fall-to-3 6-cents-p er-watt 18. Tian, B. ; Zheng, X.; Kempa, T. J. ; Fang, Y.; Yu, N.; Yu, G.; Huang, J.; Lieber, C. M. Nature 2007, 449, (71 64), 885-889. 19. Garnett, E. C.; Yang, P. J. Am. Chem. Soc. 2008, 130, (29), 9224-9225. 20. Tang, J.; Huo, Z.; Brittman, S.; Gao, H.; Yang, P. Nat. Nanotechnol. 2011, 6, (9), 568-572. 21. Kempa, T. J.; Cahoon, J. F.; Kim, S.-K. ; Day, R. W.; Bell, D. C.; Park, H.-G.; Lieber, C . M. Proc. Natl. Acad. of Sci. 2012, 109, (5), 1407-1 412. 22. Czaban, J. A.; Thompson, D . A.; LaPierre, R.R. Nano Lett. 2008, 9, (1), 148-1 54. 17 23. Goto, H.; Nosaki, K.; Tornioka, K.; Hara, S.; Hirurna, K.; Motohisa, J.; Fukui, T. Appl. Phys. Express 2009, 2, 035004. 24. Mariani, G.; Wong, P.-S.; Katzenrneyer, A. M.; Leonard, F.; Shapiro, J.; Huffaker, D. L. Nano Lett. 2011, 11, (6), 2490-2494. 25. Wei, W.; Bao, X.-Y.; Soci, C.; Ding, Y.; Wang, Z.-L.; Wang, D. Nano Lett. 2009, 9, (8), 2926-2934. 26. Shin, J. C.; Kirn, K. H.; Yu, K. J.; Hu, H.; Yin, L.; Ning, C.-Z.; Rogers, J. A.; Zuo, J.-M.; Li, X. Nano Lett. 2011, 11, (11), 4831-4838. 27. Tornioka, K.; Tanaka, T.; Hara, S.; Hirurna, K.; Fukui, T. IEEE J. Sel. Top. Quantum Electron. 2011, 17, (4), 1112-1129. 28. Hu, S.; Chi, C.-Y.; Fountaine, K. T.; Yao, M.; Atwater, H. A.; Dapkus, P. D.; Lewis, N. S.; Zhou, C. Energ. Environ. Sci. 2013, 6, (6), 1879-1890. 29. Zhu, J.; Hsu, C.-M.; Yu, Z.; Fan, S.; Cui, Y. Nano Lett. 2009, 10, (6), 1979-1984. 30. Putnam, M. C.; Boettcher, S. W.; Kelzenberg, M. D.; Turner-Evans, D. B.; Spurgeon, J.M.; Warren, E. L.; Briggs, R. M.; Lewis, N. S.; Atwater, H. A. Energ. Environ. Sci. 2010, 3, (8), 1037-1041. 31. Mariani, G.; Scofield, A. C.; Hung, C.-H.; Huffaker, D. L. Nat. Commun. 2013, 4, 1497. 32. Mariani, G.; Zhou, Z.; Scofield, A.; Huffaker, D. L. Nano Lett. 2013, 13, (4), 1632-1637. 33. Holm, J. V.; fargensen, H. I.; Krogstrup, P.; Nygard, J.; Liu, H.; Aagesen, M. Nat. Commun. 2013, 4, 1498. 34. Krogstrup, P.; Jorgensen, H. I.; Heiss, M.; Dernichel, O.; Holm, J. V.; Aagesen, M.; Nygard, J.; Fontcuberta i Morral, A. Nat. Photon. 2013, 7, (4), 306-310. 35. Cui, Y.; Wang, J.; Plissard, S. R.; Cavalli, A.; Vu, T. T. T.; van Veldhoven, R. P. J.; Gao, L.; Trainor, M.; Verheijen, M. A.; Haverkort, J. E. M.; Bakkers, E. P. A. M. Nano Lett. 2013, 13, (9), 4113-4117. 36. Gutsche, C.; Lysov, A.; Braam, D.; Regolin, I.; Keller, G.; Li, Z.-A.; Geller, M.; Spasova, M.; Prost, W.; Tegude, F.-J. Adv. Funct. Mater. 2012, 22, (5), 929-936. 37. Wallentin, J.; Anttu, N.; Asoli, D.; Huffman, M.; Aberg, I.; Magnusson, M. H.; Siefer, G.; Fuss-Kailuweit, P.; Dirnroth, F.; Witzigrnann, B.; Xu, H. Q.; Samuelson, L.; Deppert, K.; Borgstrom, M. T. Science 2013, 339, (6123), 1057-1060. 38. Yao, M.; Huang, N.; Cong, S.; Chi, C.-Y.; Seyedi, M.A.; Lin, Y.-T.; Cao, Y.; Povinelli, M. L.; Dapkus, P. D.; Zhou, C. Nano Lett. 2014, 14, (6), 3293-3303. 39. Gudiksen, M. S.; Lauhon, L. J.; Wang, J.; Smith, D. C.; Lieber, C. M. Nature 2002, 415, (6872), 617-620. 40. Glas, F. Phys. Rev. B 2006, 74, (12), 121302. 41. Sburlan, S.; Dapkus, P. D.; Nakano, A. Appl. Phys. Lett. 2012, 100, (16), 163108. 42. Chuang, L. C.; Moewe, M.; Chase, C.; Kobayashi, N. P.; Chang-Hasnain, C.; Crankshaw, S. Appl. Phys. Lett. 2007, 90, (4), 043115. 43. Ertekin, E.; Greaney, P.A.; Chrzan, D. C.; Sands, T. D. J. Appl. Phys. 2005, 97, (11), 114325. 44. Tornioka, K.; Kobayashi, Y.; Motohisa, J.; Hara, S.; Fukui, T. Nanotechnology 2009, 20, (14), 145302. 18 45. Kurtz, S. R.; Faine, P.; Olson, J.M. J. Appl. Phys. 1990, 68, (4), 1890-1895. 46. Geisz, J. F.; Friedman, D. J. Semicond. Sci. Technol. 2002, 17, (8), 769. 47. Tobias, I.; Luque, A. Progress in Photovoltaics: Research and Applications 2002, 10, (5), 323-329. 48. Green, M. A., Third generation photovoltaics: advanced solar energy conversion. Springer: 2006; Vol. 12. 19 Chapter 2 Selective area growth of nanowire arrays and scalable nanosphere lithography patterning 2.1 Introduction In recent years, semiconductor nanowires have attracted great research interest due to their unique electronic and optical properties 1 - 5 related to their large surface-to-volume ratio and potential for quantum confinement. The possibility to form functional heterostructures which is prohibited in bulk system also increases the versatility of nanowires. In addition, III-V compound semiconductor nanowires possess advantages of high mobility, direct band gap and the capability of bandgap engineering over a wide range making them promising candidacies for future electronic 6 - 14 , optoelectronic 15 - 25 and d . 26-30 energy ev10es . Unlike carbon nanotubes, the growth directionality does not come from the nature of the atomic arrangement, but instead it lies in the preferential growth in certain crystal orientation. Such preference depends on either enhancement of the growth rate in one orientation, and/or suppress10n of growth rate in the other orientations. A variety of techniques have developed to grow one-dimensional nanowire structures. The dominant category 1s growth with the assistance of catalyst particles. These particles can substantially promote the growth rate m one orientation. A model called Vapor-Liquid-Solid (VLS) 1s used to explain the growth mechanism 31 . Such system usually consists of vapor-phase precursors, a liquid alloy of both catalyst and growth 20 elements, and a solid nanowire crystal. Au nanoparticle is most widely used for this purpose since it is well developed for many materials. The main problem of using Au nanoparticles is it forms deep-level traps in the bandgap of many semiconductor materials causing undesired nonradiative recombination 32 • 33 . Atom probe measurements have confirmed trace of Au atoms in Si nanowires 34 . Other non-VLS techniques also exist for growing nanowires without the assistance of a seed particle. One method is oxide-assisted growth 35 , in which an oxide layer on the surface passivated the side wall facets. The oxide nature of the surface alters the surface chemistry and electronic state thus limits the suitability in certain application. Until recently Fukui's group developed another technique called Selective-Area-Growth (SAG) which allows to synthesis highly uniform nanowire arrays over large areas 36 . This technique uses a pre-defined mask template to define the location of nucleation process. The size of the openings in the template together with the fine-tuned growth condition favors the growth only in one direction. This can be considered as a milestone in the history of nanowire growth as it enables us, for the first time, to well control the position and dimension of nanowires so that they can be tailored for specific application and produced in a reproducible way. The first step in SAG is to generate openmg m a dielectric grow mask template. Depending on the required feature size, a variety of lithography techniques are available to produce the desired patterns. Electron beam lithography (EBL) is the most common technique due to its high precision and flexibility in pattern design. The each feature in a pattern is exposed in series by focused electron beam and thus it is a time-consuming 21 process and unsuitable for large scale manufacture. Other processes, such as photolithography and nanoimprint lithography, due to their parallel nature, are more efficient. Both of them require a pre-defined projection mask or physical mold which cannot be adjusted once fabricated. Nanosphere lithography (NSL) has emerged as a cost-effective and high throughput way to pattern large substrates 37 . NSL has been demonstrated to produce nanostructures with controlled shape, size and density 38 • 39 . In the field of VLS nanowire growth, catalyst nanoparticles were deposited in the interstitial regions among self-assembled nanospheres. The size of the nanoparticle is constrained to a tight range and the filling ration (the ratio of the area occupied by nanowires to the total area of substrate) is low 40 . In this chapter, we first compare the growth behaviors of VLS and SAG mode with the emphasis on the mechanism of SAG growth and its implementation in growth of GaAs nanowires on GaAs (lll)B substrates. A few patterning techniques are briefly introduced and compared in the context of making large area arrays of holes for SAG nanowire growth. A new NSL method that can offer flexible control of nanowire diameter and filling ratio is discussed in detail with successful demonstration of its application in SAG GaAs nanowire growth in MOCVD 41 . At the end of the chapter we study the light absmption in slightly disordered nanowire arrays produced from NSL both theoretically and experimentally. The results guarantee light absorption is not comprised m these nanow!fe arrays making NSL a promising technique to produce large scale nanow!fe arrays for photovoltaic applications. 22 2.2 Nanowire growth techniques 2.2.1 Epitaxial growth using MOCVD Epitaxy is defined as the deposition of a crystalline overlayer on a crystalline substrate, where there is registry between the overlayer and the substrate. The overlayer is called an epitaxial film. The driving force behind the crystal growth is the difference between the chemical potential of the precursor materials and the crystal to be grown. In MOCVD which is a vapor phase growth, a stable non-equilibrium condition is established with the continuous replenishing of vapor-phase precursors. The chemical potential difference can be expressed as: where µv and µs refer to the chemical potential of the vapor and solid phase, respectively. p refers to the partial pressure of the precursor in the vapor phase, and p 0 refers to the equilibrium partial pressure of that precursor over the crystalline material at the particular growth conditions. In a binary system, such as III-V materials, the chemical interaction between the precursors further complicate the estimation of chemical potential. A large number of physical and chemical processes take place in epitaxial growth, and the dominant factors to the growth behavior can been categorized into thermodynamic and kinetic ones. Thermodynamic determine the chemical potential for the crystallization to occur. The growth rate is typically governed by kinetics, which has several subcategories. Mass transport refers to the spatial movement of the source material from vapor phase to the growth front. Chemical reactions involving the precursor materials play significant role especially in MOCVD. The decomposition process involves pyrolysis in gas phase, 23 reactions on the crystal surface and even among different precursors and adducts. All these processes are influenced by mainly three conditions: temperature, overall pressure and partial pressure of the precursors. MOCVD is a subtype of vapor phase epitaxy and all the principles discussed apply. The unique feature of MOCVD is the growth of crystals involves chemical reactions unlike MBE which is purely physical deposition. The growth takes place not in a vacuum, but from gas phase at moderate pressures. At least one of the precursors (usually the group III precursor in III-V materials) is supplied in the form of metalorganic molecules. The volatile metalorganic compounds can be efficiently transported to the reaction zone by flowing carrier gas (hydrogen or nitrogen) through the metalorganic source container called bubbler. The metalorganic compounds have a wide range of vapor pressure that can be tuned by controlling the temperature of the water bath in which the bubblers sit in. The pressure around the outlet of the bubblers and the carrier gas flow rate can be precisely controlled by pressure controller and mass flow controller (MFC), thus the partial pressure of each metalorganic species in the reactor can be precisely tuned. Extremely low partial pressure is possible by adding an independent dilute line in parallel to the precursor line and control the total flow rate after they merge together. Furthermore, MOCVD can produce abrupt change of material composition due to the quick response of MFCs and solenoid manifolds. The nanowires discussed throughout this dissertation are all grown by MOCVD in Compound Semiconductor Laboratory, which is made by Thomas Swan. The reactor features a cold wall, which means the substrate is heated directly from the susceptor, so the gases do not react before they reach the hot crystal 24 surface. The reaction chamber is vertical with precursor gases entering through a shower head above the susceptor for more uniform gas distribution. 2.2.2 Nanowire growth using VLS method In 1964, Wagner and Ellis observed small Au particles, usually in hemispherical shape, on top of the Si whisker they grew. They proposed a mechanism to describe the role of these particles m the one-dimensional whisker growth, which they name the Vapor-Liquid-Solid mechanism. They suggested an alloy forms when Au is exposed to Si at the growth temperature which is above the temperature required to forma liquid alloy. The vapor phase precursor molecules, SiC1 4 in their case, stick preferentially to the surface of the liquid alloy and increase the Si concentration near the surface. With the continuous supply of over-pressured SiC1 4 , the alloy eventually become supersaturated with Si, which will precipitate out at the liquid-solid interface. The Au-Si binary phase diagram in Fig. 2.1 shows a single eutectic point at 636 K, and the Si composition of the liquid alloy at this point is 18.6 at.%. When Si concentration is higher than that, Si will precipitate out until it reaches the liquidus line. 25 Au-Si 1800 16-00 1400 ~ 1200 ~ 3 ~ :ii. 1000 E ., f- 800 600 400 300 0.0 0.2 0.4 0.6 0.8 Au X51 Figure 2. I Au-Si binary phase diagram Givmgi.zov extensively studied the details of VLS process, describing the relevant processes involved 42 . The relevant processes were specified as the dissociation of supply material at the particle surface, incorporation of growth mate1ial into the particle, diffusion through the particle, and nucleation at the interface. This work explicitly proposed that catalytic decomposition of precursor materials on the particle surface was responsible for the one-dimensional nature of such strnctures. In 1992, Hiruma et al. demonstrated small one-dimensional structures composed of GaAs by a small amount of Au 43 . Ever since then repo1ts of growth of various nanowires using metal particles boomed. Up to date, VLS remains the most popular method in nanowire synthesis because of simple sample preparation, vast applicability to many mate1ial system and :flexibility in control of growth condition. Generally, the directional growth is believed to be due to the increased local concentration of precursors. Some believe the liquid particle has catalytic effect so the decomposition activation energy is lowered. However, Wagner et al. reported the activation energy of Au assisted whisker growth is the same as that of 26 growth without Au 44 . They attributed the increased local concentration to an enhanced accommodation probability of the liquid. The growth of compound nanowires is even more difficult to understand than Si. In fact, no ternary Au-III-V phases or binary Au-V phases have been reported that are stable at temperatures below about 600 ° C. N, P and As have never been detected in the particle after growth by ex-situ techniques, which leads to the question of how the group V elements incorporate during the growth. Furthermore, Ga and In both exhibit very low melting points, far below typical growth temperatures. This means there is no solubility limit of these materials in Au at these growth temperatures. Thus these systems are not true eutectic systems which rules out thermodynamics to be the driving force of the precipitation. On the other whether particle works as catalysis is still of some controversy. It has been shown that GaAs nanowires grown by MOCVD using Trimethylgallium and arsine exhibits the same activation energy as GaAs layer growth in the same system with assistance of Au particle 45 . Since neither thermodynamics nor chemical reaction kinetics is a rigorous explanation, the role of particle in the one-dimensional growth ofIII-V nanowires is still an open question. The surface reconstruction at the liquid/solid interface might also contribute to selectivity in direction. Nevertheless, an explosion of nanow!fes grown by Au-assisted VLS has been demonstrated using MBE 46 - 48 and MOCVD 49 - 54 . On the other hand, controversy regarding the incorporation of Au during the growth 34 • 55 • 56 leads to the speculation of creating deep levels acting as recombination centers"- Group III atom self-catalyzed VLS growth has thus attracted attention since no foreign metal intermixing is involved in this method 57 -oo. 27 However, the metal droplet places constrain in the growth of axial heterostructure with different group III elements. The droplet usually remains after growth due to the VLS nature and is undesired in some application such as efficient light extraction in LEDs and good ohmic contact in solar cells. Fig. 2.2 shows III-V nanowires grown from Au- and self-catalyzed methods using both MBE and MOCVD. Figure 2. 2 III-V nanowires grown from VLS method. GaAs nanowires grown on Si substrate using (a) MBE4 7 and (b) MOCVD 50 • Ga-self catalyzed nanowires grown on GaAs (lll)B substrate using (c) MBE 59 and ( d) MOCVD by the author. 28 2.2.3 Nanowire growth using SAG method Another approach to grow nanowires is the catalyst-free selective area growth. In this method a thin mask with lithographically patterned openmg is first deposited on the growth substrate. Nanowires can be grown from the openings at controlled positions. The faceting growth mechanism ensures growth to occur preferentially in one direction by suppressing the lateral growth. In most cases, vertical nanowires surrounded by { -ll 0} sidewalls can be obtained when use (lll)A or B substrates. First developed by Fukui's th f III V · d · h G A 36 67-69 I p7o 11 group, grow o many - sem1con uctor nanow1res, sue as a s ' , n ' , In.As 72 • 73 , GaP 74 , In.GaAs 75 - 77 and GaAsP 78 have been successfully achieved using SAG. Core/shell and axial heterostructures have also been widely investigated for functional device applications. Fig 2.3 shows the process steps involved in SAG. In our case, a 20 nm thick dielectric layer is deposited on the growth substrate which is either silicon nitride by Plasma Enhanced Chemical Vapor Deposition (PECVD) or thermally grown silicon dioxide. An array of circular holes are patterned various lithography discussed in section 2.3 followed by etching. Both wet etching using buffered oxide etchant (BOE) and dry etching using CF 4 based Reactive Ion Etching have been used to transfer the pattern from resist layer into the dielectric mask layer. Wet etching is isotropic so the holes in the mask layer are usually larger than those originally in the resist. Thus in most cases we use RIE to maintain the hole size unless otherwise mentioned. After removal of resist and blow dry, samples are immediately loaded into MOCVD reactor chamber for growth. Trimethylgallium (TMG), trimethylaluminum (TMA), arsine (AsH 3 ) and phosphine (PH 3 ) 29 are used as sources for Ga, Al, As and P. Disilane is used as source of n-type dopant Si and dienthylzinc (DEZn) and tetrabromomethane (CBr 4 ) are used as sources of p-type dopant Zn and C. a ... Figure 2. 3 Schematic of SAG process. (a) Growth mask layer deposition, (b) opening generation by lithography, (c) nanowire SAG in MOCVD. Specifically for GaAs nanowires, the growth preferentially occur in (11 l)B direction. Hence GaAs (111 )B substrate is used in order to obtain vertically aligned nano wires. The Optimum growth temperature is between 700 and 790 ° C. Within this window the nanowires have well defined sidewall facets in {-110} cfu·ections due to the much lower growth rate in those directions than that in the vertical (111 )B direction. Lower temperature enhances lateral growth and suppresses vertical growth. Higher temperature results in rapid decomposition which competes with the crystallization process. ill another word, the the1modynamic dtiving force for the growth to occur becomes weaker at elevated temperature. AsH3 partial pressure (PAsHJ) also plays an important role in the growth of nanowires. The relative growth rate on different facets changes with vruying AsH3. Fig 2.4 shows 30 nanowires grown under PA>HJ from 2.J4xJ0- 5 to l.43xJ0- 3 atrn while partial pressure ofTMG is kept constant at 7.56x 10· 7 atrn. It shows at the lowest P A>HJ although the hexagonal cross-section remains, the top is no longer flat Instead of (lll)B surface, tilted {110} facets also emerge. In contrast, under the highest P A>HJ the vertical growth is largely suppressed and the cross-section turns into triangular shape. The growth in the lateral <-110> directions are dominated by the decomposition process. Hence higher P A>HJ suppresses the decomposition and enhances the lateral growth rate_ On the other hand, the growth in the (111 )B direction is dominated by the surface reconstruction that is a strong function of the As surface enrichment Ultra high vacuum study found at low Tg or high PAsH3, (2x2) reconstruction takes place which features a chemically stable As-trimer structure leading to reduced growth rate in (lll)B direction 79 - 84 _ Sarne phase transition on GaAs (lll)B surface during MOCVD growth is also reported 85 _ 31 d A .. ...-....,A • o .. A · - -. 0 • .. . · . 0 ••• ... 0 .;t, .... ... o ./;t, • ~'~ e ~ ·$.·: ¥* .. ·.¥.:v . . .¥ . Figure 2. 4 (a)-(c) 30° tilted SEM images ofGaAs nanowires grown by SAG. Growth temperature is 730 ° C, pressure is 76 Torr, TMG partial pressure is 7.56x10- 7 atm, AsH 3 paitial pressure is (a) 2.14xlo- 5 atm, (b) 2.14xl0 4 atm, and (c) l.43xI0- 3 atm. (d)As trimer model ofGaAs (lll)B (2x2) reconstmction: large open circles denote adsorbed As trimer atoms, small open circles denote first-layer As atoms, and small closed circles denote second-layer Ga atoms. (e) Detailed model ofGaAs (lll)B (-,/19x-,/I9) reconstruction: lai·ge open circles denote top As atoms, medium closed circles denote second-layer Ga atoms, and small open circles denote third-layer threefold-coordinated As atoms. ill SAG, three mass transportation processes detennine the growth rate. One is the precursor species impinging on the (111 )B surface directly :from the vapor phase, which can immediately contribute the growth in the vertical direction. Another other process is surface diffusion. Once the precursors imping on the mask layer surface due to the low reactivity, they will diffuse randomly and settle until reaching the opening in the initial stage or the nanowire surface later. So the sites of nanowires act like a sink of the source molecules. During the diffusion the precursors also go through re-evaporation of finite rate, thus there exists a diffusion length beyond which a source molecule is not able to 32 migrate to the growth location and contribute to the nanow!fe growth. Temperature, pressure, partial pressure of precursor and the surface chemistry of the mask layer surface together determine the diffusion length. Similar to the second mechanism, the third source of growth species is from those impinge on the nanowire sidewalls. Because of the low growth rate on those facets, those molecules also migrate to the top (111 )B surface. Quantitatively speaking if the first mechanism dominates the nanowire height should be proportional to the inverse of the square of nanowire diameter. In another word the total volume is constant which is determined by the surface diffusion within the diffusion length on the mask. If the second mechanism dominates, the nanowire wire should be proportional to the inverse diameter as the area of the side wall is proportional to the diameter. If the third mechanism dominates, then the nanowire height should be independent of the diameter. Figure 2.5 shows nanowires grown from the same run but with different size of initial holes due ebeam dose variation. The average lengths are 963, 881 and 574 nm and the average diameters are 110, 167 and 186 nm. Apparently the relation between the height and diameter does not comply with a single mechanism mentioned above. Multiple mechanisms might contribute to the total growth simultaneously with comparable significance. Need to mention the contribution from the nanowire sidewall and mask surface diffusion might be time variant. When the nanowires are shorter than the diffusion length on the sidewall, the contribution on the side wall is not only proportional to the diameter but also proportional to the height. Once the wires are taller than the sidewall diffusion length species migrating from the mask surface cannot reach the top surface. Further complicating the situation is the incubation time 33 required to fonn nucleus might also differ for different sizes of openings which results in the difference in the net nanowire growth time. d 1000 180 L.. ...... 900 Q) ..c ...... C) Q) Q) 800 160 E ..c co Q) 700 "O L.. '""' ~ ·3 I 140 0 600 ~ ""' 3 c 1! 0 co i 500 120 c z 500 I co z z l ~ 000002 ooooo.i oooooe oiiCXii" 400 1f01.ametef 100 Figure 2. 5 (a)-(c) GaAs nanowires grown by SAG with different diameters and heights. (d) Average diameter and height of each sample with standard deviation. Inset is height vs l/Dimater 2 , which should show linear relationship if volume is constant. 2.3 Patterning techniques for selective area growth 2.3.1 Electron beam lithography Electron beam lithography (EBL) is the practice of scanning a focused beam of electrons to draw custom shapes on a surface covered with an electron sensitive film called an e-beam resist. The electron beam changes the solubility of the resist enabling selective removal of either the exposed or non-exposed regions of the resist by immersing it in a solvent (developer) after exposure. The main advantage ofEBL is it can generate patterns 34 with sub-10 nm resolution. It does not require a physical mask, offering great flexibility to design arbitrary features during experiment. Typical EBL system consists of an electron source, electron accelerator, electron lens system, beam blanker, deflection coils, stigmation control and sample chamber. The electron column needs to be kept under high vacuum to avoid electron scattering. The EBL system installed at USC is called eLine system made by Raith. All the hardware are integrated and controlled by computers which also offer user friendly pattern design software and imaging capability. Poly(methyl methacrylate) (PMMA) and ZEP 520A have been used as positive e-beam resist. Methyl isobutyl ketone (MIBK) and ZED N50 are their developer correspondingly. PMMA offers slightly higher resolution while ZEP 520A has higher sensitivity (shorter writing time) and is more resistant to RlE. The size of a particular hole is determined by both the product of beam current and dwell time at this hole before moving to the next as well as the density of the holes. Electron from adjacent sites can spill over through secondary electron scattering which is called proximity effect and is the main mechanism to limit the resolution of EBL. Small feature displacement is realized by deflecting e-beam without actual stage movement. E-beam is deflected by adjusting the DC voltage of deflection coil and cannot be deflected above certain limit basically because beyond that limit the sample surface is significantly out of focus. The maximum area that can be reached by e-beam through deflection is called writing field. Patterns larger than the writing field are dived into several blocks and exposed field by field through stage movement. Stitching error (discontinuity) occurs at 35 the boundaries of writing fields. The biggest disadvantage of EBL 1s it 1s a time consummg process although it 1s still the most popular technique m the field of nanotechnology and nanosc1ence at research level where people care more about versatility in pattern design than high throughput. 2.3.2 N anoimprint lithography The slow speed of EBL is due to its serial process nature, meanmg each feature 1s exposed m senes m time. Several parallel processes are available to overcome this problem in scenanos patterns are not necessarily to be changed very frequently. Nanoimprint lithography (NIL) is one of these techniques that can offer low cost, high throughput and high resolution. It creates patterns by mechanical deformation of imprint resist and subsequent processes. The imprint resist is typically a monomer or polymer formulation that is cured by heat or UV light. Adhesion between the resist and the template (mold) is controlled to allow proper release. In 1995, Chou et al. named the technique and demonstrated 25 nm resolution by physically deforming a thermoplastic material in a temperature and pressure controlled printing process 86 . It was done using a silicon stamp patterned by electron beam lithography and dry etching. Fig 2.6 shows a schematic of the NIL process. A hard mold that contains that contains nanoscale surface-relief features is pressed into a polymeric material cast on a substrate at a controlled temperature and pressure, thereby creating a thickness contrast in the polymeric material. A thin residual layer of polymeric material 1s intentionally left 36 underneath the mold protrusions, and acts as a soft cushioning layer that prevents direct impact of the hard mold on the substrate and effectively protects the delicate nanoscale features on the mold surface. For most applications, this residual layer needs to be removed by an 0 2 plasma etching process to complete the pattern definition. There are two types of NIL depending on how the resist is cured. The first is called thermoplastic nanoimprint lithography (f-NIL) which is first developed by Chou et al. In this method the mold is pressed into the polymer layer heated up above the glass transition temperature of the polymer. After being cooled down, the mold is separated from the sample and the pattern is left on the rigid resist layer. The second type is called photon nanoimprint lithography (P-NIL) in which a UV curable resist is applied to the substrate. Mold is normally made of material that is transparent to UV light, such as silica. After the mold is pressed into the liquid resist layer, UV illumination turns the resist into solid. a Figure 2. 6 Schematic of selective area growth process. What makes NIL a widely accepted outstanding technology is its ability to produce ultrahigh resolution in an inexpensive way. Figure 2.7a and b shows SEM images of a mold with a pillar array of 10 nm diameter and an imprinted 10 nm hole array in PMMA 87 . Figure 2.7c and d shows successful nanowire growth on NIL patterned substrates using Au- and self-catalyzed VLS growth 64 • 88 . 37 -.t l+-10 nm ~ ~10 nm Figure 2. 7 (a) SEM image of a fabricated mold with a 10 nm diameter atrny. (b) SEM image of hole airnys imprinted in PMMA by using such a mold 87 • (c) SEM image of NIL patterned InP nanowire array grown by Au-catalyze VLS 88 • (d) SEM image ofNIL patterned GaAs nanowire array grown by Ga self-catalyzed growth 64 • 2.3.3 Nanosphere lithography Nanosphere lithography (NSL) is another patterning technique applicable to making SAG mask. It is low cost and high throughput. Compating to other lithographies, like photolithography and nanoimprint lithography, which requires fabrication of a fixed mask or mold, NSL offers some tunability in the spacing and diameter of holes by controlling the initial and final diameter of the nanospheres. The only drawback is the slight 38 disordering observed in NSL pattern due to size variation of nanospheres in the colloid. However, in section 2.4 we will show this slight ape1iodic feature does not degrade the light absorption performance which is very critical to solar cell applications. G3As(111J8 substrate Nanosphere deposition MOCVD Nanowire growth b) Oxygen Plasma Dry etch and metal removal" c) Metal Deposition d) •• •••• •••••• •••••••• ••••••••• ••••••••• ••••••• ••••• ••• • Nanosphere lift-off Figure 2. 8 Schematic diagram of nanosphere lithography and nanowire growth. (a) Spin coating nanospheres on the substrate. (b) Oxygen plasma to reduce the size of the nanospheres, which will determine the diameter of the nanowires after growth. (c) Metal (Al or Fe) deposition onto the nanospheres. (d) Dissolution of nanospheres in chloroform with the aid of sonication. ( e) Transfer of the pattern to the underlying substrate by dry etching and removal of the metal mask by HCl wet etching. (f) Nanowire growth by SA-MOCVD. Fig 2.8 shows a schematic diagram of the scalable nanowire growth process, which includes nanopatterning using nanosphere lithography and nanowire growth using the SA-MOCVD technique. The detailed procedure of scalable catalyst-free growth of semiconducting nanowires begins with PECVD deposition of30 nm ofSiN (measured by ellipsometer) on a substrate of interest. The SiN acts as a mask for the selected area 39 growth of nanowires. Next, the substrate/SiN is cleaned by oxygen plasma, making it hydrophilic. A 10 wt% aqueous (aq.) solution of polystyrene nanospheres (Thermo Scientific) of desired size is spin coated onto the hydrophilic substrate (Fig 2.8a). The spin speed was tuned to obtain uniform monolayer coverage of nanospheres on the substrate; higher spin speed resulted in islands of nanospheres, while lower spin speeds resulted in multilayer stacking of nanospheres. Next, the nanosphere size is reduced by dry etching with 0 2 plasma. The power, partial pressure and time of etching were optimized to obtain the desired size (Fig 2.8b). Aluminum or iron is then deposited on the SiN using the assembled nanospheres as a shadow mask (Fig 2.8c), and the nanospheres are dissolved in chloroform with the aid of sonication (Fig 2.8d). Finally, the pattern is transferred to the underlying substrate by dry etching SiN using CF 4 plasma followed by removal of the metal mask using wet etching in hydrochloric acid (Fig 2.8e). The pattern is then ready for nanowire growth using MOCVD. Highly aligned vertical nanowires of desired length were obtained by controlling the growth conditions (Fig 2.8f). 40 Figure 2. 9 Demonstration of scalable GaAs nanowire synthesis on a GaAs substrate using NSL technique. (a) Photograph of the spin-coated polystyrene nanospheres on a wafer scale. SEM images of nanosphere assembly at two random locations on the wafer, showing monolayer assembly of the nanospheres. (b) SEM image of nanospheres after size reduction by oxygen plasma. (c) SEM image of the pattern after metal deposition, nanosphere lift-off, and dry etching, just before the nanowire growth. (d) Vertical GaAs nano wires grown on the NS L pattern showing a hexagonal cross section having six { -110} side facets. We demonstrate this scalable nanowire synthesis approach by fabricating vertical GaAs nanowires on a GaAs (111) B substrate. The polystyrene nanospheres used in these studies were nearly monodisperse with a diameter of lOOnm, 200nm or 360nm. Stable nanosphere suspensions in water (10 wt%, used as received) were spin-coated onto a hydrophilic GaAs/SiN wafer as shown in Fig 2.9a. As the solvent (water) evaporates, capillary forces draw the nanospheres together, and the nanospheres crystallize in a hexagonally close-packed pattern. The assembly of nanospheres was inspected with a SEM, and images at two random locations are shown in Fig 2.9a. The images show monolayer assembly of the nanospheres, with -90% of the substrate successfully coated with large domains of defect-free nanosphere packings. As in all naturally occurring 41 crystals, some regions include point defects (vacancies), line defects (slip dislocations), and polycrystalline grains. Typical domain sizes are in the 10-100 µm 2 range and increase with the size of the nanospheres. Fig 2.9b shows the SEM image of the nanosphere assembly after oxygen plasma, showing uniform reduction of the nanosphere size. Following size reduction, a thin film of 30 nm Fe is deposited by thermal evaporation from a source normal to the substrate through the nanosphere mask. After metal deposition, the nanosphere mask is removed by sonicating the entire sample in chloroform, leaving behind a mesh-like pattern of Fe on the substrate. The pattern is then transferred to the underlying GaAs substrate by dry etching the SiN using CF 4 RIE, as shown in Fig 2.9c. GaAs nanowires are grown using SAG MOCVD and only form in the openings in the SiN (Fig 2.9d). Nanowires were grown both with and without a Fe mask (etched by HCl), and no difference in the morphology of the nanowires was observed. Details of the nanowire growth are similar to those in Ref. 36 and have been discussed in section 2.2.3. The total growth time was varied depending on the desired length of the nanow1res. Our technique can be contrasted with other NSL methods for nanowire growth, in which the nanosphere pattern is used as a mask to deposit metal catalyst in the interstitial regions 40 . In that case, any area with defects in the nanosphere packing (vacancies, line defects, grain boundaries, etc) is covered by metal, ultimately resulting in bulk material growth or a large distribution in nanowire diameter. In our case, the defect sites in the nanosphere pattern are covered by SiN and do not contribute to the growth of nanowires, 42 only the areas which were initially covered with nanospheres. In principle, the method of scalable nanowire growth we propose is not restricted to GaAs nanowires on GaAs substrates and can be used to grow a vaiiety of III-V nanowires on vaiious semiconducting substrates. a) GaAs Nanowire using EBL pattern b) GaAs Nanowire using NSL pattern Figure 2. 10 Structural quality comparison of GaAs nanowires grown using (a) electron beam lithography and (b) nanosphere lithography, using HRTEM. Inset in each case shows the twin planes and diffi:action pattern conesponding to Zinc-blende strncture. Both images show similai· distribution of twin planes and crystal structures. The airnw indicates the twin planes present in the nanowires. MOCVD nanowire growth is a complicated process involving migration of precursors which can be affected by the patterning defects at the grain boundaries in nanosphere lithography. As it has been shown that crystal stmcture of nanowires is closely conelated to growth conditions 57 • 89 - 91 , we believe it's important to compare the crystal structures and defects of nanowires using NSL and EBL. We used GaAs nanowires grown on GaAs (111) B substr·ates patterned by electron beam lithography to compare the stmctural quality of nanowires to those grown by NSL. High-resolution transmission electron 43 microscopy (HRTEM) measurement was carried out on the nanowires using a JEOL 200 Ke V TEM. Nanowires were separated from the substrate by sonication in methanol and were dispersed onto carbon-supporting Cu grids. Fig 2.1 Oa and b show the TEM images of the nanowires synthesized using electron beam lithography and nanosphere lithography, respectively. The bright and dark regions with 2-30 rnn periods along the growth direction show that the nanowires contain many stacking defects. The insets in Fig 2.1 Oa and b show the diffraction pattern for nanowires grown using EBL and NSL patterning, respectively. It consists of basic Bragg spots coinciding with the zinc-blende pattern observed from the direction and additional spots which are line-symmetric to the basic ones with respect to the [111] axis. The crystal structure of the GaAs nanowire is thus a rotational twin of the zinc-blende type with a [111] twin axis. Therefore, the bright and dark regions in the TEM image in Fig 2.10 result from the difference in electron diffraction conditions on the two sides of the twin boundary. The density of these rotational twins was similar in nanowires produced by both NSL and EBL patterning. 2.4 Optical absorption study of nanowire array grown from pattern generated by nanos phere lithography In order to quantify the broadband anti-reflection property of the GaAs NW arrays, we evaluated the weighted reflection loss (A.R.L.), defined as: 44 Here A, is wavelength, and Ag is the wavelength corresponding to the band gap of GaAs (867nm). !(A.) is the AMl.5 G direct and circumsolar solar irradiance spectrum, andRrE(A.) and RrM(A.) ai-e the surface reflectance for the TE and TM polruization, respectively. 1bis figure of merit gives the loss in the ultimate efficiency 92 due to the reflection off the nanowire rurny. The ultimate efficiency is defined under the assumption that each absorbed photon with energy greater than the band gap produces one and only one electron hole pair with energy hc!A.g. a) ~ ·*· y~ . x b) 100 'E 200 c: - \J 300 ,_ <1> .... E 400 ro 0 500 600 1 2 3 4 aid Figure 2. 11 (a) Schematic of the GaAs nanowire array. (b) Calculated weighted reflection loss as a function of nanowire diameter and lattice constant - diameter ratio for GaAs nanowires of 130 nm in height. A range of optimum configurations can be extracted from the graph and used to fabricate GaAs nanowires using NS L. The simulated strncture is a vertically-aligned GaAs nanowue rurny on top of a semi-infinite GaAs substrate, as shown in Fig 2.lla. The array consists ofnanowires with diameter d and lattice constant a ruTanged in a hexagonal lattice and surrounded by air. The nanowire height is 130 nm. The structure is illuminated by normally-incident sunlight, modeled by the AM 1.50 solar spectrum. The optical constants ofGaAs used in 45 the calculation are taken from Ref. 93. For GaAs solar cells, the wavelength range of interest is from 400nm, where the solar irradiance is negligibly small, to 867nm, the wavelength corresponding to the band gap of GaAs. We used the scattering matrix method implemented in the ISU-TMM simulation package 94 • 95 to calculate the average reflectance as a function of wavelength across the solar spectrum, defined as the average between the reflectance for the incidence light polarized along the rM direction and IT direction. Fig 2.11 b shows the weighted reflection loss as a function of nanowire diameter and the lattice constant - diameter ratio (aid). Two trends can be observed. Firstly, for fixed aid, there exists an optimal range of nanowire diameters to minimize the reflection loss. Secondly, for fixed nanowire diameter, there exists an optimal range of lattice constants where the weighted reflection loss is minimal. For either small or large al d, the reflection loss is very high. This trend can be qualitatively explained as follows. For small aid (large filling fraction), the nanowires are almost touching each other and form a continuous film. The optical properties of the structure resemble those of a semi-infinite GaAs substrate and exhibit high reflection loss. On the other hand, for large aid (small filling fraction), the nanowires are far apart from each other. Therefore, the incident light mainly interacts with the underlying substrate instead of the nanowires, also resulting in high reflection loss. There exists a fairly large region of parameter space, with a diameter of 150nm-400rnn and aid of 1.5-3, that can produce a reflection loss value of 7% or less with the minimum around d~300nm and aid~ 1.9. 46 1 11 2 11 3 111r11 c) 25 ---· Experiment.al measurement 20 - Simulation using SEM images ~ -;:; 15 - Simulation of periodic array C> ·= .... QI 10 ; QI 0:: 5 0 400 500 600 700 800 Wavelength (nm) b) 25 '*' 20 g 15 ·~ 10 ; ti 5 0 d) ..-. ";' E c f< ~ E ~ c 0 .. u Q) c;:: Q) 0::: 400 500 600 700 800 Wavelength (nm) a~..--------.,,,---,--.,....,---------. - Experimental measurement - Simulation usirg SEM images - Simulation c:A periodc array a25 a20 a15 a 10 Total rel Po.....,. den!ity from 400 to800rnt aos Exp Meas= 44.77Wm. 2 Sim Cl{ SEM• 38. 18W'm· 2 aoo Sim ot periodic= 46.02Wm • 2 400 500 600 700 000 Wavelength (nm) Figure 2. 12 Comparison between experimentally measured and simulated reflectance spectrum for the as-fabricated GaAs nanowire array sample (a=360nm, d=l95nm). (a) SEM image of the top view of the GaAs nanowires grown using NS L. Four regions extracted from the SEM image (1, 2, 3 and 4 shown in the inset) were selected to simulate the reflection profile across the visible spectrum. (b) Corresponding reflectance spectra using section 1, 2, 3 and 4. (c) Plot ofreflectance spectra for GaAs nanowires grown on a GaAs substrate using NSL (experiment), simulated periodic nanowire array with a=360 run and d=195 run, and average of simulated nanowire arrays in (b). (d) Reflected power density under AMl.5 spectrum showing experimental measurement of the GaAs nanowire array grown on NSL patterned substrate, simulation of the periodic nanowire array, and simulation using nanowire arrays with SEM images in (a). The total integrated reflected power density is also calculated. In the simulation, we assumed that the nanowire array is a perfectly periodic structure with uniform spacing, diameter and height. However, in reality, the fabricated sample always deviates from the ideal structure, due to variation in nanowire spacing, diameter, and height, as well as the finite grain size. In order to more accurately model the as-fubricated sample, we import the actual size, shape, and arrangement of nanowires in a 47 2µm by 2µm region of the sample from SEM pictures into the Lumerical FDTD solutions simulation package. We compare four different sections of the sample, as shown in Fig 2.12a. We use periodic boundary conditions in both X and Y directions and assume that the nanowires have a uniform height of 130 nm. Our simulations take into account the variation in the nanowire diameter and spacing, and hence are more accurate than the simulation for the perfectly periodic structure. The simulated results are plotted in Fig 2.12b and c. We can observe in Fig 2.12b that the average reflectance spectra for the four different sections in the SEM picture are similar to each other, with only a slight difference in value. Furthermore, as we can see in Fig 2.12c, the experimentally measured spectrum (using a Perkin Elmer Lambda 950 integrating sphere) shares similar features with the simulated spectrum for both the perfectly periodic structure and the as-fabricated structures, with slight differences in long wavelengths. We believe the deviation of simulation from experimental measurements is likely due to the fact that the simulation used a perfectly collimated incident light beam, while the incident beam for experimental measurements cannot be perfectly collimated. From Fig 2.14, the experimental spectra typically exhibit shallower and broader features than the theoretical spectra, which is consistent with a wider angular spread in the illumination angles. Fig 2.12d shows the reflection loss in the wavelength range between 400 and 800 nm under illumination of AMl.5 ASTMGl 73 spectrum and the integrated reflection loss is 44.77 W/m 2 for our sample, which is close to that from a periodic array of nanowires with the same dimension and pitch (46.02 W/m 2 ). 48 a) b) 40 35 - 30 e 25 c: 20 0 -~ 15 QI ;;::: QI 10 a: 5 0 c) 40 d) 35 ...... 30 E c: *- -25 c: 0 "B 20 "' & 15 - Experimental me·asurement 1< "' . E ~ c: 0 10 - Simulation using SEM images ;; u 5 - Simulation of periodic array a> i;:: a> a:: 0 400 500 600 700 800 Wavelength (nm) 400 0.6 0.5 0.4 0.3 0.2 0 .1 0.0 - sectionl - section2 - Sec-tion3 500 600 700 800 Wavelength (nm) -Experiment measurement - Simulation using SEM from 400nm 1o 800nm Exp Meas= 132.2 .3W'm " 2 Sim of SEM= 134.41W'm" 2 Sim of periodic = 130.54W'm· 2 400 500 600 700 Wavelength (nm) 800 Figure 2. 13 Comparison betvveen experimentally measured and simulated reflectance spectrum for a GaAs nanowire array sample (a = 890nm, d = 450nm) fabricated using NSL. (a) SEM image of the nanowires (top view). Three regions extracted from the SEM image (labelled 1, 2 and 3) are shown in the inset, each with dimensions of 5µm by 5µrn. (b) Corresponding reflectance spectra using section 1, 2 and 3. (c) Plot of reflectance spectra for GaAs nanowires grown on a GaAs substrate using NSL (experiment), simulated periodic nanowire array with a = 890 nm and d = 450 nm, and average of simulated nanowire arrays in Figure 6(b). (d) Reflected power density under AMl.5 spectrum showing experimental measurement of the GaAs nanowire array grown on NSL patterned substrate, simulation of the periodic nanowire array, and simulation using nanowire arrays with SEM images in Figure 6(a). The total integrated reflected power density is also calculated. We have also compared simulations and measurements of reflection for a nanowire array with larger pitch. The pattern for nanowire growth was fabricated using nanospheres with a diameter of 890nm, which were reduced to 450nm after assembly. This resulted in nanowires with a diameter of 450nrn and a pitch of 890nrn as shown in Fig 2.13a. The height of the nanowires was ~130nm. According to Fig 2.11, we would expect more reflection from this nanowire array. As evident from the SEM image, the nanowires have 49 more size and shape variation than those with 360nm pitch shown in Fig 2.12. Fig2.13b shows the simulated reflection spectra using three 5µm by 5µm sections from the SEM image of Fig 2.13a. The spectra ai-e very similar to each other. In Fig 2.13c, we show the comparison between experimental measurements, simulated values from the pe1iodic array (using a= 890nm, d = 450nm and height= 130nm), and the averaged simulated values in Fig 2.13b. The three quantities match closely and again the integrated reflection loss of our sample (132.23 W/m 2 )is ve1y close to that from a periodic array ofnanowires with the same dimension and pitch (130.54 W/m 2 ) as shown in Fig 2.13d. As expected, reflection values are higher than those in Fig 2.12. a) 20min 30min -· 35 lS - 30 30 g,, g,. -· •• c c .2 zo ·~ 20 to .. 15 ~lS ;;:: : 10 .. - Simvl:ldon a: 10 - fxperlme.n1 •• ~~ --- 400 soo 600 700 800 300 •OO soo 600 100 800 Wavelength (nm) Wavelength (nm) b) 20min 30min lS .. 30 10 - Sim.IJl.)tion - Oi:perimellt - ZS -" '/I. .... c20 ~20 ~ 15 ~lS ~ 10 .. ;: 10 .. .. a: s a: s 300 400 500 600 700 800 )<)() 400 soo 600 100 800 Wavelength (nm) Wavelength (nm) Figure 2. 14 Comparison of reflection spectra between simulation and experimental results of GaAs nanowires array grown on EBL patterned GaAs (11 l)B substrate. (a) 360 nm pitch, grown for 20 min and 30 min. (b) 890 nm pitch, grown for 20 min and 30 min. We have also compared simulations to experimental results for perfectly periodic arrays, patterned using EBL. We synthesized GaAs nanowires on EBL-pattemed GaAs (111) B 50 substrates with 360nm and 890nm pitch. Two sets of nanowires with growth times of 20 minutes (corresponding to a height of 500nm for 360nm pitch and 1350nm for 890nm pitch) and 30 minutes (corresponding to a height of 750nm for 360nm pitch and 1800nm for 890nm pitch) were produced. The experimental and simulated spectra along with corresponding SEM images are shown in Figure S2. The simulated and experimental curves show similar features, with slight variation in the absolute values. The results demonstrate that reliable and realistic predictions of reflection spectra can be obtained from simulation. Simulations can thus be used as a tool to predict optimum nanowire array designs and as guidance for nanowire fabrication and growth. 2.5 Summary To summarize, in this chapter we demonstrated SAG as a neat technique to grow nanow!fes. Nanowire density and size can be confidently controlled by the pattern geometry on the growth mask. As a catalyst-free epitaxy, SAG can absolutely avoid contamination from metal particles used in VLS. One-dimensional growth can be achieved with faceting growth mechanism and is sensitive to growth conditions including temperature, pressure, and precursor partial pressure. Different nano-patterning techniques have been discussed with their advantages and disadvantages. EBL is most widely used in research level due to its flexibility in pattern design and precision. Other parallel techniques are more suitable when high throughput is concerned. At the end of this chapter we focused on a versatile, scalable fabrication technique that used NSL 51 patterning for synthesis of vertically aligned, catalyst-free nanowires using SA-MOCVD. The presence of defects in the assembled nanosphere pattern resulted in negligible effect on the optical properties of the array, as characterized by the reflection loss spectrum. There is thus 'no need' to fabricate highly periodic structures in order to achieve good optical performance. As an alternative to costly, time-consuming patterning methods such as EBL, NSL will undoubtedly be a powerful technique to produce scalable nanowire arrays for photovoltaic applications. 52 References 1. Huang, Y.; Duan, X.; Cui, Y.; Lauhon, L. J.; Kirn, K.-H.; Lieber, C. M. Science 2001, 294, (5545), 1313-1317. 2. Gudiksen, M. S.; Lauhon, L. J.; Wang, J.; Smith, D. C.; Lieber, C. M. Nature 2002, 415, (6872), 617-620. 3. Lauhon, L. J.; Gudiksen, M. S.; Wang, D.; Lieber, C. M. Nature 2002, 420, (6911), 57-61. 4. Johnson, J. C.; Choi, H.-J.; Knutsen, K. P.; Schaller, R. D.; Yang, P.; Saykally, R. J. Nat Mater 2002, 1, (2), 106-110. 5. Parkinson, P.; Lloyd-Hughes, J.; Gao, Q.; Tan, H. H.; Jagadish, C.; Johnston, M. B.; Herz, L. M. Nano Lett. 2007, 7, (7), 2162-2165. 6. Thelander, C.; FrobergFroberg, L. E.; Rehnstedt, C.; Samuelson, L.; Wemersson, L. E. Electron Device Letters, IEEE 2008, 29, (3), 206-208. 7. Tornioka, K.; Yoshimura, M.; Fukui, T. Nature 2012, 488, (7410), 189-192. 8. Dayeh, S. A.; Aplin, D. P.R.; Zhou, X.; Yu, P. K. L.; Yu, E.T.; Wang, D. Small 2007, 3, (2), 326-332. 9. Egard, M.; Johansson, S.; Johansson, A. C.; Persson, K. M.; Dey, A. W.; Borg, B. M.; Thelander, C.; Wemersson, L. E.; Lind, E. Nano Lett. 2010, 10, (3), 809-812. 10. Kasai, S.; Hasegawa, H. Physica E: Low-dimensional Systems and Nanostructures 2002, 13, (2-4), 925-929. 11. Thelander, C.; Martensson, T.; Bjork, M. T.; Ohlsson, B. J.; Larsson, M. W.; Wallenberg, L. R.; Samuelson, L. Appl. Phys. Lett. 2003, 83, (10), 2052-2054. 12. Bryllert, T.; Wemersson, L. E.; Froberg, L. E.; Samuelson, L. Electron Device Letters, IEEE 2006, 27, (5), 323-325. 13. Ng, H. T.; Han, J.; Yamada, T.; Nguyen, P.; Chen, Y. P.; Meyyappan, M. Nano Lett. 2004, 4, (7), 1247-1252. 14. Thelander, C.; Nilsson, H. A.; Jensen, L. E.; Samuelson, L. Nano Lett. 2005, 5, (4), 635-638. 15.Svensson, C. P. T.; Thomas, M.; Johanna, T.; Christina, L.; Michael, R.; Dan, H.; Lars, S.; Jonas, 0. Nanotechnology 2008, 19, (30), 305201. 16. Guo, W.; Zhang, M.; Banerjee, A.; Bhattacharya, P. Nano Lett. 2010, 10, (9), 3355-3359. 17. Mayer, B.; Rudolph, D.; Schnell, J.; Morkiitter, S.; Winner!, J.; Treu, J.; Miiller, K.; Bracher, G.; Abstreiter, G.; Koblrniiller, G.; Finley, J. J. Nat Commun 2013, 4. 18. Saxena, D.; Mokkapati, S.; Parkinson, P.; Jiang, N.; Gao, Q.; Tan, H. H.; Jagadish, C. Nat Photon 2013, 7, (12), 963-968. 19. Yan, R.; Gargas, D.; Yang, P. Nat Photon 2009, 3, (10), 569-576. 20. Huang, M. H.; Mao, S.; Feick, H.; Yan, H.; Wu, Y.; Kind, H.; Weber, E.; Russo, R.; Yang, P. Science 2001, 292, (5523), 1897-1899. 21. Qian, F.; Li, Y.; Gradecak, S.; Park, H.-G.; Dong, Y.; Ding, Y.; Wang, Z. L.; Lieber, C. M. Nat Mater 2008, 7, (9), 701-706. 22. Duan, X.; Huang, Y.; Agarwal, R.; Lieber, C. M. Nature 2003, 421, (6920), 241-245. 23. Pettersson, H.; Triigardh, J.; Persson, A. I.; Landin, L.; Hessman, D.; Samuelson, L. 53 Nano Lett. 2006, 6, (2), 229-232. 24. Seyedi, M. A.; Yao, M.; O'Brien, J.; Wang, S. Y.; Dapkus, P. D. Appl. Phys. Lett. 2013, 103, (25), 251109. 25. Tomioka, K.; Motohisa, J.; Hara, S.; Hiruma, K.; Fukui, T. Nano Lett. 2010, 10, (5), 1639-1644. 26. Mariani, G.; Scofield, A. C.; Hung, C.-H.; Huffaker, D. L. Nat Commun 2013, 4, 1497. 27. Wallentin, J.; Anttu, N.; Asoli, D.; Huffman, M.; Aberg, I.; Magnusson, M. H.; Siefer, G.; Fuss-Kailuweit, P.; Dimroth, F.; Witzigmann, B.; Xu, H. Q.; Samuelson, L.; Deppert, K.; Borgstrom, M. T. Science 2013, 339, (6123), 1057-1060. 28. Cui, Y.; Wang, J.; Plissard, S. R.; Cavalli, A.; Vu, T. T. T.; van Veldhoven, R. P. J.; Gao, L.; Trainor, M.; Verheijen, M. A.; Haverkort, J. E. M.; Bakkers, E. P. A. M. Nano Lett. 2013, 13, (9), 4113-4117. 29. Hu, S.; Chi, C.-Y.; Fountaine, K. T.; Yao, M.; Atwater, H. A.; Dapkus, P. D.; Lewis, N. S.; Zhou, C. Energy & Environmental Science 2013, 6, (6), 1879-1890. 30. Yao, M.; Madaria, A. R.; Chi, C.; Huang, N.; Lin, C.; Povinelli, M. L.; Dapkus, P. D.; Zhou, C. In Scalable synthesis of vertically aligned, catalyst-free gallium arsenide nanowire arrays: towards optimized optical absorption, SPlE Defense, Security, and Sensing, Baltimore, 2012; Baltimore, p 837314. 31. Wagner, R. S.; Ellis, W. C. Appl. Phys. Lett. 1964, 4, (5), 89-90. 32. Bullis, W. M. Solid-State Electronics 1966, 9, (2), 143-168. 33. Breuer, S.; Ffiiller, C.; Flissikowski, T.; Brandt, O.; Grahn, H. T.; Geelhaar, L.; Riechert, H. Nano Lett. 2011, 11, (3), 1276-1279. 34. Allen, J.E.; Hemesath, E. R.; Perea, D. E.; Lensch-Falk, J. L.; LiZ.Y; Yin, F.; Gass, M. H.; Wang, P.; Bleloch, A. L.; Palmer, R. E.; Lauhon, L. J. Nat Nano 2008, 3, (3), 168-173. 35. Shi, W. S.; Zheng, Y. F.; Wang, N.; Lee, C. S.; Lee, S. T. Adv. Mater. 2001, 13, (8), 591-594. 36. Noborisaka, J.; Motohisa, J.; Fukui, T. Appl. Phys. Lett. 2005, 86, (21), 213102. 37. Haynes, C. L.; Van Duyne, R. P. The Journal of Physical Chemistry B 2001, 105, (24), 5599-5611. 38. Hulteen, J. C.; Van Duyne, R. P. Journal of Vacuum Science & Technology A 1995, 13, (3), 1553-1558. 39. Hulteen, J. C.; Treichel, D. A.; Smith, M. T.; Duval, M. L.; Jensen, T. R.; Van Duyne, R. P. J. Phys. Chem. B 1999, 103, (19), 3854-3863. 40. Fuhrmann, B.; Leipner, H. S.; Hoche, H.-R.; Schubert, L.; Werner, P.; Gosele, U. Nano Lett. 2005, 5, (12), 2524-2527. 41. Madaria, A. R.; Yao, M.; Chi, C.; Huang, N.; Lin, C.; Li, R.; Povinelli, M. L.; Dapkus, P. D.; Zhou, C. Nano Lett. 2012, 12, (6), 2839-2845. 42. Givargizov, E. I. J. Cryst. Growth 1975, 31, (0), 20-30. 43. Yazawa, M.; Koguchi, M.; Muto, A.; Ozawa, M.; Hiruma, K. Appl. Phys. Lett. 1992, 61, (17), 2051-2053. 44. Levitt, A. P., Whisker technology. Wiley-Interscience: 1970. 45. Borgstrom, M.; Deppert, K.; Samuelson, L.; Seifert, W. J. Cryst. Growth 2004, 260, 54 (1-2), 18-22. 46. Ihn, S.-G.; Song, J.-I.; Kirn, Y.-H.; Lee, J. Y. Appl. Phys. Lett. 2006, 89, (5), 053106. 47. Ihn, S.-G.; Song, J.-I.; Kirn, T.-W.; Leern, D.-S.; Lee, T.; Lee, S.-G.; Koh, E. K.; Song, K. Nano Lett. 2007, 7, (1 ), 39-44. 48. Paek, J. H.; Nishiwaki, T.; Yamaguchi, M.; Sawaki, N. physica status solidi (c) 2008, 5, (9), 2740-2742. 49. Martensson, T.; Svensson, C. P. T.; Wacaser, B. A.; Larsson, M. W.; Seifert, W.; Deppert, K.; Gustafsson, A.; Wallenberg, L. R.; Samuelson, L. Nano Lett. 2004, 4, (10), 1987-1990. 50. Bao, X.-Y.; Soci, C.; Susac, D.; Bratvold, J.; Aplin, D. P.R.; Wei, W.; Chen, C.-Y.; Dayeh, S. A.; Kavanagh, K. L.; Wang, D. Nano Lett. 2008, 8, (11), 3755-3760. 51. Aamoud, L. R.; Marcel, A. V.; Olaf, W.; Stacey, S.; Harry, W.; Erik, P. A. M. B. Nanotechnology 2006, 17, (11), S271. 52. Huang, H.; Ren, X.; Ye, X.; Guo, J.; Wang, Q.; Yang, Y.; Cai, S.; Huang, Y. Nano Lett. 2009, 10, (1 ), 64-68. 53.Kang, J. H.; Gao, Q.; Joyce, H.J.; Tan, H. H.; Jagadish, C.; Kirn, Y.; Choi, D. Y.; Guo, Y.; Xu, H.; Zou, J.; Fickenscher, M.A.; Smith, L. M.; Jackson, H. E.; Yarrison-Rice, J. M. Nanotechnology 2010, 21, (3), 035604. 54. Mohseni, P. K.; Maunders, C.; Botton, G. A.; LaPierre, R. R. Nanotechnology 2007, 18, (44), 445304. 55. Bar-Sadan, M.; Barthel, J.; Shtrikrnan, H.; Houben, L. Nano Lett. 2012, 12, (5), 2352-2356. 56. Perea, D. E.; Allen, J. E.; May, S. J.; Wessels, B. W.; Seidman, D. N.; Lauhon, L. J. Nano Lett. 2005, 6, (2), 181-185. 57. Krogstrup, P.; Popovitz-Biro, R.; Johnson, E.; Madsen, M. H.; Nygard, J.; Shtrikrnan, H. Nano Lett. 2010, 10, (11), 4475-4482. 58. Munshi, A. M.; Dheeraj, D. L.; Todorovic, J.; van Helvoort, A. T. J.; Wernan, H.; Firnland, B.-0. J. Cryst. Growth 2013, 372, (0), 163-169. 59. Cirlin, G. E.; Dubrovskii, V. G.; Sarnsonenko, Y. B.; Bouravleuv, A. D.; Durose, K.; Proskuryakov, Y. Y.; Mendes, B.; Bowen, L.; Kaliteevski, M.A.; Abram, R. A.; Zeze, D. Phys. Rev. B 2010, 82, (3), 035302. 60. Sandra, J. G.; Jonathan, P. B.; Ray, R. L. Semicond. Sci. Technol. 2013, 28, (10), 105025. 61. Fauzia, J.; Vincenzo, G.; Silvia, R.; Faustino, M. Nanotechnology 2008, 19, (27), 275711. 62. Paek, J. H.; Nishiwaki, T.; Yamaguchi, M.; Sawaki, N. physica status solidi (c) 2009, 6, (6), 1436-1440. 63. Plissard, S.; Dick, K. A.; Wallart, X.; Caroff, P. Appl. Phys. Lett. 2010, 96, (12), -. 64. Munshi, A. M.; Dheeraj, D. L.; Fauske, V. T.; Kirn, D. C.; Huh, J.; Reinertsen, J. F.; Ahtapodov, L.; Lee, K. D.; Heidari, B.; van Helvoort, A. T. J.; Firnland, B. O.; Wernan, H. Nano Lett. 2014, 14, (2), 960-966. 65. Plissard, S.; Dick, K., A.; Larrieu, G.; Godey, S.; Addad, A.; Wallart, X.; Caroff, P. Nanotechnology 2010, 21, (38), 385602. 66. Kwoen, J.; Watanabe, K.; Iwamoto, S.; Arakawa, Y. J. Cryst. Growth 2013, 378, (0), 55 562-565. 67. Motohisa, J.; Noborisaka, J.; Takeda, J.; Inari, M.; Fukui, T. J. Cryst. Growth 2004, 272, (1-4), 180-185. 68. Ikejiri, K.; Noborisaka, J.; Hara, S.; Motohisa, J.; Fukui, T. J. Cryst. Growth 2007, 298, (0), 616-619. 69. Ikejiri, K.; Sato, T.; Yoshida, H.; Hiruma, K.; Motohisa, J.; Hara, S.; Fukui, T. Nanotechnology 2008, 19, (26), 265604. 70. Mohan, P.; Motohisa, J.; Fukui, T. Nanotechnology 2005, 16, (12), 2903. 71. Ikejiri, K.; Kitauchi, Y.; Tomioka, K.; Motohisa, J.; Fukui, T. Nano Lett. 2011, 11, (10), 4314-4318. 72. Tomioka, K.; Mohan, P.; Noborisaka, J.; Hara, S.; Motohisa, J.; Fukui, T. J. Cryst. Growth 2007, 298, (0), 644-647. 73. Akabori, M.; Sladek, K.; Hardtdegen, H.; Schiipers, T.; Griitzmacher, D. J. Cryst. Growth 2009, 311, (15), 3813-3816. 74. Tomioka, K.; Ikejiri, K.; Tanaka, T.; Motohisa, J.; Hara, S.; Hiruma, K.; Fukui, T. J. Mater. Res. 2011, 26, (17), 2127-2141. 75. Akabori, M.; Takeda, J.; Motohisa, J.; Fukui, T. Nanotechnology 2003, 14, (10), 1071. 76. Sato, T.; Motohisa, J.; Noborisaka, J.; Hara, S.; Fukui, T. J. Cryst. Growth 2008, 310, (7-9), 2359-2364. 77. Sato, T.; Kobayashi, Y.; Motohisa, J.; Hara, S.; Fukui, T. J. Cryst. Growth 2008, 310, (23), 5111-5113. 78. Fujisawa, S.; Sato, T.; Hara, S.; Motohisa, J.; Hiruma, K.; Fukui, T. Jpn. J. Appl. Phys. 2011, 50, (4S), 04DH03. 79. Biegelsen, D. K.; Bringans, R. D.; Northrup, J. E.; Swartz, L. E. Phys. Rev. Lett. 1990, 65, (4), 452-455. 80. Thornton, J.M. C.; Woolf, D. A.; Weightman, P. Appl. Surf Sci. 1998, 123-124, (0), 115-119. 81. Rajkumar, K. C.; Chen, P.; Madhukar, A. J. Appl. Phys. 1991, 69, (4), 2219-2223. 82. Woolf, D. A.; Westwood, D. I.; Williams, R. H. Appl. Phys. Lett. 1993, 62, (12), 1370-1372. 83. Avery, A. R.; Tok, E. S.; Jones, T. S. Surf Sci. 1997, 376, (1-3), L397-L402. 84. Chen, P.; Rajkumar, K. C.; Madhukar, A. Appl. Phys. Lett. 1991, 58, (16), 1771-1773. 85. Nishida, T.; Uwai, K.; Kobayashi, Y.; Kobayashi, N. Jpn. J. Appl. Phys. 1995, 34, (12R), 6326. 86. Chou, S. Y.; Krauss, P.R.; Renstrom, P. J. Science 1996, 272, (5258), 85-87. 87. Chou, S. Y.; Krauss, P. R.; Zhang, W.; Guo, L.; Zhuang, L. J. Vac. Sci. Technol., B 1997, 15, (6), 2897-2904. 88. Martensson, T.; Carlberg, P.; Borgstrom, M.; Montelius, L.; Seifert, W.; Samuelson, L. Nano Lett. 2004, 4, (4), 699-702. 89. Dick, K. A.; Caroff, P.; Bolinsson, J.; E. Messing, M.; Johansson, J.; Deppert, K.; Wallenberg, L. R.; Samuelson, L. Semicond. Sci. Technol. 2010, 25, (2), 024009. 90. Joyce, H.J.; Wong-Leung, J.; Gao, Q.; Tan, H. H.; Jagadish, C. Nano Lett. 2010, 10, (3), 908-915. 56 91. Uccelli, E.; Arbiol, J.; Magen, C.; Krogstrup, P.; Russo-Averchi, E.; Heiss, M.; Mugny, G.; Morier-Genoud, F. o.; Nygard, J.; Morante, J. R.; Fontcuberta i Morral, A. Nano Lett. 2011, 11, (9), 3827-3832. 92. Shockley, W.; Queisser, H.J. J. Appl. Phys. 1961, 32, (3), 510-519. 93. Palik, E. D., Handbook of Optical Constants of Solids (Academic, Orlando, Fla., 1985). 94. Li, M.; Hu, X.; Ye, Z.; Ho, K.-M.; Cao, J.; Miyawaki, M. Opt. Lett. 2006, 31, (23), 3498-3500. 95. Li, M.; Li, Z. Y.; Ho, K.-M.; Cao, J. R.; Miyawaki, M. Opt. Lett. 2006, 31, (2), 262-264. 57 Chapter 3 GaAs nanowire array solar cells with axial p-i-n junctions 3.1 hdroduction Jn recent years, solar cells based on semiconductor nanowires have been a topic of intense research and development for next generation photovoltaics 1 - 21 • One approach to swpass the Shockley-Queisser efficiency limit is to use multi-junction solar cells containing several p-n junctions in series 22 - 26 • Each junction is designed to absorb a specific wavelength range of sun light, reducing thermalisation losses, and thereby increasing efficiency (Fig 3.1 a). a 2500 2000 ~ 1500 e 1000 - 500 s: Figure 3. 1 Multi-junction solar cells. (a) Multi-junction solar cells consist of mate1ials with different band-gaps. Mate1ials with larger band-gaps stack on top so different parts of the solar spectmm can be preferentially absorbed at different depths from the surface. (b) A solar cell made from ill-V nanowire p-n junctions grown on Si substrate. (c) A three-junction solar cell made from heterogeneous ill-V nanowires grown on Si solar cells. 58 Traditional, monolithically integrated multi-junction solar cells consist of sequentially stacked thin films. The lattice constants of the materials used are matched to allow high quality epitaxial growth. GaAs and Germanium are currently the most widely used substrates since there are lattice matched materials with larger bandgaps suitable for forming a current matched set of junctions. These substrates contribute a large portion of the total material cost and prevent large scale terrestrial implementation. The small footprint of nanowires allows them to accommodate lattice-mismatch induced strain through elastic relaxation at the edges 27 - 32 . This capability offers great freedom in choosing growth substrates, which are otherwise infeasible in the planar thin film scenario. For instance, dislocation-free InP or GaAs nanowires can be grown on a silicon substrates 32 - 35 , significantly reducing substrate cost and enabling use of the existing silicon industry infrastructure (Fig 3.lb). Meanwhile, the freedom to use multiple, stacked, non-latttice-matched materials within each nanowire allows the construction of multi-junction cells with optimal band gap combinations (Fig 3.lc). In addition, both simulation and experiments have shown that nanowires are inherently excellent light absorbers due to light scattering and the effect of resonant modes, even without additional . fl . . 11 36-43 anti-re ecbon coatmgs · . Among various nanowires, those made from III-V compound semiconductor materials are considered as some of the most prom1smg platforms for photovoltaics due to high absmption, direct band gap, supenor earner mobility, and well-developed synthesis techniques. Significant progress with InP nanow!fe solar cells has recently been 59 d . II . h . I . . 18 21 44 I G A . h reporte especrn y wit axrn p-1-n Junctions · · . n contrast, a s 1s anot er important material with suitable bandgap and has the advantage that gallium 1s more abundant than indium. However progress in the development of GaAs nanow!fe solar cells has been relatively slow and their potential has not been fully exploited. Previous work on GaAs nanowire solar cells focused on radial p-n or p-i-n junctions. For example, in 2008, LaPierre et al. 5 reported pioneering work of a GaAs nanowire array based solar cell with core/shell radial junction and power conversion efficiency (PCE) of 0.83%. In 2009, i Morral et al. 45 reported a single GaAs nanowire solar cell with PCE of 4.5% by molecular beam epitaxy, and their V 0 , close to IV indicates high-quality radial p-i-n junctions. In 2011, Mariani et al. 7 demonstrated solar cells made from lithographically patterned GaAs nanorod arrays with radial junctions and exhibited a PCE of 2.54%, which represented excellent progress towards large area realization. Recently Tegude et al. 20 reported a single GaAs/InGaP/GaAs with a radial p-i-n junction with PCE of 4. 7%. Recent efforts on in situ passivation of GaAs nanowires with lattice-matched wide band gap materials revealed encouraging improvements in the short circuit current. By growing a lattice-matched InGaP passivation layer, quantum efficiency close to unity has been observed leading to reportedPCEbetween 4% and 7.43% 15 • 1 9 . In contrast, solar cells based on axial p-n or p-i-n junctions have not been fully studied, although it is intriguing to compare their performance to that of radial junction devices and study the underlying science. Our recent theoretical study (Huang et al. 37 ) indicates that axial junctions are more sensitive to the presence of surface states, but axial junctions 60 provide higher V 0 , and more flexibility in the design of junction structure such as the thickness of base and emitter layers than radial junctions. Furthermore, a single axial junction is of considerable technical importance, as it is an important building block for multi-junction solar cells with three or more junctions as shown in Fig 3.lc. Nanowires with junctions stacked in the axial direction provide an intuitive analogue to the traditional thin film multi-junciton solar cells where the incident light can pass from materials with larger bandgap to those with lower bandgap sequentially. Here, we report solar cells based on GaAs nanow!fes with axial p-i-n junctions that achieve a PCE of 7.58%. This value is so far the highest reported PCE for any solar cells made from GaAs nanowire arrays. We carry out optical simulations to systematically study the light absorption properties of GaAs nanowire arrays. Theoretical comparison between radial and axial junction solar cells is conducted, which reveals that the axial junction design is more tolerant to doping variation. For a particular doping range, axial junction designs can offer better performance. Under the guidance of these simulations, we experimentally demonstrate solar cells based on uniformly patterned GaAs nanowire arrays with axial p-i-n junctions via SAG in MOCVD. Two key design parameters are studied to optimize device characteristics. Our results indicate that nanowires with shallower junctions offer better performance than deeper junctions, and large Wife diameter can lead to a short circuit current density (Jsc) as high as 23.28 mA/cm 2 in the absence of any surface passivation treatment. The junction depth dependence is further investigated and explained by cathodoluminescence measurements. 61 3.2 Theoretical study of optical and electrical properties of GaAs nanowire solar cells a b 400 500 600 700 •{"'11) d LO 0.8 c 0.6 .Q ~ 0 "' 0.4 .0 <( d = lOOnm 0.2 d = 250nm d = 350nm 0.0 300 400 SCIO 600 700 Wavelength (nm) 800 800 ,, . ,, - - - - /,, ,, SiN, mAJcm 1 2900 27 00 25.00 B.00 2LOO 1900 17.00 15.00 100 200 300 400 Diameter (nm) r jµmj '(µm) r iµml Figure 3. 2 Device structure and optical absorption simulation. (a) Schematic of a solar cell made from GaAs nanowires with axial junctions. Carrier flow direction in solar cell operation under sunlight is shown in the zoomed-in graph of the junction. (b) Maximum achievable Jsc (mA/cm 2 ) versus pitch a & diameter/pitch ratio dla. Nanowires arn 3 µm tall and grown on a GaAs substrate. (c) Maximum achievable Jsc versus nanowire diameter for a fixed pitch of 600nm and height of 3 µm. (d) Absorption spectra of nanowire aITays with 600 nm pitch and 100 nm (black), 250 nm (red) and 350 nm (blue) diameter. ( e) Carrier generation rate profile under AM 1.5G solal' spectrum for nanowires embedded in BCB and capped by ITO for diameters of 100 nm, 250 nm and 350 nm. 62 Fig 3.2a is a schematic of a solar cell made from a vertically aligned GaAs nanowire array grown on a GaAs substrate. Nanowires are embedded in transparent insulating polymer BCB, for mechanical support and covered by a transparent conductive indium tin oxide (ITO) front contact to let sunlight pass through. Incident light generates electrons and holes which flow toward the n-type and p-type region, respectively. The absorption properties of periodic nanowire arrays strongly depend on the array structure. In order to determine the optimal structure for absorption, we first carry out full-vectorial electromagnetic simulations to calculate the absorption of nanowire arrays with different diameters and pitch. We calculate the maximum achievable short circuit current from the optical absorption by assuming unity external quantum efficiency. Fig 3.2b shows Jsc as a function of pitch, a, and diameter-to-pitch ratio, dla, for a nanowire height of 3µm. Calculations were performed using the AM l.5G solar spectrum assuming an infinitely thick GaAs substrate. Two local maxima are observed on this map. One is at a ~ 300 nm and d/a ~ 0.55, and the other one is around a ~ 650 nm and d/a ~ 0.6. We focus on the maximum point associated with pitch of 600 nm due to the fact that, without sacrificing too much absorption (<l mA/cm 2 ), a larger pitch eases fabrication requirements. Furthermore, we will show later in the paper that the small diameter of the other maximum point is unfavorable for the electrical properties. In Fig 3.2c, we plot Jsc versus nanowire diameter for fixed pitch of 600 nm and height of 3 µm. We can achieve Jsc as high as 28 mA/cm 2 for a nanowire array with diameter of 350 nm, which is close to the theoretical limit of a single junction GaAs solar cell. Comparing the absorption spectra of arrays with diameters of 100 nm, 250 nm and 350 nm (600 nm pitch, 3 µm height) shown 63 in Fig 3.2d, the 350 nm one enhances the absorption by filling in the low absorption dips between 400 nm and 550 nm and between 700 nm and 900 nm that appear in the spectra of the 100 nm and 250 nm diameter wire arrays. We also plot the spatial carrier generation rate distribution within the realistic nanowire array structure (with the BCB polymer and ITO caps) for nanowires with height of 2.5 µm and diameters of 100, 250 and 350 nm, respectively (Fig 3.2e). The maximum achievable short circuit currents are also indicated in each profile. 64 a c 0.82 0.80 > u 0.78 0 > 0.76 0.74 2.0 1.5 E :i. -;::; 1.0 0.5 0.0 -1 ' ' ' I I Axial junction Radial junction 101& 1017 1018 n type base doping (cm -3) A 2.0 A 1.5 i I E ' ' A' :::1. B' -;:;-1.0 0.5 ' I ; 0.0 -- - ' 0 1 ·1 0 1 Energy (eV) A ' Energy (eV) ·C' B' 10!! ::r 1.0 ~ > 0 .8 lO 8 ~ 0.6 10 :i > C"I ~ 0.4 (1) Q) a c 0.2 iil w 10s 5· 0 .0 :i -0.2 - - - - - -~- c~ - - - - - - - n- +-~~~~~~~~~~ 3 -0.10 -0.05 0.00 0.05 0.10 10° - °' r (µ m) Figure 3. 3 Comparison between nanowire solar cells with axial junctions and radial junctions. (a) Schematic diagram of axial (upper) and radial (lower) junctions in nanowires. (b) Jsc, ( c) V 0 c and ( d) efficiency of nanowire solar cells with axial junctions (black) and radial junctions (red) versus doping concentration in n-type base. Surface recombination velocity is 3x10 5 cm/s. See text for details of junction geometry and doping concentration. (e) 30 mapping of hole concentration in the dark and in thermal equilibrium for the axial junction (left) and the radial junction (right). Band diagrams under AM 1.5 shoit circuit conditions along the center of each nanowire (A-.A: and B-B') are shown on the left together with quasi Fermi levels of electrons (blue dashed line) and holes (red dashed line). (f) Band diagram along C-C' line in the radial junction in (e) together with quasi Fermi levels of electrons and holes. 65 In terms of electrical properties, the design of the p-n junction governs the overall performance of a solar cell. We thus carefully examined the operational conditions of nanowire solar cells with radial and axial junctions. Fig 3.3a shows schematics of these two junctions in the nanowires. We numerically solve the current density-voltage (J-V) response of the nanowire solar cells with 600 nm pitch, 250 nm diameter and 2.5 µm height. We use Synopsys Sentaurus to solve the drift-diffusion equations for carrier transport within the nanowires. The nanowires are subject to AM l.5G solar irradiation. The position-dependent carrier generation rate is determined from the optical absorption simulations above and is shown in Fig 3.2e. For the axial junction, the p-type segment is on top, with a thickness of 100 nm, and for the radial junction, the p-type region is on the outside, with a thickness of 30 nm. Both intrinsic regions are 50 nm thick. The thicknesses of the p-type regions were chosen based on our previous study 37 . Thicker shell lead to degraded performance. We assume that the doping dependent mobilities for electrons and holes are the same as in the bulk, with a Shockley-Read-Hall (SRH) recombination lifetime of 1 ns. Both donor-like and acceptor-like surface state densities are fixed to be l.5 xl0 12 cm- 2 , corresponding to a surface recombination velocity (SRV) of 30,000 cm/s. We fix the doping concentration of the p-type emitters to be 10 18 cm- 3 for the purpose of forming a low resistance ohmic contact, and vary the n-type base doping in both the axial and radial junctions. 10 nm thick minority carrier reflectors with doping concentrations of 10 19 cm- 3 are placed right below the top contact and above the bottom contact to reduce recombination loss. In the real devices, heavily doped n+ substrate (2-3x10 18 cm- 3 ) serves as bottom minority reflector and the p-doping near the tip is 66 increased to serve as top minority reflector. In Fig 3.3b, c and d we show the Jm Voe and PCE as functions of n-type base doping concentration. For the axial junction, increasing the doping concentration reduces Jsc. Nanowires usually exhibit shorter minority diffusion length than bulk, so the built-in electric field in the junction depletion region contributes significantly to the total collection of carriers. With increasing doping in then-type base of axial junction, both the junction depletion region width, which defines the carrier drift zone, and the mobility which determines carrier diffusion, are reduced. The open circuit voltage however increases with increasing doping concentration due to the larger built-in potential. The overall efficiency does not change much with the increasing doping concentration. For the radial junction, the short circuit current is very high when the base doping is higher than 10 17 cm- 3 . However it drops sharply when the doping is smaller than 10 17 cm- 3 . The open circuit voltage and the efficiency of the radial junction have similar trends. The precise measurement of spatial distribution of dopant in nanowires remains a great challenge in this field and has not been fully investigated since most techniques and physics used for bulk material could not be directly applied to nano-scale materials. However people did observe doping 46 and material composition 16 variation along a nanowire over one order of magnitude. Emerging technologies such as atom probe tomography 47 or single wire Hall Effect measurement 48 could provide us more powerful tools to obtain detailed doping information in a nanowire. 67 We can gain insight into the difference between the J-V characteristics of axial and radial junctions in the low doping region from Fig 3.3e and f. Fig 3.3e shows the band diagrams across the center of the wires (A-A' for the axial junction and B-B' for the radial junction) under the AM l.5G solar spectrum and short circuit condition and the corresponding hole concentration mapping in the dark under thermal equilibrium conditions with a base doping of 10 16 cm- 3 . In the axial junction, the n-type region is fully depleted by surface states under thermal equilibrium, as one can tell from the fact that the hole concentration is close to the intrinsic carrier concentration of GaAs (2.lxl0 6 cm- 3 , 300K). However, under illumination, the nanowire behaves as a normal p-i-n junction solar cell (see band diagram) because the surface states are largely filled by light-generated carriers. There is a gradual band bending along the axial direction near the junction at the heights between 2 µm and 2.4 µm (intrinsic segment is between 2.35 and 2.4 µrn). This causes an electric field pointing upwards, which helps the extraction of the light-generated carriers near the tip of the nanowire. In a radial junction, when the core region of the nanowire is lightly doped, the p-type shell will cause an inversion of the core carrier type to also be p-type. The structure is hence equivalent to a p-type nanowire with a very thin n-type emitter at the bottom as can be seen in the band diagram along B-B' in Fig 3.3e. This can further be confirmed by checking the cross-sectional (C-C') band diagrams in Fig 3.3f, which shows the band is almost flat. The minority carriers thus need to diffuse through the entire length of nanowire to reach the bottom and get extracted, under which circumstance an extremely low Jsc is expected. 68 In principle the heavily doped radial junction device can outperform the axial junction due to its excellent carrier collection efficiency and its high tolerance to surface effects. However, high doping in the base is usually undesirable because it reduces mobility and diffusion length. State-of-the-art thin film GaAs solar cells normally use a base doping concentration of the order of magnitude of 10 17 cm- 3 . The possibility of using lower doping in the axial junction provides us with more optimization space and a more robust design. Moreover, as mentioned before, the axial junction is an indispensable intermediate step toward a multi-junction nanowire solar cell device with three or more junctions. The systematic experimental study of the GaAs axial junction nanowire solar cell has not previously been performed. 69 3.3 Fabrication and optimization of GaAs nanowire array with axial p-i-n junction a b ITO sputtering c MOCVD SAG RIE etching d l BCB infiltration Figure 3. 4 Solar cell fabrication process and SEM images. (a) Fab1ication steps of GaAs nanowire anay solar cells with axial junction: (i) electron beam lithography to fonn hole array in silicon nitride mask, (ii) SAG of p-i-n GaAs nanowire using MOCVD, (iii) BCB infiltration, (iv) RIE to expose nanowire tips, and (v) ITO deposition. (b) 30° tilted SEM image of as grown vertical GaAs nanowire anay on GaAs (lll)B substrate. (c) SEM image after nanowires are embedded in BCB and etched by RIE to expose short tips. (d) SEM image after coating of ITO film by sputtering. A conformal dome-like cap is formed on the tips of nanowires. To expe1imentally test the performance predicted by simulation we fabricated and measured solar cells with 1 mm x 1 mm area GaAs nanowire anays. The schematic diagram of device structure has been shown in Fig 3.2a and the fabrication process is shown in Fig 3.4a. Then-type, undoped and p-type segments are grown sequentially on a 70 n+ (111 )B substrate usmg SAG m MOCVD. The growth detail has been mentioned earlier in section 2.3.2. This growth method doesn't require metal catalyst as used in vapor-liquid-solid (VLS) method which is believed to incorporate along the growth and cause deep level traps. The vapor phase epitaxial nature of SAG can also avoid reservoir effects often encountered in VLS and achieve an abrupt junction interface. The background doping for the i-region is not known at this point, but is believed to be much lower than the p and n region, and may be in the range of 10 14 -10 15 cm- 3 based on literature 49 . We believe i-region with optimized length could help carrier collection to certain degree given the short diffusion length in nanowires. Nanowires are separated by 600 nm from center to center and are about 2.5 µm tall and 320 nm in diameter, unless otherwise stated, in order to achieve nearly optimized absorption of AM l.5G solar irradiation. Although the nanowires only cover less than 20% of the total area, they could potentially absorb close to 90% of the incident sun light. It's been widely studied that resonant modes would allow light absorption in nanowires to significantly exceed ray optics limit given just a fraction of the material consumed in a bulk device 11 • 21 . Fig 3.4b is SEM image taken at 30° tilted angle. Nanowires distribute uniformly in the EBL defined template and exhibit six-fold symmetric cross section consisting of sidewalls parallel to {-110} family planes which indicate high crystal quality. After growth, nanowire arrays are planarized and etched to only expose the tip of p-type emitter. Eventually ITO is deposited as transparent front contact (Fig 3.4a) while AuGe is alloyed to the backside of the substrate to form ohmic contact. The sheet resistance of 71 planar ITO film is about 10 Ohm/o measured by four-probe method. The highly conductive ITO film allows us to achieve low series resistance without forming additional metal fingers on top of ITO as will be shown later. Sputtering provides a conformal layer of ITO on the non-planar top surface and forms a dome shape cladding over the nanowire tip which helps to concentrate the incident light near the junction region as has been seen in Fig 3.2e and also pointed out by Mariani et al. 15 . Fig 3.4c shows the 30° tilted SEM image of nanowire array after BCB infiltration and reactive ion etching (RIE). About 100 nm tip is exposed for contacting. Fig 3.4d shows ITO film conformally wrapping the nanowire tips to ensure good conductivity at the front electrode. Basically the high uniformity is inherited from EBL-made pattern by the subsequent process and demonstrates the superior capability of SAG in controlling the morphology and location of nanowires. 72 a c 0.65 0.60 - 0.55 > a.so u g 0.45 - 0 .40 0.35 ::...._o c-07~ 'O><c-- c -o-250nm • Voe -o- 250nm • fF 3 Sample# -i:r- IOOnmf d•rt (A) -0- IOOnmf I sun --6- 2sonmt darlc (8) ...J\1-250nmf 1 sun .().2 00 0.1 04 0.6 d Voltage (V) 140 ....---------------, 065 0 .60 o. ss 0 ~ 80 o.so v ~ (! g 60 O.• S ;:: ">:: u. ·~ 40 c 040 .. "C 20 035 0.0 lOOnm ~ Dark J/V vs V 0.2 0.4 0.6 Voltage (V) 0.8 1.0 Figure 3. 5 Comparison between performance of nanowires with 100 nm diameter and 250 nm diameter. (a) 30° tilted SEM images of nanowires with diameter of 100 nm (sample A) and 250 nm (sample B). (b) J-V curves of typical devices made from 100 nm and 250 nm thick nanowires in dru·k and under AM 1.5 solar spect:ium. (c) V 0 c (black) and FF (blue) of six devices from each batch. (d) J!V-V cmve of the device with 100 nm nanowires shown in (b) which shows lineru· region marked by the blue line. Insets are schematics of a fully depleted thin wire and of a partially depleted thick wire that still has conductive channel in the center. Because of high density of surface states, carrier t:i·ansport in GaAs nanowires is known to be significantly affected by surface-to-volume ratio. For nanowires with surface Fenni level pinned at mid gap, Chang et al. 50 pointed out majority carriers would be neru·ly depleted if wire diameter is less than 100 nm, resulting in a less conductive channel which is also observed in our simulation. Fmthennore a considerable portion of minority cru1iers would be captmed by surface states and subsequently annihilate through recombination for a thin wire. Thus one expects significant loss of .fsc when nanowire diameter gets ext:i·emely small. Fig 3.5a shows SEM images of the 30° tilted nanowire 73 arrays used for two sets of devices. Both samples are around 1.5 µm tall with 300 nm deep junction. Sample A consists of nanowires about 100 nm in diameter while the wires in sample B are close to 250 nm in diameter. J-V characteristics are measured under dark condition as well as under AM 1.5G solar spectrum at 1 sun illumination intensity (100 m W/cm 2 ) using a solar simulator (Photo Emission Tech). Distinct J-V curves can be seen in Fig 3.5b for representative devices made from these two batches. Remarkably improved performance is obtained for nanowires of 250 nm than nanowires of 100 nm diameter due to both larger Jsc and higher Voe· Another important parameter reflecting how closely a device resembles an ideal diode is fill factor (FF) defined by the ratio of maximum power output to the product of Isc and Voe. Fig 3.5c shows the Voe and FF of all the samples from batch A and Band each batch contains six individual devices. From this figure, the FF of sample B is much higher than sample A on average following the same trend as V oc· In larger diameter nanowires, fewer carriers will recombine at the surface leading to higher Jsc· On the other hand if the entire n-type base is depleted as a result of a small diameter, then the built-in potential across the p-i-n junction will definitely be smaller than an ideal junction because the Fermi-level in the n-type region is deeper into the bandgap than it would be in an ideal case. Moreover, plots of current density divided by voltage (J/ V) versus V for sample A shows, after the diode turns on, J/V is proportional to Vindicating strong quadratic dependence on voltage for dark current as shown in Fig 3.5d. This phenomenon can be attributed to space charge limited (SCL) transport in a fully depleted crystal which was recently modeled and discussed for gallium nitride nanowires 51 and indium arsenide nanowires 52 . 74 -0.2 0.0 c -Dark =- -14 -50t:W c -500uW ....J -16 -lmW -5mW -18 -lOmW -20 -22 -0.5 0.0 O.Z Voltage {V) 0.5 1.0 Voltage (V) 0.4 0.6 1.5 2.0 = c -2 --' -4 -6 -8 d lE-3 -;:(' ::; lE-4 c ~ ::; u lE-5 -1.0 D Dark 1 Sun 320nm diameter 19 lOOnm junction depth -0.5 0.0 Voltage (V) 100 1000 Pump Power {mW) 0.5 1.0 10000 Figure 3. 6 Junction depth dependency and J-V under varied light intensity. (a) J-V curves of devices with different junction depths between 100 nm and 600 nm under AM 1.5 solar spectrum. For the device with 100 nm junction depth, diameter is increased to 320 nm. (b) Dark and AM 1.5 J-V curves of the device with 100 nm junction depth and 320 nm diameter shown in (a) plotted in semi-logarithmic scale. Ideality factor is found to be 2.34. (c) J-V curves of the device with 150 nm junction depth and 220 nm diameter shown in (a) under different illuminating power between 50 µWand 10 mW from an 850 nm laser. V 0 e of0.716 Vis observed under the highest light intensity. Insets shows the curve of V 0 e v.s. ln(Ise) with extracted ideality factor to be 1.72. (d) Current versus pump power for the device shown in ( c) at -2 V and 0 V bias. In addition to nanowire diameter we studied another important design parameter which is the length of the p-emitter region, also mentioned as junction depth (DJe) interchangeably in this paper. As has been shown in Fig 3.2e, upon illumination of solar spectrum the carrier generation hot spot is located very close to the nanowire tip due to the high absorption coefficient of GaAs over most of the solar spectrum as well as the concentrating effect from the ITO cap. On the other hand, to achieve low resistance 75 ohmic contact between ITO and GaAs, the p-doping near the top is usually sufficiently high leading to relatively shorter minority carrier lifetime. Thus, we want to minimize carrier loss in the p-region while keeping a good contact to ITO. Experimentally we varied the length of p-segment between 600 nm and 100 nm by adjusting the growth time. As shown in Fig 3.6a a steady increase in short circuit current can be observed when we decrease the junction depth. Typical devices show Jsc of 1.07, 8.52 and 11.58 mA/cm 2 for devices with 220 nm diameter and nominally 600 nm (sample C), 300 nm (sample D) and 150 nm (sample E) deep junctions respectively. Shallower junctions also feature higher V oc in general. The batch with 600 nm deep junction (sample C) exhibits relatively low Voe around 350 mV and, in contrast, devices in the batch with 150 nm deep junction (sample E) have V oc approaching 650 m V. In batch F we further reduce junction depth to 100 nm and also increase diameter to around 320 nm. We observe a tremendous increase of Jsc to above 20 mA/cm 2 with the highest being 23.28 mA/cm 2 . Although both nanowire diameter and junction depth affect the solar cell performance, we believe nanowire diameter is the more dominant factor, especially for nanowires with junction depth less than 300 nm. This can be better seen from Fig 3.7 which shows the performance Sample G with 320 nm diameter and 300 nm junction depth, so it differs from Sample D only in diameter and differs from Sample F only in junction depth. We can find that the Sample G performance is closer to Sample F performance and farther away from Sample D performance, which tells us that the nanowire diameter affects the solar performance more significantly than the junction depth. 76 +-' c 5 -Dark - Light 300 nm D;, 320 nm ¢ J = 21.21mAlcm' SC V =0.511V oc ~ -15 FF= 0 6075 ::J 0 ~ = 6 56% -20 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 Voltage (V) Figure 3. 7 Dark and light J-V curves of device G with 320 nm diameter and 300 nm junction depth. Fig 3.6b shows the dark and 1 Sun J-V curve of the best device from batch F plotted as a semi-logarithmic plot. The dark characteristic shows a good rectifying behavior with an on-off ratio of l.89xl0 5 at ± 1 V. The dark current at -1 V is only about 100 nA for a 1 mm 2 area also indicating a good junction. The small leakage current could be attributed to the small total junction area possessed by axial junction geometry. Vapor phase growth of the junction should, in principle, produce a sharp junction with low leakage current. The ideality factor extracted from the intermediate forward bias regime is around 2.34. We believe, the greater-than-unity ideality factor is due to the existence of recombination current through surface states and space charge region m the undoped part. Further optimization of the length of the undoped segment is ongoing. This particular device shows Jsc of 21.08 mA/cm 2 and V 0 c of 0.565 V . With a fill factor of 0.6365 the overall PCE is 7.58%. In previous publication 37 , we have simulated and experimentally studied the role of substrate on photon absorption. Even though the substrate may absorb a small portion of the photons, because carrier generation is relatively far away from the p-i-n 77 junction, the contribution of substrate to Jsc would be negligible. We thus expect the power conversion capability mainly comes from the nanowire array. The slightly lower V oc here in sample F compared to sample D and E is due to the logarithmic dependence of V oc on fsc for an ideal diode which will be shown later, thus the increase of V 0 c with shallower junction depth is not as pronounced as Jsc, and sometimes can be masked by the device-to-device variation as manifested by the red, blue and pink curves in Fig 3.6a. We note that in the device showing efficiency of 7.58%, the ITO may be in contact with the i-region for some of the nanowires as both the junction depth and the exposed nanowire tip length after BCB etching are about 100 nm. By further optimizing the process and control of exposed tip length we may be able to achieve even better performance. All the devices discussed so far are summarized in Table 1. Table 1 Characteristics of devices with different diameters and junction depths Diameter Junction depth Device# J,c (mA/cm 2 ) Voe (V) Efficiency (% ) (nm) (nm) A 100 300 4.20 0.412 0.68 B 250 300 7.2 1 0.642 3.03 c 220 600 1.07 0.410 0.21 D 220 300 8.72 0.600 3.30 E 220 150 11.58 0.592 4.62 F 320 100 21.08 0.565 7.58 G 320 300 21.21 0.511 6.56 78 When a non-ideal p-n junction with a series resistance is considered with respect to the forward bias voltages, the current across a p-n diode is given by In order to accurately determine the series resistance we can differentiate both side of the current equation and get (Cheung's method 53 ): d V = nkT +JR d(lni) q s A plot of dV/d(lnJ) versus I will be linear and gives Rs as slope and nkT/q as the y-axis intercept. The extracted values of ideality factor and series resistance are 2.2 and 41 n (Fig 3.8), respectively. 0.35----------------- 0 0 .30 0 0 .25 0 .20 ~ 0 .15 ~ -0 0.10 0.05 0 .00 -0.05-+--~-~-~-~-~-~-~---,-' 0.001 0.002 0.003 0.004 I (A) Figure 3. 8 dV/d(lnJ) versus I curve of the device showing the highest efficiency. Based on Cheung's method, Rs is the slope and nkT/q is the y-axis intercept. The extracted values of ideality factor and series resistance are 2.2 and 41 n, respectively Finally, we measured device response under varied illumination power. Fig 3.6c shows the J -V curves of device with 300 nm junction depth and 220 nm diameter (red curve in 79 Fig 3.6a) in dark and at various illumination powers from an 850 nm laser plotted in semi-logarithmic coordinates. The short circuit current increases steadily when power is increased from 50 µ W to 5 mW and starts to saturate at higher powers which indicates reduced quantum efficiency. This behavior is due to increased voltage drop across series resistance under high illumination as well as carrier screening effect which reduces internal electrical field when carrier concentration is sufficiently high. Similar trend was observed in photo detector devices made from similar nanowires recently in our group 54 . By applying a larger reverse bias we can rebuild the strength of electrical field so we can see incremental change of current scales with illumination power at -2 V applied bias (Fig 3.6d) while the current at short circuit fails to track under high illumination intensity. The open circuit voltage increases steadily with increasing pump power. V 0 , for the highest illumination reaches 0.72 V. Neglecting the small series resistance the diode 1-V behavior under illumination could be described as I =10 ( exp(::r )-1 )-IL ~10 (exp(::r )-1)-1" So we can extract V oc as a function of short circuit current: Inset of Fig 3.6c is the plot of Voe versus ln(Isc) and the slope is nkT/q, from which we extracted the ideality factor to be 1. 72. 80 a A.=350 nm 2000 1000 0 0 200 r(nm) 650 nm 0 200 r (nm) 500nm 0 200 r(nm) 800nm 0 200 r(nm) 1a2l 10 21 10 20 b 1.0 0.9 0.8 0.7 0.6 UJ Cf 0.5 UJ 0.4 0.3 0.2 0.1 c - 600nm Djc• 220nm <!> - 300nm Djc• 2ionm <I> - 150nm Djc• 220nm <I> - 100nm Djc· 320nm <!> -- AMl. SG , 350 400 450 500 550 600 650 700 750 800 850 Wavelength (nm) xl0 18 - E c: ..... x10" ..S > ..., ·;;; c: CIJ l -c xlO 6 .... 0 .s:;; 0. Cl intensity Cl intensity Figure 3. 9 Spectrum response and Cathodolmninescence mapping. (a) Ca.trier generation rate profile under monochromatic light of different wavelengths between 350 nm and 800 nm. All the light intensities ai·e 100 mW/cm 2 . Nanowires are 2.5 µm tall and 350 nm in diameter with 600 nm pitch. Nanowires ai·e sunounded by BCB and covered by ITO on top. (b) External quantmn efficiency (left y-axis) v.s. wavelength of the four devices shown in Fig 3.6a. Dash-dot curve is the photon intensity of AM 1.50 solai· spectrum (right y-axis). Jsc of each device are calculated by integrating the product of EQE and photon intensity over the range of 350 nm to 880 nm and are indicated in the plot. (c) SEM images and normalized cathodolmninescence mapping at wavelength of 865 nm for nanowires with 400 nm deep junction (left) and 100 nm deep junction (right). Red curves ai·e relative CL intensity along the center of each nanowire. From the results shown in Fig 3.6a, it's apparent that a shallow junction is essential for unpassivated GaAs nanowire solar cells to capture minority ca.triers generated close to the tip which would otherwise recombine quickly in heavily doped p-emitter region. To better understand the underlying physical mechanism we simulated the cat1iei· generation 81 profile under monochromatic light of different wavelengths and measured the spectrum response of the four devices shown in Fig 3.6a. The integration of product of EQE and AM l.5G photon density resulted in Jsc similar to the value measured experimentally in Fig 3.6a. From Fig 3.9a we can see most of the shorter wavelength light (400 nm) is absorbed near the wire surface while a bigger portion of longer wavelength light (800 nm) is absorbed deeper into the bulk of the nanowire. Carriers generated by shorter wavelength thus are more likely to recombine at the surface and annihilate. That explains why the EQE of shorter wavelength is always lower than the EQE of longer length for all four devices shown in Fig 3.9b. The surface-to-volume ratio decreases with increasing diameter so a smaller portion of the total generated carriers are distributed in the vicinity of surface for thicker nanowires which leads to higher EQE in 320 nm thick nanowire than in 220 nm thick nanowire with similar junction depth over a broad range of wavelength. To interpret the junction depth dependency of Jsc we conduct cathodoluminescence (CL) measurement (Horiba) built in Hitachi S-4800 SEM. To excite luminescence from a single nanowire the sample was cleaved so that electron beam can be focused and scanned normal to the sidewall surface. Acceleration voltage was kept at 5 kV to ensure that no electrons penetrate the wire to be examined and cause luminescence from other wires behind. SEM micrographs and normalized CL intensity maps at 865 nm from two single nanowires, one with 400 nm long p-region (left) and the other 100 nm long p-region (right), are plotted in Fig 3.9c and the line scans from the center of the two wires are also superimposed. The top 300 nm of the wire with a 400 nm deep junction does not emit efficiently. The intensity gradually increases with distance from the top of 82 the nanowire and eventually saturates when the electron beam moves towards the bottom of the wire. This trend qualitatively agrees with the aforementioned model that non-radiative recombination processes are dominant in the p-doped region due to the high impurity concentration introduced to achieve good ohmic contact. On the other hand the luminescent intensity is much more uniform for the wire with only 100 nm long p-region. The luminescence intensity gradient along the wire is due to the fact that electron beam with 5kV acceleration has 200-300 nm interaction volume upon impinging at the surface although the junction itself is considered more abrupt given the vapor phase epitaxy employed. In addition, those holes generated close to the junction have a higher probability of being swept by the junction before they radiatively recombine, while those holes generated far from the junction are more likely to recombine in the n-layer. The short radiative recombination life time and high nonradiative recombination rate observed in the p-emitter may well explain the increasing EQE with decreasing junction depth seen in Fig 3.9b as less carriers would be lost in the p-emitter when its length is reduced. Last but not least, the V 0 , measured from our devices is still considerably lower than the values reported for GaAs planar p-n junctions, which can be attributed to several factors. In nanowire solar cells because all the nanowires are isolated from each other, we need to form a front contact that can access all of them. This has been shown to be detrimental to V oc in planar solar cells where one tries to passivate most of the emitter surface and to only make contact on area as small as possible 55 . Intrinsic surface states together with surface defects introduced by processes such as etching and sputtering can lead to 83 significant recombination at the nanowire/ITO interface leading to lower V oc· The presence of band bending and doping variation m the radial direction will cause inhomogeneous barrier at the junction. Since the lower barrier part dominates the current conduction, V 0 e is also largely determined by the lowest barrier height across the intersection. In principle, a window layer with larger band gap and low surface states density could passivate the GaAs nanowire surface. This part of work is still on going and we are studying the passivation effect of shells consist of AlGaAs, GaAsP and InGaP. One issue that needs to be paid extra care to is that the presence of passivation layer should not introduce a significant shunting path between p- and n-region which requires low unintentional doping in the shell and low interface trap density. Furthermore a rigorous assessment of the doping concentration within the nanowire is necessary to determine the build-in potential which sets the upper limit of Voe· 3.4 Summary In summary, we carried out optical simulation to predict the optimized wire array geometry for maximum light absorption. We then compared the advantages and limitations of both axial junction and radial junction design. We experimentally demonstrated solar cells realized by arrays of GaAs nanowires with axial p-i-n junction which are grown by versatile selective area growth method using mass production compatible MOCVD technique. Nanowire array with low filling ratio turns out to be highly absorptive. Systematic study on effect of diameter reveals that thicker nanowires 84 are favorable because of the high surface recombination velocity on the bare GaAs nanow!fe surface. Junction depth also plays a significant role in earner collection efficiency. By reducing junction depth to around 1 OOnm and keeping diameter at 320nm we are able to achieve efficiencies as high as 7.58%. Under concentrated 850nm light a Voe as high as 0.716 V has been obtained. The results demonstrate that GaAs nanowires are good candidates for high-efficiency and low-cost solar energy conversion and open up great opportunities for the next generation photovoltaics based on multi-junction devices composed of lattice mismatched material systems. 85 References 1. Tian, B.; Zheng, X.; Kempa, T. J.; Fang, Y.; Yu, N.; Yu, G.; Huang, J.; Lieber, C. M. Nature 2007, 449, (7164), 885-889. 2. Garnett, E. C.; Yang, P. J. Am. Chem. Soc. 2008, 130, (29), 9224-9225. 3. Tang, J.; Huo, Z.; Brittman, S.; Gao, H.; Yang, P. Nat. Nanotechnol. 2011, 6, (9), 568-572. 4. Kempa, T. J.; Cahoon, J. F.; Kirn, S.-K.; Day, R. W.; Bell, D. C.; Park, H.-G.; Lieber, C. M. Proc. Natl. Acad. of Sci. 2012, 109, (5), 1407-1412. 5. Czaban, J. A.; Thompson, D. A.; LaPierre, R.R. Nano Lett. 2008, 9, (1), 148-154. 6. Goto, H.; Nosaki, K.; Tornioka, K.; Hara, S.; Hirurna, K.; Motohisa, J.; Fukui, T. Appl. Phys. Express 2009, 2, 035004. 7. Mariani, G.; Wong, P.-S.; Katzenrneyer, A. M.; Leonard, F.; Shapiro, J.; Huffaker, D. L. Nano Lett. 2011, 11, (6), 2490-2494. 8. Wei, W.; Bao, X.-Y.; Soci, C.; Ding, Y.; Wang, Z.-L.; Wang, D. Nano Lett. 2009, 9, (8), 2926-2934. 9. Shin, J.C.; Kirn, K. H.; Yu, K. J.; Hu, H.; Yin, L.; Ning, C.-Z.; Rogers, J. A.; Zuo, J.-M.; Li, X. Nano Lett. 2011, 11, (11), 4831-4838. 10. Tornioka, K.; Tanaka, T.; Hara, S.; Hirurna, K.; Fukui, T. IEEEJ. Sel. Top. Quantum Electron. 2011, 17, (4), 1112-1129. 11. Hu, S.; Chi, C.-Y.; Fountaine, K. T.; Yao, M.; Atwater, H. A.; Dapkus, P. D.; Lewis, N. S.; Zhou, C. Energ. Environ. Sci. 2013, 6, (6), 1879-1890. 12. Zhu, J.; Hsu, C.-M.; Yu, Z.; Fan, S.; Cui, Y. Nano Lett. 2009, 10, (6), 1979-1984. 13. Putnam, M. C.; Boettcher, S. W.; Kelzenberg, M. D.; Turner-Evans, D. B.; Spurgeon, J.M.; Warren, E. L.; Briggs, R. M.; Lewis, N. S.; Atwater, H. A. Energ. Environ. Sci. 2010, 3, (8), 1037-1041. 14. Mariani, G.; Scofield, A. C.; Hung, C.-H.; Huffaker, D. L. Nat. Commun. 2013, 4, 1497. 15. Mariani, G.; Zhou, Z.; Scofield, A.; Huffaker, D. L. Nano Lett. 2013, 13, (4), 1632-1637. 16. Holm, J. V.; J0fgensen, H. I.; Krogstrup, P.; Nygard, J.; Liu, H.; Aagesen, M. Nat. Commun. 2013, 4, 1498. 17. Krogstrup, P.; Jorgensen, H. I.; Heiss, M.; Dernichel, O.; Holm, J. V.; Aagesen, M.; Nygard, J.; Fontcuberta i Morral, A. Nat. Photon. 2013, 7, (4), 306-310. 18. Cui, Y.; Wang, J.; Plissard, S. R.; Cavalli, A.; Vu, T. T. T.; van Veldhoven, R. P. J.; Gao, L.; Trainor, M.; Verheijen, M.A.; Haverkort, J.E. M.; Bakkers, E. P.A. M. Nano Lett. 2013, 13, (9), 4113-4117. 19. Nakai, E.; Yoshimura, M.; Tornioka, K.; Fukui, T. Jpn. J. Appl. Phys. 2013, 52, 055002. 20. Gutsche, C.; Lysov, A.; Braam, D.; Regolin, I.; Keller, G.; Li, Z.-A.; Geller, M.; Spasova, M.; Prost, W.; Tegude, F.-J. Adv. Funct. Mater. 2012, 22, (5), 929-936. 21. Wallentin, J.; Anttu, N.; Asoli, D.; Huffman, M.; Aberg, I.; Magnusson, M. H.; Siefer, G.; Fuss-Kailuweit, P.; Dirnroth, F.; Witzigrnann, B.; Xu, H. Q.; Samuelson, L.; Deppert, K.; Borgstrom, M. T. Science 2013, 339, (6123), 1057-1060. 86 22. King, R.R.; Law, D. C.; Edmondson, K. M.; Fetzer, C. M.; Kinsey, G. S.; Yoon, H.; Sherif, R. A.; Karam, N. H. Appl. Phys. Lett. 2007, 90, (18), 183516. 23. Geisz, J. F.; Kurtz, S.; Wanlass, M. W.; Ward, J. S.; Duda, A.; Friedman, D. J.; Olson, J.M.; McMahon, W. E.; Moriarty, T. E.; Kiehl, J. T. Appl. Phys. Lett. 2007, 91, (2), 023502. 24. Geisz, J. F.; Friedman, D. J.; Ward, J. S.; Duda, A.; Olavarria, W. J.; Moriarty, T. E.; Kiehl, J. T.; Romero, M. J.; Norman, A. G.; Jones, K. M. Appl. Phys. Lett. 2008, 93, (12), 123505. 25. Guter, W.; Schone, J.; Philipps, S. P.; Steiner, M.; Siefer, G.; Wekkeli, A.; Weiser, E.; Oliva, E.; Bett, A. W.; Dimroth, F. Appl. Phys. Lett. 2009, 94, (22), 223504. 26. Derkacs, D.; Jones-Albertus, R.; Suarez, F.; Fidaner, 0. Journal of Photonics for Energy 2012, 2, (1), 021805-1. 27. Gudiksen, M. S.; Lauhon, L. J.; Wang, J.; Smith, D. C.; Lieber, C. M. Nature 2002, 415, (6872), 617-620. 28. Glas, F. Phys. Rev. B 2006, 74, (12), 121302. 29. Sburlan, S.; Dapkus, P. D.; Nakano, A. Appl. Phys. Lett. 2012, 100, (16), 163108. 30. Chuang, L. C.; Moewe, M.; Chase, C.; Kobayashi, N. P.; Chang-Hasnain, C.; Crankshaw, S. Appl. Phys. Lett. 2007, 90, (4), 043115. 31. Ertekin, E.; Greaney, P.A.; Chrzan, D. C.; Sands, T. D. J. Appl. Phys. 2005, 97, (11), 114325. 32. Tomioka, K.; Kobayashi, Y.; Motohisa, J.; Hara, S.; Fukui, T. Nanotechnology 2009, 20, (14), 145302. 33. Martensson, T.; Svensson, C. P. T.; Wacaser, B. A.; Larsson, M. W.; Seifert, W.; Deppert, K.; Gustafsson, A.; Wallenberg, L. R.; Samuelson, L. Nano Lett. 2004, 4, (10), 1987-1990. 34. Moewe, M.; Chuang, L. C.; Crankshaw, S.; Chase, C.; Chang-Hasnain, C. Appl. Phys. Lett. 2008, 93, (2), 023116. 35. Yao, M.; Madaria, A. R.; Chi, C.; Huang, N.; Lin, C.; Povinelli, M. L.; Dapkus, P. D.; Zhou, C. In Scalable synthesis of vertically aligned, catalyst-free gallium arsenide nanowire arrays: towards optimized optical absorption, SPIE Defense, Security, and Sensing, Baltimore, 2012; Baltimore, p 837314. 36. Huang, N.; Lin, C.; Povinelli, M. L. J. Opt. 2012, 14, (2), 024004. 37. Huang, N.; Lin, C.; Povinelli, M. L. J. Appl. Phys. 2012, 112, (6), 064321. 38. Wen, L.; Zhao, Z.; Li, X.; Shen, Y.; Guo, H.; Wang, Y. Appl. Phys. Lett. 2011, 99, (14), 143116. 39. Madaria, A. R.; Yao, M.; Chi, C.; Huang, N.; Lin, C.; Li, R.; Povinelli, M. L.; Dapkus, P. D.; Zhou, C. Nano Lett. 2012, 12, (6), 2839-2845. 40. Lin, C.; Povinelli, M. L. Opt. Express 2009, 17, (22), 19371-19381. 41. Lin, C.; Povinelli, M. L. Opt. Express 2011, 19, (SS), All48-All54. 42. Lin, C.; Huang, N.; Povinelli, M. L. Opt. Express 2012, 20, (Sl), Al25-Al32. 43. Kelzenberg, M. D.; Boettcher, S. W.; Petykiewicz, J. A.; Turner-Evans, D. B.; Putnam, M. C.; Warren, E. L.; Spurgeon, J.M.; Briggs, R. M.; Lewis, N. S.; Atwater, H. A. Nat. Mater. 2010, 9, (3), 239-244. 44. Heurlin, M.; Wickert, P.; Falt, S.; Borgstrom, M. T.; Deppert, K.; Samuelson, L.; 87 Magnusson, M. H. Nano Lett. 2011, 11, (5), 2028-2031. 45. Colombo, C.; Hei~, M.; Griitzel, M.; Fontcuberta i Morral, A. Appl. Phys. Lett. 2009, 94, (17), 173108. 46. Dufouleur, J.; Colombo, C.; Garma, T.; Ketterer, B.; Uccelli, E.; Nicotra, M.; Fontcuberta i Morral, A. Nano Lett. 2010, 10, (5), 1734-1740. 47. Allen, J.E.; Hemesath, E. R.; Perea, D. E.; Lensch-Falk, J. L.; LiZ.Y; Yin, F.; Gass, M. H.; Wang, P.; Bleloch, A. L.; Palmer, R. E.; Lauhon, L. J. Nat. Nanotechnol. 2008, 3, (3), 168-173. 48. Storm, K.; Halvardsson, F.; Heurlin, M.; Lindgren, D.; Gustafsson, A.; Wu, P. M.; Monemar, B.; Samuelson, L. Nat. Nanotechnol. 2012, 7, (11), 718-722. 49. Kushibe, M.; Eguchi, K.; Funamizu, M.; Ohba, Y. Appl. Phys. Lett. 1990, 56, (13), 1248-1250. 50. Chang, C.-C.; Chi, C.-Y.; Yao, M.; Huang, N.; Chen, C.-C.; Theiss, J.; Bushmaker, A. W.; LaLumondiere, S.; Yeh, T.-W.; Povinelli, M. L.; Zhou, C.; Dapkus, P. D.; Cronin, S. B. Nano Lett. 2012, 12, (9), 4484-4489. 51. Talin, A. A.; Leonard, F.; Swartzentruber, B. S.; Wang, X.; Hersee, S. D. Phys. Rev. Lett. 2008, 101, (7), 076802. 52. Katzenmeyer, A. M.; Le; x; onard, F.; Talin, A. A.; Toimil-Molares, M. E.; Cederberg, J. G.; Huang, J. Y.; Lensch-Falk, J. L. IEEE. T Nanotechnol. 2011, 10, (1), 92-95. 53. Cheung, S. K.; Cheung, N. W. Appl. Phys. Lett. 1986, 49, (2), 85-87. 54. Seyedi, M.A.; Yao, M.; O'Brien, J.; Wang, S. Y.; Dapkus, P. D. Appl. Phys. Lett. 2013, 103, (25), 251109. 55. Zhao, J.; Wang, A.; Altermatt, P. P.; Wenham, S. R.; Green, M.A. Sol. Energy Mater. Sol. Cells 1996, 41-42, 87-99. 88 Chapter 4 Heteroepitaxial Growth of GaAs Nanowires on Si Substrates and Twin Formation Mechanism 4.1 Introduction In recent years, semiconductor nanowires have attracted great research interest due to their umque electronic and optical properties 1 - 5 _ III-V compound semiconductor nanowires possess advantages of high mobility, direct band gap and the capability of bandgap engineering over a wide range, making them promising candidacies for future 1 . 6-14 l . 15-25 d d · 26-30 o h h h d s· · ·11 h e ectron1c , optoe ectron1c an energy ev1ces . n t e ot er an , 1 1s st1 t e dominant material in today's semiconductor industry because of its relatively low cost and mature processing technology. It thus has been a natural and everlasting desire to integrate III-V semiconductors with Si to enable novel functional devices which can take advantage of the benefits offered by both ends. This technology can revolutionize the integrated circuit industry by the monolithic integration of III-V and Si circuitries and high-speed on-chip optical communication. Of no less importance to the energy frontier research is the development of multijunction solar cells using III-V materials grown on 1 d . .d s· b 31 32 ow-cost an ng1 1 su strates · . To achieve such integration m a thin fihn fashion represents a great technological challenge caused by several problems existing at the interface of the two dissimilar materials. Lattice constant and thermal expans10n coefficient mismatch would lead to 89 misfit dislocation or even non-epitaxial growth 33 ' 34 , Polar nature of III-V materials makes the unidirectional growth on nonpolar Si not so straightforward and causes anti-phase boundaries (APBs ) 35 - 38 _ Native oxide layer commonly observed on Si surface is another major factor hindering the epitaxial growth 39 - 41 _ Benefiting from the small lateral dimension of nanowires, strain due to lattice mismatch can effectively relieve in the lateral directions 42 - 45 and chance of forming APBs is remarkably reduced_ High-quality single-crystal III-V nanowires grown on Si substrate have been successfully demonstrated 3941 46-10 A 'd f l' - h 1 b 1 d - 1 d' over recent years ' . w1 e range o app 1cat1ons ave a so een exp ore inc u 1ng high speed transistors 7 , tunnel diodes 71 ' 72 , light emitting diodes (LED) 15 ' 16 ' 25 , room temperature lasers 17 and solar cells 73 _ Vapor-liquid-solid (VLS) growth method was first developed by Wagner and others in 1960s and 1970s 74 , which has become the most prevalent approach to grow III-V nanowires on Si substrates_ Au is widely used as the catalyst due to the ease of preparation and low reactivity, Vertical nanowire arrays have been successfully obtained through Au-catalyzed VLS in molecular beam epitaxy (MBE) 57 ' 67 ' 70 and metalorganic chemical vapor phase epitaxy (l'v!OCVD) 39 ' 54 ' 56 ' 58 ' 59 ' 69 _ On the other hand, controversy regarding the incorporation of Au during the growth 75 - 77 leads to the speculation of creating deep levels acting as recombination centers 78 _ Group III atom self-catalyzed VLS growth has thus attracted attention since no foreign metal intermixing is involved in this h d 47-51 53 60-63 H h 1 d 1 1 - - h h f - 1 met o ' ' _ owever, t e meta rop et p aces constram m t e growt o axrn heterostructure with different group III elements_ The droplet usually remains after 90 growth due to the VLS nature and is undesired in some application such as efficient light extraction in LEDs and good ohmic contact in solar cells. Non-VLS methods were achieved until recently by oxide-assisted self-induced growth 64 • 65 and selective area growth (SAG) 40 · 41 . In SAG, nanowires are located precisely at a pre-defined array of holes and no catalyst is required. The high uniformity of nanowire morphology allows the fabrication of versatile devices at a controllable level and photodetectors 24 , LEDs 79 and solar cells 26 · 30 have been achieved using this method. Nevertheless rotational twins and polytypism are common defects observed in both VLS and SAG grown nanowires. These defects are believed to be directly accountable for altered band structure in nanowires compared to their bulk counterpart. Carrier scattering, local quantum confinement and non-intrinsic carrier transition process are incurred due to the band discontinuity between zincblende (ZB) and wurtzite (WZ) regions 80 - 84 . Understanding of the driving force of twins and control of its presence has become a topic under intense research. Substantial progress has been achieved for VLS mode with a general agreement that the h d · h · I h · rf 1· · h d · f 47 49 85-93 Th t ermo ynam10s at t e tnp e-p ase mte ace me 1s t e ommant actor · · . e understanding of twin formation mechanism m SAG mode is emerging and was interpreted from several different perspectives including faceting growth 94 and h d . f I . 83 95 t ermo ynam1cs o nuc eat1on ' . Here, we report direct heteroepitaxial growth of GaAs nanowires on Si (l ll) substrates using SAG. In spite of 4.1 % lattice mismatch between GaAs and Si, highly uniform nanowire arrays are obtained on electron beam lithography (EBL) patterned substrates. 91 No complex buffer growth or surface conditioning is required, except for hydrogen annealing for removal of native oxide. Study of early nucleation found that unlike the growth on GaAs (111 )B substrate, the initial growth at the heterointerface adopts Vohner - Weber (VW) mode and the growth rate among different crystal orientations depends on growth temperature and group V to group III species (V /III) ratio. Transmission electron microscopy (TEM) of nanowire cross sections reveals rotational twins parallel to the (111 )B plane. Twin density shows strong temperature dependence. Based on the morphology of nanowires grown at high temperature we propose a model to illustrate the growth process involving twin formation. The presence of twin defects is understood in a coherent context of competing facets growth and thermodynamics of nucleation process that is sensitive to growth conditions. 4.2 GaAs nanowire grown on Si (111) substrate using SAG Silicon nitride layer is first deposited on double side polished epi-ready Si (Ill) substrates using PECVD. EBL or photolithography (ASML) followed by dry (CF 4 ) or wet etching (BOE) open up an array of holes in the silicon nitride layer. No difference in vertical nanowire yield is observed between dry and wet etching. Uniform nanowire array can even be obtained on substrates with CF 4 over-etched into Si by 5 nm. After resist strip-off and oxygen plasma cleaning, samples are quickly dipped in 7:1 BOE for 3 seconds, blown dry with nitrogen and immediately loaded into reactor. Nanowires are then grown in a vertical Thomas Swan MOCVD with shower head at 0.1 atm pressure. 92 TMG, and AsH 3 are used as the precursors for Ga and As. The total flow rate of carrier gas is 7 standard liters per minute (SLM) and partial pressure for TMG and AsH 3 are 7.56x10- 7 atm and 2.14xl0-4 atm mtless otherwise stated. Prior to actual nanowire growth, the substrate is annealed in hydrogen ambient for 5 minutes at 925 °C. This step is found to be essential because growth with no hydrogen annealing or annealing at lower temperature results in increased number of irregular growth (Fig 4.1 ). Figure 4. 1 Top view SEM images of GaAs nanowire arrays grown on Si with different hydrogen annealing times. Other conditions are all the same as described in text. Lower annealing time reduces vertical nanowire yield. Then the temperature was decreased to nanowire growth temperature between 700 and 790 °C and TMG and AsH 3 are opened simultaneously. After growth TMG is closed while AsH 3 is kept on until temperature is below 300 °C to prevent decomposition at high temperature. Typical growth condition is summarized in Fig 4.2a. Under this condition nanowues grow in the vertical direction which conesponds to the GaAs (lll)B 93 orientation. Fig 4.2b shows SEM image of nanowires grown at 760 °C taken at a 30° tilted angle. All the wires are located uniformly in the pre-defined template and exhibit 6-fold symmetric cross section consisting of {1-10} sidewall facets. Yield of vertical nanowires is 100% (Fig 4.2c ). a b e - 92 AsH3 P 16 - ~ TMG ~ ~ ., c. ~ ~y ~ ~~ ~ ,· I . E ~30 ~ ~ ~~ I . ' ~ ~ ~ ium 1 Time c • <I •• • ....... •• •• •• < ~ ••• • •• ·.-··.::·.·.·· • ..... . .............. :. . .............. . ·. ·:. ': · :. ·: ·. ·. ··:· .. .. · .. · ... · ... · .. . ·. . . . . . . . ..... · ....... ·. ::·. ·. ··:·. ·. ' : : .. · .. ·. ·. ··:· .. ·. ·. . . . . . . . . -~\%}{'.t\@~~~~~;m '··:·.:·. · ...... ··:· ........ · ...... · •. · ......... ': ·.~ .. · .. ~:·. :. ·• •• :: ............................. • •• • •• • •• • •• ·: : •• •• . .µ, . • • d Figure 4. 2 (a) Temperature profile of GaAs nanowire growth on Si (111). Red and blue blocks indicate the time AsH 3 and TMG are supplied, respectively. (b) 30°tilted and ( c) top view SEM images of uniform GaAs nanowire arrays grown on Si. ( d) Schematic diagram of crystal lattice at GaAs/Si interface, viewed from [1-1 O] orientation. ( e) HRTEM image taken at the GaAs/Si interface. Arrows indicate sites with misfit dislocation. The lattice constants of Si and GaAs are different by 4.1 %. GaAs nanowires with small enough footprint can however be grown on Si coherently without generating misfit dislocations (MFDs) at the interface. Such MDF free interface was observed by Tomioka et al. when they grew GaAs on Si with openings smaller than 10 nm 40 . For larger nanowires, the increased strain energy would lead to MFDs once the nanowires are beyond a critical size. Previous calculations based on continuum elasticity have estimated 94 the critical height and diameter of NW for MFD generation and shown considerable stress relief through MFDsoo. The nanowires discussed in this paper fall into this category. To exam the crystal structure of the nanowires and the heteroepitaxial interface we cut a 70 nm thin slice that is parallel to the (1-10) crystal planes from the center of a nanowire using focused ion beam (FIB, JEOL). High resolution TEM (HRTEM) images are taken with [1-1 OJ zone axis. In Fig 4.2e high density of twins are observed immediately after the growth starts for nanowires grown at 760 °C. The size of our nanowires is above the critical diameter and height for MFD free growth so periodic MFDs are observed at the GaAs/Si interface together with periodic contrast variation on the GaAs side due to strain distribution. Average period ofMFD is 8.45 nm which agrees with calculated value based on lattice constant of GaAs and Si. As shown in Fig 4.2d the distance between two adjacent As atoms in the [11-2] direction is 3.47 A and the lattice mismatch is 4.1 % so the period ofMFD is 3.47 A /0.041 ~8.46 nm. The measured period also agrees with the value calculated by Yuan et al. using molecular-dynamics and quantum-mechanics combined simulations% 95 Figure 4. 3 Top view SEM images of nanowire arrays grown with and without low temperature nucleation step prior to the nanowire growth. Nucleation is 4 minutes ifthere IS. To initiate coherent epitaxial growth of GaAs nanowires on Si, previous studies indicate a nucleation step is often required prior to the nanowire growth. For the Au-catalyzed VLS growth, Kang et al. found vertical nanowires can only be obtained with high temperature nucleation followed by low temperature wire growth 59 . ill the case of SAG, Tomioka et al. pointed out Si surface needs to be soaked in As ambient first to form Si (lll):As surface reconstruction which is an analogue of GaAs or InAs (lll)B surface. To prevent this surface from As evaporation, low temperature nucleation has to been conducted p1ior to ramping up temperature for nanowire growth 40 • 41 . However, in our case we used single step growth without As treatment or nucleation step. As shown in Fig 4.3 nucleation step at lower temperature only decreases the vertical nanowire yield. Furthermore, high yield vertical nanowires can be grown even on Si surface damaged by dcy etching showing the 96 robustness of our method (Fig 4.4). Figure 4. 4 TEM image at the nanowire/Si interface. The struiing point of nanowire is deeper than the Si substrate surface indicated by red dashed line because of over-RIE. One possible mechanism that allows us to grow without nucleation could be the relatively lai-ger size of holes in our study compared to those rep01ted in other papers. The size of the holes for EBL and photolithography prepru·ed samples are ru·ound 100 and 200 nm respectively. Lruger holes might help to collect more sources. Because the initial growth adopts VW (island) mode, as will be shown later, the abundant sources in the larger holes can ensure adequate local supply of adatoms toward the nucleus and prevent it from decom posi ti.on. 97 a b d (-1-12) "'- (-211) (1-21) r ! {-12-1) AsH 3 partial pressure (atm) e (10-1) {11-2) Figure 4. 5 30° tilted SEM images at the holes on (a) GaAs (lll)B and (b) Si (111) substrates after 2 minutes growth at 790 °C. ( c) Top view SEM images of nuclei after 2 minutes growth under different temperature and AsH 3 partial pressure. TMG partial pressure is kept constant at 7.56x10- 7 atm. (d) Schematic of relevant low-index planes viewed from (111 )B direction. ( e) Projection of relevant low-index planes on (1-10) plane in GaAs zincblende lattice. To better understand the growth behavior at the initial stage we grow nanowires for only 2 minutes and exam the nucleus morphology under SEM. Fig 4.5a and b are 30° tilted SEM images at the holes on GaAs (lll)B and Si (111) substrates after 2 minutes growth at 790 °C and AsH 3 partial pressure of 2.14x10- 5 atm (equivalent V/III ratio= 283). The 98 material deposited on GaAs substrate fills an entire hole with flat surface. In contrast, the nuclei on Si substrate appear to be islands and one can tell the crystal facets. The classical picture to depict different film growth mechanisms is 97 : layer-by-layer growth (Frank - van-der-Merve, or FM, mechanism), islanding growth (Volmer - Weber, or VW, mechanism), and the Stranski - Krastanov (SK) (an initially continuous film becomes islanded but with a thin continuous 'wetting' layer left). The interplay among surface energies of the substrate (y~), the film (yr), and their interface (yi) decide the actual growth mechanism. If the substrate surface energy is smaller than the sum of the two other surface energies, the island growth (VW) occurs because exposed substrate surface is energetically favorable: Y~ <Yr+ Yi, whereas the reverse inequality: y~>rr+yi 1s associated with layer-by-layer growth (FM). Early study found GaAs-on-Si growth occurs in VW mode 98 - 100 . As passivated Si (e.g. Si (lll):As surface mentioned earlier) was found to be highly inert (low y~) due to the existence of As long pair states and is even shown to be almost unaffected by exposure to oxygen or air 1 m In addition, forming 3D islands also decreases free energy due to the decrease in elastic deformations at the island tops which was first proposed by Asaro and Tiller 102 and Grinfel'd 103 . We then further vary growth temperature (Tg) between 670 and 790 °C. AsH 3 partial pressure (P A>HJ) is varied between 2.14x10· 5 and l.43xl0- 3 atm while TMG partial pressure is kept constant at 7.56xl0- 7 atm. Fig 4.5c shows top view SEM images of typical nuclei of each 99 growth condition. All the conditions result in VW growth and hexagonal cross section emerges after such a short time. However, the growth rates along different orientations show strong dependence on both Tg and P A>HJ· In general, lateral growth is suppressed under lower P A>HJ or higher Tg while the vertical growth is enhanced under these conditions. We believe this is related to the different Tg and P A>HJ dependence of the growth rate for different crystal planes. Fig 4.5d shows relevant low-index planes viewed in the (lll)B direction and Fig 4.5e shows their orientations in relation to the zincblende GaAs crystal lattice viewed from [1-1 OJ direction. In crystal growth, the visible facets appear because of their slow growth rate. At higher Tg and lower P A>HJ (those within the blue background in Fig 4.5c), the growth in (lll)B direction is faster than in the <1-10> directions, so adatoms migrate to the top of a nucleus and contribute to the vertical growth. The lateral growth is instead suppressed and clear { 1-10} sidewall facets can be seen due to the slow growth. The tilted {-1-10} facets have the same surface atomic configuration as that of { 1-10} sidewalls but because of adatoms accumulate at the top, the growth in tilted <-1-1 O> directions is still faster than in the horizontal <1-1 O> directions and top surface is usually flat. However, in the extreme case of highest Tg and lowest P A>HJ the nuclei tend to be pyramid enclosed by tilted {-1-10} facets meaning the growth rate in those directions are significantly reduced also. In contrast, when Tg decreases and P A>HJ increases, the growth rate in (lll)B direction decreases and that along the <1-1 O> increases. The nuclei under these conditions (those within the pink background in Fig 4.5c) extend further laterally and show much smaller height than those grown under higher Tg and lower P A>HJ· Due to the increased growth rate in <1-1 O> 100 directions the hexagonal cross section is less obvious and multiple nuclei are observed. Such strong correlation between growth rate and growth condition is consistent with the morphology change observed by Yoshida et al. when they also varied Tg and P A>HJ for nanowire growth 94 . Variation of growth rate in the (111 )B direction is believed to be dominated by the surface reconstruction that is related to As surface enrichment and a strong function of Tg and P A>HJ· Ultra high vacuum study found at low Tg or high P A>HJ, (2x2) reconstruction takes place which features a chemically stable As-trimer structure leading to reduced growth rate in (lll)B direction and at high Tg or low P A>HJ (>ll9x>ll9) reconstruction appears 104 - 109 . Nishida et al. reported the same phase transition on GaAs (lll)B surface during MOCVD growth 110 . 4.3 Twin formation mechanism in SAG and effect of growth condition Not only does the growth condition changes the nucleation behavior but also affects the crystal structure in terms of twin defects. The highest growth temperature that allows nanowire to grow by the aforementioned direct heteroepitaxy is 800 °C. Above this temperature, no GaAs can be deposited inside the holes due to increased Ga diffusion length. However we are able to grow nanowires at 850 °C by starting growth at 760 °C for 5 minutes and then quickly ramping up the temperature to 850 °C. However, as shown in Fig 4.6a, nanowires become much less uniform in height and uneven nanowire tops appear with tilted {-1-10} facets and other low-index facets to various degrees. Some nanowires have a completely pinched-off tetrahedron tip while some others have a 101 triangulaT-shaped thin mesa sUJTounded by three tilted { -1-10} facets. Based on the various nanowire tip morphologies (Fig 4.6b) we then propose a model to demonstrate the evolution of the growth front and the mechanism driving the twin formation. a b c vi i vii Figure 4. 6 (a) 30° tilted SEM images of nanowires grown at 850 °C, the initial part is grown at 760 °C to help nucleate. (b) SEM of images of nanowire tips with different morphologies. Images are 01ganized in a sequence according to the proposed twin fonnation model in (c). (c) Schematic of twin formation process during nanowire growth. (d) SEM image of a nanowire tip. After even number of twins form the wire pinches off (c) SEM image of a nanowire tip with odd number of twins. (f) TEM image of a region consists of even number of twins. The crystals at two sides of transitional region shaTe the same atomic registry. (g) TEM image of a region consists of odd number of twins. The crystal at two sides of transitional region can be considered as rotated by 180 ° compared to each other. At temperature as high as 850 °C, the growth rate in <-1-10> directions is extremely low so when a nanowire grows in the vertical (lll)B direction the three tilted {-1-10} facets extend toward the center from three corners at the tip. Hence, the top (1 ll)B facet is intersected by the these tluee facets and appears to be a triangle (Fig 4.6c, stage i). This 102 triangle keeps shrinking while the {-1-10} facets further extend (Fig 4.6c, stage ii) till it reaches certain critical dimension, around which the probability of forming a rotational twin, upon the deposition of the next rnonolayer, would be dramatically increased (Fig 4.6c, stage iii). If no twin forms, then the tip will pinch off and the nanowire growth terminates. Otherwise the triangular (lll)B facet will stop shrinking but instead expand laterally as a mesa after a twin is introduced (Fig 4.6c, stage iv). When a twin is formed at the interface of the nucleus and the substrate, the crystal lattice of the nucleus can be considered to be rotated by 60° (or 180° given the three-fold symmetry) azimuthally in relation to that of the substrate. The tilted <-1-1 O> directions of the substrate are thus no longer the same slow-grow directions for the nucleus and lateral growth rate in those directions become considerable. The twined mesa will fill up all the space above the original {-1-10} facets making the nanow!fe top flat (Figure 3c, stage v to vii). Sometimes another slow-grow facet, (lll)A facet, will appear temporarily before the mesa forms a complete flat hexagonal top (Fig 4.6c, stage vii). When the growth proceeds, three { -1-10} facets of the newly grown crystal above the twin will start to emerge at the three comers different from those where the original { -1-10} facets stern. At this point the (111 )B facet starts to shrink again and the growth will repeat the whole process mentioned above. Through this repeating growth sequence, semi-periodic twins are introduced during the nanowire growth. We also notice that around the critical size of (111 )B triangle the twined mesa is metastable. After the first twin forms, multiple consecutive twins are also highly likely to take place. If even numbers of the twins form before the mesa eventually stabilize (Fig 4.6f), the new stable crystal share the same 103 lattice orientation of the original crystal underneath the first twin. Thus three {-1-1 O} facets emerge :from the edges of the triangle mesa and will finally pinch off the tip (Fig 4.6d). Similar to the effect of a single twin, odd number of twins (Fig 4.6g) will rotate the lattice and allow the stabilized mesa to expand laterally (Fig 4.6e). g 0.40 Length distribution of twin 0 .35 free segment s 0 .30 - 95nm - 180nm i 0 .25 - 280nm 0.20 0 .15 0 .10 0.05 0 .00 10 20 30 40 50 60 70 80 Length (nm) Figure 4. 7 (a)-(c) TEM images of nanowire grown at different temperatures: (a) 760 °C, (b) 790 °C and (c) 850 °C. The initial segment in (c) is grown at 760 °C in order to nucleate. (d)-(f) TEM images of nanowires grown at 850 °C with different diameters: (d) 95 nm, (e) 180 nm and (f) 280 nm. (g) Histogram showing length distribution of twin :free segments of nanowires grown at 850 °C with different diameters. 104 According to this model, twin density or length of a twin-free segment is determined by how soon a flat hexagonal (111 )B top surface shrinks to the triangle of critical dimension. In another word, the critical triangle size and the nanowire diameter together determine the length of a single twin-free segment. Fig 4.7a to c are TEM images of nanowires grown at different temperatures and the different crystal orientation between two segments separated by a twin is manifested by the brightness contrast. Distinct twin densities indicate strong temperature dependence of the growth characteristics. For nanowire grown at 760 °C (Fig 4.7a), twin forms every few mono layers while twin-free segments can extend longer than I 0 rnn in nanowires grown at 790 °C (Fig 4. 7b ). The difference is even more striking in Fig 4. 7c where we have the initial part grown at 760 °C and later part grown at 850 °C. An abrupt interface exists, across which the twin density changes dramatically. In fact, the twin in the initial part is so dense that in many cases it forms every mono layer which makes the crystal to be effectively in wurtzite structure. As will be discussed in detail in the next section, higher temperature features a smaller critical dimension while in the case of lower temperature the probability of forming twin is considerable even when the (111 )B facet is pretty large. Because the probability of forming twin increases when the tip is around the critical dimension, larger diameter would favor less twins as it takes longer time for the tip to reach the critical size. Our observation on nanowires with different diameters but grown at the same temperature agrees with this trend. Fig 4.7d to fare TEM images of nanowire with 95, 180 and 280 nm diameter controlled by the hole size in the silicon nitride mask. Although they are all grown at 850 °C their distributions of length of twin-free segments are substantially 105 different. As shown by the statistics in Fig 4.7g, the length of twin-free segment shifts toward the longer side with increasing diameter. The average lengths for nanowires of 95, 180 and 280 nm diameters are 17.9, 39.1 and 53.1 nm respectively. Gaussian peak fittings are superimposed in the histogram which show the peaks shift to the right side with increasing diameter. The nature of the twin peaks for 180 nm nanowrre is not fully understood at this point. It might suggest at this transitional size twin density features a broader distribution. Based on the model mentioned earlier, the maximum length of a twin-free segment can only be reached when the tip is completely pinched off. Given that and the tetrahedron enclosed by three {-1-10} facets, the maximum length would be 33.6, 63.6 and 99 nm for nanowires of 95, 180 and 280 nm diameters. All the lengths we measure are below the theoretical limit with only one exception for the 95 nm nanowire as can be seen in Fig 4.7g (one data falls into between 40 and 50 nm). This discrepancy could be due to the fact that the TEM sample might not be cut right through the center of a nanowire so the actual diameter is larger than what we observed under TEM. Also we assumed zero growth rate in <-1-1 O> directions while it can be inegligible in the real situation. 4.4 Thermodynamic model of twin probability To calculate the probability distribution P(h) of the twin-free segment length h based on the mechanism outlined in Figure 3, we use a simple nucleation-growth model. 1 We consider a hexagonal GaAs nanowire with diameter D, in which a new twin-free segment 106 starts to grow at height h = 0 to form a tetrahedral island on the (111 )B top surface. At height h, the next GaAs bilayer of area A(h) can grow with either the zinc-blende (zb) or fault (f) stacking. The change of the Gibbs free energy for the growth of the next bilayer of area A is given by (Jc= zb or f), (Eq 4.1) Where s;'" and £~gn are the areal and peripheral energy densities of the bilayer and µgm ( T) is the chemical potential of the vapor, which depends on the temperature Tin addition to the vapor pressures of the reactant gases. o;(A) takes a maximum value, G, 1 =G;(A.), at a critical area A'. The Gibbs free energy used to obtain the twin statistics is then G' ( h) = {~;(A( h )) (A(h)s:A.) (A(h)>A.) (Jc= zb orf). In the case of fault bilayer, we also take into account the twin-twin interaction, Gf (h) = Gf (h) +sm, (h)A(h), (Eq 4.2) (Eq 4.3) where £'", (h) is the twin-twin interaction energy density. The probability to find a twin at height h is calculated as p(h)= exp[-Gr(h)/ kBT] , exp[-Gr(h)/ kBT]+exp[-G'b(h)/ kBT] (Eq 4.4) where k 8 is the Boltzmann constant. The probability for the first occurrence of a stacking fault at the n-th bilayer is then n-1 P(nf'..) = f1[1- p(i/'..)]p(nf'..), (Eq 4.5) 1=1 where L1 is the distance between consecutive bilayers. Fig 4.8 shows the calculated P(h) for three diameters, D = 95, 180 and 280 nm. At 107 temperature T = 850 °C, the calculated distribution in Fig 4.8a exhibits a sharp peak for every diameter, where the peak position is an increasing function of D. Accordingly, the average twin-free segment length is 19, 37 and 53 nm for D = 95, 180 and 280 nm, in reasonable agreement with the experimental observation. In contrast, at 760 °C, the distribution is nearly independent of D and decays rapidly to zero within the first few bilayers (see Fig 4.8b). . . 0.07 i- (a) . 850°c 0.06 .. 95nm 180nm 0.05 I- . '.C' 0.04 .. 280nm Q ~ 0.03 ... 0.02 .. 0.01 I- . ) J ) 0 \ 0 20 40 60 80 100 h(nm) 0.6 (b) 160°c 0.5 ~ + 760-280 0.4 x 760-180 L 760-95 '.C' 0.3 Q * 0.2 0.1 tJs tJs * 0 0 0.5 1 1.5 2 h(nm) Figure 4. 8 Probability distribution of the twin-free segment length in GaAs nanowires of the diameter 95, 180 and 280 nm at temperature (a) 850 °C and (b) 760 °C. 4.5 Summary To summarize we report direct heteroepitaxial growth of GaAs nanowires on Si (111) substrate. Highly uniform nanowire arrays can be obtained with careful tuning of growth 108 condition. The temperature for pre-growth annealing in hydrogen ambient is found to be critical for high yield of vertical nanowires. Also affecting the nucleation process 1s growth temperature and partial pressure of arsme. Distinct Wife mmphologies are observed for nanowires grown at lower and higher temperatures based on which a model is proposed to demonstrate the twin formation during the growth. The temperature dependence as well as diameter dependence is well interpreted by a mechanism of competing facet growth and thermodynamics of nucleation process. The result opens up great opportunity of heterostructure devices requiring epitaxial III-V material on Si substrate and successfully deepens our understanding of twin defects frequently observed . . 1n nanow1res. 109 References 1. Huang, Y.; Duan, X.; Cui, Y.; Lauhon, L. J.; Kirn, K.-H.; Lieber, C. M. Science 2001, 294, (5545), 1313-1317. 2. Gudiksen, M. S.; Lauhon, L. J.; Wang, J.; Smith, D. C.; Lieber, C. M. Nature 2002, 415, (6872), 617-620. 3. Lauhon, L. J.; Gudiksen, M. S.; Wang, D.; Lieber, C. M. Nature 2002, 420, (6911), 57-61. 4. Johnson, J. C.; Choi, H.-J.; Knutsen, K. P.; Schaller, R. D.; Yang, P.; Saykally, R. J. Nat Mater 2002, 1, (2), 106-110. 5. Parkinson, P.; Lloyd-Hughes, J.; Gao, Q.; Tan, H. H.; Jagadish, C.; Johnston, M. B.; Herz, L. M. Nano Lett. 2007, 7, (7), 2162-2165. 6. Thelander, C.; FrobergFroberg, L. E.; Rehnstedt, C.; Samuelson, L.; Wemersson, L. E. Electron Device Letters, IEEE 2008, 29, (3), 206-208. 7. Tornioka, K.; Yoshimura, M.; Fukui, T. Nature 2012, 488, (7410), 189-192. 8. Dayeh, S. A.; Aplin, D. P.R.; Zhou, X.; Yu, P. K. L.; Yu, E.T.; Wang, D. Small 2007, 3, (2), 326-332. 9. Egard, M.; Johansson, S.; Johansson, A. C.; Persson, K. M.; Dey, A. W.; Borg, B. M.; Thelander, C.; Wemersson, L. E.; Lind, E. Nano Lett. 2010, 10, (3), 809-812. 10. Kasai, S.; Hasegawa, H. Physica E: Low-dimensional Systems and Nanostructures 2002, 13, (2-4), 925-929. 11. Thelander, C.; Martensson, T.; Bjork, M. T.; Ohlsson, B. J.; Larsson, M. W.; Wallenberg, L. R.; Samuelson, L. Appl. Phys. Lett. 2003, 83, (10), 2052-2054. 12. Bryllert, T.; Wemersson, L. E.; Froberg, L. E.; Samuelson, L. Electron Device Letters, IEEE 2006, 27, (5), 323-325. 13. Ng, H. T.; Han, J.; Yamada, T.; Nguyen, P.; Chen, Y. P.; Meyyappan, M. Nano Lett. 2004, 4, (7), 1247-1252. 14. Thelander, C.; Nilsson, H. A.; Jensen, L. E.; Samuelson, L. Nano Lett. 2005, 5, (4), 635-638. 15.Svensson, C. P. T.; Thomas, M.; Johanna, T.; Christina, L.; Michael, R.; Dan, H.; Lars, S.; Jonas, 0. Nanotechnology 2008, 19, (30), 305201. 16. Guo, W.; Zhang, M.; Banerjee, A.; Bhattacharya, P. Nano Lett. 2010, 10, (9), 3355-3359. 17. Mayer, B.; Rudolph, D.; Schnell, J.; Morkiitter, S.; Winner!, J.; Treu, J.; Miiller, K.; Bracher, G.; Abstreiter, G.; Koblrniiller, G.; Finley, J. J. Nat Commun 2013, 4. 18. Saxena, D.; Mokkapati, S.; Parkinson, P.; Jiang, N.; Gao, Q.; Tan, H. H.; Jagadish, C. Nat Photon 2013, 7, (12), 963-968. 19. Yan, R.; Gargas, D.; Yang, P. Nat Photon 2009, 3, (10), 569-576. 20. Huang, M. H.; Mao, S.; Feick, H.; Yan, H.; Wu, Y.; Kind, H.; Weber, E.; Russo, R.; Yang, P. Science 2001, 292, (5523), 1897-1899. 21. Qian, F.; Li, Y.; Gradecak, S.; Park, H.-G.; Dong, Y.; Ding, Y.; Wang, Z. L.; Lieber, C. M. Nat Mater 2008, 7, (9), 701-706. 22. Duan, X.; Huang, Y.; Agarwal, R.; Lieber, C. M. Nature 2003, 421, (6920), 241-245. 23. Pettersson, H.; Triigardh, J.; Persson, A. I.; Landin, L.; Hessman, D.; Samuelson, L. 110 Nano Lett. 2006, 6, (2), 229-232. 24. Seyedi, M. A.; Yao, M.; O'Brien, J.; Wang, S. Y.; Dapkus, P. D. Appl. Phys. Lett. 2013, 103, (25), 251109. 25. Tomioka, K.; Motohisa, J.; Hara, S.; Hiruma, K.; Fukui, T. Nano Lett. 2010, 10, (5), 1639-1644. 26. Mariani, G.; Scofield, A. C.; Hung, C.-H.; Huffaker, D. L. Nat. Commun. 2013, 4, 1497. 27. Wallentin, J.; Anttu, N.; Asoli, D.; Huffman, M.; Aberg, I.; Magnusson, M. H.; Siefer, G.; Fuss-Kailuweit, P.; Dimroth, F.; Witzigmann, B.; Xu, H. Q.; Samuelson, L.; Deppert, K.; Borgstrom, M. T. Science 2013, 339, (6123), 1057-1060. 28. Cui, Y.; Wang, J.; Plissard, S. R.; Cavalli, A.; Vu, T. T. T.; van Veldhoven, R. P. J.; Gao, L.; Trainor, M.; Verheijen, M. A.; Haverkort, J. E. M.; Bakkers, E. P. A. M. Nano Lett. 2013, 13, (9), 4113-4117. 29. Hu, S.; Chi, C.-Y.; Fountaine, K. T.; Yao, M.; Atwater, H. A.; Dapkus, P. D.; Lewis, N. S.; Zhou, C. Energ. Environ. Sci. 2013, 6, (6), 1879-1890. 30. Yao, M.; Huang, N.; Cong, S.; Chi, C.-Y.; Seyedi, M.A.; Lin, Y.-T.; Cao, Y.; Povinelli, M. L.; Dapkus, P. D.; Zhou, C. Nano Lett. 2014, 14, (6), 3293-3303. 31. LaPierre, R.R. J. Appl. Phys. 2011, 110, (1), 014310. 32. Huang, N.; Lin, C.; Povinelli, M. L. J. Appl. Phys. 2012, 112, (6), 064321. 33. Fang, S. F.; Adomi, K.; Iyer, S.; Morko9, H.; Zabel, H.; Choi, C.; Otsuka, N. J. Appl. Phys. 1990, 68, (7), R31-R58. 34. Ishiwara, H.; Hoshino, T.; Katahama, H. Mater. Chem. Phys. 1995, 40, (4), 225-229. 35. Cohen, D.; Carter, C. B. Journal of Microscopy 2002, 208, (2), 84-99. 36. Komninou, P.; Stoemenos, J.; Dimitrakopulos, G. P.; Karakostas, T. J. Appl. Phys. 1994, 75, (1), 143-152. 37. Yu, B. B.; Pchelyakov, 0. P. Physics-Uspekhi 2008, 51, (5), 437. 38. Kawanami, H. Sol. Energy Mater. Sol. Cells 2001, 66, (1-4), 479-486. 39. Martensson, T.; Svensson, C. P. T.; Wacaser, B. A.; Larsson, M. W.; Seifert, W.; Deppert, K.; Gustafsson, A.; Wallenberg, L. R.; Samuelson, L. Nano Lett. 2004, 4, (10), 1987-1990. 40. Tomioka, K.; Kobayashi, Y.; Motohisa, J.; Hara, S.; Fukui, T. Nanotechnology 2009, 20, (14), 145302. 41. Tomioka, K.; Motohisa, J.; Hara, S.; Fukui, T. Nano Lett. 2008, 8, (10), 3475-3480. 42. Glas, F. Phys. Rev. B 2006, 74, (12), 121302. 43. Chuang, L. C.; Moewe, M.; Chase, C.; Kobayashi, N. P.; Chang-Hasnain, C.; Crankshaw, S. Appl. Phys. Lett. 2007, 90, (4), 043115. 44. Sburlan, S.; Dapkus, P. D.; Nakano, A. Appl. Phys. Lett. 2012, 100, (16), 163108. 45. Cirlin, G. E.; Dubrovskii, V. G.; Soshnikov, I. P.; Sibirev, N. V.; Samsonenko, Y. B.; Bouravleuv, A. D.; Harmand, J. C.; Glas, F. physica status solidi (RRL) - Rapid Research Letters 2009, 3, (4), 112-114. 46. Hertenberger, S.; Funk, S.; Vizbaras, K.; Yadav, A.; Rudolph, D.; Becker, J.; Bolte, S.; Diiblinger, M.; Bichler, M.; Scatpa, G.; Lugli, P.; Zardo, I.; Finley, J. J.; Amann, M.-C.; Abstreiter, G.; Koblmiiller, G. Appl. Phys. Lett. 2012, 101, (4), -. 47. Krogstrup, P.; Popovitz-Biro, R.; Johnson, E.; Madsen, M. H.; Nygard, J.; Shtrikrnan, 111 H. Nano Lett. 2010, 10, (11), 4475-4482. 48. Munshi, A. M.; Dheeraj, D. L.; Todorovic, J.; van Helvoort, A. T. J.; Wernan, H.; Firnland, B.-0. J. Cryst. Growth 2013, 372, (0), 163-169. 49. Cirlin, G. E.; Dubrovskii, V. G.; Sarnsonenko, Y. B.; Bouravleuv, A. D.; Durose, K.; Proskuryakov, Y. Y.; Mendes, B.; Bowen, L.; Kaliteevski, M.A.; Abram, R. A.; Zeze, D. Phys. Rev. B 2010, 82, (3), 035302. 50. Sandra, J. G.; Jonathan, P. B.; Ray, R. L. Semicond. Sci. Technol. 2013, 28, (10), 105025. 51. Fauzia, J.; Vincenzo, G.; Silvia, R.; Faustino, M. Nanotechnology 2008, 19, (27), 275711. 52. Mandi, B.; Stangl, J.; Martensson, T.; Mikkelsen, A.; Eriksson, J.; Karlsson, L. S.; Bauer, G.; uurnl; nther; Samuelson, L.; Seifert, W. Nano Lett. 2006, 6, (8), 1817-1821. 53. Paek, J. H.; Nishiwaki, T.; Yamaguchi, M.; Sawaki, N. physica status solidi (c) 2009, 6, (6), 1436-1440. 54. Bao, X.-Y.; Soci, C.; Susac, D.; Bratvold, J.; Aplin, D. P. R.; Wei, W.; Chen, C.-Y.; Dayeh, S. A.; Kavanagh, K. L.; Wang, D. Nano Lett. 2008, 8, (11), 3755-3760. 55. Wei, W.; Bao, X.-Y.; Soci, C.; Ding, Y.; Wang, Z.-L.; Wang, D. Nano Lett. 2009, 9, (8), 2926-2934. 56. Aamoud, L. R.; Marcel, A. V.; Olaf, W.; Stacey, S.; Harry, W.; Erik, P. A. M. B. Nanotechnology 2006, 17, (11), S271. 57. Ihn, S.-G.; Song, J.-I.; Kirn, Y.-H.; Lee, J. Y. Appl. Phys. Lett. 2006, 89, (5), 053106. 58. Huang, H.; Ren, X.; Ye, X.; Guo, J.; Wang, Q.; Yang, Y.; Cai, S.; Huang, Y. Nano Lett. 2009, 10, (1 ), 64-68. 59.Kang, J. H.; Gao, Q.; Joyce, H.J.; Tan, H. H.; Jagadish, C.; Kirn, Y.; Choi, D. Y.; Guo, Y.; Xu, H.; Zou, J.; Fickenscher, M. A.; Smith, L. M.; Jackson, H. E.; Yarrison-Rice, J. M. Nanotechnology 2010, 21, (3), 035604. 60. Plissard, S.; Dick, K. A.; Wallart, X.; Caroff, P. Appl. Phys. Lett. 2010, 96, (12), 121901. 61. Munshi, A. M.; Dheeraj, D. L.; Fauske, V. T.; Kirn, D. C.; Huh, J.; Reinertsen, J. F.; Ahtapodov, L.; Lee, K. D.; Heidari, B.; van Helvoort, A. T. J.; Firnland, B. O.; Wernan, H. Nano Lett. 2014, 14, (2), 960-966. 62. Plissard, S.; Dick, K., A.; Larrieu, G.; Godey, S.; Addad, A.; Wallart, X.; Caroff, P. Nanotechnology 2010, 21, (38), 385602. 63. Kwoen, J.; Watanabe, K.; Iwamoto, S.; Arakawa, Y. J. Cryst. Growth 2013, 378, (0), 562-565. 64. Koblrniiller, G.; Hertenberger, S.; Vizbaras, K.; Bichler, M.; Bao, F.; Zhang, J. P.; Abstreiter, G. Nanotechnology 2010, 21, (36), 365602. 65. Rudolph, D.; Hertenberger, S.; Bolte, S.; Paosangthong, W.; Spirkoska, D. e.; Diiblinger, M.; Bichler, M.; Finley, J. J.; Abstreiter, G.; Koblrniiller, G. Nano Lett. 2011, 11, (9), 3848-3854. 66. Tateno, K.; Hibino, H.; Gotoh, H.; Nakano, H. Appl. Phys. Lett. 2006, 89, (3), 033114. 67.Ihn, S.-G.; Song, J.-I.; Kirn, T.-W.; Leern, D.-S.; Lee, T.; Lee, S.-G.; Koh, E. K.; Song, K. Nano Lett. 2007, 7, (1 ), 39-44. 112 68. Martensson, T.; Wagner, J. B.; Hilner, E.; Mikkelsen, A.; Thelander, C.; Stangl, J.; Ohlsson, B. J.; Gustafsson, A.; Lundgren, E.; Samuelson, L.; Seifert, W. Adv. Mater. 2007, 19, (14), 1801-1806. 69. Mohseni, P. K.; Maunders, C.; Botton, G. A.; LaPierre, R.R. Nanotechnology 2007, 18, (44), 445304. 70. Paek, J. H.; Nishiwaki, T.; Yamaguchi, M.; Sawaki, N. physica status solidi (c) 2008, 5, (9), 2740-2742. 71. Tomioka, K.; Fukui, T. Appl. Phys. Lett. 2011, 98, (8), 083114. 72. Yang, T.; Hertenberger, S.; Morkotter, S.; Abstreiter, G.; Koblmuller, G. Appl. Phys. Lett. 2012, 101, (23), 233102. 73. Heurlin, M.; Wickert, P.; Falt, S.; Borgstrom, M. T.; Deppert, K.; Samuelson, L.; Magnusson, M. H. Nano Lett. 2011, 11, (5), 2028-2031. 74. Wagner, R. S.; Ellis, W. C. Appl. Phys. Lett. 1964, 4, (5), 89-90. 75. Bar-Sadan, M.; Barthel, J.; Shtrikrnan, H.; Houben, L. Nano Lett. 2012, 12, (5), 2352-2356. 76. Perea, D. E.; Allen, J. E.; May, S. J.; Wessels, B. W.; Seidman, D. N.; Lauhon, L. J. Nano Lett. 2005, 6, (2), 181-185. 77. Allen, J.E.; Hemesath, E. R.; Perea, D. E.; Lensch-Falk, J. L.; LiZ.Y; Yin, F.; Gass, M. H.; Wang, P.; Bleloch, A. L.; Palmer, R. E.; Lauhon, L. J. Nat. Nanotechnol. 2008, 3, (3), 168-173. 78. Breuer, S.; Ffiiller, C.; Flissikowski, T.; Brandt, O.; Grahn, H. T.; Geelhaar, L.; Riechert, H. Nano Lett. 2011, 11, (3), 1276-1279. 79. Fukui, T.; Yoshimura, M.; Nakai, E.; Tomioka, K. AMBIO 2012, 41, (2), 119-124. 80. Parkinson, P.; Joyce, H. J.; Gao, Q.; Tan, H. H.; Zhang, X.; Zou, J.; Jagadish, C.; Herz, L. M.; Johnston, M. B. Nano Lett. 2009, 9, (9), 3349-3353. 81. Spirkoska, D.; Arbiol, J.; Gustafsson, A.; Conesa-Boj, S.; Glas, F.; Zardo, I.; Heigoldt, M.; Gass, M. H.; Bleloch, A. L.; Estrade, S.; Kaniber, M.; Rossler, J.; Peiro, F.; Morante, J. R.; Abstreiter, G.; Samuelson, L.; Fontcuberta i Morral, A. Phys. Rev. B 2009, 80, (24), 245325. 82. Thelander, C.; Caroff, P.; Plissard, S. b.; Dey, A. W.; Dick, K. A. Nano Lett. 2011, 11, (6), 2424-2429. 83. Yuan, Z.; Nomura, K.-i.; Nakano, A.Appl. Phys. Lett. 2012, 100, (16), -. 84. Ikejiri, K.; Kitauchi, Y.; Tomioka, K.; Motohisa, J.; Fukui, T. Nano Lett. 2011, 11, (10), 4314-4318. 85. Algra, R. E.; Verheijen, M. A.; Borgstrom, M. T.; Feiner, L.-F.; Irnmink, G.; van Enckevort, W. J. P.; Vlieg, E.; Bakkers, E. P. A. M. Nature 2008, 456, (7220), 369-372. 86. Algra, R. E.; Verheijen, M. A.; Feiner, L.-F.; Immink, G. G. W.; Enckevort, W. J.P. v.; Vlieg, E.; Bakkers, E. P.A. M. Nano Lett. 2011, 11, (3), 1259-1264. 87. CarotIP; Dick, K. A.; JohanssonJ; Messing, M. E.; DeppertK; SamuelsonL. Nat Nano 2009, 4, (1), 50-55. 88. Dick, K. A.; Bolinsson, J.; Messing, M. E.; Lehmann, S.; Johansson, J.; Caroff, P. J. Vac. Sci. Technol., B 2011, 29, (4), -. 89. Johansson, J.; Karlsson, L. S.; Patrik T. Svensson, C.; Martensson, T.; Wacaser, B. A.; 113 Deppert, K.; Samuelson, L.; Seifert, W. Nat Mater 2006, 5, (7), 574-580. 90. Johansson, J.; Dick, K. A.; Caroff, P.; Messing, M. E.; Bolinsson, J.; Deppert, K.; Samuelson, L. The Journal of Physical Chemistry C 2010, 114, (9), 3837-3842. 91. Joyce, H. J.; Wong-Leung, J.; Gao, Q.; Tan, H. H.; Jagadish, C. Nano Lett. 2010, 10, (3), 908-915. 92. Glas, F.; Harmand, J.-C.; Patriarche, G. Phys. Rev. Lett. 2007, 99, (14), 146101. 93. Dubrovskii, V. G.; Sibirev, N. V.; Harmand, J. C.; Glas, F. Phys. Rev. B 2008, 78, (23), 235301. 94. Yoshida, H.; Ikejiri, K.; Sato, T.; Hara, S.; Hiruma, K.; Motohisa, J.; Fukui, T. J. Cryst. Growth 2009, 312, (1), 52-57. 95. Joshua, N. S.; Andrew, L.; Christian, R.; Huffaker, D. L. Nanotechnology 2013, 24, (47), 475601. 96. Yuan, Z.; Shimamura, K.; Shimojo, F.; Nakano, A. J. Appl. Phys. 2013, 114, (7), -. 97. Toumie, E.; Ploog, K. H.J. Cryst. Growth 1994, 135, (1-2), 97-112. 98. Zinke - Allmang, M.; Feldman, L. C.; Nakahara, S. Appl. Phys. Lett. 1988, 52, (2), 144-146. 99. Biegelsen, D. K.; Ponce, F. A.; Smith, A. J.; Tramontana, J. C. J. Appl. Phys. 1987, 61, (5), 1856-1859. 100.Hull, R.; Fischer - Colbrie, A.; Rosner, S. J.; Koch, S. M.; Harris, J. S. Appl. Phys. Lett. 1987, 51, (21), 1723-1725. 101.Uhrberg, R. I. G.; Bringans, R. D.; Olmstead, M.A.; Bachrach, R. Z.; Northrup, J.E. Phys. Rev. B 1987, 35, (8), 3945-3951. 102.Asaro, R. J.; Tiller, W. A. Metal!. Trans. 1972, 3, (7), 1789-1796. 103.Grinfel'd, M.A. Dok!. Akad. Nauk 1986, SSSR 290, 1358-1363. 104.Biegelsen, D. K.; Bringans, R. D.; Northrup, J. E.; Swartz, L. E. Phys. Rev. Lett. 1990, 65, (4), 452-455. 105. Thornton, J. M. C.; Woolf, D. A.; Weightman, P. Appl. Surf Sci. 1998, 123-124, (0), 115-119. 106.Rajkumar, K. C.; Chen, P.; Madhukar, A. J. Appl. Phys. 1991, 69, (4), 2219-2223. 107. Woolf, D. A.; Westwood, D. I.; Williams, R. H. Appl. Phys. Lett. 1993, 62, (12), 1370-1372. 108.Avery, A. R.; Tok, E. S.; Jones, T. S. Surf Sci. 1997, 376, (1-3), L397-L402. 109.Chen, P.; Rajkumar, K. C.; Madhukar, A. Appl. Phys. Lett. 1991, 58, (16), 1771-1773. 110.Nishida, T.; Uwai, K.; Kobayashi, Y.; Kobayashi, N. Jpn. J. Appl. Phys. 1995, 34, (12R), 6326. 114 Chapter 5 tandem solar cells made by GaAs nanowires grown on Si 5.1 Introduction As has been discussed in previous chapters, several approaches are being pursued in order to surpass the Shockley-Queisser limit of conventional solar cells. One idea is to use multiple materials with different bandgaps to preferentially absorb different portion of the solar spectrum so that low energy photons can be utilized without causing significant thermalisation loss for high energy photons. Such devices are called multijunction solar cells. The concept was first introduced by the Research Triangle Institute and by Varian Research Center in the late 1970s to mid 1980s when dual junction devices were formed from an AlGaAs junction stacked or grown on top of a GaAs junction, and interconnected by a semiconductor tunnel junction'· 2 . In the same decade, another tandem device was proposed by the Solar Energy Research Institute (now the National Renewable Energy Laboratory) that consisted a Galn.P junction formed on top of a GaAs junction both grown on an inactive Ge substrate which formed a dual junction solar cell 3 . Later in 1990s, changes in the top cell thickness led to record efficiencies for dual junction and triple junction solar cells with Galn.P and GaAs both grown on top of an active Ge bottom cell substrate 4 - 6 . Terrestrial application under concentration was successfully demonstrated by Spectrolab with 32.2% efficiency 7 . The encouraging result then sparked a surge in efficiencies led by new device designs at Spectrolab. With metamorphic triple junction GalnP/Galn.As/Ge concentrator cell, they demonstrated a record efficiency of 40. 7%, the 115 first solar cell of any type to reach over the 40% efficiency millstone 8 . Even more important is the increase in performance has never slowed down. By using a combination of Galn.NAsSb/GaAs/In.GaP with bandgap of (1/1.4/1.9 e V), Solar Junction successively demonstrated a lattice matched triple junction solar cell with 44% efficiency. Recently Sharp hits a Fraunhofer-Institute-confirrned, world-record 44.4% for its triple junction solar cell made by inverted metamorphic technique. The top InGaP (1.9 e V) and GaAs (1.4 e V) cells are lattice matched to Ge or GaAs substrates while the bottom Galn.As cell is grown through a transparent buffer layer to accommodate the lattice difference between GaAs and Galn.As 9 . This year, Soitec reported 44. 7% efficiency from a wafer bonded 4-junction Gain.P/GaAs/Galn.AsP/Galn.P solar cell under 297X concentration 10 . This is the first time a 4-junction cell achieves efficiency higher than that of any triple junction solar cells which might lead to another surge of efficiencies later on. In large scale terrestrial implementation, system cost would be the key factor. III-V multijunction solar cells itself has no advantage in this sense. However at system level, multijunction concentrated PV could offer a low-cost path towards large scale installation. By using light concentration and high efficiency, the usage of expensive semiconductor materials is minimized. Assuming 500X concentration and three times higher efficiency, the area of multijunction solar cells themselves would be only 1/1500 of the total area of the Si cells producing the same power. Currently, multijunction solar cells of 37% efficiency cost between $8 and $10/cm 2 and more technical advances need to occur for multijunction solar cells to completely overcome cost barriers. 116 So far we have seen the wide application of III-V semiconductor nano wires. In particular, they have shown great potential in photovoltaics applications 11 - 16 . To dissociate light generated electron-hole pairs, p-njunctions need to be established within nanowires. Both . 1 14 " 16 d d" 1 12 13 . . . h b d d . h bl ax1a an ra 1a ' JUnct1on geometnes ave een emonstrate wit compara e success. In terms of multijunction solar cells, axial junction is definitely more promising. Similar to conventional multijunction solar cells, light can pass sequentially from large bandgap material to small bandgap material. The current match condition can be reached by tuning the length of each segment. Axial junction can also accommodate more strain between lattice-mismatched materials than radial junction due to the smaller heterogeneous interface. Strain relaxation between nanowire and substrate further allows the usage of low-cost substrate such as Si which would incredibly lower the overall cost of PV modules. In fact, as discussed in Chapter I, the optimum bandgap combination for a dual junction solar cell requires the bottom cell with 1.1 e V bandgap which makes Si an ideal active substrate. In Chapter 2 we have demonstrated GaAs nanow!fe solar cell with axial junction. In Chapter 3 we showed uniform GaAs nanow!fe arrays can be grown on Si substrate through SAG. In this chapter, we report on further developments in the GaAs nanowire-on-Si dual junction solar cells. Tunnel junction is a very important component in any monolithically integrated multijunction solar cells. Tunnel junctions are usually made by heavily doped p-n junction between two subcells so that the reverse biased 117 p+-n+ junction can conduct sufficient current during the normal operation of a solar cell. In section 5.2, we demonstrate the characteristics of n+ GaAs nanowire/p+ Si heteroepitaxial tunnel junction. We report a non-destructive optical method to characterize the doping level in n-type GaAs. In section 5.3, we discuss the fabrication and characterization of Si solar cell. In section 5.4, we show the performance of integrated dual junction solar cells. 5.2 Doping characteristics of n-type GaAs nanowire doped by Si A solar cell in operation works at forward bias. When two solar cells are connected in series the junction between emitter of the bottom cell and the base of the top cell is in however in reverse bias condition which will hinder the transport of light current. A good Ohmic connection with high conductivity and low optical absorption is required to pass through the current between two adjacent cells 17 . In 1958, Esaki measured the current of degenerately doped Ge diodes and showed that charge transport at small biases is completely governed by the laws of quantum tunneling through the space charge region of the junction 18 . This band to band tunneling effect has attracted particular interest in multijunction solar cell field. In most of today's high efficiency multijunction solar cells different sub-cells are connected by tunnel junctions made from degenerately III-V compound semiconductor. Typical GaAs tunnel junction requires doping concentration higher than 10 19 cm- 3 to have efficiency band to band tunneling 4 • 5 • 1 9 . While this is easy to achieve in thin film it poses a big challenge when the junction needs to be grown within a 118 nanow!fe. High density surface states often times reduce the carrier concentration in a nanowire to a level much lower than that required abovew When a GaAs nanowire on Si tandem solar cell is concerned tunnel junction should be formed at the interface of GaAs and Si. On the other hand a degenerately doped Si p+/n+ junction has low peak current since the tunneling efficiency in Si is low 21 . But promising results based on heterojunction band engineering and incorporation of IIIN materials have recently been reported. Considering a Si/III-V TFET, the combination of a Si drain channel with an InAs source is predicted to give the highest current 22 . Recent progress on heavily doped p-Si/n-InAs NW tunnel junction indicated very high tunnel currents, demonstrating their suitability as key components for high performance TFETs 23 - 25 . However InAs is a low bandgap material it will absorb light and reduce the light current from Si bottom cell. We thus try to use heavily doped n+ GaAs to replace InAs to form heteroepitaxial tunnel junction with p+ Si. Unlike InAs which always has Fermi level pinned in conduction band, GaAs need to be heavily doped to form tunnel junction. One big challenge in the field of nanowire solar cells field is to have an accurate assessment of the doping distribution within a nanowire. Due to the sub-micron size of nanowires standard characterization techniques used for bulk material is not directly applicable 26 . Most of the times people use the doping concentration on a film grown under same condition as nanowire to estimate the doping in nanowires. However doping incorporation depends strongly on the crystal orientation, growth rate and surface landscape which make this estimation questionable. Previous 119 studies have relied on field effect mobility measurements to extract the carrier density and carrier mobility from semiconductor nanowires. Unfortunately, field-effect measurements suffer from limitations in the geometries that are accessible and have many inherent uncertainties, of which the estimation of gate capacitance is of primary concern. Furthermore, performing spatially resolved field-effect mobility measurements presents serious technical challenges. Another option is to use atom probe tomography 27 • 28 to map out the dopant distribution of the nanowires. Although this method offers superior advantages in some regards, the technique is both time-consuming and complex and is difficult to implement on a routine basis. It has been well studied that heavy impurity incorporation results in notable changes in the semiconductor band structure. One effect is called band gap narrowing due to the formation of tails in density of states as a result of inhomogeneous dopant distributions 29 • 30 . The random distribution of impurities also disturbs the original periodic energy potential. Because the translational symmetry is broken, indirect k-nonconserving transitions become possible. Another important effect is the blue-shift of interband transition energy due to band filling in material with degenerate electron distribution, which is named Burstein-Moss effect after they reported in 1954 31 • 32 . These effects on the band structure are directly reflected in absorption and spontaneous emission spectra. Therefore photoluminescence (PL) analysis will be a useful and quantitative tool to provide insight into the doping concentration. To avoid the donor-acceptor transition, which is dominant in low-temperature PL 33 • 34 , our study is based on room temperature PL. 120 Significant PL spectra evolution is observed for nanowires doped with Si of different levels. Based on the analysis of spectra profile, information regarding band tail, Fermi level and carrier density can be deduced. Nanowires are grown by SAG MOCVD with condition mentioned Chapter 3. TMG, and AsH 3 are used as the precursors for Ga and As. The total flow rate of carrier gas is 7 standard liters per minute (SLM) and partial pressure for TMG and AsH 3 are 7.56xl0- 7 atm and 2.14x10· 4 atrn. Prior to actual nanowire growth, the substrate is annealed in hydrogen ambient for 5 minutes at 925 "C. Disilane is used as precursor of Si, which is n-type dopant for GaAs nanowires. The partial pressure of disilane is varied between 0 and 3.6x10· 9 atm. PL spectra are collected using a lOOX objective lens with 0.8 numerical aperture, an 1800 I/mm grating, and a silicon charged-couple device (CCD) detector to detect the photons in the range of 750 nm to 1000 nm. Samples are excited by 532 nm semiconductor diode laser with relatively low power ( ~ 1 W/cm 2 ) to avoid optical heating. 121 a b c - ••• • 1@fi1@@ 1 000® 1··••1®00$ 1000© ••• •1®00~ (000© ••••1@00@ c O®@© 20000 - 0 seem disilane d - 0 .01 e - 0 .1 - 0 .2 15000 -0.3 nooo ,-----------,.0.115 - 0 .4 c - 0 .5 :::J -0.6 0 10000 - 0 .7 (.) - 0 .8 _J a... - 1 5000 0 1.3 1.4 1.5 1.6 Photon energy (eV) 2<1000 tlOOO 150'° ~ "' 0 UllOO (,) ,2(100 ~ CD 1000G Q. --' 0.. "'" .... .... ,. .. 1.7 ---~-~~-~---.-' •• ;1 oi i>• oe 1 o Oisi ane now (seem) 0 110 0.105 0.100 0.095 >c 0.090 i 0.085 ~ 0.080 u. 0.075 0.070 Figure 5. 1 Top view SEM images of n-type GaAs nanowues grown with different disilane flow rate: (a) 0.2, (b) 1 and (c) 2.5 seem. Scale bat" is 1 µm. (d) PL spectra of GaAs nanowire grown with different disilane flow rate. (e) Peak intensity and width (at halfmaximmn) of spectra in (d). Fig 5.1 (a) to (c) show top view SEM images of GaAs nanowires grown on Si (111) susbstrates with disilane flow rate of0.2, 1 and 2.5 seem. ill all the cases unifo1m vertical nanowires arrays can be obtained. Nanowires show hexagonal cross section enclosed by { 1-10} facets. However the wire diameter increases with increasing disilane flow rate. The average diameters measured between two diagonal vertices are 256, 321 and 340 nm respectively, showing enhanced growth rate on {l-10} facets. Fig 5.ld shows measured PL spectra between 1.24 and 1.65 eV for nanowires grown with disilane flow rate from 0 to 1 seem. 1 seem disilane corresponds to partial pressure of 1.4x 10- 9 atm. Line shape changed from asymmetiical for lower disilane flow to more symmetrical for higher ones. 122 Absolute intensity increases before disilane flow rate reaches 0.4 seem and decreases after (Fig 5.le). The initial increase of PL intensity is due to the increased free carrier concentration. In low injection condition, where the light excited carriers Lin are much less than the free carrier in equilibrium n (which is our case, Lln<l0 14 cm- 3 assuming 1 ns carrier life time), the radiative recombination rate is proportional to the doping: Free carrier concentration saturates when further increasing disilane flow rate. As will be discussed later, other Si-induced non-radiative recombination centers arise so the overall PL intensity starts to decrease after disilane flow rate reaching 0.4 seem. The broadening of peak shown in Fig 5 .1 e is related to both band filling effect and formation of band tails extending into the forbidden gap. 123 a E VB /'"]' c 1.0 - fitDng of direct B-8 - experimental non-doped 0.8 ..J n. Goe i . !:! ~ 0.4 0 z 0.2 1.40 1.45 150 1.55 1.60 Photon energy (eV) d ..J n. "C Q) .!:! ii E 0 z 1.65 b 1.0 - Cumulative Peak Flt - FnPeak t - FnPeak2 0 · 8 - Fn Peak 3 0.6 0.4 0.2 - experiment 828 nm 760 780 Band edge I 800 820 640 860 Photon energy (eV) 880 900 Figme 5. 2 (a) Schematic energy band diagram for direct band to band transition with k conse1vation. (b) Schematic energy band diagram for indirect band to acceptor transition without k conservation. (c) Experimentally measmed PL spectrum of non-doped nano wires and theoretical fitting based on direct band to band transition in Eq 5 .1. ( d) PL spectrum of single non-doped nanowire measmed at 4K. Depending on the impmi.ty concentration, or in another word whether the translational symmetry is destroyed, two kinds of transition mechanisms exists: (1) When k-conse1vation is required, direct band to band transition occurs (Fig 5.2a), thus intensity ofPLI(E) can be described as: ( ) 1/2 [ ( m * E - E I(E) oc E 2 x E -Eg x 1 +exp • h • x g me +mh kT E;k;E• )J' E;;/• )J}~ .. I' x 1 - 1 + exp :- e * x g { [ ( m· E-E me +mh kT (Eq. 5.1) where the first square term refers to the simple photon density of states in bulk, the second square root tenn refers to the density of states of electron (same dependence 124 for the corresponding hole states in vertical transition). The next term in brackets refers to the probability of a state in the conduction band is occupied by electron based on Fermi-Dirac distribution while the term in braces is the probability of a corresponding state in the valence band is unoccupied. The last term is the transition matrix element determining the oscillator strength of the transition. (2) When translational symmetry is broken, indirect band to acceptor transition without k-conservation dominates (Fig 5.2b). Most photocreated minority carriers are thermalized completely before radiative recombination and are energetically situated at the extremum of the band tail. Thus the PL spectra directly reflects the earner population in the conduction band: (Eq. 5.2) where p, is the density of states in conduction band. In sufficiently heavily doped semiconductors band edge fluctuates spatially. By assuming a Gaussian distribution of the edge fluctuation and local DOS still follow a parabolic relationship, Kane gave the analytic form of distorted density of states 29 : (Eq. 5.3) where E, is the integral variable to take into account all the possible band edge position and IIE is the root mean square of the energy fluctuation of the band edge. From Fig 5.ld we can see the nondoped one shows clear asymmetrical line shape and lowest peak energy among all. Since there's no intentional doping introduced, the un-doped one thus is the one most likely to follow the first discussed direct band to band 125 transition with k-conservation. Fig 5.2c shows the experimentally measured PL spectrum ofun-doped nanowires and best fitting using Eq. 5.1 with Fermi-level as fitting parameter and unity matrix element. The theoretical curve predicts a sharp cut-off at the conduction band edge of 1.424 e V. Although the high energy side matches well with fitted curve, significant tail below the band edge is observed in the experimentally measured spectrum. Since no disilane is supplied during the growth, we suspect the band tail is caused by unintentionally incorporated carbon from methyl radical in TMG during the pyrolysis process 35 • 36 , and is enhanced for As rich surfaces 37 . FTIR studies of TMG decomposition, carried out under UHV conditions, report the formation of strongly bound CH 2 groups on GaAs surface at low surface coverage 38 . Kinetic modeling studies 39 • 40 have further developed and refined this carbon incorporation mechanism through a detailed description of gas phase and surface reactions occurring during growth. Carbon incorporates through surface reactions of adsorbed GaCH 3 , leading to gallium carbine (GaCH 2 ) that reacts with an exposed Ga atom to place carbon on an As site. Fig 5.2d is the PL spectra measured from a single un-doped nanowire. Peak-fitting through Lorentzian shape are used to deconvolute individual peaks. Two dominant peak are located at 828 and 834 nm, well below the bandgap energy of 1.519 eV. These two peak have been previously reported to be related to electron-to-carbon acceptor (e, A 0 ) transition and carbon donor-to-carbon acceptor (D 0 , A 0 ) transition 41 . This observation agrees with the room temperature PL tail below the bandgap in Fig 5.2c. 126 a -' CL -0 <ll -~ ro E 0 z 1~ 1M 1M 1M1M 1M 1~1fil1~ 1~ 1~1~ Photon energy (eV) b 1.50 1.48 > 1A6 ~ w ..... 1.44 1.42 C 2.50E+018 C?-; 2.00E+018 .::. ~ 1.50E+018 ·u; c: ., "'C 1.00E+018 a; "E ca 5.00E+017 (.) 0.0 0.2 0.4 0.6 O.B 1.0 Disilane flow (seem) 0.0 0.2 0.4 0.6 0.8 1.0 Disilane flow (seem) Figure 5. 3 (a) Normalized measured PL spectra of nanowires grown with different disilane flow rate and curve fitting based on indirect band to acceptor transitions Given the presence of tail states we try to interpret the measured PL spectra through the second proposed transition mechanism. We used Eq 5.2 and Eq 5.3 to fit the experimental data using Fermi-level E 1 and mean square root of band edge fluctuation (JE as fitting parameters. Eg is fixed at the nominal band gap of 1.424 eV for GaAs at room temperature. Best fitting are shown in 5.3a as the black curves superimposed on the measured ones. All the curves are normalized between 0 and 1 and are offseted to better illustrate the line shape evolution with increasing disilane flow rate. The standard deviation for each curve is on the order of 10- 4 _ The fitting is intentionally limited between 1.38 eV and 1.66 eV because at energy lower than 1.38 eV another deep-level peak around 1.35 eV starts to appear for those with high disilane flow rate (Fig 5.4). 127 E 1 and uE of best fitting for each doping level is plotted in Fig 5.3b, based which carrier concentration is calculated as: where p, is the density of states in conduction band expressed in Eq 5.3. The calculated carrier concentration is shown in Fig 5.3c. Carrier concentration increases approximately linearly till it reaches about 2xl0 18 cm- 3 . It then slows down when further increasing disilane flow rate and eventually dropped after it reaches 2.25xl0 18 cm- 3 . Previous work on Si doped GaAs also found electron concentration increases linearly up to 3x10 18 cm- 3 with increasing disilane-to-TMG ratio 42 . According to Hall and SIMS measurements, below 3xl0 18 cm- 3 , the electron and atomic Si concentrations agree well, indicating that incorporated Si atoms are fully ionized as donors. Above 3x10 18 cm- 3 , the electron concentration tends to saturate, while the atomic Si concentration continues to mcrease linearly. This deviation is commonly observed in heavily Si-doped GaAs and is caused by the self-compensation of Si 43 - 45 . Si can incorporate in GaAs on both Ga and As lattice sites as donor and acceptor, respectively. This amphotericity is known to reduce the doping efficiency of Si. Additional models have also been developed to explain the compensation mechanism, such as the formation of Si pairs 46 , Si clusters 47 , and complexes of Si with a speculative native defect 48 • 49 , the existence of a nonhydrogenic Si level resonant with conduction band, and a variety of other mechanisms. Domke et al., atomically resolved various compensation mechanism in real-space by scanning tunneling microscopy 44 . They found with increasing Si concentration the Si 0 , donors are consecutively electrically deactivated by SiA~ acceptors, Si clusters, and Si 0 ,-Ga-vacancy 128 complexes. Siaa and SiAs are shallow traps and Sia.-Ga-vacancy complex is reported to be a deep acceptor trap. a 77 K --30n\IV 1.0 n4sccm --300nW _J 0.8 --LSµW "- µVV v .~ 0.6 ro E 0 0.4 z 0.2 0.0 1.3 1.4 LS 1.6 1.7 El'ergy (eV) c 77 K --30n\V 1.0 0.8 seem 300 nW _J 0.8 --1.SµW "- µW u ?J 0.6 ro §o.4 0 z 0.2 0.0 1.3 L4 1.5 L6 1.7 Energy (eV) bLO 0.8 0: "C .f) 0.6 8 Eo4 0 . z 0.2 0.0 1.3 dlO _J- 0.8 "- v .~0.6 ro §o.4 z 0.2 0.0 1.3 77 K --30nVI/ 0.6 seem --300r.w --1.S~t\IV 3µW 1.4 LS 1.6 1.7 Energy (eV) 77 K 1 seem --30n\V 300 nW --1.5µ\'V µVV L4 1.5 L6 1.7 Energy (eV) Figure 5. 4 Normalized PL spectra measured at 77 K for nanowires with different disilane flow rate and laser excitation power. In the room temperature PL spectra shown in Fig 5.ld, we see an increased intensity in the tail luminescence with increasing disilane flow rate. To further study the properties of tail states we carry out low temperature PL measurement. Fig 5.4 shows the PL spectra taken at 77 K of nanowires grown with different disilane flow rate. Four different excitation power were used for each sample. The curves are normalized so that the intensity of the near-band edge peak is unity. At low temperature the luminescence peak associated with deep impurity band becomes much stronger compared to that in room temperature. The relative intensity of impurity band related emission increases with 129 increasing disilane flow rate and surpasses the near-band edge peak intensity for 1 seem sample under low excitation. Furthermore the intensity of the impurity band related peak decreases with increasing excitation power. a 10 77 K --30nW 1 seem 300 0W 0.8 a'. 3 ~tVV v .~0.6 "' 504 z 0.2 0.0 1.3 1.4 1.5 1.6 1.7 Energy (eV) 173 K --30n\IV c 1.0 300 r.W _J 0.8 --1.5;tW OL 3 µVV v ~0.6 ro Eo4 0 z 0.2 0.0 L3 1.4 1.5 1.6 1.7 El'ergy (eV) b 1.0 a' 0.8 v .~ 0.6 G Eo4 0 . z 0.2 0.0 d 1.0 _J 0.8 Q_ v .~0.6 ro §04 0 . z 0.2 0.0 1.3 1.3 123K --30n'vV --300nW --1.SµW 3 µVJ 14 1.5 1.6 1.7 Er.ergy (eV) 223K --30r.VI/ 300 nW --1.SµVv' µVV 1.4 1.5 1.6 1.7 Energy (eV) Figure 5. 5 temperature dependence of PL spectra of nanowires grown with 1 seem disilane flow rate. We also vary the temperature to study how the relative intensity of the two peaks evolves. As shown in Fig 5.5 the relative intensity of the impurity band related peak decreases with increasing temperature and show less excitation power dependence. Unlike the band edge peak, the peak energy of the impurity peak position does not shift when changing temperature, which is always around 1.35 e V. This peak was reported to be related to a deep acceptor level when high Si concentrations are present (> 10 18 cm- 3 ) in Si doped GaAs grown from Ga solution 50 - 52 . These are presumed to arise at some complex, possibly a Sioa-SiAs pair 50 • 53 • 54 . Redfield et al. report long radiative recombination 130 lifetime of this deep acceptor state, up to microsecond at 77 K 55 . The long carrier lifetime is attributed to the fact that these acceptor states have very localized wave function and are spatially separated from conduction band tail states so the matrix element is significantly damped. The density of states associated with deep acceptor state is limited, and is likely to saturate under strong excitation due to the long radiative recombination lifetime. After saturation, excess photo generated holes will occupy states in the valence band and contribute to the near band edge luminescence. The higher disilane is supplied during the growth the larger density of states of deep acceptors is. Thus we observe in Fig 5.4, the deep acceptor peak starts to saturate for excitation less than 300 n W for nanowires grown with 0.4 seem disilane, while it starts to saturate around 3 µ W for those grown with 1 seem disilane. The temperature dependence shown in Fig 5.5 can be explain by the fact that at higher temperature some states are thermally excited from deep states to band tails and recombine more rapidly in either radiative recombination or non-radiative recombination because of the reduced localization of excited states. 131 5.3 GaAs nanowire on Si dual junction solar cell a b 0. 04 c 0.03 - disile.ne 1sccm - 0.Ssccm O.o2 - 0.2sccm g 001 c ~ 000 ::J () .()01 .0. 02 .()03 -1 5 -1.0 .() 5 00 0.5 10 1.5 Bias (VJ Figure 5. 6 (a) Schematic of n+-GaAs/p+Si heteroepitaxial tunnel junction device. (b) 30° tilted SEM images of the device proposed in (a) at each step: (i) as grown n+-GaAs nanowires on p+-Si (111) substrate; (ii) after nanowire a.nay being planarized and etched by RIE, wire tips are exposed; (iii) after deposition of AuGe/Ni/Au on top; (iv) after rapid thermal annealing at 380 °C with 5 °C/s ramping rate. After assessing the doping characteristics of n-type GaAs nanowires, we try to fabricate the GaAs nanowire on Si dual junction solar cell proposed in section 5 .1. The first step is to fonn an appropriate interconnecting junction between p+-Si and n+-GaAs nanowires. To measure the I-V characteristics of the heterojunction, we fabricate the device shown in Fig 5.6a. Three devices are made with different disilane flow rate of 0.2, 0.5 and 1 seem. Nanowires are first grown on Si (111) substrate based on the condition discussed in Chapter 4. The Si wafer is implanted by Boron ion followed by thermal activation. So the surface layer heavily doped p-type with dopant concentration arnund lxl0 21 cm- 3 . The nanowire anays are then planarized by spin coating casting of BCB and cured at 250 °C. The bmied nanowires are exposed by RIE using CF 4 /0 2 plasma. AuGe/Ni/Au metal 132 tri-layer with thickness of 100/30/100 nm are deposited using e-beam evaporation. As deposited film does not form good ohmic contact with n+-GaAs nanowires and 1s granular. After rapid thermal annealing at 380 °C for 5 second at ramping rate of 5 °C/s good ohmic contact can be formed (has been previously separately on n+-GaAs nanow!fes grown on n+-GaAs substrates) and the metal surface is smoothen. I-V characteristics of heterojunction with different disilane flow rate are shown in Fig 5.6c. The nanowires grown with 1 seem disilane forms a close-to-Ohmic junction with the p+-Si substrate while the other two shows symmetrical I-V curves with certain barriers. Early we have shown the carrier concentration for 1 seem disilane is lower than that of 0.5 seem disilane (Fig 5.3c), however the overall conductivity of the former is higher than the later when forming a heterojunction with p+-Si. We believe the deep acceptor levels introduced in higher Si concentration effectively reduced the valence band discontinuity between GaAs and Si, lowering the barrier for hole current. More deep level states can also facilitate trap assisted tunneling process at the heterointerface. Assuming Jsc of 20 mA/cm 2 for the dual junction solar cell, the voltage drop across the heterojunction would be just around 0.01 V based on the resistance extracted from Fig 5.6c. Thus the heterojunction is a good candidate of interconnecting scheme for the proposed dual junction solar cell architecture. 133 a U1 ... ... .. "' c ~ .. , .. ~Str-Jtte ~pth fSO -6 al d '..... .... ·a.• ... 10---------r~ - Dark N~ • l-- - - Logll -+ \ - --'---- """"''------1--1 i - -10 ~ 1h iii "O ·20 ~ FF= 0.6518 ~ = 10.4% 8 .JO I----+---- .02 oo 02 o~ Voltage (V) o• Figure 5. 7 (a)Boron impurity profile after implantation with ion energy of 40 keV and ion dose of 8x 10 15 cm- 2 . (b) Phosphorus impurity profile after implantation with ion energy of 85 keV and ion dose of2.5xl0 16 cm- 2 . (c) Photography of a completed Si solar cell with SiN AR coating and two front bus lines. ( d) Dru·k and light J-V curve of a typical Si solar cell. The Si bottom cells are made from 280 µm thick :floating-zone phosphorus doped n-type Si (111) wafers with resistivity of 3±2 Ohm-cm. p+ emitter is made by boron ion implantation with ion energy of 30 keV and ion dose of 1.5x10 16 cm- 2 . The profile line shape is close to error function with peak at 250 nm deep and peak concentration of lxl0 21 cm- 3 (Fig 5.7a). Back surface field (BSF) is made by phosphorus ion implantation with ion energy of 40 keV and ion dose of 8x10 15 cm- 2 (Fig 5.7b). Samples are annealed at 950 °C in nitrogen for 30 minutes to activate the implanted impmities. Front contact is made from e-beam evaporated Ti/Pd/Ag and back contact is made from thermal evaporated AL For a typical 1 cm cell shown in Fig 5.7c, two-bus line front contact 134 geometry is used. The width of each bus line is 500 µm and that of each horizontal line is 100 µm. Dark and light J-V curves of a typical device is shown in Fig 5.7d. This particular device showsJsc of29.67 mA/cm 2 , V 0 c of0.541V, FF of0.6518 and efficiency of 10.4%. a -0.2 0.0 0.2 0.4 0.6 0.8 Voltage 0/) - Dark s - Light z. -5 S11'1<f\ 3 s \<. ·~ <IJ "'C -10 c <IJ ~ -15 0 0.0 v = 0.909 v oc FF= 0.5777 11 = 9.56% 0.2 0.4 0.6 Voltage 0/) 0.8 1 .0 --··· 900 20.64 0.956 0.5784 11.4 1800 9.865 0.746 0.3611 2.65 2300 5.73 0.743 0.1343 0.569 1.0 Figure 5. 8 (a) Schematic of dual junction solar cell with GaAs nanow1re top cell epitaxially grown on top of Si planar bottom cell. (b) Dark and light J-V curve of device A mentioned in text, disilane flow rate in n+ GaAs is 1 seem. ( c) Dark and light J-V curves of device C to E mentioned text. Nanowire lengths are 900, 1800 and 2300 nm for C, D and E, respectively. Disilane flow rate in n+ GaAs is 1.5 seem. Performance specifications are tabulated. Eventually we are able to build the dual junction solar cells with GaAs nanowire top cell epitaxially grown on a planar Si bottom cell. The schematic of device structure is shown in Fig 5.8a. Prior to the growth of GaAs nanowire solar cell, a thin segment with heavy silicon doping is grown first in order to form tunnel junction discussed above. The growth 135 condition of the GaAs nanowire solar cell part is the same as that discussed in Chapter 3. For the first device attempted, we used I seem disilane to grow the n+-GaAs part. The nanowires are 1.3 µm tall and 290 nm in diameter. J-V characteristic of a typical device is shown in Fig 5.8b. It shows Jsc of 18.27 mA/cm 2 , Voe of 0.909V, FF of 0.5777, and efficiency of 9.56%. Our results of GaAs nanowire solar cells in Chapter 3 and the results of stand-alone Si solar cells indicate Voe should be around 1.1 e V, which is higher than the value we obtain from the tandem cell. We suspect some voltage drop might occurred across the hetero-tunnel junction. We thus increase disilane flow rate during the growth of n+-GaAs to 1.5 seem. We also varied the length of the nanowires by changing the growth time of the n-type part. Three batches of devices are fabricated with average nanowire length of 900, 1800 and 2300 nm. The J-V curves and tabulated performance specifications are displayed in Fig 5.8c. Best device comes from shortest nanowires, which shows Jsc of 20.64 mA/cm 2 , Voe of 0.956V, FF of 0.5784, and efficiency of 11.4%. The other two devices with taller nanowires show lower Jsc and significantly degraded FF. The decreased Jsc is due to mis-matched absmption between GaAs and Si, which is shown by the simulation in Fig 5.9. The low FF and kinked light current curves for the two devices with taller nanowires might be due to additional barriers formed during the prolonged growth although the exact mechanism is still unknown. 136 a I 06 ~ 0.4 0.2 0 400 J GaAsNW SI Substrate Total 600 800 1000 Wavelength (nm) d h= 1 .0um a:.60Qnm dlaSCl.46 it-=32.6341% 30 - GaAoNW - Si Substrate Totlll 'I ~ I 0 o 0.5 1 t.5 V oltage (VJ j 0.6 ~ ! 0 .4 0 .2 h=1 .811'11 a=600nm dla=0.5 1000 h•I Sum a"600nm dla•0.5 ~"26.256% e Jo~-------~ - GaAsNW 125,_ ____ - Si SUbstrete '> Totlll g20 ,.. ! 15 ~ ~ 10 a 5 0.5 , 1.5 Vo11ago (V) c j 0.6 ~ ~ 0.4 0.2 0 400 SI Substrata Total 600 800 t OOO Wavelength (nm) f 30 h=3.0um a=600nm dla=0.5 ~·23.959~ 'f 2s i20 ,., ~ 15 ~ ~ 1 0 !! 8 s - GaAsNW ' - SI SUbSlrate Total l 0,5 t 1.5 V oltage (V) Figure 5. 9 (a) to (c) absorption in GaAs nanowire, Si substrate and all stmcture for nanowires with different indicated nanowire lengths. (d) to (f) corresponding J-V curves, which shows nearly matched current between GaAs and Si for the first geometry and largely mis-matched currents for the taller nanowires. 5.4 Summary In this chapter, we first discussed an optical, non-destructive method to investigate doping charncteristics in then-type GaAs nanowire. In-direct transition without k-conservation mechanism fits the measured PL spectra vety welL We exb:acted Fermi-level position and the amplitude of band tail from which we furthe1· calculated carrier concentration. The result show cru1ie1· concenb:ation increases with increasing disilane flow rate before it reaches 2x10 18 cm- 2 . The increase of carrier concentration then slows down and eventually decreased for disilane flow rate beyond 1 seem. This phenomenon is explained by self-compensation of Si through SiAs, Si pair, Si clusters and SiGa-Ga-vacancy. These defects caused a luminescent peak well below the band gap, and is clearly resolved at low temperature. The relative intensity of this peak to the intensity of the band edge emission 137 shows dependence on excitation energy and temperature which can be explained by the lower radiative recombination efficiency of these states. We fabricate and measured heterojunction between heavily doped n+-GaAs nanow!fe and p+-Si substrate which shows nearly ohmic behavior for nanowires grown with 1 seem disilane which is used as tunnel junction for tandem solar cells. Finally we made dual junction tandem solar cells with GaAs nanowire top cell epitaxially grown on Si planar bottom cell. Under current match condition the best devices show V 0 , up to 0.985V and efficiency of 11.4%. This is the first demonstration of tandem solar cell using nanowires grown on lattice-mismatched bottom cell and opens up great opportunity for future nanowire-based low-cost high-efficiency multijunction solar cells. 138 References 1. Bedair, S. M.; Lamorte, M. F.; Hauser, J. R. Appl. Phys. Lett. 1979, 34, (1), 38-39. 2. Virshup, G. F.; Chung, B. C.; Werthen, J. G. 20th IEEE Photovoltaic Specialists Conference, 1988; pp 441-445. 3. Olson, J. M.; Gessert, T.; Al-Jassim, M. 18th IEEE Photovoltaic Specialists Conference 1985; pp 552-555. 4. Olson, J. M.; Kurtz, S. R.; Kibbler, A. E.; Faine, P. Appl. Phys. Lett. 1990, 56, (7), 623-625. 5. Bertness, K. A.; Kurtz, S. R.; Friedman, D. J.; Kibbler, A. E.; Kramer, C.; Olson, J. M. Appl. Phys. Lett. 1994, 65, (8), 989-991. 6. Yamaguchi, M.; Wakamatsu, S. 25th IEEE Photovoltaic Specialists Conference, 1996; pp9-ll. 7. Cota!, H. L.; Lillington, D. R.; Errner, J. H.; King, R. R.; Karam, N. H.; Kurtz, S. R.; Friedman, D. J.; Olson, J. M.; Ward, J. S.; Duda, A.; Emery, K. A.; Moriarty, T. 28th IEEE Photovoltaic Specialists Conference, 2000; pp 955-960. 8. King, R.R.; Law, D. C.; Edmondson, K. M.; Fetzer, C. M.; Kinsey, G. S.; Yoon, H.; Sherif, R. A.; Karam, N. H. Appl. Phys. Lett. 2007, 90, (18), 183516. 9. Green, M. A.; Emery, K.; Hishikawa, Y.; Warta, W.; Dunlop, E. D. Progress in Photovoltaics: Research andApplications 2014, 22, (7), 701-710. 10. Dimroth, F.; Grave, M.; Beutel, P.; Fiedeler, U.; Karcher, C.; Tibbits, T. N. D.; Oliva, E.; Siefer, G.; Schachtner, M.; Wekkeli, A.; Bett, A. W.; Krause, R.; Piccin, M.; Blanc, N.; Drazek, C.; Guiot, E.; Ghyselen, B.; Salvetat, T.; Tauzin, A.; Signamarcheix, T.; Dobrich, A.; Hannappel, T.; Schwarzburg, K. Progress in Photovoltaics: Research and Applications 2014, 22, (3 ), 277-282. 11. Czaban, J. A.; Thompson, D. A.; LaPierre, R.R. Nano Lett. 2008, 9, (1), 148-154. 12. Mariani, G.; Scofield, A. C.; Hung, C.-H.; Huffaker, D. L. Nat. Commun. 2013, 4, 1497. 13. Nakai, E.; Yoshimura, M.; Tomioka, K.; Fukui, T. Jpn. J. Appl. Phys. 2013, 52, 055002. 14. Wallentin, J.; Anttu, N.; Asoli, D.; Huffman, M.; Aberg, I.; Magnusson, M. H.; Siefer, G.; Fuss-Kailuweit, P.; Dimroth, F.; Witzigmann, B.; Xu, H. Q.; Samuelson, L.; Deppert, K.; Borgstrom, M. T. Science 2013, 339, (6123), 1057-1060. 15. Cui, Y.; Wang, J.; Plissard, S. R.; Cavalli, A.; Vu, T. T. T.; van Veldhoven, R. P. J.; Gao, L.; Trainor, M.; Verheijen, M. A.; Haverkort, J. E. M.; Bakkers, E. P. A. M. Nano Lett. 2013, 13, (9), 4113-4117. 16. Yao, M.; Huang, N.; Cong, S.; Chi, C.-Y.; Seyedi, M.A.; Lin, Y.-T.; Cao, Y.; Povinelli, M. L.; Dapkus, P. D.; Zhou, C. Nano Lett. 2014, 14, (6), 3293-3303. 17. Takamoto, T.; Agui, T.; Kamimura, K.; Kaneiwa, M. Proceedings of 3rd World Conference on Photovoltaic Energy Conversion, 2003; pp 581-586. 18. Esaki, L. Physical Review 1958, 109, (2), 603-604. 19. Takamoto, T.; Ikeda, E.; Kurita, H.; Ohmori, M. Appl. Phys. Lett. 1997, 70, (3), 381-383. 20. Casadei, A.; Krogstrup, P.; Heiss, M.; Rohr, J. A.; Colombo, C.; Ruelle, T.; Upadhyay, 139 S.; Sorensen, C. B.; Nygard, J.; Fontcuberta i Morral, A. Appl. Phys. Lett. 2013, 102, (1), 013117-4. 21. Choi, W. Y.; Park, B.-G.; Lee, J.-D.; Liu, T.-J. K. Electron Device Letters, IEEE 2007, 28, (8), 743-745. 22. Verhulst, A. S.; Vandenberghe, W. G.; Maex, K.; De Gendt, S.; Heyns, M. M.; Groeseneken, G. Electron Device Letters, IEEE 2008, 29, (12), 1398-1401. 23. Tornioka, K.; Fukui, T. Appl. Phys. Lett. 2011, 98, (8), 083114. 24. Bessire, C. D.; Bjork, M. T.; Schmid, H.; Schenk, A.; Reuter, K. B.; Riel, H. Nano Lett. 2011, 11, (10), 4195-4199. 25. Yang, T.; Hertenberger, S.; Morkotter, S.; Abstreiter, G.; Koblrnuller, G. Appl. Phys. Lett. 2012, 101, (23), 233102. 26. Storm, K.; Halvardsson, F.; Heurlin, M.; Lindgren, D.; Gustafsson, A.; Wu, P. M.; Monernar, B.; Samuelson, L. Nat. Nanotechnol. 2012, 7, (11), 718-722. 27. Perea, D. E.; Hemesath, E. R.; Schwalbach, E. J.; Lensch-Falk, J. L.; Voorhees, P. W.; Lauhon, L. J. Nature Nanotech. 2009, 4, (5), 315-319. 28. Perea, D. E.; Allen, J. E.; May, S. J.; Wessels, B. W.; Seidman, D. N.; Lauhon, L. J. Nano Lett. 2005, 6, (2), 181-185. 29. Kane, E. O.PhysicalReview 1963, 131, (1), 79-88. 30. Casey, H. C.; Stem, F. J. Appl. Phys. 1976, 47, (2), 631-643. 31. Burstein, E. Physical Review 1954, 93, (3), 632-633. 32. Moss, T. S. Proceedings of the Physical Society. Section B 1954, 67, (10), 775. 33. De - Sheng, J.; Makita, Y.; Ploog, K.; Queisser, H. J. J. Appl. Phys. 1982, 53, (2), 999-1006. 34. Lideikis, T.; Treideris, G. Semicond. Sci. Technol. 1989, 4, (11), 938. 35. Kushibe, M.; Eguchi, K.; Funarnizu, M.; Ohba, Y. Appl. Phys. Lett. 1990, 56, (13), 1248-1250. 36. Kuech, T. F.; Redwing, J.M. J. Cryst. Growth 1994, 145, (1-4), 382-389. 37. Kuech, T. F.; Veuhoff, E. J. Cryst. Growth 1984, 68, (1), 148-156. 38. Annapragada, A. V.; Salim, S.; Jensen, K. F. MRS Online Proceedings Library 1991, 222, null-null. 39. Mountziaris, T. J.; Jensen, K. F. J. Electrochem. Soc. 1991, 138, (8), 2426-2439. 40. Masi, M.; Sirnka, H.; Jensen, K. F.; Kuech, T. F.; Poternski, R. J. Cryst. Growth 1992, 124, (1-4), 483-492. 41. Bose, S. S.; Lee, B.; Kirn, M. H.; Stillman, G. E.; Wang, W. I. J. Appl. Phys. 1988, 63, (3), 743-748. 42. Shirnazu, M.; Karnon, K.; Kimura, K.; Mashita, M.; Mihara, M.; Ishii, M. J. Cryst. Growth 1987, 83, (3), 327-333. 43. Drurninski, M.; Wolf, H. D.; Zschauer, K. H.; Wittrnaack, K. J. Cryst. Growth 1982, 57, (2), 318-324. 44. Domke, C.; Ebert, P.; Heinrich, M.; Urban, K. Phys. Rev. B 1996, 54, (15), 10288-10291. 45. Gebauer, J.; Krause-Rehberg, R.; Domke, C.; Ebert, P.; Urban, K. Phys. Rev. Lett. 1997, 78, (17), 3334-3337. 46. Kolm, C.; Kulin, S. A.; Averbach, B. L. Physical Review 1957, 108, (4), 965-971. 140 47. Muto, S.; Takeda, S.; Hirata, M.; Fujii, K.; The, K. Philos. Mag. A 1992, 66, (2), 257-268. 48. Maguire, J.; Murray, R.; Newman, R. C.; Beall, R. B.; Harris, J. J. Appl. Phys. Lett. 1987, 50, (9), 516-518. 49. Schubert, E. F., Doping in 111-V Semiconductors. Cambridge University Press: 2005. 50. Kresse!, H.; Dunse, J. U.; Nelson, H.; Hawrylo, F. Z. J. Appl. Phys. 1968, 39, (4), 2006-2011. 51. Kresse!, H.; Nelson, H.J. Appl. Phys. 1969, 40, (9), 3720-3725. 52. Rossi, J. A.; Holonyak, N.; Dapkus, P. D.; Burnham, R. D.; Williams, F. V. J. Appl. Phys. 1969, 40, (8), 3289-3293. 53. Spitzer, W. G.; Kahan, A.; Bouthillette, L. J. Appl. Phys. 1969, 40, (8), 3398-3400. 54. Spitzer, W. G.; Panish, M. B. J. Appl. Phys. 1969, 40, (10), 4200-4202. 55. Redfield, D.; Wittke, J.P.; Pankove, J. I. Phys. Rev. B 1970, 2, (6), 1830-1839. 141 Chapter 6 Conclusion and outlook 6.1 Conclusion In this dissertation we have studied the growth, characterization of GaAs nanowires and its application in solar energy conversion in detail. Starting from material synthesis, we systematically developed and evaluated each component involved in a GaAs-on-Si tandem solar cell and successfully demonstrated a pilot device with the proposed concept and architecture. In specific, we first achieved growth of highly uniform GaAs nanowire array using non-catalytic selective-area-growth that allows the precise control of nanowire size and location. This growth technique then was applied to demonstrate GaAs nanowire homojunction solar cell devices with a champion efficiency of 7.58%, highest among all the GaAs nanowire array solar cells so far. The result encourages us to take one step forward to make a nanowire-on-Si tandem solar cell. In order to achieve this structure, we first solved the issue of growing nanowires on Si (Ill) substrate. This is a nontrivial advancement from growth on GaAs (ll l )B substrate given the dissimilar crystal registry of GaAs and Si as well as complicated surface chemistry. Nevertheless we successfully obtained epitaxially grown GaAs nanowires on Si with I 00% vertical yield, and even more importantly, the growth does not require complicated nucleation or growth of buffer layer. Heavily doped n-type GaAs nanowires were then grown on p+ Si substrate usmg this technique. The doping concentration was assessed innovatively through a non-destructive method usmg photoluminescence measurement based on Burstein-Moss effect. Ohmic transport 1s observed across this heavily doped 142 heterojunction which paves the way toward the tandem solar cells. An 11.4% efficiency tandem solar cell monolithically integrating GaAs nanowire top cell and Si planar bottom cell was eventually delivered by optimizing the length of nanowire. This is the first time the concept of epitaxial III-V on Si solar cell is materialized and sets a milestone in the progress of multijunction solar cells. It opens up opportunity for optimized bandgap combination and allows low cost by using inexpensive Si substrate. 6.2 Outlook Yet this is by no means a superior device in terms of efficiency, and there is still plenty room to improve the device performance. With regard to the GaAs-on-Si tandem solar cells, the GaAs nanowire top cell needs to be significantly improved. The open circuit voltage obtained so far is way below that of a state-of-art GaAs single junction solar cell. Doping in the n-type segment can be further increased. As presented in chapter 5, the doping concentration corresponding to 0.1 seem disilane is around mid of 10 17 cm- 3 . Further increasing the doping can increase the build-in potential across the p-n junction and also alleviate the surface depletion effect 1 . GaAs is known for its high density of surface states so surface passivation is important. Successfully surface passivation using large band gap material has been successfully applied to radial junction devices 2 - 3 . Passivation in axial junction is more complicated as one need to avoid creating shunting path through the passivation shell. We have attempted to use AlGaAs as passivation. However, in spite of increased Jsc, V 0 , is always lower in a passivated device compared 143 with that of non-passivated device. Careful band alignment design is thus required. High density of twin defects could be another reason causing degraded performance. The discontinuity in band structure 1s known to scatter earners and to reduce minority diffusion length 4 - 5 . Several other aspects could be optimized to improve the performance of the tandem cell other than the GaAs nanowire top cell. The Si planar bottom cell used in chapter 5 is still far from the state-of-art Si solar cell in terms of efficiency. V 0 , is relatively low, but can be improved through careful configuration of junction properties and surface passivation. The doping in the emitter is now around 10 21 cm- 3 , and might be too high and lead to significant nonradiative recombination. Junction depth is around 100 nm and should be further optimized to achieve low series resistance. Front surface is simply passivated by PECVD silicon nitride and back surface is not passivated. Both sides need better passivation strategies to reduce surface recombination and to increase V oc· The transport mechanism across the n+-GaAs/p+-Si heterojunction needs to be better understood to interpret the observed ohmic behavior. As discussed in chapter 5, deep acceptor-assisted tunneling might dominate the carrier transport. To fully enable the potential of III-V-on-Si two junction solar cell, we should also explore nanowires with 1. 7 e V bandgap as the top cell because this constructs the optimal bandgap combination as discussed in chapter 5. AlGaAs, In.GaP and GaAsP could be candidates, to just name a few. The more friendly surface chemistry of these materials compared with that of GaAs makes the performance of the optimized tandem cell even more promising. 144 References 1. Chang, C.-C.; Chi, C.-Y.; Yao, M.; Huang, N.; Chen, C.-C.; Theiss, J.; Bushmaker, A. W.; LaLumondiere, S.; Yeh, T.-W.; Povinelli, M. L.; Zhou, C.; Dapkus, P. D.; Cronin, S. B. Nano Lett. 2012, 12, (9), 4484-4489. 2. Mariani, G.; Scofield, A. C.; Hung, C.-H.; Huffaker, D. L. Nat. Commun. 2013, 4, 1497. 3. Nakai, E.; Yoshimura, M.; Tomioka, K.; Fukui, T. Jpn. J. Appl. Phys. 2013, 52, 055002. 4. Parkinson, P.; Joyce, H. J.; Gao, Q.; Tan, H. H.; Zhang, X.; Zou, J.; Jagadish, C.; Herz, L. M.; Johnston, M. B. Nano Lett. 2009, 9, (9), 3349-3353. 5. Shimamura, K.; Yuan, Z.; Shimojo, F.; Nakano, A. Appl. Phys. Lett. 2013, 103, (2), 022105 145

Abstract (if available)

Abstract This dissertation covers the growth, characterization of GaAs nanowires and application in photovoltaic solar energy conversion. Nanowires, due to their unique electronic, optical and crystallographic properties, are promising candidate for the next-generation high-efficiency and low-cost solar cells. They could be grown on inexpensive substrates regardless of the lattice mismatch. Better bandgap combinations for multijunction solar cells become possible. The strong interaction between nanowire arrays and sunlight allows significantly reduced material usage. ❧ In this dissertation, we systematically demonstrated how we progressively approached the goal of III-V nanowire-on-Si tandem solar cells. Chapter 1 first presents a big picture of the global energy problem and justifies solar energy would be a promising renewable energy source followed by some discussion of how to surpass the Shockley Queisser limit using multijunction solar cells. Then I discuss the motivation of using nanowire as the building blocks of multijunction solar cells. Chapter 2 presents the technique we used to synthesis GaAs nanowires in a non-catalytic and highly uniform fashion which is selective area growth using MOCVD. Patterning techniques associated with selective area growth are discussed with some emphasis on nanosphere lithography that is suitable for high-throughput patterning. GaAs nanowire homojunction solar cell is demonstrated in chapter 3. The device shows champion efficiency of 7.58% after junction depth and diameter optimization. Chapter 4 is an extension of the growth technique reported in chapter 2. Here we transfer the growth to Si (111) substrate and solved the lattice mismatch issue for the heteroepitaxy. Along the way we also come up with a model to coherently interpret twin formation during the growth which agrees with experimental observations and atomistic simulations. In chapter 5, we first demonstrate heavily doped GaAs/Si heterojunction that shows ohmic transport behavior. The doping is characterized by photoluminescence measurement based on Burstein-Moss effect. Eventually we are able to integrate the GaAs nanowire top cell with the Si planar bottom cell and to deliver a working tandem cell with 11.4% efficiency. This is the first time the concept of epitaxial III-V on Si solar cell is materialized and sets a milestone in the progress of multijunction solar cells.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

GaAs nanowire optoelectronic and carbon nanotube electronic device applications

PDF

Light management in nanostructures: nanowire solar cells and optical epitaxial growth

PDF

Optical, mechanical, and electrical properties of nano-structured materials

PDF

Nanostructure electronic band modeling for solar cell and lighting applications

PDF

Nanophotonic light management in thin film silicon photovoltaics

PDF

The growth and characterization of III-V semiconductor nanowire arrays by nanoscale selective area metalorganic chemical vapor deposition

PDF

Multilayer grown ultrathin nanostructured GaAs solar cells towards high-efficiency, cost-competitive III-V photovoltaics

PDF

Zero-dimensional and one-dimensional nanostructured materials for application in photovoltaic cells

PDF

Photodetector: devices for optical data communication

PDF

GaN nanostructures grown by selective area growth for solid-state lighting

PDF

One-dimensional nanomaterials for electronic and sensing applications

PDF

Synthesis, characterization, and device application of two-dimensional materials beyond graphene

PDF

Gallium arsenide nanosheets: study of stacking-fault-free nano-structure and its applications on semiconductor devices

PDF

Semiconducting metal oxide nanostructures for scalable sensing applications

PDF

Efficient ternary blend bulk heterojunction solar cells with tunable open-circuit voltage

PDF

Nanorod-based InGaN/GaN core-shell nanoLEDs

PDF

Chemical vapor deposition of graphene and two-dimensional materials: synthesis, characterization, and applications

PDF

Porphyrin based near infrared‐absorbing materials for organic photovoltaics

PDF

Gallium arsenide based semiconductor laser design and growth by metal-organic chemical vapor deposition

PDF

Flame synthesis of TiO₂ nanoparticles and its application in dye sensitized solar cells

Asset Metadata

Creator Yao, Maoqing (author)

Core Title Growth, characterization of gallium arsenide based nanowires and application in photovoltaic cells

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Electrical Engineering

Publication Date 11/05/2014

Defense Date 08/26/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag gallium arsenide,heteroexpitaxial,MOCVD,nanowires,OAI-PMH Harvest,photovoltaics,tandem solar cells

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Dapkus, Paul Daniel (committee chair), Zhou, Chongwu (committee chair), Nakano, Aiichiro (committee member)

Creator Email maoqingy@usc.edu,yaomq06@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-515317

Unique identifier UC11297544

Identifier etd-YaoMaoqing-3064.pdf (filename),usctheses-c3-515317 (legacy record id)

Legacy Identifier etd-YaoMaoqing-3064.pdf

Dmrecord 515317

Document Type Dissertation

Format application/pdf (imt)

Rights Yao, Maoqing

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA