A Dissertation
entitled
Spectroscopic Ellipsometry Studies of Thin Film Si:H Materials in Photovoltaic
Applications from Infrared to Ultraviolet
by
Laxmi Karki Gautam
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the
Doctor of Philosophy Degree in Physics
______Dr. Nikolas J. Podraza, Committee Chair
______Dr. Robert W. Collins, Committee Member
______Dr. Randall Ellingson, Committee Member
______Dr. Song Cheng, Committee Member
______Dr. Rashmi Jha, Committee Member
______Dr. Patricia R. Komuniecki, Dean College of Graduate Studies
The University of Toledo
May, 2016
Copyright 2016, Laxmi Karki Gautam
This document is copyrighted material. Under copyright law, no parts of this document may be reproduced without the expressed permission of the author.
An Abstract of
Spectroscopic Ellipsometry Studies of Thin Film Si:H Materials in Photovoltaic Applications from Infrared to Ultraviolet
by
Laxmi Karki Gautam
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Physics
The University of Toledo
May 2016
Optimization of thin film photovoltaics (PV) relies on the capability for characterizing the optoelectronic and structural properties of each layer in the device over large areas and correlating these properties with device performance. This work builds heavily upon that done previously by us, our collaborators, and other researchers. It provides the next step in data analyses, particularly that involving study of films in device configurations maintaining the utmost sensitivity within those same device structures. In this Dissertation, the component layers of thin film hydrogenated silicon (Si:H) solar cells on rigid substrate materials have been studied by real time spectroscopic ellipsometry (RTSE) and ex situ spectroscopic ellipsometry (SE). Growth evolution diagrams has been used to guide deposition of materials with good optoelectronic properties in the actual hydrogenated amorphous silicon (a-Si:H) PV device configuration. The nucleation and evolution of crystallites forming from the amorphous phase were studied using near infrared to ultraviolet spectroscopic ellipsometry in situ, during growth for films prepared as a function of hydrogen to reactive gas flow ratio R =
[H2] /{[SiH4] + [Si2H6]. Furthermore, the major challenge in Si:H manufacturing is that iii
quantitative analysis, characterization, and control of the relative nanocrystalline and amorphous volume fractions within mixed-phase films were covered during these studies.
In conjunction with higher photon energy measurements, the presence and relative absorption strength of silicon-hydrogen infrared modes were measured by infrared extended ellipsometry measurements to gain some insight into chemical bonding. Structural and optical models have been developed for the back reflector (BR) structure consisting of sputtered undoped zinc oxide (ZnO) on top of silver (Ag) coated glass substrates. Characterization of the free-carrier absorption properties in Ag and the interface formed when Ag is over-coated with ZnO were also studied by infrared extended spectroscopic ellipsometry. Measurements ranging from 0.04 to 5 eV were used to extract layer thicknesses, composition, and optical response in the form of complex dielectric function spectra (ε = ε1 + iε2) for undoped a-Si:H layers in a substrate n-i-p a-
Si:H based PV device structure and on TCO coated glass for p-i-n configurations.
iv
To my Family.
Acknowledgements
First of all, I would like to express my sincere gratitude to my supervisor Dr.
Nikolas J. Podraza for his valuable guidance, unlimited support, expert mentoring, advice and encouragement throughout this work. His continuous guidance and support helped me to broaden my view and knowledge in subject matter.
I would also like to extend my sincere gratitude to my committee members, Dr.
Robert W. Collins, Dr. Randall Ellingson, Dr. Song Cheng and Dr. Rashmi Jha, for being in my committee and giving me valuable advices whenever needed. The laboratory support staff also has been very helpful and I would like to acknowledge support and the help I got from Terrance Kahle, Carl Salupo, and Nirupama Adiga.
I would like to express my sincere gratitude to Dr. Patrick Hurley and Dr. Robert
Ridgeway, and financial support from Air Products and Chemicals Inc. and Department of Energy (DE-EE0000580). I also thank Matthew L. Herold at the Air Force Institute of
Technology Wright-Patterson Air Force Base for collaboration providing ZnO films.
My deepest appreciation goes to all of my fellow research group members of the ellipsometry group and colleagues at the University of Toledo for motivating me and ensuring me to work in a very pleasant and comfortable environment. Most of all, my heartiest gratitude goes to my parent, my husband, Madhav, and daughter, Manasi, for supporting me under any circumstances.
v
Table of Contents
Abstract ...... iii
Acknowledgements ...... v
Table of Contents ...... vi
List of Tables ...... ix
List of Figures ...... xi
1 Introduction……...... 1
1.1 Motivation and Background ...... 1
1.2 Dissertation Organization ...... 9
2 Experimental Techniques used in Optical Measurements ...... 12
2.1 Deposition of Si:H Thin Films ...... 12
2.1.1 Cluster Tool Deposition System ...... 14
2.1.2 Sputtering Process ...... 16
2.1.3 Plasma Enhanced Chemical Vapor Deposition ...... 19
2.2 Material Characterization Techniques ...... 21
2.2.1 Spectroscopic Ellipsometry ...... 21
2.2.1.1 Theoretical Formalism ...... 23
2.2.1.2 Experimental Methods in Spectroscopic Ellipsometry ...... 33
2.2.1.3 Data Analysis Strategies in Spectroscopic Ellipsometry ...35
2.2.2 Experimental Methods in Solar Cell Characterization ...... 45 vi
3 Growth Evolution of Si:H Prepared with Disilane Additives as Studied by Real
Time Spectroscopic Ellipsometry ...... 49
3.1 Introduction and Motivation ...... 49
3.2 Overview of Deposition Processes and Microstructural evolution...... 50
3.2.1 Deposition Processes of Si:H Films ...... 50
3.2.2 Microstructural Evolution and Phase Diagram ...... 52
3.3 Experimental Details ...... 59
3.4 S = 0, and variable R: Si:H Growth Evolution Baseline ...... 61
3.5 S = 0.12 and 1 and verses R: Effect of Disilane on Si:H Growth ...... 63
3.6 Summary… ...... 67
4 Si:H Layer Optimization for Substrate Configuration of Solar Cell ...... 69
4.1 Introduction ...... 69
4.2 Experimental Details ...... 72
4.3 Ag/ZnO BR Substrate ...... 73
4.4 n-layer Si:H on Ag/ZnO Substrate ...... 82
4.5 i-layer Si:H on Ag/ZnO/n-Si:H Substrate ...... 85
4.6 p-layer Si:H on glass/i-layer Substrate ...... 89
4.7 Summary… ...... 98
5 Growth and Analysis of Amorphous to Nanocrystalline Transition in
Hydrogenated Silicon films ...... 99
5.1 Introduction and Motivation ...... 99
5.2 Overview of Microstructural Evolution (a-Si:H to nc-Si:H) ...... 102
5.3 Experimental Details ...... 106
vii
5.4 Microstructural evolution of nc-Si:H on c-Si ...... 107
5.5 Microstructural evolution of n-type Si:H ...... 115
5.6 Microstructural evolution of Si:H on (n-type a-Si:H)/BR ...... 126
5.7 Summary… ...... 129
6 Infrared Extended Spectroscopic Ellipsometry Applied to Characterization of
Thin Films and PV Devices Structures ...... 131
6.1 Overview…...... 131
6.2 Experimental Details ...... 133
6.3 Data Analysis and Results ...... 135
6.3.1 Si:H Layers on Ag/ZnO Back-Reflectors ...... 135
6.3.2 Si:H Layers on Transparent Conducting Oxides (TCO) ...... 145
6.4 Summary…...... 159
7 Summary and Future Works ...... 161
7.1 Summary…...... 161
7.2 Future Works ...... 167
References ...... 171
viii
List of Tables
4.1 Deposition conditions for the individual layers of the a-Si:H n-i-p solar cells
configuration deposited on 6"x 6" borosilicate glass substrates ...... 74
4.2 Parameters describing and microstructural parameters for a semi-infinite silver
(Ag) film on a borosilicate glass over coated by chromium (Cr) ...... 77
4.3 Parameters describing and microstructural parameters for a ZnO film on Ag
(back reflector) in visible region (0.734 to 5.0 eV) ...... 80
4.4 Deposition conditions for p-type Si:H layers of the a-Si:H n-i-p solar cells
configuration deposited on 6"x 6" borosilicate glass substrates coated with
intrinsic layer ...... 94
6.1 Complex dielectric function spectra and structural parameters for a ZnO film on
Ag from 0.04 to 5 eV at RT ...... 136
6.2 Parameters describing and structure for a Ag/ZnO BR coated with n-type and i-
type a-Si:H in mid-IR to near UV energy range (0.04 to 5 eV)...... 141
6.3 Optical parameters tabulated for and structure for the three incorporated SnO2:F
layers in TEC™-15 substrate as determined in SE analysis in the energy region
0.04 to 5 eV collected at room temperature ...... 150
ix
6.4 Optical parameters tabulated for and structure for the three incorporated SnO2:F
layers in TEC™-15 substrate as determined in SE analysis in the energy region
0.04 to 5 eV collected at 200oC ...... 152
TM 6.5 Drude parameters of the SnO2:F TCO layers in the TEC -15 substrates are
tabulated as measured by spectroscopic ellipsometry at RT, 200oC, and plasma
modified ...... 155
6.6 Structural model along with the values of the fixed parameters and the best fit
parameters obtained from the SE analysis (0.04 to 5 eV) of a R =10 a-SiH layer
on TECTM-15 substrate ...... 156
x
List of Figures
2-1 A typical single junction a-Si:H solar cell in the substrate configuration ...... 13
2-2 Eight chamber cluster tool deposition system ...... 15
2-3 Schematic of the rf magnetron sputtering process for metal and TCO...... 16
2-4 Schematic of the PECVD chamber used in the deposition of hydrogenated silicon
(Si:H) thin films layers...... 19
2-5 Schematic drawing of a rotating-compensator multichannel ellipsometer used in
RTSE analysis ...... 22
2-6 Schematic representation for the variation of the state of polarization with phase
difference ...... 26
2-7 Schematic of the plane of incidence along with the propagation vectors that
demonstrates the definition of p-polarized and s-polarized fields ...... 27
2-8 Schematics of plane wave reflection from a multilayer structure of n-i-p
configuration single junction a-Si:H PV device at non-normal incidence A...... 30
2-9 IR-VASE data acquisition system ...... 35
2-10 Schematic of a two layer sample structure consisting of a bulk layer and surface
roughness layer on top of a known substrate ...... 39
2-11 Schematic of n-i-p configuration a-Si:H solar cell with the electronic components
connected and incident light used in measurements of J-V curve...... 46
xi
2-12 Maps of n-i-p a-Si:H solar cell performance parameters of open circuit voltage,
Voc; short circuit current, Jsc; fill factor, FF; and power conversion efficiency ..47
3-1 Surface roughness layer thickness and unbiased estimator, σ, versus bulk layer
thickness for a Si:H film prepared with S = 0.63 at R = 23 ...... 54
3-2 Surface roughness layer thickness and unbiased estimator, σ, versus bulk layer
thickness for a Si:H film prepared with S = 0.12 at R = 33 that nucleates
crystallites ...... 57
3-3 Growth evolution of Si:H in the form of the surface roughness layer thickness as a
function of bulk layer thickness for films prepared at fixed S = 0, and variable H2-
dilution ...... 61
3-4 Growth evolution diagrams for Si:H prepared at S = 0 and variable R depicting the
aa, a(a+nc), and (a+nc)nc structural transitions ...... 62
3-5 Growth evolution diagrams for Si:H prepared at S = 0.12 and 1 as functions of R
depicting the aa, a(a+nc), and (a+nc)nc structural transitions ...... 64
3-6 Film growth rates for Si:H prepared at S = 0, 0.12, and 1 as functions of R ...... 66
4-1 Schematic n-i-p structure with optimized R values and intended thicknesses for
different layers in devices ...... 71
4-2 Spectra in as a function photon energy for a semi-infinite Ag film deposited on
Cr coated glass ...... 76
4-3 Spectra in as a function photon energy for Ag+ZnO interface layer between Ag
and ZnO ...... 79
4-4 Spectra in as a function photon energy extracted over a range 0.734 to 5 eV for
ZnO deposited on semi-infinite Ag substrate...... 81
xii
4-5 Phase diagram for Si:H n-layers deposited on 6” x 6” borosilicate glass coated
with a Cr/Ag/ZnO structure ...... 82
4-6 Deposition rates of the n-layer as a function of hydrogen dilution deposited on
Cr/Ag/ZnO/n-layer coated borosilicate 6”x 6” glass ...... 84
4-7 The band gap energy (or Cody gap) plotted as a function of the hydrogen dilution
ratio R for amorphous n-layers on back-reflector coated glass at 200oC ...... 84
4-8 Phase diagram for Si:H i-layers deposited on 6” x 6” borosilicate glass coated
with a Cr/Ag/ZnO/n-layer structure...... 86
4-9 Deposition rates of the i-layer as a function of hydrogen dilution deposited
on Cr/Ag/ZnO/n-layer coated borosilicate 6”x 6” glass ...... 87
4-10 The band gap energy (or Cody gap) plotted as a function of the hydrogen dilution
ratio R for amorphous i-layers deposited on Cr/Ag/ZnO/n coated glass substrates
at a substrate temperature of 200oC ...... 88
4-11 Structural evolution comparison for R = 10 intrinsic layer deposited on
Cr/Ag/ZnO/(R=50 n-layer) coated 6”x 6” borosilicate glass and for R =10
intrinsic layer deposited on 6”x 6” borosilicate glass substrate only...... 90
4-12 Complex dielectric function spectra for bulk R = 10 i-layers on (BR/R = 50 n-
layer) and borosilicate glass ...... 92
4-13 Phase diagram for p-type Si:H layers deposited on 6” x 6” borosilicate glass
coated with R = 10 intrinsic layer ...... 95
4-14 Deposition rates of the p-layer as a function of hydrogen dilution deposited on
Cr/Ag/ZnO/n-layer coated borosilicate 6”x 6” glass ...... 96
xiii
4-15 The band gap energy as a function of hydrogen dilution ratio R for p-type Si:H
layers deposited on glass coated with R = 10 intrinsic layer ...... 97
5-1 Results from RTSE analysis using a two-layer optical model for a Si:H film
deposition in which the anc transition is observed ...... 102
5-2 Schematic of a four-medium optical model used in VIA ...... 105
5-3 Spectrally average MSE versus assumed surface roughness layer thickness values
in a VIA by least squares regression for a Si:H film prepared at R = 30 on native
oxide covered c-Si substrate ...... 109
5-4 Results of the VIA compared to results obtained in RTSE analysis using the two-
layer model for a R = 30 and S = 0 undoped Si:H film ...... 112
5-5 Mean square error (), surface roughness, nanocrystalline volume fraction and
void fraction in the top 10 Å of the bulk layer, plotted versus db for a R = 30 and S
= 0.12 undoped Si:H film ...... 114
5-6 Results of the VIA applied to RTSE data for Si:H film deposited on thermal oxide
covered c-Si substrate ...... 117
5-7 Comparison of the experimental ellipsometric spectra in Ν, Ϲ, Ѕ and model fit for
ZnO films on thermal oxide coated c-Si ...... 118
5-8 Time dependence of bulk layer thickness, surface roughness, relative
nanocrystallite fraction compared to voids within the bulk film layer and nc
fraction compared to voids in the initial nucleation stage with two component
model for ZnO films on thermal oxide coated c-Si (sample A) ...... 120
5-9 Results of two component and 3 component EMA as determined by RTSE VIA of
R = 100 n-type Si:H film for ZnO films on thermal oxide coated c-Si ...... 121
xiv
5-10 Results of the VIA applied to RTSE data for a R = 80 n-type Si:H film deposited
on Ag/ZnO back reflector coated glass ...... 123
5-11 Comparison of spectra in 2 for samples Ag/ZnO/R=80 n-type Si:H, c-
Si/SiO2/R=100 n-type Si:H and c-Si/SiO2/ZnO/ R=100 n-type Si:H ...... 124
5-12 Spectra in of a-Si:H and nc-Si:H reference material used in VIA as functions of
photon energy extracted over a spectral range from 2.75 to 5.0 eV ...... 127
5-13 Mean square error, surface roughness, nanocrystalline volume fraction and void
fraction in the top 10 Å of the bulk layer, plotted versus the accumulated bulk
layer thickness for undoped R = 50 Si:H on R = 50 ntype a-Si:H / Ag/ZnO ....128
6-1 Comparison of the experimental ellipsometric spectra in Ν, Ϲ, Ѕ and model fit for
R = 10 a-Si:H films on BR over-coated with a R = 50 n-layer ...... 134
6-2 Spectra in as a function photon energy over a spectral range from 0.04 to 5.0 eV
for ZnO film on Ag ...... 137
6-3 Optical model with structural parameters used in the IR extended analysis of a-
Si:H based n-i-p substrate ...... 139
6-4 Comparison of 2 between Ag/ZnO back reflector samples with and without a-
Si:H overdeposition ...... 142
6-5 Spectra in as a function photon energy extracted over the energy spectral range
from 0.04 to 5 eV for R = 10 a-Si:H on R = 50 a-Si:H n-layer ...... 144
6-6 Optical model showing different layers for the TECTM-15 substrates at RT ...... 146
TM 6-7 Spectra in for soda lime glass/SnO2/SiO2 of TEC -15 glass at RT ...... 147
6-8 Spectra in showing three different sublayers of SnO2:F layer of TEC™-15 from
0.734 to 5 eV at RT ...... 148 xv
6-9 The IR spectra in showing three different sublayers of SnO2:F layer of TEC™-
15 from 0.04 to 0.73 eV at RT ...... 149
6-10 Spectra in showing a plasma modified SnO2:F layer of TEC™-15 extracted
from RTSE data collected at 200oC ...... 153
6-11 Spectra in for R = 10 a-Si:H films on TEC™-15 showing SiH vibrational
modes as a part of IR-extended SE analysis 2 ...... 158
xvi
Chapter 1
Introduction
1.1. Motivation and Background
In contemporary age, semiconductors are the basis of modern electronics devices and technology with the use of transistors, solar cells, diodes, and amplifiers as well as numerous different digital and analog integrated circuits (Sze et al., 2006; James et al.,
2010). The never-ending research on semiconducting materials, behavior, and fabrication processes have made constant progresses in the complexity and the speed of semiconductor devices, an effect known as Moore’s Law (Mack, 2011). As the world’s population increases, the necessity for cost-effective energy will also increase.
Renewable energy sources, such as hydropower, wind, geothermal, solar thermal, and photovoltaics (PV) are some of the promising alternatives to replace conventional fossil fuels. Renewable energy production has improved, but the portion it plays in the total energy supply has not increased considerably. These sources comprise a very small amount of the whole energy consumption. The U.S. Energy Information Administration
(EIA) estimates that about 11% of world marketed energy consumption is from renewable energy sources (biofuels, biomass, geothermal, hydropower, solar, and wind) 1
with a projection for 15% by 2040. Also, EIA estimates that about 21% of world electricity generation was from renewable energy in 2011 with a projection for nearly
25% in 2040 (USEIA, 2014). With all the persuasive benefits and drive for renewable energy, the reality is that there is a long way to go before it meaningfully impacts the world energy supply (Boyle, 2004).
Among these environment friendly alternative energy sources, solar energy derived sources are considered the most favorable as these are readily available globally wherever human civilization is concerted. Solar energy is, in fact, the ultimate source of energy and has been used in various forms since the beginning of human civilization.
Edmund Becquerel in 1839 (Becquerel 1839) discovered the effect of converting light directly into electrical energy, known as PV. Due to the developments in the twenty-first century of low cost PV modules, through which photons are converted directly into electrical energy via the generation of electron-hole pairs, this technology has become one of the most appreciated approaches for utilizing the sun's energy.
The first generation PV technology comprising bulk crystalline or multicrystalline silicon occupies ~90% of the solar module market share and exhibits
~25% submodule efficiency and ~22.9 % module efficiency (Green et al., 2014).
Production of monocrystalline solar cells fabricated by the Czochralski process is expensive due to the price of solar-grade silicon feedstock and the large energy requirements. Although these solar cells have higher conversion efficiencies, they can be costly to fabricate and face difficulties in competition with conventional sources of energy. In order for PV to be an economically feasible option as an energy source, PV device efficiency must increase and cost of PV device fabrication must decrease. 2
Thin film solar cells, known as second generation solar cells, have drawn the most commercial attention nowadays (Green et al., 2014). These include hydrogenated amorphous silicon (a-Si:H), cadmium telluride (CdTe), and copper indium-gallium diselenide (CuIn1-xGaxSe2 or CIGS) as well as other new developing technologies such as those based upon organic-inorganic semiconductors (CH3NH3PbI3 perovskites) (Kojima et al., 2009; Park, 2013). Although thin film devices have lower conversion efficiencies in comparison with first generation crystalline silicon (c-Si) solar cells, they hold the potential for large scale production at much lower cost. Thin film layers can be fabricated onto low-cost substrates at much lower temperatures using simpler techniques compared to c-
Si fabrication technology. Since very little material is used in solar modules, comparative to silicon in bulk c-Si solar modules, the high output thin film production processes such as magnetron sputtering, closed space sublimation (CSS), vapor transport deposition (VTD), plasma enhanced chemical vapor deposition (PECVD) is an added advantage, and enables further reductions in the cost of module fabrication.
Hydrogenated silicon (Si:H) solar cells, the focus of this dissertation, belong to the “second generation” or thin film class of PV devices that have the potential for cost effective solar power. These Si:H PV devices have many benefits over “first generation” c-Si PV and other thin film PV devices, thereby making it an attractive option for further research and development. The materials used in the production of these Si:H solar panels are not toxic and are readily obtainable. The high optical absorption coefficient of a-Si:H allows the use of thinner absorber layers within the PV device. Therefore, less material is consumed during fabrication. As a result, the manufacturing cost for these solar cells can be reduced significantly. Amorphous Si:H (a-Si:H) was one of the first
3
thin film materials incorporated into a solar cell structure and the record efficiency for a single-junction amorphous silicon minimodule is currently at 10.2% (Green et al., 2014).
Unfortunately this material has not achieved commercial success because of low efficiency and the low deposition rate (Collins et al., 2003) required to achieve the optimum material quality. The low efficiency is result of degradation in electronic properties due a continuous distribution of localized states in the band gap region and a light induced degradation effect over time (Staebler and Wronski, 1977).
Most of the layers in the Si:H solar cell are deposited using PECVD in a well- regulated manner. The precursor gas molecule in this method is exposed to some form of energy (other than thermal) that breaks it into smaller constituents (radicals), which can adsorb on the substrate. PECVD utilizes a plasma as an energy source to convert precursors into radicals. Energetic ions from the plasma may also contribute to radical- substrate reactions (Luque et al., 2011). In this dissertation, the deposition of doped and undoped Si:H thin films in solar cell structures of the n-i-p configuration will be the main concern.
In the late 1960’s, interest in a-Si:H as an optoelectronic material began with the pioneering work of Chittick et al., that demonstrated the semiconducting properties of a-
Si:H prepared by glow discharge of silane (Chittick et al., 1969). Several years later around mid-1970’s, doping of a-Si:H was reported by Spear and Lecomber (Spear et al.,
1975), and soon after, the first PV devices were developed by Carlson and Wronski
(Carlson et al., 1976). Even as these progresses were being made, the critical role of hydrogen in passivating Si dangling bonds in a-Si:H was not yet understood (Fritzsche,
2001). Hydrogen aids in reducing the electronic defects associated with dangling bonds, 4
and reducing the disorder through the formation of an a-Si:H alloy by relaxing the tetrahedral network as was explored in 1980’s. The elimination of defect states in the band gap results in a better ordered thin film material with improved electronic properties. In fact, hydrogen atoms generated from decomposition of H2 along with SiH4 or Si2H6, if incorporated in the gas phase in appropriate quantity, leads to the growth of a nanocrystalline Si:H phase by relaxing strained Si-Si bonding (Lu et al., 1991).
Furthermore, the process promotes dangling bond passivation eliminating possible states in the band gap and increasing carrier mobility (Sakata et al., 1993). The intrinsic i-layers in devices are prepared from a-Si:H and its alloys with germanium (Si1-xGex:H) where the doped p- and n- layers can be either a-Si:H or nanocrystalline silicon (nc-Si:H) (Yang et al., 1997; Deng et al., 2003; Stoke et al., 2008). At present, both a-Si:H and μc/nc-Si:H component materials are used in state-of-the-art Si:H based PV devices. A critical necessity exists for measuring, monitoring, and controlling the thickness, structure, phase, and composition of solar cell component layers in the same configuration as is used for solar cell manufacturing, especially for devices processed over large areas.
Doped and undoped Si:H thin films are used in single, tandem, and multijunction solar cell applications in both substrate and superstrate configurations (Koh et al., 1999; Deng et al., 2003; Guha et al., 2011; Yan et al., 2011; Matsui et al., 2013; Sai et al., 2014).
Because of the limitation on solar spectrum collection by single junction PV devices, a-
Si:H based multijunction solar cells have been mass-produced (Ahn et al., 2012).
Development of devices incorporating thinner layers reduces not only deposition time and production costs, but also light-induced degradation (LID) in a-Si:H. Thicker absorber layers exhibit larger Staebler–Wronski Effect (SWE) which negatively impacts 5
solar cell performance (Deng et al., 2003). Si:H solar cells have an advantage in this respect because devices can be constructed to incorporate multiple junctions or
‘micromorph tandem’ structures (Shah et al., 2002). For example, micromorph tandem structures are produced by combining an a-Si:H top cell with a (μc/nc)-Si:H bottom cell in the superstrate configuration (Shah et al., 2002). Most commonly, a-Si:H and a-Si1- xGex:H films are deposited by PECVD of silane (SiH4) or disilane (Si2H6), germane
o (GeH4), and hydrogen (H2) at temperatures lower than ~300 C. By changing the deposition parameters in the PECVD process, the band gap of the intrinsic layer can be changed to maximally collect a wider range of the solar spectrum according to other layer deposited (Dahal et al., 2014).
In order to analyze the properties of thin films, such as thickness, optical band gap, and phase, a powerful non-destructive method is essential. The prime characterization tools in this work for optimization of the layers for a-Si:H solar cell applications include real time spectroscopic ellipsometry (RTSE). Being a non-invasive technique, RTSE is useful for exploring the fundamental relationships among the preparation and the structural parameters of thin film materials (Kajzar et al., 1986). This technique has been employed to analyze the microstructural evolution of doped and undoped Si:H layers deposited on various bulk and thin film substrates ranging from those which are microscopically smooth to rough, including native and thermal oxide coated crystalline silicon, transparent conducting oxide (TCO) coated glass, as well as a-
Si:H and nc-Si:H underlying layers (Koh 1995; Collins et al., 2000; Ferlauto et al.,
2001). nc-Si:H, either as individual component layers or comprising all layers in a PV junction, has significant enhancement in near infrared (IR) absorption of the solar 6
spectrum and has high stability against the prolonged light illumination in contrast to its amorphous counterpart (Vetterl et al., 2000; Shah et al., 2003). The quantitative analysis, characterization, and control of the relative nanocrystalline and amorphous volume fractions within mixed-phase films is also a major challenge in Si:H manufacturing. Most often the nanocrystalline fraction is estimated from x-ray diffraction or Raman spectroscopy, which can yield values ranging an order of magnitude (Vetterl et al., 2000;
Shah et al., 2003; Deng et al., 2003). Although these measurements are valuable, limitations exist. Typically ex situ x-ray diffraction measurements average information over the full depth of a thin film sample, and ex situ Raman spectroscopy averages information over a finite penetration depth into the sample that is dependent upon the wavelength of the probing laser, its power, and the absorption coefficient of the material.
There is a greater challenge to profile these materials non-invasively due to penetration depth of the probes in the material and likely non-uniform crystallite fraction with depth into films. Si:H films may be inhomogeneous with thickness as crystallites nucleate from the amorphous phase and while the amorphous and nanocrystalline phases coexist.
Deconvolving gradients in crystallinity from ex situ x-ray diffraction and Raman spectroscopy measurements requires multiple samples, while in situ RTSE measurements applied during film deposition have been used to quantify structural gradients in crystallinity within a single film. Also, this type of non-invasive analysis has also led to an extensive set of deposition phase diagrams designed to guide optimization of Si:H based films and devices.
In the mid-IR range of the spectrum, there exists a large optical contrast between the optical properties in the form of the complex dielectric function ( = 1 + i2) spectra 7
of the semiconductor and the TCO layers. Free carrier responses of the zinc oxide (ZnO) and fluorine doped tin oxide (SnO2:F) layers may be modified by over-deposition of the semiconducting layers and by different processing steps. Thus, the objective of building a realistic optical model of the solar cell structure at various stages of processing is further enabled by characterizing the TCO layers via IR-spectroscopic ellipsometry (IR-SE). In addition, for a-Si:H extracted from IR-SE data is sensitive to the silicon-hydrogen bonding configuration.
The two techniques (in situ RTSE and ex situ IR-SE) mentioned above can provide different types of information about the film and therefore, they can complement one another. For instance, in situ RTSE is useful in extracting the nucleation and growth characteristics of a thin film including phase evolution through an effective medium theory, as well as very accurate optical properties of materials at the temperature of the deposition process i.e., in the absence of sample oxidation or contamination. On the other hand, ex situ IR extended analysis of same sample also retains sensitivity to the multilayer structure of thin film material stacks, but can detect IR absorption in back reflector (BR) components (related to electrical transport properties and phonon modes) and the growth of oxide layers on the surface upon exposure of materials to the laboratory ambient medium. Also, silicon-hydrogen bonding in Si:H and modification of
TCO after Si:H deposition can be assessed. These techniques have better control on understanding of growth processes and alteration of structure of Si:H in different configurations of devices.
8
1.2 Dissertation Organization
This dissertation presents a comprehensive analysis of SE studies of layers in Si:H based solar cells. The first section of Chapter 1 provided an overview of thin film PV, especially a-Si:H, and the motivation for employing SE techniques. The second section, i.e. the present one, refers to the structure of the thesis, and in this section, an outline will be provided.
Chapter 2 is an overview of the basic principles used in SE and RTSE data analysis, including the propagation of light through thin film stacks and the description of composite materials through effective medium theories. The subsequent part of this chapter is a description of the experimental setup and data collected by both the RTSE and IR-SE equipment. The last part of this chapter explains the sample arrangement and deposition system used in the fabrication of the solar cell devices.
Chapter 3 is devoted to a comprehensive explanation of RTSE data analysis used in this research. This chapter determines spectra in and monolayer structural changes using a two-layer (bulk film)/(surface roughness) model, numerical inversion, and least- squares regression. In this work we study the effects of adding disilane (Si2H6) to the standard silane (SiH4) precursor on the growth evolution of Si:H by using in situ RTSE during film growth. The results of these RTSE studies indicate the potential impact of these additives on the evolution of crystallinity, material deposition rate, and electronic quality of the material.
Chapter 4 represents the phase diagram development for n-type, intrinsic, and p- type Si:H depositions on Ag/ZnO back reflector overcoated glass substrates based on
9
RTSE data collection and analysis. The growth regimes of interest are those achieving the most ordered protocrystalline n-, i- and p-layers in the n-i-p configurations as identified from the deposition phase diagram. The band gap trends of n-, i- and p-layer deposited at different conditions are presented.
Chapter 5 discusses the microstructural evolution including a surface roughening and smoothening onsets associated with the amorphous-to-nanocrystalline transition.
RTSE of nc-Si:H deposited on different substrates is used to observe crystallites nucleation from the amorphous phase and the amorphous-to-nanocrystalline transition versus accumulated thickness. Virtual interface techniques are used to further analyze
RTSE data for films exhibiting preferential growth of the crystalline phase. This information on the evolution of crystallinity can be correlated with literature results describing desirable characteristics of Si:H layers in nc-Si:H based solar cells.
In Chapter 6, IR-SE studies are performed for ZnO deposited on a silver back
TM reflector (BR), and SnO2:F coated (TEC -15) glass. The studies are conducted to probe the modification of the underlying TCO layer optical properties due to the deposition of the overlying semiconductor layers. Any modification of the underlying TCO layer is important for solar cell operation, as this layer serves as an electrode and any parasitic absorption may result in less light incident upon the intrinsic absorber layer. The associated parameters, such as resistivity () and scattering time () for the component layers of Ag, ZnO, and SnO2:F are determined and reported in this chapter. Furthermore, silicon-hydrogen bonding in a-Si:H layers in the n-i-p and p-i-n solar cell configurations are studied.
10
In Chapter 7, we conclude the dissertation by highlighting some of the important results along with the suggestions for future research direction. In addition to a description of immediate plans, a description of the longer range future research activities is presented in this Chapter. Such activities are designed to elevate the scientific understanding of the results presented in this dissertation and to effect improvements in
Si:H solar cell performance as a consequence.
11
Chapter 2
Experimental Techniques used in Optical Measurements
2.1. Deposition of Si:H Thin Films
Thin film Si:H technology has been explored for years as an alternative to c-Si PV and solar cells have been made from thin film a-Si:H, a-Si1-xGex:H, and nc-Si:H in single or multiple junction configurations. The solar cells made from c-Silicon are highly efficient, as high as 25% in the laboratory, but is limited in lowest production cost achievable due to the quantity of high quality silicon material required in high efficiency modules. a-Si:H materials exhibit a much higher optical absorption coefficient in the visible range of the solar spectrum than c-Si enables devices with absorber layers less than 1 μm thick as compared to 50 μm or greater for c-Si (Shah et al., 1999; Nelson,
2003). The optical transitions that are forbidden in indirect gap single c-Si, due to the requisite of crystal momentum conservation, become permitted in a-Si:H as a result of the easing of conservation laws for a disorder structured material.
In thin film Si:H cells, reducing recombination losses due to defects related to the disordered structure is a main concern. Specifically, the doped forms of Si:H are highly defective and do not support high mobility of minor carriers. For this reason, Si:H-based
12
solar cells are fabricated in the substrate n-i-p structures and superstrate p-i-n configurations different than the conventional p-n junction solar cell. The substrate n-i-p configuration typically uses rigid or flexible substrate with back reflector whereas the superstrate p-i-n configuration typically uses TCO coated glass as the starting superstrate for rigid cell or module fabrication. Irrespective of the configuration, substrate n-i-p or superstrate p-i-n, the structure is designed so that the light enters through the thin p-layer compared to intrinsic layer. Figure 2-1 shows the schematic of a typical single junction a-
Si:H solar cell in the substrate configuration on Ag/ZnO back reflector coated glass.
Figure 2-1. A typical single junction a-Si:H solar cell in the substrate configuration on a back reflector (BR) Ag/ZnO coated glass. Substrate devices are illuminated from the p-type Si:H layer and the metal contact Ag and TCO (Indium doped Tin oxide) act as the two electrodes that carry current to the external circuit.
13
This design is due to the lower mobility of holes which does not travels far on average as electrons (Deng et al., 2003). Si:H thin films can be deposited by methods such as
PECVD, vhf glow discharge deposition, and hot-wire glow discharge deposition.
The structural and optical properties of deposited thin films depend sensitively on the deposition process. This section provides a brief description of the experimental methods that has been applied to fabricate the various layers of the thin film Si:H solar cell in the n-i-p configuration. The description starts with the cluster tool deposition system for sputtering as well as for PECVD of component layers.
2.1.1. Cluster Tool Deposition System
The computer controlled multi-chamber cassette cluster tool system at The
University of Toledo consists of eight modular process zones (MPZs) or chambers. The system has three sputtering chambers, one for metal, another TCOs, and the third currently with a mapping ellipsometer installed. The system also has four PECVD chambers for standard Si:H n-, i-, and p- layer depositions and one chamber specifically for nc-Si:H deposition at high temperature as shown in Figure 2-2. These chambers are connected through a central transfer zone, called the isolation and transfer zone (ITZ) so that the thin film Si:H solar cell layers can be deposited without a vacuum break. The load lock in the system is used for the entry and exit of the substrate carrier with the option of preheating / degassing. Each chamber is equipped with ellipsometry ports for in situ RTSE measurement. The system also includes: pumping, electronics cabinet, computer and gas manifold. The high vacuum pumping and the process gas pumping of
MPZs are provided via turbo molecular pumps backed by rotary vane pumps. The system 14
can be partially operated locally via the control panel on the electronics rack or via the computer interface in manual, semi-automatic, or fully automatic modes. The system is comprised of proper disposal of toxic/pyrophoric gases as well as emergency gas off system.
Figure 2-2. Eight chamber cluster tool deposition system with the capabilities of load lock system, magnetron sputtering of metals and TCOs in three different chambers; rf PECVD of Si:H n-, i-, and p- layers in three separate chambers; and high temperature PECVD of Si:H layer deposition in a single chamber. The chambers are designed for rigid glass substrates as well as flexible substrates using a roll-to-roll cassette configuration. Also shown is a spectroscopic ellipsometer connected to the undoped Si:H chamber for RTSE data acquisition. In fact, all chambers are fitted with ellipsometry ports and one of them is capable of mapping of SE.
15
2.1.2. Sputtering Process
In this research, all the component thin films of the BR and contacts, comprising metals and TCOs, such as Ag, ZnO, and ITO are deposited by rf magnetron sputtering. A schematic of the sputtering apparatus setup is shown in Figure 2-3. The deposition conditions of the BR layers deposited by sputtering are provided in the applicable chapters. The ejection of atoms from the surface of a material (the target) by bombardment with energetic particles is called sputtering.
Figure 2-3. Schematic of the rf magnetron sputtering process for metal like chromium (Cr), silver (Ag) and TCO like zinc oxide (ZnO) and Indium doped Tin Oxide (ITO). The deposition was done on respective substrates used in Si:H solar cells mounted above the rectangular target. The rf power leads to dc self-biasing of the target relative to the substrate; (adopted and modified from Dahal, 2013).
16
Sputtered atoms travel until they strike a substrate, where they chemisorb, bonding to the substrate atoms and accumulating to form the film (Wasa et al., 1992). The modification of deposition process and conditions of thin films and their interfaces in real time during deposition can result in a range of desirable properties. Consequently, sputtered thin films are used extensively in integrated circuits, electronic, optical devices, multilayer optical coatings, and hard and decorative coatings.
Among the different types of sputtering, one of the most popular radio frequency
(rf) magnetron sputtering, is applied in the present work. In magnetron sputtering, the density of the plasma used in the process is boosted by a magnetic field oriented parallel to the surface of the target (or cathode), trapping energetic electrons in this region
(Rossnagel et al., 1990). For both metal and TCO sputtering in this research, alternating rf power at 13.56 MHz is applied to the cathode with the chamber walls grounded. rf frequency is special since the voltage can be coupled through any kind of impedance, and therefore the electrodes (target and substrate plane) do not need to be conductors.
Specifically, rf magnetron sputtering is required for TCOs, while either rf or dc sputtering is possible for the metal layer depositions. In rf magnetron sputtering, the target develops a self-biased negative potential relative to ground such that it behaves like dc sputtering whereby positive ion bombardment removes target atoms for consequent deposition. The dissimilarity in electron and ion mobilities infers that the positively charged electrode draws more electron current than the negatively charged electrode draws positive ion current. Therefore, a large primary electron current flows during the positive half cycle of the rf power supply while a small positive ion current flows during the negative half cycle. A net current is averaged over each cycle, charging the capacitive plasma through 17
which the rf power is dissipated via connections to target and (normally grounded) substrate holder. The target is self-biased and has an average negative potential each cycle. The high voltage applied across the electrodes before the self-biasing creates a plasma which consists of high energy electrons and ions (Chapman, 1980). In order to avoid any chemical reaction between the sputtered atoms and the sputtering gas, commonly an inert gas is used such as argon. However, in some depositions, such as of oxides and nitrides, a reactive gas is purposely added to inert gas argon so that the deposited film is a chemical compound. This type of sputtering is called “reactive sputtering.” Ar with a very small flow ratio of O2 is used as the sputtering gas for ITO deposition. The role of O2 in this reactive sputtering process is to provide sufficient oxygen for a nearly stoichiometric TCO layer. Otherwise the layer has high concentration of metal interstitials so that it becomes more conducting, but less transmitting, thus not a good transparent conductor.
The plasma emits radiation with a spectral distribution that depends on the nature of the gas used for the sputtering; as a result the plasma is sometimes called a “glow discharge”. The substrate is generally placed on conducting holder so that the desired rf power can be supplied to the plasma between the magnetron electrode and the substrate holder. The magnetic field strength is adjusted in such a way that it will influence the electrons significantly but not the ions. In the current work, Ar is used in the sputtering process for Cr, Ag, and ZnO. Sputtering also allows the deposition of films having the same composition as the target source. This is the primary reason for the widespread use of sputtering as a thin-film deposition technique giving more uniform and reproducible results. 18
2.1.3. Plasma Enhanced Chemical Vapor Deposition (PECVD)
PECVD is an excellent technique for depositing a variety of thin films at lower temperatures than other CVD reactors without compromising film quality. A schematic of the PECVD apparatus including RTSE setup is shown in Figure 2-4. The Si:H thin film layers for PV (n-, i-, and p- layers) are fabricated using rf PECVD chambers of the cluster tool.
Figure 2-4. Schematic of the PECVD chamber used in the deposition of Si:H thin films layers. The graphite heater is attached to the cathode inside the chamber under vacuum whereas the heater well is fixed outside the chamber at atmospheric pressure. The heater well runs radiatively heating the chamber as well as the substrate; (adopted and modified from Dahal, 2013).
19
The source gases including SiH4, Si2H6, and H2 flows as well as dopant carrying gases are injected into the deposition chamber. The plasma produced by the strong oscillating electric field between the two electrodes causes the precursor gas molecules to dissociate into radicals, which diffuse and adsorb onto the substrate or growing film surface. This deposition process is more confined than sputtering because of the uniform electric field established between the cathode plate and the grounded substrate holder leading predominant deposition on the inner surfaces of the electrodes. The variable parameters that can be adjusted are the cathode configuration, rf plasma power, substrate temperature, total chamber pressure, the gas flows, and gas flow ratios. The spacing between the cathode and the substrate is 1.5 cm whereas the dimension of the cathode is
6.86" x 6.86", yielding a surface area of 47 inch2 or 298 cm2 above the substrate holder.
The heater well operates radiatively and is controlled by a solid state controller. It turns on/off when the actual temperature is below/above the set point, respectively. When the flow of H2 is much higher than the flow of silicon or dopant carrying gases, plasma ignition must be stimulated by a hot wire mounted at the corner of the cathode injecting electrons into the plasma region. However, the plasma in low H2 flow generally ignites when the operating pressure and the rf power is reached. Since the electric field is confined between the cathode and the flat substrate holder and the gas pressure is high, most radicals either react in the gas phase or at the surfaces of the cathode or substrate. It has been reported that the highest concentration species in the plasma in case of Si:H films from SiH4 that contributes to film growth is SiH3 (Chapman, 1980; Matsuda et al.,
1986).
20
2.2. Material Characterization Techniques
2.2.1. Spectroscopic Ellipsometry
Ellipsometry is a non-destructive, optical measurement technique which uses the fact that light undergoes some change in polarization when it is reflected off the surface of a material. The polarization change is characteristic of the structure of sample, and when this polarization change is analyzed can yield various information about the material. Some of the advantages of using ellipsometry for extracting film properties are that it requires neither special sample preparation nor a special measurement environment. It can measure samples of any size and the vast majority of films are unaffected by the process (Fujiwara 2007).
In the beginning, ellipsometers were single wavelength instruments (Fujiwara,
2007). With the emergence of improved detector and light source technology, ellipsometers have been designed to collect spectroscopic data. If the process is designed to collect spectroscopic data from a wide of energy in time periods short enough to track sub-monolayer changes in materials systems during in situ dynamic process like growth, it is referred to as RTSE. This makes ellipsometry a powerful optical tool for monitoring a wide range of surface and thin film processes in diverse environments (Collins, 1990;
Collins et al., 2001). In particular, RTSE has proven to be robust to study the real time growth of amorphous and nanocrystalline semiconductor thin films, providing information regarding many film properties, such as chemical composition (Fujiwara,
1998; Podraza et al., 2006a), microstructural evolution (An et al., 1990; Koh et al., 1999;
21
Ferlauto et al., 2004), optical band gap (An et al., 1991; Koh et al., 1995; Kim et al.,
1996) and void volume fractions (Kim et al., 1996).
In what follows, we will introduce the theoretical formalism of ellipsometry and polarized light from Maxwell’s Equations to theoretical basis of effective medium approximation in order to describe the complex optical response of homogeneous composites of different materials in the polarization state of light after a plane wave of light with a known polarization state interacts with the planar surface/interfaces of a sample (Azzam and Bassara, 1977). The general schematic drawing of a single rotating- compensator multichannel ellipsometer apparatus used in data collection of Si:H thin films is shown in Figure 2-5.
Sample (S)
s s p p
i rp tan e
rs
Rotating Compensator Analyzer (A) (CR)
Polarizer (P)
Detector
Xe Light source
Figure 2-5. Schematic drawing of a rotating-compensator multichannel ellipsometer used in RTSE analysis in this thesis. The figure illustrates the arrangement of components from the light source to the detector.
22
Structural properties such as bulk layer and surface roughness layer thicknesses on the reflecting sample surface can be obtained with atomic layer sensitivity when the polarization state of the light wave reflected from the sample can be measured accurately during the dynamic process. The analysis of reflected light beam with certain polarization state can be used to extract the sample’s structural and optical properties. At each wavelength, two wave-matter interaction parameters, which is a relative amplitude ratio, and Δ, which is a phase shift difference, can be deduced from the polarization state of a beam before and after its specular reflection from the sample.
2.2.1.1 Theoretical Formalism
i) Propagation of Light in Thin Films
The study of the propagation of electromagnetic waves in free space and their interaction in media can be explained by Maxwell’s equation in terms of SI unit as: .,E (2.1) 0
B E (2.2) t .B 0 (2.3)
E BJ µ00 (2.4) t
Here, E, B, J and ρ are the macroscopic quantities, namely electric field vector, magnetic
induction vector, current density vector, and charge density, respectively. 0 and µ0 are the permittivity and permeability of free space, respectively (Jackson, 1998).
23
To describe the propagation of electromagnetic radiation in matter, two additional vectors, namely displacement vector D and magnetic field intensity vector H are introduced as:
DE ˆ 0 (2.5) and,
BH µ µˆ 0 (2.6)
Here, ˆ and µˆ are known as dielectric permittivity and magnetic permeability of the medium. They are generally expressed as tensors to describe the optical response of the medium (Wooten 2013).
If we consider the medium to be isotropic and homogeneous, 0 and µˆ reduce to scalars
(ˆ , µˆ μ) and do not depend on the position vector r . Furthermore, if the medium is non-magnetic (μ 1) and if there are no external charges or currents (i.e., charges and currents other than those of the neutral solids), then the solution of Maxwell’s Equations yield a transverse polarized electromagnetic wave propagating at phase speed c /[Re(N)], where N is the complex index of refraction of the medium as:
2 12 i N . (2.7)
If the electromagnetic wave is propagated along z-axis, then the electric field of the electromagnetic wave can be represented by:
Nz EE(z, t)0 exp[iω t ] . (2.8) c
The complex refractive index N can be expressed as N n ik, where the real part is
24
the real index of refraction and the imaginary part k is the extinction coefficient of the material., Expanding N2 from Eqn. (2.7) and equating both real and imaginary part of
N n ik , we get,
22 1 n k , (2.9)
2 2nk . (2.10)
The complex vector E0 determines the polarization state of the wave. In Cartesian coordinate system, can be expressed as:
E0E 0x exp iφ x x E 0y exp iφ y y , (2.11)
where, φx and φy represent the absolute phase of E0x and E0y at z = 0 and t = 0. For the wave propagation in isotropic media, the electric filed is used to describe the polarization state. Polarization of light refers to the time evolution of vector fields that describe the wave at a given fixed point in space. For a general polarization state, the electric field vector as described in Equation (2.11) traces out an ellipse in the x-y plane as a function
of time at a fixed value z z0 as shown in Figure 2-6. There are two limiting cases: (i)
light is linearly polarized when φx = φy and (ii) circularly polarized when φxx φ π / 2 and amplitudes of electric field vector at z = 0 and t = 0 in x-and y- coordinates are equal,
i.e. EE0x 0y . The polarization ellipse is generally defined by orthogonal field components in terms of the plane of incidence, rather than Cartesian coordinates (x, y).
Polarized electric field vector components can be chosen as parallel (p) and perpendicular
(s) to the plane of incidence as shown in Figure 2-7. The optical response of a secularly reflecting surface can be described by the complex amplitude reflection coefficients as:
25
Er p r |r|exp(i ) , (2.12) p p p Ei p
Er r | r | exp(i ) s , (2.13) s s s Ei s
As in Figure 2-7, Ep(s) and φp(s) represent the parallel component of complex electric field amplitude and the phase shift of light after reflection, respectively. Subscripts “r” and “i” represent the reflected and incident beams. Ellipsometry measures the ratio of the “p” to the “s” reflection coefficient rather than individual values. Thus, the complex amplitude reflection ratio combines the two measured parameters in ellipsometry ( , ).
Figure 2-6. Schematic representation for the variation of the state of polarization with
phase difference as xyor yx. In this figure, the amplitudes of the
waves in the x and y directions are the same, i.e. EE0x 0y , for circularly polarized light, adapted from (Fujiwara, 2007).
26
It is defined as:
rp rp ρr exp[iφφ p s ] (2.14) rrss
Here, ρr can be described in terms of the ellipsometric angles ( , ) equivalent to the relative amplitude tanΨ exp(i ),and the phase difference ( ) between the p and s waves:
ρr tanΨ exp(i ), (2.15) where
rp tanΨ, rs
φφps .
Figure 2-7. Schematic of the plane of incidence along with the propagation vectors that demonstrates the definition of p-polarized and s-polarized fields. Here,k i , k t and kr represent the incident, transmitted and reflected propagation vectors, respectively. 27
ii) Reflection from a Single Interface
The simplest example of the single interface consists of an interface between a semi-infinite ambient and a semi-infinite isotropic, homogeneous, and uniform medium
of complex index of refraction, N1 (Azzam and Bassara, 1977). The Fresnel equations
describe the reflection of a plane electromagnetic wave incident at an angle 1 on a
planar interface between two materials with complex refractive indices N0 and as,
N1 Cos(θ i ) N 0 Cos(θ t ) rp (2.16) N1 Cos(θ i ) N 0 Cos(θ t )
N0 Cos(θ i ) N 1 Cos(θ t ) rs (2.17) N0 Cos(θ i ) N 1 Cos(θ t )
Here, is the real index of refraction of ambient, i and t are angle of incidence and complex angle of transmission, respectively. Complex index of refraction can be
expressed in terms of dispersion and extinction coefficient as N1 n 1 ik 1 . One of the extra variables in these two equations can be solved with the help of Snell’s law where
and N1 related as,
N0 Sinθ i N 1 Sin θ t (2.18)
With the use of Equations (2.16), (2.17) an (2.18), we can express ellipsometric angle ( , ) in terms of , and . The final result, which relates the measured ellipsometry parameters to the complex dielectric function spectra of the substrate medium is given as: 2 2 2 2 2 1ρr 1 NN 1 0 Sin θ i 1 tan θ i (2.22) 1 ρ r
28
iii) Reflection from a Multilayer Stack
In the case of a thin film stack, i.e., one or more stacked layers on top of a substrate, each layer is characterized by a thickness and different thickness and different optical properties. The reflection and transmission behavior can be calculated using the
S matrix formalism (Azzam and Bassara, 1977). The matrix formalism allows us to simulate measurable quantities such as transmittance, reflectance, and ellipsometric angles in terms of thickness, index of refraction, n and extinction coefficient, k for each layers of stack. Figure 2-8 illustrates a plane wave that impinges on a multilayer stack having 5 layers between a transparent ambient and semi-infinite substrate. Here, each
layer is characterized by its complex index N j and thickness d j . A 2 x 2 scattering matrix for both p and s polarization should be determined. The matrix can be expressed as a product of the interface matrices, I and layer matrices, L that describe the effect of individual interfaces and layers of the entire film structure. The matrix can be expressed as,
6 SILI 01 j j( j 1) (2.23) j1
Here, subscript, “0” refers to the ambient and “6” refers to the semi-infinite substrate in the schematic. Now, we define interface matrix and layer matrix separately. Interface
matrix I j( j 1) at interface j/(j+1) can be defined in terms of Fresnel reflection and
transmission coefficients, rj( j 1) and, t j( j 1) respectively as:
29
1 1rj( j 1) I . (2.24) j( j 1) r1 t j( j 1) j( j 1)
Figure 2-8. Schematic representation of plane wave reflection from a multilayer structure of n-i-p configuration single junction a-Si:H PV device at non- normal incidence. The structure consists of 5 layers plus the semi-infinite
Ag substrate (6) and the ambient (0). 0 is the angle of incidence at the
interface between the ambient 0 and layer 1, and 6 is the angle of refraction into the substrate. The wave vector S within the ambient and at the (j, j+1) interfaces are shown as arrows.
30
For s-polarized wave, the complex Fresnel Coefficients are given by:
qqj j 1 rj( j 1) (2.25) qqj j 1
2q j t j( j 1) . (2.26) qqj j 1
Here, q j can be related to incident angle j as:
qj N j Cos( j ). (2.27)
The incident angle for the interface between jth layer and its adjacent layer, j+1th, can be obtained from successive application of Snell’s law as:
N0 Sin 0 N 1 Sin 1 N 2 Sin 2 N5 Sin5 N 6 Sin 6 . (2.28) For p-polarized wave, the complex Fresnel Coefficients are given by:
Nj Cos j N j 1 Cos j 1 rj( j 1) (2.29) Nj Cos j N j 1 Cos j 1
2Njj Cos t j( j 1) . (2.30) Nj 1 Cos j N j Cos j 1
The layer matrix L j describes the propagation of light through the layer j. It is also called the phase matrix as the matrix incorporates the changes in phase angle as the light traverses through the thickness of jth layer. It is expressed in terms of thickness of jth layer
d j as:
31
Z j 0 L j 1 . (2.31) 0 Z j
Here,
2πid Z exp(j N Cos( )) . (2.32) jλ j j
With Eqns. (2.24) through (2.31), scattering matrix S as defined in Eqn. (2.23) can be evaluated for a structure with multiple films stacked together. Once the scattering matrix
is evaluated, the complex amplitudes of the electric fields E0 at the top of the structure
can be related to the electric field vector Em1 through scattering matrix as:
E0 SE m 1 . (2.33)
In terms of electric field vector travelling in forward (+) and backward (-) z- direction,
Equation (2.32) can be written as:
E0 SS11 12 Em1 . (2.34) E0 SS21 22 Em1
As both interface and layer matrices are 2x2 matrices, S is also a 2x2 matrix. It is now necessary to relate the S matrix to the measured quantities, such as the complex Fresnel reflection coefficients of the entire structure:
Sp s 21 rp(s) . (2.35) Sp s 11
Thus, the ellipsometry angles are determined from the ratio of the reflection coefficients:
rSp p21 Ss11 ρr tanΨexp i . (2.36) rs S p11 S s21
32
2.2.1.2 Experimental Methods in Spectroscopic Ellipsometry
i) Real Time and Exsitu Spectroscopic Ellipsometry
The major optical characterization techniques in this work are in situ RTSE over a spectral range from 0.73 to 5.88 eV. Both tools employ a rotating compensator multichannel ellipsometer for high accuracy and high speed as shown in Figure 2-5.
RTSE was performed in situ at a single spot during deposition and provides ellipsometric spectra (in the form of N = cos 2, C = sin 2 cos , S = sin 2 sin ) from 0.734 to
5.88 eV with a minimum data acquisition time of 50 ms (Lee et al., 1998; Johs et al.,
1999). This type of instrument collects all photon energies in parallel by a combination of a one dimensional linear detector array and serial pixel readout. Dual detectors are required to access this spectral range, and consist of a silicon based charged coupled device (CCD) and indium gallium arsenide photodiode array (PDA).RTSE measurements were collected at the respective deposition temperature at angles of incidence near 70o and spectra obtained from single optical cycles were averaged over 1.5 second intervals to reduce the signal-to-noise ratio. Analysis of experimentally collected RTSE data was performed using J. A. Woollam Co. CompleteEASE software. The time evolution of the bulk and surface roughness layer thicknesses, as well as the spectroscopic ε of the bulk
Si:H layers, extracted from RTSE data used a global Σ-minimization procedure as explained later in section 2.2.1.3. For Si:H films, global Σ-minimization analysis of
RTSE involves using test db and ds values for the Si:H layer being deposited on top of a pre-defined substrate stack to numerically solve for of the Si:H layer (Oldham, 1969).
The test values of are then used to fit other spectra collected at different times when the
33
film is relatively homogeneous, typically near 100-200 Å in accumulated material thickness for Si:H films where structural transitions, namely the nucleation of crystallites from the amorphous phase, have not yet had time to mature. The approach is applied in the regime prior to crystallite nucleation and is iterated in order to obtain numerically inverted yielding the lowest average error, , over the multiple time measurements selected and then used to determine structural parameter variations over the full set of
RTSE data. A similar tool is applied to spatially resolve variations in structure and across samples. In this work we primarily fix and consider variations in bulk and surface roughness thicknesses. ii) Infrared Spectroscopic Ellipsometry
IR-SE incorporates a Fourier Transform Infrared (FTIR) interferometer source with a rotating compensator ellipsometer to provide accurate ellipsometric measurements in the near to mid-IR range as shown in Figure 2-9. Wide-band polarizers and a patented compensator are combined with an optimized beam splitter, collimators, and detector to provide the widest available IR-extended spectral range. IR-SE was performed ex situ on the different samples, including before and after Si:H deposition on TCOs. The IR-SE
VASE ellipsometer is a commercial product of J. A. Woollam Co. The instrument covers the mid-IR wavelength range from 1.4 to 33 microns (0.04 eV to 0.89 eV in photon energy). IR ellipsometry has potentially higher sensitivity to ultrathin films than FTIR reflection / absorbance due to its capability of detecting changes to both amplitude and phase. As such it retains sensitivity to layer thicknesses, complex optical response of materials, and also can be used to assess chemical composition of studied materials via
IR vibrational modes. 34
Figure 2-9. IR-VASE is a variable angle spectroscopic ellipsometer that acquires ellipsometric data in the near to mid infrared range of the electromagnetic spectrum. The ellipsometer is consists of FTIR spectrometer, input unit that focuses the light exiting the spectrometer and establishes the polarization state relative to the plane of incidence, sample stage and polarization analyzer and detector unit that converts the intensity of the reflected beam to an electrical signal and detects the polarization state.
The IR-VASE offers non-contacting measurements of many different material properties including thickness, optical constants, material composition, chemical bonding, and doping concentration.
2.2.1.3 Data Analysis Strategies in Spectroscopic Ellipsometry
Spectroscopic ellipsometry is an indirect technique to obtain the physical and optical properties of material. It uses the incident light of varying energy and the changes in the polarization state of the light beam upon reflection are recorded in terms of observable quantities ( , ). To retrieve the desired information about the films and
35
stack, the measured spectra ( , ) can be analyzed by employing various methods such as least-squares regression analysis, mathematical inversion, and global minimization techniques for RTSE data. The most basic structure that can be modeled and analyzed using ellipsometry is an isotropic, atomically smooth, opaque material. An example of such a structure is a single interface where light enters from semi-infinite ambient environment to semi-infinite medium. Equation 2.22 is a closed form solution to extract for such a situation. If other layers such as surface roughness and native oxides are introduced into the model, the model diverts from its simplicity. Furthermore, if the medium is not semi-infinite, the layer must be included into the model with the variable parameters associated with the layer thickness. If the sample structure is known, e.g., the thicknesses of the layers, including the interface and surface roughness, as well as the optical properties of all but one of the constituent materials, then it is possible to apply the equations based on multilayer optical analysis as described in detail in Section 2.2.1.1 under “reflection from a multilayer stack” to extract the spectra in of the unknown material from the ellipsometric angles. As the known parameters in the equations involved in the sections are coupled together, it is impossible to find the closed form analytical solutions. Therefore, a numerical method is used to extract optical functions of the unknown material through an iterative algorithm. This process is called “numerical inversion” (Aspnes et al., 1982). For the analysis of ellipsometric spectra collected in this work, we utilize least squares regression and global minimization coupled with numerical inversion.
36
i) Least-Square Regression Analysis
Least-squares regression analysis is the most commonly applied procedure for the interpretation of spectroscopic ellipsometry data obtained for a thin film sample. If the component materials of the sample structure and their spectra in are known, photon energy independent parameters, such as layer thickness and material volume fractions, can be used to create a model for the film structure designed to fit the spectral data. Also, if the complex dielectric function spectra for different films are also known as a function of energy (parameterized ), these parameters can also be used in the least squares regression as well.
The first step in performing a least squares regression analysis is to create a physically realistic model for the sample structure. Once the model is established, the second step is to guess the values of the various energy independent parameters such as bulk, surface, and interface layer thicknesses and also the volume fraction of the constituent materials of composites as “initial guess”. In addition to these parameters, there may also be some other variables as initial guesses used to parameterize of a material. The third step is to generate a trial “simulated spectra” based on the first guesses. Next, the fourth step is to adjust the guessed parameters and variables to minimize the difference between the simulated spectra and experimentally observed spectra from the device using an automated iterative approach. The quantity minimized is the square of the unbiased estimator of the mean square deviation. Mean square error
(MSE) is calculated from the N = cos2, C = sin2 cos, and S = sin2 sin values and no weighting is applied.
37
2 cos 2mod cos 2 exp jj 2 1 N mod mod exp exp sin 2 j cos j sin 2 j cos j . (2.37) 3N M j1 2 mod mod exp exp sin 2 sin sin 2 sin j j j j
Here, N is the number of measured values and M is the number of fit parameters, “exp” denotes experimental spectra, and “mod” denotes the spectra generated from the model.
For single point ellipsometry experiments, M equals to 2 corresponding to the two experimentally obtained angles E,E . The minimum of the unbiased estimator
2 yields the best fit components. The degree of correlation among the parameters can be determined by inspecting the correlation matrix. This statistical information is used to evaluate the validity of the model and its fit. For example, two parameters are strongly correlated and both parameters are not in the range of accepted values after the unbiased estimator minimum is found, then it may be necessary to determine one of the parameters from a technique other than ellipsometry. The parameter can be fixed and the fit repeated.
ii) Global -minimization Method for RTSE Data Analysis
Spectra in of the resulting film depends on the deposition conditions for any thin film deposition processes which may impact film internal structure. This dynamic process of optical function evolution allows for extracting depth profiles in the material’s and phase composition by RTSE. From the analysis of RTSE data versus time during growth, two types of parameters can be obtained describing the thin film: (i) the time independent parameters (or at least parameters that can be considered time independent over a
38
selected time interval), such as which are functions of photon energy and (ii) the photon energy independent microstructural parameters, such as the bulk layer and surface roughness thicknesses which are functions of time. In order to analyze the multiple sets of ellipsometric spectra collected versus time, a method combining mathematical inversion solving for (Oldham, 1969) and least squares regression analysis has been developed (An et al., 1990).
The first step in this type of analysis is to determine the optical properties and structural parameters of the substrate and the properties and structure of any layers underlying the film under investigation before thin film deposition. Figure 2-10 depicts a two layer optical model that is used to analyze bulk film growth on a smooth substrate.
Figure 2-10. Schematic of a two layer sample structure consisting of a bulk layer and surface roughness layer on top of a known substrate. The optical response of the surface roughness layer is modeled by the Bruggeman effective medium approximation as a 0.5/0.5 bulk material/void volume fraction mixture.
39
If the experimental setup allows, it is best to take ellipsometric spectra of each layer individually and perform least-squares analysis on each layer of the underlying stack as it is built up. After the substrate’s optical properties and structure are understood sufficiently well, a suitable model for the analysis of thin film growth is adopted and properties retrieved are held fixed while analyzing the next overlying layer. This process is repeated until the ambient above the nth deposition is reached. The variable microstructural parameters include the bulk layer thickness db and the surface roughness layer thickness ds for the sample structure in Figure 2-10. The bulk layer has the characteristic spectra in to be obtained from the so called “global minimization method” whereas, the optical response of the surface roughness layer is represented by a
Bruggeman effective medium approximation (Aspnes, 1982; Fujiwara et al., 2000) with a spectra in calculated assuming a 0.5/0.5 volume fraction mixture of ambient and bulk layer media. Prior to forming a continuous film, the layer may be represented by an effective medium approximation of variable material and void fractions. After the film becomes continuous, the void volume fraction often stabilizes near 0.5 and this layer is incorporated as the surface roughness layer above a continuous bulk layer (An et al.,
1990).
iii) Cody-Lorentz Oscillator and Parameter Coupling of the Cody-Lorentz Oscillator
The Cody-Lorenz oscillator model, developed by Ferlauto et al, (Ferlauto et al.,
2002), is also designed for amorphous solids. Spectra in for all a-Si:H n-, i-, and p- layers in this thesis research are modeled using the Cody-Lorentz oscillator.
40
It is similar to the Tauc-Lorentz in that it defines the optical band gap energy Eg and a
Lorentzian absorption peak. However, the two models behave differently in the absorption onset region. The Tauc-Lorentz model follows the Tauc Law formula
22 2 2g (EE)E, whereas Cody-Lorentz model follows 2g (EE), which is Cody's modification of the Tauc Law. The Ferlauto et al publication also includes an Urbach term, which describes the absorption below the band gap primarily due to structural and thermal disorder in the material. The Tauc-Lorentz model is based on the assumptions of parabolic bands and a constant momentum matrix element (CM-ME), whereas the Cody-
Lorentz model is based on the assumptions of parabolic bands and a constant dipole matrix element (CD-ME).The equation for 2 defining the Cody-Lorentz oscillator is given by:
E1 EE t exp 0 E Et EEu 2 CL (2.38) (EE) 2 AE E g .0 E E (EE)E(EE)E2 2 2 2 2 2 2 t g p 0
The parameter Et is the energy at which the Urbach tail absorption transitions to the band- to-band excitations. The parameter Eu is the Urbach absorption tail slope that gives the rate at which the Urbach absorption exponentially decays with energy below Et. The parameter Ep allows the user to define the energy Eg+Ep at which the function transitions from gap-like behavior to Lorentz oscillator-like behavior. The quantities A, E0, and are the amplitude, energy, and broadening parameters of the Lorentzian peak, and Eg is the band gap energy. Applying the continuity condition at E = Et, the parameter E1 becomes E1 = Et G(Et) L(Et). The real part of the complex dielectric function spectra (ε1)
41
in the Cody Lorentz oscillator model is obtained from the Kramers-Kronig transformation of ε2:
2 2 C L P d (2.39) 1 22 Eg E
Equation 2.39 can be solved analytically as described in Ferlauto et al. 2002. In this thesis research the Urbach tail absorption is neglected for all amorphous materials as reflection mode ellipsometry measurements are relatively insensitive to this feature.
Measurements more suited to characterizing low values of the absorption coefficient, such as transmittance spectroscopy and dual beam photoconductivity, were used in
Ferlauto et. al., 2002 to obtain sensitivity to the Urbach tails.
Although only five parameters (A, , E0, Eg and Ep) were needed to describe spectra in for a-Si:H, these parameters were correlated with optical band gap Eg(T&R) as measured by transmission and reflection spectroscopy. This correlation is used to generate of high quality PV materials at room temperature from a single parameter
(Ferlauto et al., 2002). This correlation covers the range of Eg (T&R) from 1.30 eV to
1.95 eV. The mathematical expressions of these correlations are reprinted from Ferlauto et al. 2002:
A = 64.7 + 7.05 Eg (T&R) for Eg (T&R) ≤ 1.8 eV (2.40)
A = -435 + 289 Eg (T&R) for Eg (T&R) > 1.8 eV (2.41)
E0 = 3.32 + 0.280 Eg (T&R) for Eg (T&R) ≤ 1.8 eV (2.42)
E0 = 5.31 - 0.825 Eg (T&R) for Eg (T&R) >1.8 eV (2.43)
= 3.79 - 0.906 Eg (T&R) for Eg (T&R) ≤ 1.8 eV (2.44)
42
= -6.68 + 4.88 Eg (T&R) for Eg (T&R) > 1.8 eV (2.45)
Ep = -0.777 + 108 Eg (T&R) for Eg (T&R) ≤ 1.8 eV (2.46)
Ep = -14.7 + 8.82 Eg (T&R) for Eg (T&R) > 1.8 eV (2.47)
Eg = 0.343 + 0.770 Eg (T&R) (2.48)
By replacing Eg(T&R) with the Cody optical band gap (Eg) through Equation (2.48), the new coupling equations between the Cody-Lorentz parameters and Eg become:
A = 61.56 + 9.16 Eg for Eg ≤ 1.73 eV (2.49)
A = -563.736 + 375.325 Eg for Eg > 1.73 eV (2.50)
E0 = 3.195 + 0.364 Eg for Eg ≤ 1.73 eV (2.51)
E0 = 5.689 - 1.104 Eg for Eg > 1.73 eV (2.52)
= 4.194 - 1.177 Eg for Eg ≤ 1.73 eV (2.53)
= -8.854 + 6.338 Eg for Eg > 1.73 eV (2.54)
Ep= -1.258 + 1.403 Eg for Eg ≤ 1.73 eV (2.55)
Ep = -18.629 + 11.455 Eg for Eg > 1.73 eV (2.56)
These coupling equations, from (2.48) through (2.55), have been used to model the amorphous i-layer and p-layer optical properties at room temperature in several situations in this thesis research. These expressions have become very useful in modeling spectroscopic ellipsometry data to fit multilayer structures with a reduced number of fitting parameters—and subsequently reduced correlations between parameters.
Moreover, these coupling relations, together with the temperature coefficients of the
Cody-Lorentz parameters as given by Podraza et al., (Podraza et al., 2006), can be used to introduce both optical band gap and temperature as free parameters. The constant, dispersionless contribution, ε∞, to 1 can deviate from unity due to absorption features 43
well outside the spectral range under study. This constant is a parameter of interest that can contribute to in addition to the parameters of each of the mathematical expressions.
iv) Effective Medium Theories
Thin films consisting of homogeneous materials have well-defined optical and electronic responses. In reality, films may be inhomogeneous with non-uniform compositions of different materials like void and grain boundary regions, phase segregated materials such as an amorphous matrix with nanocrystalline inclusions, and even interfacial layers (top + bottom layers) (Aspnes, 1982; Collins et al., 1998;
Fujiwara et al., 2000). Effective medium theories (EMT) are applied in modeling the optical responses of these more complicated materials by using and volume fractions of the components as input. Surface and interface roughness is modeled similarly by replacing the rough region with a layer of well-defined thickness and with optical properties determined as an effective medium mixture of the underlying and overlying materials.
In general, EMT with isotropic screening, appropriate for a composite materials consisting of spherical inclusions embedded in a host materials, can be written as:
eff h ih fi . (2.57) eff 2 hi i 2 h
Here, εeff , εh , εi and fi are the effective medium complex dielectric function spectra, the host complex dielectric function spectra, and the complex dielectric function spectra and volume fraction of ith component, respectively. The Lorentz-Lorenz (LL) effective
44
medium expression assumes that materials are small and mixed on an atomic scale in void and it is more useful for gas mixtures. Therefore, Equation (2.57) reduces to LL
approximation with εh = 1. If one of the component material is comparable to the host medium, Eqn. (2.57) reduces to Maxwell-Garnett approximation (MG) with
εεh k where index k designates the dominant phase ( fi < fk for all i). If the fraction of different materials are similar and the material types are comparable, of the host
medium will be equal to the effective medium . In such case, εεeff h , and the approximation is called Bruggeman approximation (EMA). According to Fujiwara et al.,
2000, the Bruggeman approximation offers the best fit in the analysis of surface roughness layers that evolve during a-Si:H film growth. This is the most self-consistent choice and it is adopted for the full scope of this work. With this approximation, Equation
(2.57) reduced to:
i eff fi 0. (2.58) i i 2 eff
2.2.2. Experimental Methods in Solar Cell Characterization
A PV device or solar cell is a semiconductor junction that converts the energy of sunlight into electrical energy. Most conventional solar cells are p-n junctions in which on approximately visible range photon absorbed near the junction region produces an electron- hole pair. The minority carriers diffuse to the junction region due to their concentration gradient and then across the junction by drift due to a strong electric field created by ions at
45
the junction. A current is then generated as majority carriers in the p / n layers move to the adjacent contact (Deng et al., 2003; Honsberg and Bowden, 2010).
The thin film Si:H solar cell design adopted here is different than the conventional p- n junction solar cell in that a thick intrinsic layer is sandwiched between a thin p-type layer on the sun side and an n-type layer on the opposite side. The resulting thin Si:H solar cell deposited on an opaque substrate is often called the n-i-p or substrate configuration according to the order in which it is deposited as shown in Figure 2-11. This design should be distinguished from the p-i-n or superstrate configuration fabricated on a transparent substrate.
In any solar cell, however, the performance is measured from the current-voltage (J-V) characteristic of the device in the presence of photon flux that simulates the solar radiation spectrum and its total irradiance.
Figure 2-11. Schematic of n-i-p configuration a-Si:H solar cell with the electronic components connected and incident light used in measurements of J-V curve. The light that the simulator shines on the sample is spectrally similar to the light from the sun passing through an air mass of 1.5. An air mass of 1.0 corresponds to the path that sunlight would take to travel to sea level if the sun were positioned at the zenith. 46
The device is connected to a variable voltage source, a voltmeter, and a current meter so that the applied voltage and the induced current can be measured. The solar simulator system
2 exposes the solar cell to a standard 1000 W/m spectrum, and the electronic response to light is measured as the induced current at a given voltage step. The dark J-V characteristic, providing a measure of the diode behavior of the solar cell without illumination is obtained by the same procedure but without illumination.
Figure 2-12. Maps of n-i-p Si:H solar cell performance parameters of open circuit
voltage, Voc; short circuit current, Jsc; fill factor, FF; and power conversion efficiency,.
47
The measurement starts with zero voltage applied across the cell under illumination as indicated by the source meter, and the induced current is recorded. The measured current density at zero applied voltage is the short circuit current density (Jsc) of the device. Then, the induced current from the device is measured at increasing steps of the applied voltage in forward bias until the current vanishes. The voltage at which the total current in the circuit, consisting of photocurrent and forward bias diode current, vanishes is the open circuit voltage
(Voc). In addition to Jsc and Voc, the fill factor (FF) and efficiency () are critical parameters used to evaluate the solar cell performance. FF is determined by the ratio of the current and voltage at the point of maximum output power from the device to the product of Jsc and Voc.
The maximum output power point on the J-V characteristic is the point at which the product of the current and the voltage (or equivalently the output power) is a maximum.
The solar cell efficiency is obtained from the following relation:
J V FF =,sc oc (2.59) Pin
2 where Pin is the power flux in the incident beam typically 1000 W / m under air mass 1.5 illumination. Figure 2-12 shows maps of PV device performance parameters collected from a n-i-p a-Si:H single junction solar cell. The 6” x 6” sample was divided into an array of 16 x 16 small area (0.0707 cm2) devices.
48
Chapter 3
Growth Evolution of Si:H Prepared with Disilane Additives as Studied by Real Time Spectroscopic Ellipsometry
3.1. Introduction and Motivation
The optoelectronic and structural properties of Si:H and the associated device performance can vary strongly according to fabrication methods and deposition conditions. Optimized Si:H layers can exhibit relatively high efficiencies, however there is always the economical need to make them faster and cheaper. Use of different reactive gas mixtures can simultaneously manipulate film structure and growth rate within the film. The motivation is to explore the variations in the growth and properties of Si:H films that arise due to additive gas incorporation into the plasma for materials appropriate for integration into thin film Si:H based PV devices.
This chapter describes experimental results and analysis of in situ RTSE applied for PECVD Si:H growth evolution with disilane (Si2H6 ) added to the standard silane
(SiH4). RTSE is a non-invasive, non-destructive, non-contacting optical measurement technique as described in Chapter 2 and, when applied during film deposition, allows for the extraction of the time dependence of the structural (bulk layer thickness db, surface roughness ds) and optical properties of a film. The results of these RTSE studies explain
49
the potential impact of these additives on the evolution of crystallinity, material deposition rate, and electronic quality of the material. Furthermore, these features are incorporated into deposition phase diagrams or growth evolution diagrams that guide optimization principles for Si:H based PV and will be helpful in explaining correlations between device structure and performance in coming chapters. Growth evolution is tracked as function of hydrogen to silicon carrying gas ratio, R = [H2] / {[SiH4] +
[Si2H6]} and silane-to-disilane gas ratio, S = [Si2H6] / {[SiH4] + [Si2H6]}.
3.2. Overview of Deposition Processes and Microstructural Evolution
3.2.1. Deposition Processes of Si:H Films
The deposition processes of Si:H by PECVD consists of three interlinked stages: i) gas phase/plasma processes, ii) surface processes and iii) subsurface processes. The primary processes occurs as gas-phase which is controlled by different deposition parameters such as gas flow, partial pressure, the plasma excitation power, electrode geometry and reactor geometry. It involves the electron impact induced dissociation, excitation, or ionization of SiH4 giving out neutral radicals (SiHn, n = 0, 1, 2, 3). The secondary processes deals with the reaction of reactive species generated in gas-phase with the supplied gases like SiH4, Si2H6 and H2 molecules. High reactivity radicals such as SiH2 are consumed in these reactions leading to an increase in the fraction of low reactivity radicals SiH3 (Matsuda 1998). These low reactivity SiH3 radicals are weakly absorbed onto the growing surface, with this surface diffusion loading to the formation of
50
high electronic quality a-Si:H (Capezzuto et al., 1995; Gallagher 1988, Matsuda 1998;
Takai et al., 2000). Compared to SiH4, Si2H6 as a source gas has higher dissociation rate in plasma and results in different concentration of SiHn species. The interest of using disilane is also related to the introduction of large amount of more reactive species especially SiH2 into the gaseous phase. These species are responsible for a higher activation energy, higher deposition rates, and different optimal level of R for a given temperature (Doyle et al., 1992; Temple-Boyer et al., 2010). a-S:H growth can be thought of as a two-step process, whereby a dangling bonds on the surface is created from
H abstraction by SiH3 or H so that a second diffusing SiH3 radical bonds to the surface
(Doughty et al., 1990).
This surface layer formed by SiH3, shows greater H content compared to the bulk.
Removal of excess H in the bulk and formation of predominating silicon network may be explained by chemical annealing or thermodynamic equilibrium concentration of H
(Shirai et al., 1991; Robertson 2000). The dilution of silicon carrying gases with H2 results in the formation of a film with H-passivation of dangling bonds as well as enhanced diffusion of surface precursors to form a more stable network, although at lower deposition rates usually. In fact, higher H2 dilution can compensate for lower deposition temperatures, which typically would produce inferior electronic quality material. The precise mechanisms controlling H-incorporation in the bulk and diffusion are not fully understood, but the correlations between higher R, a more stable a-Si:H network, and improved PV device stability are well-known.
51
3.2.2. Microstructural Evolution and Phase Diagrams
The microstructure of Si:H based thin films of interest for PV applications may vary with deposition parameters or thickness (Collins et al., 2003). Si:H prepared by
PECVD may exhibit a-Si:H, nc-Si:H, and mixed-phase (amorphous+nanocrystalline)
(a+nc-Si:H) growth regimes—often within the same film as a function of thickness.
Several structural transitions may occur during the growth evolution of Si:H.
(i) Nucleation
Primarily, clusters composed of Si and H atoms nucleate on the substrate used.
The shape and density of these clusters are controlled by the substrate nature and deposition conditions (Collins and Yang, 1989; An et al., 1990; Li et al., 1992). Once the first bulk-like monolayer forms by making contact with the surfaces of neighboring clusters, the nucleation-generated roughness coalesces to form a smoother film. In the coalescence process, the residual effect of the clustering is a surface roughness layer that evolves in thickness as the bulk layer accumulates. Therefore, the height of the clusters in the nucleation stage and the resultant surface roughness layer thickness in the bulk film are described by a single parameter, ds, that can be obtained in RTSE analysis (An et al.,
1990; Li et al., 1992). The value of ds increases with time in the nucleation regime as the clusters increase in size, until it reaches a maximum at the time when the first bulk-like monolayer forms as shown in Figure 3-1. In this case, the nucleation density is controlled by the surface diffusion length of the precursor on the substrate, which is influenced by the intrinsic nature of the substrate surface and its associated free energy. For heterogeneous nucleation, clusters form at defects on the substrate surface and thus the nucleation density is controlled by the defect density, i.e. the extrinsic nature of the substrate surface. In 52
heterogeneous nucleation the defect density on the substrate can be modified in the initial stages of growth, such as by ion bombardment to increase H2-dilution can passivate these defects. It may be difficult to distinguish nucleation via homogeneous and heterogeneous processes on the basis of variations in the maximum ds value versus a given deposition parameter.
(ii) Coalescence of initial amorphous nuclei
The coalescence of nucleation induced microstructure is observed as a surface smoothening effect during a-Si:H growth within the first ~100-200 Å of bulk layer thickness. Such an effect can be observed in Figures 3-1 and 3-3. These figures demonstrate this effect for films that initially nucleate in the amorphous phase, regardless of the deposition conditions and eventual phase evolution. This type of smoothening is believed to be the result of diffusion of precursors on the film surface and have been applied to both physical and chemical vapor deposition processes by using continuum models (Mazor et al., 1988; Palmer and Gordon,
1988; Bales et al., 1989; Thiart et al., 2000). In this analysis, the surface roughness is considered to be a small perturbation from a perfectly flat surface (or from a straight line profile in a 1-D simulation), allowing one to linearize the differential equation that describes the surface profile evolution. Then the time evolution of the amplitude of a surface perturbation with spatial period in the in-plane direction is tracked. This approach has been tested in RTSE studies of a-Si:H growth on microscopically rough transparent conducting oxide surfaces having well defined in-plane periods (Collins and Yang, 1989). In a-Si:H, a competition between atomic scale roughening and capillary driven diffusion of precursors across the sample surface dictates the evolution of roughness (Podraza et al.,
2006; Kryukov et al., 2009). Previous studies have shown that a-Si:H films with the largest
53
decrease in ds exhibit the best electronic performance for applications as intrinsic layers (i- layers) in solar cells, as long as these films remain amorphous throughout growth (Li et al.,
1992; Lu et al., 1994).
Figure 3-1. Surface roughness layer thickness, ds, and unweighted error, σ, versus bulk
layer thickness for a Si:H film prepared with S = [Si2H6] / {[SiH4] +
[Si2H6]} = 0.63 at R = 23 that grows in the amorphous phase throughout deposition. 54
(iii) Amorphous stable surface regime
After nucleation and coalescence of a-Si:H clusters on the surface under defined set of deposition conditions, the surface roughness of the material will decrease with increasing thickness until reaching a minimum or a stable surface regime. The duration of this stable surface regime is controlled by the diffusion length of precursors across the surface, interactions with atomic hydrogen in the plasma, and the nature of the precursors. If the diffusion length of precursors across the surface is sufficiently large and the film remains amorphous, then the film will remain smooth with a stable surface roughness value. The a-Si:H surface remains smooth and stable such that < 1 Å change occurs in the roughness layer thickness from the end of coalescence through the growth of a specified bulk layer thickness. a-Si:H has been shown to exhibit this regime of growth from 200 to 1000 Å for a R = 15 film in Figure 3-3.
(iv) Amorphous roughening transition [aa]
For deposition conditions different from the optimum associated with the stable surface regime, an onset in roughening is observed at a well-defined db value. This onset identifies the amorphous roughening transition as can be seen in Figure 3-1. The growing film is amorphous on both sides of this transition to roughening as determined from the spectra in of the film extracted at different thicknesses. When the surface diffusion length of precursors is smaller than the in-plane feature size of the surface roughness, the value of surface roughness will increase from the minimum stable surface regime while the material remains in the amorphous phase. Correlations between a-Si:H device performance and associated i-layer material structural evolution reveal that reductions in
55
performance and stability of solar cells occur when the incorporated i-layers are prepared under conditions yielding an aa roughening transition at reduced db (Ferlauto, 2001).
(v) Amorphous-to-mixed-phase nanocrystalline transition [a(a+nc)] and mixed-phase (a+nc) growth regime
In the intermediate H2-dilution regime, a different type of roughening transition is observed in which crystallites nucleate from the accumulating amorphous phase and
grow preferentially. Because the crystallite nucleation density is usually low (109-1011 cm-2) (Ferlauto et al., 2000a; Fujiwara et al., 2001), the crystalline protrusions generate a roughness layer that increases rapidly in thickness with db due to preferential growth of the crystalline phase. The onset of roughening identifies a transition to the mixed-phase
(a+nc) film growth regime. For the example in Figure 3-2, this transition occurs at db =
250 Å. During the mixed phase (a+nc) growth regime, the crystallites increase in size forming inverted conical structures with increasing bulk layer thickness until the cone bases make contact and coalescence occurs. In this case, the changes in the phase composition after the roughening transition are accompanied by changes in the film optical properties, in contrast to the case of the amorphous roughening transition. Details on this type of film evolution will be provided in Chapter 5.
(vi) (Mixed-phase)-to-single-phase nanocrystalline transition [(a+nc)nc]
For thin films that have already undergone an a(a+nc) transition, a second transition is possible that occurs at even greater bulk layer thicknesses. An example of such a transition is shown in Figure 3-2 for the S = 0.12 at R = 33 film at a bulk layer thickness of db= 1030 Å. Crystallites nucleating from the amorphous phase grow preferentially over the surrounding material, until the point at which the crystallites cover the surface.
56
The same effects play a role in the growth of hydrogenated nanocrystalline silicon (nc-
Si:H) as in case of a-Si:H, but competitive growth of neighboring crystalline grains also impacts the roughness evolution. This process is expressed in the data as a transition from surface roughening to smoothening.
Figure 3-2. Surface roughness layer thickness, ds, and unbiased estimator, σ, versus bulk layer thickness for a Si:H film prepared with S = 0.12 at R = 33 that nucleates crystallites.
57
Once the crystallites have coalesced to cover the growing film surface completely, single- phase nanocrystalline growth proceeds and surface smoothening occurs, ultimately with a resumption of surface roughening if the film is grown sufficiently thick.
These types of structural transitions discussed above and growth evolution have been observed for Si:H prepared under a variety of deposition conditions (Collins et al.,
2003) and its alloys with germanium Si1-xGex:H (Podraza et al., 2006) and carbon Si1- xCx:H (Lu et al., 1994). The comparison of the behavior of these transitions as a function of single deposition parameters have been used to produce deposition phase diagrams or growth evolution diagrams which have guided the development of optimization principles in Si:H based PV (Tsu et al., 1997; Collins et al., 2003; Stoke et al., 2008). For example, the structural evolution can be controlled by the dilution of reactive silicon carrying gases with H2 during the deposition process. Films prepared at low H2-dilution remain amorphous throughout their total thickness, while those prepared at higher dilutions nucleate crystallites. The [aa] transition thickness has been observed to increase with H2-dilution for films that remain amorphous, and the optimum a-Si:H based
PV devices incorporate layers prepared under conditions where the highest H2-dilution is used without nucleating crystallites within the thickness desired for the amorphous layer
(Tsu et al., 1997; Collins et al., 2003; Stoke et al., 2008). In the case of a-Si:H, the additional hydrogen dilution improves ordering in the a-Si:H network, while for nc-Si:H low hydrogen dilution ensures that hydrogen etching does not occur and the grain boundaries remain well-passivated (Vetterl et al., 2000; Cao et al., 2008; Stoke et al.,
2008). In this chapter we will explore how the growth evolution and deposition rate of
58
Si:H thin films may be impacted by varying the gas chemistry—namely through using mixtures of silicon carrying gases with hydrogen.
3.3. Experimental Details
The Si:H films for phase diagram development were deposited on native-oxide coated c-Si wafers in a load-locked rf (13.56 MHz) PECVD chamber of the cluster tool as described in Chapter 2. Native oxide covered c-Si wafers were chosen in order to maintain the highest sensitivity to the evolution of surface roughness and bulk layer thicknesses with time. Fixed deposition parameters included a low substrate temperature
o of Ts = 200 C, a low total pressure of p = 0.8 Torr, and a low plasma power density of P
= 0.04 W/cm2. The nominal substrate temperature of 280oC, which corresponds to a true temperature of the substrate surface of 200oC was used for the deposition. This substrate temperature was calibrated by detecting shifts in the critical point features of c-Si and applying an established temperature dependence (Lautenschlager et al., 2004). Variable parameters include the hydrogen-to-reactive-gas ratio, R = [H2] / [SiH4], to control the microstructure and the silane-to-disilane gas ratio, S = [Si2H6] / {[SiH4] + [Si2H6]} to impact the growth rate and film structure. The 6 inch diameter native-oxide coated c-Si wafer was pre-heated in the intrinsic-layer chamber at a true temperature of 200oC for 3 hours prior to Si:H deposition. Series of films were prepared at fixed S = 0, 0.12, and 1 as functions of R.
RTSE was executed during each Si:H deposition using a rotating-compensator multichannel instrument (Lee et al., 1998) that can provide ellipsometric spectra from
59
0.75 to 5.88 eV. The RTSE measurement system was used as described in Chapter 2. To improve accuracy, ellipsometric spectra were collected within a time of ~ 1.5 s and averaged. The angle of incidence for all depositions was fixed at a value within the range of 70.0o ± 0.5o. Before acquiring RTSE data during Si:H film deposition, single scan data were collected for the well-known c-Si substrate, both at room temperature and 200oC, to determine the oxide thickness and then fixing that thickness and applying numerical inversion to extract spectra in for room and high temperature c-Si. Reference optical properties for c-Si and native oxide were initial applied (Herzinger et al., 1998) . In order to study the phase diagram for the film deposited, the optical properties at a bulk layer thickness of ~ 200 Å thicknesses were extracted by global -minimization. In this analysis, a known set of a-Si:H optical properties were taken as the first guess values to execute the analysis further. Here, the optical properties of the underlying substrate were obtained from the single scan data and analysis through an iterative least-squares regression algorithm. The thickness of the native oxide was deduced from the room temperature data before heating the Si wafer and depositing the Si:H film. After the thickness of native oxide was determined for the deposition, numerical inversion of the data collected at 200°C was then performed in order to extract the spectra in of c-Si at this temperature. In this analysis, the native oxide thickness from the room temperature
(RT) analysis was used, and this thickness was assumed to remain constant upon heating the substrate. The global -minimization routine (Section 2.2.1.3 of Chapter 2) was employed for detection of the phase transition and growth evolution of the Si:H layer on top of c-Si (Ferlauto 2001; Fujiwara et al., 2001; Stoke et al., 2008).
60
3.4. S = 0, Variable R: Si:H Growth Evolution Baseline
The baseline for structural transitions and growth in Si:H was prepared under a variety of deposition conditions for S = 0. Figure 3-3a shows the change in ds, as a function of bulk layer thickness, db, for films prepared at S = 0 and low values of R.
a)
b)
Figure 3-3. Growth evolution of Si:H in the form of the surface roughness layer
thickness, ds, as a function of bulk layer thickness, db, for films prepared at
fixed S = 0, and variable H2-dilution. a) Films that remain amorphous throughout the growth. b) Films that nucleates crystallites. 61
This set of depositions will only use SiH4 as the silicon carrying gas and will serve as a basis of comparison for films prepared under similar conditions, however with some or all of the SiH4 replaced with Si2H6. A growth evolution diagram developed for films prepared at S = 0 and variable R is shown in Figure 3-4. The aa roughening transition for R = 15 is not observed probably because it occurs at thicknesses greater than the final thickness deposited here. Deterioration of electronic properties in a-Si:H is correlated with a decrease in the magnitude and rate of smoothening. The RTSE measurements of films for the high R in Figure 3-3b, showing amorphous-to-nanocrystalline evolution, are characterized by two regimes. First, surface roughness associated with the nucleation of crystallites from the amorphous phase and their consequent favored growth.
Figure 3-4. Growth evolution diagrams for Si:H prepared at S = 0 and variable R depicting the aa, a(a+nc), and (a+nc)nc structural transitions. Arrows pointing upward indicate the transition occurs beyond the maximum thickness measured. 62
These films show amorphous to mixed-phase [a(a+nc)] followed by a (mixed-phase amorphous plus nanocrystalline) to single-phase nanocrystalline [(a+nc)nc] transitions as the film thickness evolves and that optimum nc-Si:H may be produced at slightly higher R. For these films, the initial appearance of crystallites shifts to lower db with increasing R. Crystallite coalescence follows a similar trend.
3.5. S = 0.12 and 1 versus R: Effect of Disilane on Si:H Growth Evolution
The impact of adding additive gas Si2H6 to the source gas may alter crystallite evolution as observed in the a(a+nc) and (a+nc)nc transition thicknesses shown in
Figure 3-5a and 3-5b. For the S = 0.12 and 1 series, the highest H2-dilutions prior to crystallite nucleation for a ~1000 Å thick layer occur at R = 22 and 54, respectively. The corresponding value for S = 0 occurs at R = 15. These microstructural variations indicate that the presence of Si2H6 requires more H2-dilution to improve ordering in the amorphous film and nucleate crystallites. This behavior is expected as each Si2H6 molecule possesses double the silicon as SiH4, so the average silane radical to hydrogen ratio in the plasma is increased when disilane gas is added. Comparison between the S =
0.12 and S = 1 films indicate that the maximum aa transition thickness is lower for films made with only Si2H6. The transitions for a-Si:H films prepared with R less than the optimum for the respective series shows that the aa transition thickness appears to reach higher values at R lower than the optimum for S = 0.12 and 1 compared to S = 0.
This behavior indicates that there is potential for widening the process window over which improved, but not fully-optimized, a-Si:H may be deposited. With sufficiently high 63
hydrogen dilution, nanocrystallites were observed to nucleate at all values of S in this study.
a)
b)
Figure 3-5. Growth evolution diagrams for Si:H prepared at S = 0.12 and 1 as functions of R depicting the aa, a(a+nc), and (a+nc)nc structural transitions. Arrows pointing upward indicate the transition occurs beyond the maximum thickness measured.
64
Comparing the baseline, i.e. S = 0 and S = 0.12 films prepared at the lowest R where crystallites appear, the addition of Si2H6 appears to cause initial crystalline nucleation at lower bulk layer thickness. The transition in a(a+nc) thickness for the films made with pure disilane gas as the reactive gas (S = 1 series) increases relative to other two series S
= 0 and S = 0.12. At significantly higher hydrogen dilution, all film samples exhibit prompt crystallite nucleation, growth, and coalescence within the first 400 Å of film growth. The decrease in the thickness at which crystallites coalesce with increasing R for each S series indicates that initial nucleation density of crystallites increases, assuming crystallites evolve using a cone growth model as has been observed for other Si:H based materials (Collins et al., 2003; Podraza et al., 2008). At the minimal R for which crystallites nucleate, the a(a+nc) transition occurs much more abruptly for the S = 0.12 films compared to S = 0 and 1. This behavior indicates that proper tailoring of S can be used to control crystallinity.
Unfortunately, as is common for Si:H materials, the deposition rate decreases with increasing R. Increased atomic hydrogen present in the plasma expected from the increase in H2-dilution may etch weakly bonded material at the film surface. This type etching behavior leads to the removal of potentially defect rich material but slows the deposition rate. Figure 3-6 shows the deposition rate for the S = 0, 0.12, and 1 films prepared as functions of R. Optimized a-Si:H at R = 15 for S = 0, R = 22 for S = 0.12, and
R = 54 for S = 1 have deposition rates at 0.91, 0.89, and 0.84 Å/s, respectively. These similar rates indicate that similar etching by atomic hydrogen may be occurring regardless of the precursor gas. The small decrease in growth rate with S may be consistent with more weakly bonded material present in the films prepared with any 65
addition of Si2H6. The relative quality of optimized a-Si:H prepared with S < 0.12 remains unclear as the aa transition thickness has not been observed for S = 0. The lower aa transition thickness for S = 1 compared to S = 0.12 indicates that optimized material made with pure Si2H6 under these conditions may be of lesser quality. The deposition rate for H2-diluted, but not optimized, material is overall higher with increasing S. This behavior, combined with the observed higher aa transition thicknesses at R below the optimum for films prepared with S > 0, indicates that some improvement in both a-Si:H film quality and growth rate can be achieved for non- optimum conditions. nc-Si:H deposition rates show improvement when comparing films among the different series.
Figure 3-6. Film growth rates for Si:H prepared at S = 0, 0.12, and 1 as functions of hydrogen dilution R. The solid black line indicates the growth rate of a-Si:H prepared at the maximum R prior to crystallite nucleation.
66
For example, comparing the R = 107, S = 1 film with structurally similar material prepared at R = 40, S = 0 indicates that nc-Si:H made with Si2H6 has a 0.13 Å/s (25%) higher growth rate. This structural similarity is induced by the thickness at which crystallites initially nucleate then coalesce. Differences in crystallite nucleation density at the a(a+nc) transition thickness make direct comparisons difficult. This increase in rate with S observed for (a+nc)-Si:H materials is coupled with comparable or prompter nucleation of crystallites. Similar growth rate material at S = 0 and 1 indicates that material prepared at S = 1 exhibits prompter crystallite nucleation. Overall, the comparison indicates that varying S may serve as another parameter in manipulating and optimizing nc-Si:H growth, in this case promoting crystallinity while improving the film growth rate.
3.6. Summary
A series of Si:H thin films has been analyzed in which each film exhibits growth in the amorphous, mixed-phase (a+nc), and single-phase nanocrystalline regimes. RTSE has been used to study the growth evolution of Si:H films prepared under variable H2- dilution and silane-to-disilane ratio deposition conditions. a-Si:H quality at low R degrades with increasing S, however R can be increased to maintain film quality. The film growth rate of optimized a-Si:H material appears relatively stable, or slightly decreasing (~6%) with S. However, a wider processing window for improved, but not fully-optimized, material is available at lower R with higher rates. Increasing S suppresses crystallinity at low R, while promoting crystallite growth at lower thicknesses
67
in high R, S = 0.12 films. (a+nc)-Si:H films can be deposited at higher rates with addition of Si2H6, even though the H2-dilution must be increased to cause crystallite nucleation.
This behavior is expected to be useful in the fabricating nc-Si:H PV devices at higher rates. Further analysis methods such as cross sectional transmission electron microscopy and atomic force microscopy would be useful, however excellent correlations between these measurements and RTSE results have previously been demonstrated (Fujiwara et al., 2001; Podraza et al., 2008).
68
Chapter 4
Si:H Layer Optimization for Substrate Configuration of Solar Cell
4.1. Introduction
Thin film PV are reliant on the capability for characterizing the opto-electronic and structural properties of each layer over large areas and correlating these properties with electrical performance of the device. Growing and characterizing each layer in complete devices facilitate comparison of the substrate dependent growth process with individually grown layers. Thin film Si:H based solar cell performance can be limited by many factors like the optical loss that occurs within the BR structure (Sainju et al., 2006), relative disorder in films, and lack of optimization in doped and undoped Si:H layers to obtain the best electronic quality material. Single, tandem and multijunction cells both in substrate or superstate configuration need optimization in the band gap of each layer so that each junction provides maximum photon collection in different parts of spectrum.
Most commonly studied, the structural evolution of Si:H can be controlled by dilution of
SiH4 with H2 during deposition. The microstructural evolution of Si:H has been studied for layers deposited on different bulk and thin film substrates, with varying degrees of surface roughness, including native and thermal oxide coated c-Si, TCO coated glass,
69
polyethylene naphthalate (PEN) polymer, as well as underlying structurally distinct Si:H layers prepared under different deposition conditions.
A primary technique for studying this growth evolution is the use of near infrared
(NIR) to ultraviolet (UV) in situ RTSE applied during Si:H thin film growth (Koh et al.,
1995; Koh et al., 1999; Rovira et al. 1999; Robertson, 2000; Rovira et al., 2000; Fujiwara et al., 2001; Koval et al., 2002; Collins et al., 2003; Ferreira et al., 2004; Stoke et al.,
2008; Dahal et al., 2010). Koh et al (Koh et al., 1995) reported the growth evolution of intrinsic Si:H on native oxide covered c-Si, a-Si:H films prepared without additional hydrogen dilution, and on newly deposited 200 Å p-type nc-Si:H. These results depicted that the nature of the underlying material influences nucleation of crystallites, suppressing nucleation with underlying a-Si:H and promoting nucleation with underlying nc-Si:H. The growth evolution of p-layers on specular ZnO coated glass and the ability to promote a high nucleation density of nc-Si:H were studied by Rovira et al. (Rovira et al.,
1999). Based on the study of growth evolution of p-type Si:H on ZnO coated glass and
ZnO over-coating tin oxide (SnO2), Koval et al. (Koval et al., 2002) reported that valid correlations between material properties and device performance can be better realized for any given material when the properties are obtained from deposition on similar substrates with similar thicknesses as those used in the respective device. Si:H films prepared by PECVD may exhibit several structural transitions during growth in the n-i-p solar cell configuration. The generation of so-called “deposition phase diagrams” or
“growth evolution diagrams” for vhf and rf PECVD of Si:H films reported that vhf
PECVD shows significant differences in structural evolution with processing conditions including plasma excitation frequency (Ferreira et al., 2004). Growth evolution diagrams 70
have been developed by Stoke et al. for intrinsic a-Si:H, amorphous silicon germanium alloys, and nc-Si:H for top, middle, and bottom cell i-layers used in triple junction devices (Stoke et al., 2008b). Dahal et al. reported growth evolution diagrams for intrinsic and p-type Si:H deposited on unoptimized silver (Ag/ZnO/n-layer coated PEN in n-i-p configuration PV devices (Dahal et al., 2010). Figure 4-1 shows a schematic structure of optically characterized layers in a n-i-p a-Si:H solar cells thicknesses and R for a-Si:H layers are obtained from growth evolution diagrams.
In this chapter, a detailed explanation is provided for the amorphous-to nanocrystalline phase transitions of Si:H layers as measured by RTSE in the substrate n-i- p solar cell configuration. Growth evolution diagrams as functions of R are developed for doped and updoped materials to identify the transitions from amorphous to nanocrystalline growth.
Indium Tin Oxide
p: a-Si:H (R = 150 – 175,100 Å) i: a-Si:H (R = 10, 3000 Å)
n: a-Si:H (R = 50, 200 Å) ZnO (3000 Å)
Ag (Semi-infinite)
Figure 4-1. Schematic structure of the optically characterized layers in a n-i-p a-Si:H solar cells thicknesses and R for a-Si:H layers are obtained from growth evolution diagrams. 71
The prime goal of developing these deposition phase diagrams is to ensure deposition of the most ordered amorphous, called “protocrystalline,” on a commonly used ZnO covered Ag BR for optimized single junction solar cell performance (Koval et al., 2002;
Wronski et al., 2004).
4.2. Experimental Details
The phase diagrams for doped and undoped Si:H were developed using multichamber rf-PECVD and RTSE as described earlier in Chapters 2. All Si:H films were deposited onto 6" x 6" borosilicate glass in the chromium (Cr)/Ag/ZnO/n/i device configuration except p-type Si:H which was deposited onto 6" x 6" borosilicate glass substrates initially pre-coated with ~3000 Å undoped a-Si:H. Thus, the RTSE studies of p-type Si:H closely mimic how the material is expected to grow in the device. PECVD was used for the deposition of Si:H layers, whereas magnetron sputtering deposition technique was used for Cr, Ag, and ZnO films.
Hydrogen dilution was the main process variable used in the development of phase diagrams. For the n- and p-layers, the dopant gas ratio D (D = [PH3] / [SiH4] and D
= [B2H6] / [SiH4] respectively) was also a critical parameter of interest. The dopant gas ratio plays a significant influence on the structural as well as the electronic properties of these layers. For the n- and i-layer, the nominal substrate temperature was 200oC whereas for the p-layer, the temperature was 100oC. Deposition conditions for the individual layers of the a-Si:H n-i-p solar cells configuration deposited on 6"x 6" borosilicate glass substrates were shown in Table 4.1. All PECVD chambers used were calibrated to the true substrate temperature on the basis of the a-Si:H band gap shift as characterized by
72
spectroscopic ellipsometry (Podraza et al., 2006). The calibration was done by Dr. Lila
Raj Dahal (Dahal, 2013) by depositing ~2000 Å a-Si:H on a borosilicate glass substrate.
The ellipsometric spectra taken by stabilizing the temperature at increasing set points starting from room temperature and then extracting the true temperature at the set point based on the known temperature shift of the a-Si:H band gap (Podraza et al., 2006a). The temperature calibration was performed for all three chambers independently on the basis of the temperature shift of the a-Si:H band gap instead of on the basis of the shifts of the
E1 and E2 critical points c-Si with temperature (Lautenschlager et al., 1987). The reason for this choice is that the former gives the temperature calibration relevant for the borosilicate glass, which is the substrate used for the actual device configuration. In situ
RTSE data from 0.73 eV to 5.88 eV was collected for each deposited layer and were analyzed by using aforementioned methods.
4.3. Ag/ZnO BR Substrate
The Ag layer acts as the primary reflector so that light not absorbed in the a-Si:H layers, incident from the p-layer side, maybe reflected then absorbed in the a-Si:H layers.
ZnO acts as a diffusion barrier to prevent Ag atoms from migrating into the a-Si:H layers.
Cr acts as an adhesion layer between Ag and the glass supporting substrate here. This BR structure helps improve the IR response of a-Si:H based solar cells. Due to the low minority carrier mobility in doped a-Si:H layers, carrier collection is difficult in thick absorber layers, ultimately limiting the intrinsic absorber layer thickness. As the above- gap absorption onset of the absorber i-layer of a-Si:H is not very sharp compared to that of a direct band gap crystalline semiconductor, the penetration depth is high.
73
Table 4.1. Deposition conditions for the individual layers of the a-Si:H n-i-p solar cells configuration deposited on 6"x 6" borosilicate glass substrates. The 5% dopant gas in H2 is by volume.
Layer Substrate Deposition RF Gas flow Intended Growth Temperature Pressure, Plasma (sccm) Thickness Rate (oC) (mTorr) Power (Å) (Å/s) (W/cm2) Cr RT 5 0.92 [Ar] = 10 2000 0.84 Ag RT 5 0.92 [Ar] = 10 5000 2.78 ZnO RT 5 0.92 [Ar] = 10 3000 1.35 R = 20, D = 0.0125 250 1.28 [SiH4] = 2, 5% [PH3] in [H2] =0.5 R = 40, D = 0.0125 250 0.68 [SiH 4] = 2, 5% [PH3] in [H2] =0.5 n-layer 200 1500 0.032 R = 50, D = 0.0125 250 0.62 [SiH4] = 2, 5% [PH3] in [H2] =0.5 R = 60, D = 0.0125 250 0.52 [SiH4] = 2, 5% [PH3] in [H2] =0.5 R = 80, D = 0.0125 250 0.43 [SiH4] = 2, 5% [PH3] in [H2] =0.5
R = 10, [H2] = 50, 3000 1.301 [SiH4] = 5
R = 15, [H2] = 75, 3000 0.96 [SiH4] = 5
R =20, [H ] = 100, 3000 0.58 i-layer 200 800 0.04 2 [SiH4] = 5
R = 50, [H2] = 250, 3000 0.32 [SiH4] = 5
74
Without a BR structure, long wavelength light not initially absorbed by the a-Si:H intrinsic layer is lost and does not contribute to Jsc. The BR can possibly become more effective if the metal layer is textured so that the photons are scattered into oblique angles, i.e. through diffuse reflection, or scattering. It has been shown that the Jsc can increase up to 30% with a Ag/ZnO back reflector in the a-Si1-xGex:H solar cell on a flexible stainless steel substrate (Yang et al., 2003).
Deposition conditions were set such that the Cr layer was deposited to a thickness of ~2000 Å, the Ag layer was deposited to a thickness of ~5000 Å, and the intrinsic ZnO layer was grown to a thickness of ~3000 Å, all in succession at room temperature in a sputtering chamber without any intended heating. Here, Cr was used as an adhesion interlayer to avoid delamination of Ag from the borosilicate glass surface. The Ag layer was used as a BR and ZnO as diffusion barrier and dielectric spacer between Ag and n- type Si:H. Spectra in and microstructural parameters (db, ds) were extracted using a least squares regression analysis and an unweighted error function to fit the experimental ellipsometric spectra to an optical model. Data collected for semi-infinite optically opaque Ag substrate were taken before ZnO layer deposition at room temperature and were analyzed in the energy range 0.734 to 5.88 eV. Figure 4-2 shows spectra in for Ag parameterized by a combination of a Drude oscillator (Tiwald et al., 1998) for the free electrons in the low photon energy, NIR region; and a pair of critical point parabolic band
(CPPB) resonances (Aspnes, 1980) for the bound electrons in high photon energy UV region; and a constant additive term to 1 denoted as follows:
75
μ Γ n in n ε(E) =εε +D (E) + A e (4.1) n 2E 2E iΓ n nn
The third term, which is the interband term, is described as a sum of CPPB oscillators where An, En, Γn, n, and n are the amplitude, resonance energy, broadening, exponent, and phase of each CPPB, respectively. All parameters describing semi-infinite bulk Ag and its surface roughness are listed in Table 4.2. The exponent can assume the values of
1/2, 0, and -1/2 depending on 1-D, 2-D, or 3-D nature of CP. In this work, only the 1-D CP oscillator has been used, and so its value is fixed at = 0.5.
Figure 4-2. Spectra in as a function photon energy extracted using Drude and critical point parabolic band (CPPB) oscillators over a spectral range from 0.734 to 5.88 eV for a semi-infinite Ag film deposited on Cr overcoated borosilicate glass.
76
Table 4.2. Parameters describing complex dielectric function spectra ( = 1 + i2) and structure for a semi-infinite Ag film on a borosilicate glass over coated by Cr before ZnO deposition. Experimental ellipsometric spectra were collected in situ after deposition at room temperature in the spectral range from 0.734 to 5.88 eV and fit using least squares regression analysis with an unweighted estimator error function, = 5 x 10-3. For bulk Ag, the parameterization of consisted of a Drude oscillator, two oscillators assuming critical point parabolic bands (CPPB), and a constant additive term to 1 denoted as . Spectra in for the 30 ± 2 Å surface roughness layer were parameterized with two
Lorentz oscillators and = 1.
Ag Surface Roughness
Oscillator A (Unitless) (eV) E0 (eV) Lorentz 4.2 ± 0.2 2.5 ± 0.1 5.17 ± 0.02 Lorentz 1.0 ± 0.3 0.06 ± 0.03 3.61 ± 0.01
Bulk Ag
A (eV) En (eV) Ө (degrees) CPPB 5.29 ± 0.09 0.70 ± 0.03 3.845 ± 0.008 180.306 ± 0.002 0.5 CPPB 10.39 ± 0.07 0.87 ± 0.01 4.025 ± 0.001 7.0 ± 0.4 0.5 Drude ( cm) (fs) 3.02 ± 0.03 x 10-6 16.7 ± 0.1
1.632 ± 0.008
The second term is given by the Drude oscillator having quantities and as the fitting parameters as follows:
2 ε (E) = , (4.2) D 2 0 E i E where is Planck’s constant divided by 2, 0 is the permittivity of free space, is the scattering time, and is the resistivity given by = m*/Nq2 = 1/qN. In the expression
77
for the resistivity, m* is the effective mass of the charge carrier, which is the electron mass in this case, q is electronic charge, and N are the carrier mobility and concentration, respectively. A resistivity of 3.02 ± 0.03 x 10-6 cm and a scattering time of 16.7 0.1 fs were determined from the Drude oscillator parameters of the Ag film.
The surface roughness is represented by two Lorentz oscillators (Collins and
Ferlauto, 2005) with = 1. Each Lorentz oscillator is represented by:
AΓE ε(E) = 0 (4.3) E22 E iΓE 0 where A, and E0 represent amplitude, broadening, and resonance energy respectively.
When ZnO film growth is initiated, the depositing material starts filling the void space in the surface roughness layer of the underlying film. At the same time, the peaks in the surface roughness of the Ag substrate film are coated with the depositing material, generating a roughness layer associated with the growing film. Initially, the volume fraction of depositing material in this new surface roughness layer is small. With the deposition progress, the void space at the interface is essentially replaced by the depositing material (with the possibility of some voids remaining) as the depositing material roughness layer increases in thickness and in material volume fraction. Analysis of ellipsometric spectra collected for the Ag/ZnO BR applies a Lorentz oscillator to represent the interface between Ag and ZnO because separation of Ag protrusions by less conductive ZnO generates screening and a plasmonic resonance feature. The Ag/ZnO interface layer formed with overlying ZnO deposition is typically thicker than that of the
Ag surface roughness. The structural model for the Ag/ZnO BR in the energy range 0.73
78
to 5 eV consisted of a semi-infinite Ag metal layer deposited onto glass, a 108 ± 10 Å
Ag/ZnO interfacial layer, a 3059 3 Å bulk ZnO layer, and a 80 1 Å surface roughness represented using Bruggeman effective medium approximation of 0.5 ZnO and 0.5 void volume fractions and an air ambient. Parametric expressions were used to describe for
Ag, ZnO, and the Ag/ZnO interface and are listed in Tables 4.2 and 4.3. Figure 4-3 shows spectra in obtained for the Ag/ZnO interface modeled by Lorentz oscillator and a Drude oscillator in the near IR to near UV range (0.734 to 5 eV) with = 1.
Figure 4-3. Spectra in as a function photon energy extracted over a range 0.734 to 5 eV for the 107.67 10 Å thick Ag/ZnO interface layer parameterized with a Lorentz and a Drude oscillator. The Ag/ZnO interface layer formed with overlying ZnO deposition is typically thicker than that of the Ag surface roughness and behaves optically different than the optical properties of Ag and ZnO.
79
Table 4.3. Parameters describing and structure for a ZnO film deposited on Ag and the Ag /ZnO interface formed. Experimental ellipsometric spectra were collected in situ after deposition at room temperature in the spectral range from 0.734 to 5.0 eV and fit using least squares regression analysis with an unweighted estimator error function, = 7 x 10- 3. Parameters describing for Ag were fixed from Table 4.2. For ZnO, the parameterization of consisted of two CPPB oscillators, a Sellmeier oscillator, and . For the Ag/ZnO interface, the parameterization of consisted of a Drude oscillator, a
Lorentz oscillator, and .
Layer Oscillators
CPPB (µ = 0.5) = 2.27 ± 0.01
ZnO A (eV) En (eV) Ө (degrees) db = 3060 ± 3 Å 2.63 ± 0.02 0.199 ± 0.002 3.363 ± 0.001 20.1 ± 0.5 ds = 80 ± 1 Å 1.41 ± 0.02 3.83 ± 0.08 4.36 ± 0.03 0 (fixed) 2 A (eV ) (eV) E0 (eV) Sellmeier 0.080 ± 0.002
Ag/ZnO Drude Lorentz, = 1 Interface (cm) (fs) A (Unitless) (eV) E0 (eV) 3.7 ± 0.5 x10-5 2.7 ± 2.8 ± 0.2 0.57 ± 0.05 2.83 ± 0.01 di = 108 ± 11 Å 0.3
The Ag/ZnO interface exhibits a clear localized or particle plasmon absorption feature which can be modeled by using Lorentz oscillator with a resonance energy at 2.83 0.01 eV, and behaves optically different than the optical properties of Ag and ZnO (Sainju et al., 2006). A resistivity of 3.7 0.5 x 10-5 cm and a scattering time of 2.7 0.3 were determined from the Drude oscillator parameters of the Ag/ZnO interface. These values indicate the interface is less conductive than the bulk material, as expected due to the 80
higher resistivity of undoped ZnO, and potentially has greater disorder as indicated by the lower scattering time.
Over this spectral range, for intrinsic ZnO layer as shown in Figure 4-4 was initially fitted using two CPPB oscillators, , and a zero-broadened Sellmeier oscillator (Collins and Ferlauto, 2005) represented by:
A ε(E) = (4.4) 22 EEn where A and En represent the amplitude and resonance energy, respectively. The analysis of ZnO will be discussed in Chapter 6.
Figure 4-4. Spectra in as a function photon energy extracted over a range 0.734 to 5 eV for the 3060 3 Å thick ZnO deposited on semi-infinite Ag substrate.
for intrinsic ZnO layer was fitted using two CPPB oscillators, , and a zero-broadened Sellmeier oscillator.
81
4.4. n-type Si:H on a Ag/ZnO BR Substrate
The n-layer of single junction solar cell was deposited on to a BR consisting of
Cr/Ag/ZnO coated borosilicate glass substrate. The back reflector, Cr/Ag/ZnO, coated borosilicate glass as discussed in section 4.3 was pre-heated in the n-layer chamber at a real temperature of 200oC for 2 hours in a hydrogen gas environment. After that, n-type
Si:H at variable R was grown by PECVD using the conditions given in Table 4.1. The result of these data analysis is the phase diagram shown in Figure 4-5.
Figure 4-5. Phase diagram for Si:H n-layers deposited on 6” x 6” borosilicate glass coated with a Cr/Ag/ZnO structure. The a(a+nc) transition thickness represented by the lower line with solid squares and (a+nc)nc transition thickness is represented by the upper dotted line with solid triangles.
82
In this analysis, it was assumed that the BR structure and optical properties do not change after growth of the PECVD n-layer. The hydrogen dilution ratio R of five n-type Si:H depositions was varied from 20 to 80. RTSE data and a global Σ-minimization analysis procedure has been used to track the behavior of structural transitions in n-type Si:H deposited as functions. The dopant gas ratios for n-type Si:H was fixed at D = 0.0125.
For R < 50 the n-layer remains amorphous at least to a thickness of 500 Å. At R = 50 nanocrystallites nucleate in the n-type Si:H at about 450 Å of bulk layer thickness. The amorphous material prior to the a(a+nc) transition of these depositions is protocrystalline (Collins et al., 2003). A ~200 Å thick n-layer is typical for n-i-p configuration devices, and the best R for optimized n-i-p a-Si:H solar cells with a protocrystalline n-layer is identified near R = 50. As R is further increased, nanocrystallites nucleate within the amorphous phase at decreasingly lower thicknesses as indicated by the a(a+nc) transition thicknesses.
The slope of the db, r(t) = d(db(t))/dt, was used to determine the deposition rate of each film as plotted in Figure 4-6. The deposition rate shows familiar trend that is it decreases with increasing R. Increased atomic hydrogen present in the plasma expected from the increase in hydrogen dilution may etch weakly bonded material leading to the removal of potentially defect rich material and slowing the deposition rate. With the deposition rate r(t) and reference for a-Si:H and nc-Si:H, a virtual interface RTSE analysis can be easily performed to track evolution of crystallinity as discussed in
Chapter 5. The band gap energy, specifically the Cody gap, plotted as a function of R for amorphous n-layer deposited on Cr/Ag/ZnO coated glass are shown in Figure 4-7.
83
Figure 4-6. Deposition rates of the n-layer as a function of hydrogen dilution deposited on Cr/Ag/ZnO/n-layer coated borosilicate 6”x 6” glass for phase diagrams.
Figure 4-7. The Cody gap with its confidence limits plotted as a function of R for amorphous n-layers deposited on Cr/Ag/ZnO coated glass substrates at a o temperature of 200 C. The gaps are from extracted from parameterized ε2 obtained by inversion at thicknesses in the range of 200-300 Å. 84
These band gaps were extracted from parameterized ε2 obtained by numerical inversion at thicknesses in the range at ~200-300 Å of the deposited film as obtained from global Σ- minimization analysis (Equations 2.38 and 2.39). Only five parameters A, , E0, Eg and
Ep of the Cody-Lorentz oscillator expression were varied for a given R to obtain parameterized . The increase in a-Si:H band gap with the increase of R is mostly due to the alloying effect of the amorphous Si matrix with hydrogen. Higher R suggests that more hydrogen atoms are incorporated that relax compressive strain in the network and increase the band gap.
4.5. i-layer Si:H on Ag/ZnO/n-Si:H Substrate
The i-layer serves as the absorber of the thin film a-Si:H devices. The use of a relatively thick ~3000 Å (~300 nm) intrinsic layer sandwiched between very thin doped layers creates a departure from the conventional solar cell designs, which are more often based on the p-n structure. The diffusion lengths of the minority photo generated carriers in doped a-Si:H, (holes in n-type a-Si:H and electrons in p-type a-Si:H) are exceedingly small.
Therefore, the p-n structure only collects the electrons and holes generated by photons absorbed in the very thin regions closest to the junction (Deng et al., 2003). In both n-i-p substrate and the p-i-n superstrate configurations, this intrinsic absorber layer generates a built-in electric field. This sandwiched intrinsic layer arrangement enables absorption of most photons of the incident sunlight with efficient carrier collection. The built-in field separates the photo-generated electrons and holes, which are transported to the contacts. Hence, i-layer is critical for operation, and its optical response and phase composition tremendously
85
impact the performance of the device. The growth evolution diagram for intrinsic Si:H as a function of variable hydrogen dilution 10 ≤ R ≤ 50 onto n-layer coated BRs has been developed and is shown in Figure 4-8. The hydrogen dilution and thickness of n-layer was fixed at R = 50 and ~200 Å, based on protocrystallinity observed in the n-layer growth evolution diagram. The experimental details of i-layer produced at 10 ≤ R ≤ 50 and other fixed conditions are given in Table 4.1. For the intrinsic layer, R = 15 is the lowest hydrogen dilution ratio at which the a(a+nc) transition is observed within 3000
Å of layer growth. The decrease in the (a+nc)nc thickness with R may indicate higher nucleation density of crystallites for higher hydrogen dilution.
Figure 4-8. Phase diagram for Si:H i-layers deposited on 6” x 6” borosilicate glass coated with a Cr/Ag/ZnO/n-layer structure. The a(a+nc) transition thickness represented by the lower line with solid squares and (a+nc)nc transition thickness is represented by the upper dotted line with solid triangles.
86
R = 10 is identified as best for optimized n-i-p a-Si:H solar cells incorporating a ~3000 Å thick protocrystalline absorber (Collins et al., 2003). On the other hand, at R = 15 and 20, the i-layer initially grows in the amorphous phase and nucleates nanocrystals, becoming mixed-phase. Finally, the growing R = 50 film transitions to single-phase nanocrystalline silicon at db ~300 Å. For these samples, virtual interface analysis described in Chapter 5 track the nanocrystallite fraction evolution with thickness, not just the “end points” reported here. The optimum i-layer from the growth evolution diagram is identified as R
= 10 for ~3000 Å thick layers for the n-i-p configuration. Figure 4-9 shows a decrease in growth rate with the increasing R. The intrinsic layer of the Si:H based n-i-p solar cell is the only source of photocurrent in the solar cell.
Figure 4-9. Deposition rates of the i-layer as a function of hydrogen dilution deposited on Cr/Ag/ZnO/n-layer coated borosilicate 6”x 6” glass. 87
Therefore, with the understanding of optical properties of optimized Si:H i-layer and its optimum thickness, along with the corresponding information for all other layer components of the stack, the external quantum efficiency (EQE) measurement can be predicted. The EQE can predict Jsc assuming that all photogenerated electrons and holes in the i-layer are collected (Aryal et al., 2011). The structural evolution and optical properties of the Si:H i-layer in the PECVD process are closely related to the electronic properties and defect density, with the latter two being optimized by varying the critical deposition parameters. Figure 4-10 shows a comparison of the band gap energy for R =
10 and 15 material.
Figure 4-10. The Cody gap, Eg, plotted as a function of R for amorphous i-layers o deposited on BR/n-layer coated glass substrates at 200 C. Eg for each
amorphous i-layer was extracted at db ~200 Å amorphous i-layer optical properties were smoothed by using the Cody-Lorentz oscillator expression. 88
Protocrystalline layers (Collins et al., 2003) with principally complete passivation of dangling bonds suppress recombination losses and degradation due to light induced defect generation (Staebler and Wronski, 1977). Minimization of recombination losses has a significant impact on cell performance and it has been reported that the best solar cell performance is obtained with protocrystalline i-layers (Collins et al., 2003; Wronski and Collins, 2004). The optical properties of the amorphous i-layers were obtained by fitting the RTSE 0.734 to 5 eV by using global Σ-minimization analysis routine. The inverted amorphous i-layer optical properties obtained at each R were parameterized using the Cody-Lorentz oscillator expression. The lower value of band gap of R = 15 could be a result of enhanced density expected from extra H-etching or could be a stronger absorption near the band edge of a-Si:H due to the presence of small crystallites.
At R > 15, the i-layer transitions to mixed-phase Si:H at a very early stage of the deposition.
4.6. p-layer Si:H on Glass/i-Si:H Substrate
The thickness of the p-layer should be thin enough to maximize transparency but thick enough to generate an electric field in the intrinsic layer. Typical p-layer thicknesses are ~100-150 Å, and a large optical band gap assists in minimizing parasitic absorption of incident light within this layer. The intrinsic layer, p-layer, and their interface are most directly responsible for open circuit voltage optimization, which can be guided using growth evolution diagrams.
The growth evolution diagram for PECVD p-type Si:H was developed by RTSE.
In order to mimic the solar cell configuration, ~3000 Å thick R = 10 intrinsic a-Si:H
89
layers were deposited on the borosilicate glass substrate prior to p-type Si:H deposition.
The deposition conditions of these intrinsic a-Si:H layers are identical to those identified as optimal for n-i-p solar cells as described in the Section 4.5. These layers therefore eliminate any contributions to the microstructural evolution from the underlying substrate. Figure 4-11 shows the evolution of ds versus db of R = 10 for i-layers deposited on two different substrates. One is borosilicate glass precoated with BR/ ~ 200 Å R = 50
(n-type a-Si:H ) and the other is glass only.
Figure 4-11. Structural evolution comparison for R = 10 intrinsic a-SiH layer deposited on Cr/Ag/ZnO/(R = 50 n-layer) coated 6”x 6” borosilicate glass and for R =10 intrinsic layer deposited on 6”x 6” borosilicate glass substrate only. The surface roughness at the end of the deposition tends to coincide besides the different structural evolution during the growth of the films.
90
These films were used in the growth evolution and phase diagram development of i-layer as described in Section 4.4 and p-type Si:H of the device. After nucleation on substrate surface for the R = 10, i-layer on Cr/Ag/ZnO/n-layer, ds decreases and shows no further roughening before i-layer deposition, thick microscopic roughness layer exists on the n- layer surface, as a 0.5/0.5 volume fraction mixture of n-layer/void.
The initial i-layer growth is modeled by considering a transition from n/i interface filling to bulk i-layer growth on the resulting n/i interface roughness layer (Koh et al.,
1995). At the onset of i-layer deposition, the void volume is filled by i-layer material.
During this time, the surface roughness on i-layer increases to the same value of the underlying n-layer, indicating conformal coverage. As a result, the initial roughness of the film is determined by the roughness on the underlying n-type film, ~40 Å. For R = 10 on glass, a lower amplitude roughness is observed throughout the growth of ~3000 Å film consistent with a-Si:H growth. Figure 4-12 depicts the spectra in 2 for the R = 10 films prepared under these conditions. Ultimately, both films have the same surface roughness thickness of ~10 Å at ~3000 Å thick.
Highest quality a-Si:H materials for the i-layers of PV devices are achieved when the surface smoothening during coalescence exhibits its maximum magnitude as discussed in Collins et al (Collins et al., 2003). The amorphous roughening transition
(aa) in intrinsic R = 10 Si;H occurs greater than the maximum possible thickness >3000
Å. This kind of result reflects deposition processes yielding the longest diffusion lengths of radical precursors on the growing film surface and the lowest density of defects on the surface. The amplitude of spectra in for a-Si:H is associated to material density, and to a first approximation, the lower the amplitude, the less dense is the material. Figure 4-12 91
supports that the i-layer grown on borosilicate glass is denser than that grown on the
Cr/Ag/ZnO/R = 50 n-layer coated glass. This suggests that the former deposition has fewer voids in the a-Si:H material compared to latter sample. To support this fact, the relative void volume fraction (fvoid) of bulk i-layer was calculated. The void volume fraction is obtained by fitting the spectra in for each of the different films assuming a mixture of a dense reference material and voids in the Bruggeman effective medium approximation (Aspnes, 1982; Fujiwara et al., 2000). In this case, spectra in of the reference material is obtained from RTSE measurements of ~200 Å of R = 10 a-Si:H on glass under optimized conditions.
Figure 4-12. Spectra in for ~ 200 Å bulk R = 10 i-layers grown at same deposition conditions on (BR/R = 50 n-layer) and borosilicate glass.
92
The fvoid range was 0.025 ≤ fvoid ≤ 0.033, indicating that the films are relatively dense on glass. For R = 10 on BR/(R = 50 n-layer), the high value of fvoid ~0.12 indicates much lower density films.
To construct a deposition phase diagram of the p-layer, R of six p-type Si:H depositions was varied from 50 to 200. The doping gas ratio D = [B2H6]/[SiH4] was fixed at D = 0.0125. Additional fixed deposition conditions include a rf power density, a total gas pressure, and a nominal substrate temperature as described in Table 4.4. RTSE data obtained as a function of time for each deposition were analyzed using the global - minimization routine. RTSE data collected for the p-layer deposited at the maximum R value (R = 200) was studied using a-Si:H optical properties extracted from R = 150. Even though spectra in for R = 200 are not accurate due to the phase evolution, the growth rate can be determined. Typical p-layer thicknesses are ~100-150 Å, and a large optical band gap assists in minimizing parasitic absorption of incident light within this layer.
Within the amorphous and protocrystalline phase, the band gap of the p-layer increases with increasing R. The layers and their interface are most directly responsible for Voc optimization, which can be guided using growth evolution diagrams. Figure 4-13 shows deposition phase diagrams deduced by RTSE for p-type doped Si:H on freshly-deposited
R = 10 a-Si:H intrinsic 3000 Å, a structure mimicking the n-i-p solar cell configuration.
It is observed that the p-layer deposited with R ≥ 150 transitions to mixed-phase.
Whereas, depositions with R ≥ 200 transitions to single phase nanocrystalline Si:H within the thickness range studied. This diagram demonstrates that for R = 110, the film grows initially as a-Si:H and crystallites begin to nucleate from the amorphous phase after a bulk layer thickness of 545 Å. 93
Table 4.4. Deposition conditions for the individual intrinsic and p-type Si:H layers of the a-Si:H mimicking n-i-p solar cells configuration deposited on 6"x 6" borosilicate glass substrates coated with ~3000 Å R = 10 intrinsic layer. The 5% dopant gas in H2 is by volume.
Layer Substrate Deposition RF Gas flow Intended Growth Temp. (oC) Pressure, Plasma (sccm) Thickness Rate (mTorr) Power (Å) (Å/s) (W/cm2) i-layer 200 800 0.04 R = 10, [H2] = 50, 3000 0.55 [SiH4] = 5
R = 50, D = 0.0125 120 0.24 [SiH4] = 2, 5% [B2H6] in [H2] = 0.5
R = 70, D = 0.0125 120 0.21 [SiH4] = 2, 5% [B2H6] in [H2] = 0.5 R = 100, D = 0.0125 120 0.14 [SiH4] = 2, 5% [B2H6] in [H2] = 0.5 R = 110, D = 0.0125 120 0.12 p-layer 100 1500 0.066 [SiH4] = 2, 5% [B2H6] in [H2] = 0.5
R = 130, D = 0.0125 120 0.09 [SiH4] = 2, 5% [B2H6] in [H2] = 0.5
R = 150, D = 0.0125 120 0.07 [SiH4] = 2, 5% [B2H6] in [H2] = 0.5 R = 200, D = 120 0.02 0.0125 [SiH4] = 2, 5% [B2H6] in [H2] = 0.5
94
The lower range from R = 50 and 100 indicates that this transition must occur for thicknesses greater than the deposited 650 Å. In contrast, at R = 200, the transition from predominantly a-Si:H to mixed-phase occurs at a bulk thickness of 40 Å, and the transition to single phase nc-Si:H occurs within 200 Å. The p-layer should be deposited at the maximum R that can be sustained without crossing the a(a+nc) transition boundary throughout the desired thickness of 100-150 Å here. This p-layer growth evolution diagram is comparable to previously published diagrams (Rovira et al., 2000;
Koval et al., 2002, Wronski et al., 2004, Pearce et al., 2007).
Figure 4-13. Phase diagram for p-type Si:H layers deposited on 6” x 6” borosilicate glass coated with R = 10 intrinsic layer. The a(a+nc) transition thickness represented by the lower line with solid squares and (a+nc)nc transition thickness is represented by the upper dotted line with solid triangles.
95
The growth rates for each film are given in Table 4.4. The p-layers growth rates plotted in
Figure 4-14. The growth rates were lower than that reported for n-layers and i-layers in
Section 4.3 and 4.4 due to fabrication at higher R. The resulting trend of band gap energies obtained at thickness of ~200 Å versus R is shown in Figure 4-15. The films grown with R = 50 to 120 shows a increase in band gap with increasing R. At R = 130 and 150, the values decrease, then exhibit a much smaller slope with R. This behavior may be due to incorporation of small nanocrystallite fractions within the top-most portion of the first 200 Å deposited bulk material. This is consistent with the presence of a small fraction of isolated crystallites in the upper part of the p-layers used in some optimized
PV devices (Pearce et al., 2007) as compared to original protocrystalline p-layer material
(Rovira 2000; Collins et al., 2003; Wronski et al., 2004).
Figure 4-14. Deposition rates of the p-layer as a function of R deposited on Cr/Ag/ZnO/n-layer coated borosilicate 6”x 6” glass for phase diagram studies. 96
This behavior is also similar to the reduction in band gap observed between R = 10 and
15 intrinsic a-Si:H in Figure 4-10. Overall, a larger band gap of the p-layer is desirable for better-quality solar cell performance, assuming the required electrical properties are preserved. The reason is that photons with energy less than the p-layer band gap are transmitted through it without absorption, and thus with an increase in this band gap, a larger fraction of photons can be absorbed in the i-layer. An increase observed in Voc with increasing R is credited to the widening of the p-layer mobility gap without the decrease in material quality (Collins et al., 2003; Wronski et al., 2004; Pearce 2007).
Figure 4-15. The band gap energy, Eg, as a function of R for p-type Si:H layers deposited on glass coated with ~3000 Å thick R = 10 intrinsic layer. Band gap information was extracted from the first ~200 Å of bulk layer deposition with a true substrate temperature of 100 °C.
97
4.7. Summary
RTSE measurement of Si:H films have been applied to develop growth evolution diagrams that provide guidance for a-Si:H film optimization for n-i-p PV devices. The deposition phase diagram illustrates that the structure of Si:H depends on PECVD process conditions including R and the accumulated material thickness. In particular, the values of R at which crystallites nucleate at given thicknesses is a delicate function of the substrate material. Therefore, the actual solar cell configuration was used to determine these transitions.
The most ordered protocrystalline n, i, and p layers have been reported as the best possible a-Si:H layers for the optimum stabilized thin film solar cells. A schematic of the device with optimized parameters adapted from phase diagrams were shown in Figure 4-
1 in Introduction. The schematic n-i-p structure has been labeled with optimized R values and intended thicknesses for the different layers in devices. These values have yielded guiding principles for cell fabrication. Deposition rates have been tracked as functions of
R to ensure that the appropriate layer thickness is generated.
98
Chapter 5
Growth and Analysis of Amorphous to Nanocrystalline Transition in Si:H films
5.1. Introduction and Motivation
nc-Si:H has been the topic of scientific and technological attention in recent years due to its higher electrical conductivity and greater doping efficiency compared to a-Si:H
(Saleh and Nickel, 2003; Yang et al., 2003; Shah et al., 2004). The band gap of nc-Si:H is somewhat tunable by controlling the deposition parameters enabling capture of sufficient solar radiation without deteriorating the series resistance of the PV device
(Meier et al., 1994; Kitao et al., 2001). Furthermore, nc-Si:H based solar cells have also been tested with noticeable enhancement in NIR absorption of the solar spectrum and high stability against the prolonged light illumination in contrast to it’s a-S:H (Prasad et al., 1991; Meier et al., 1996). For both substrate (n-i-p) and superstrate (p-i-n) PV devices with a nanocrystalline intrinsic layer, high initial conversion efficiency were reported. Also, nc-Si:H is a way to replace a-SiGe:H multijunction thin film solar cells
(Meier et al., 1996).
Electrical, optical, and the structural properties of nc-Si:H films are greatly influenced by deposition parameters (Amor et al., 2014). Through optimization of the
99
deposition parameters, such as the amount of hydrogen, substrate temperature, and the rate of change in hydrogen dilution ratio, nc-Si:H solar cell performance can be considerably improved. Two main techniques employed to obtain nc-Si:H are re- crystallization of a-Si:H films and direct deposition methods for nanocrystalline growth.
The growth of nc-Si:H through re-crystallization of a-Si:H films has many difficulties with respect to accurate control of crystalline volume fraction and grain size and generally requires post deposition annealing at high temperatures for crystallization
(Cheng et al., 2000; Wan et al., 2011), laser-melt re-crystallization (LMR) (Dyer et al.,
1993), and aluminum-induced crystallization (Aberle et al., 2001). Such high annealing temperature inevitably limits application to optoelectronic devices. On the other hand, a variety of direct chemical vapor deposition techniques have been used to yield materials with good optoelectronic properties (Scott et al., 1982; Hsiao et al., 1999; Vetterl et al.,
2000; Nasuno et al., 2002; Shah et al., 2002; Acciarri et al., 2005; Fonrodona et al.,
2006). PECVD is proven for industrial applications and control film quality is feasible for large area devices. It is apparent that, in order to improve cell efficiency, an in-depth understanding of nc-Si:H and its evolution in device configuration is needed based on the inhomogeneity in the growth of such films as demonstrated in the Chapter 4 for material nucleating crystallites from the amorphous phase.
Many recent studies have analyzed nc-Si:H formation using in situ characterization, such as RTSE and infrared attenuated total reflection (Fujiwara et al.,
2000; Collins et al., 2003). The results obtained by these studies support the surface diffusion model, which suggests that the diffusion of Si radicals is increased on the growth surface with the presence of additional hydrogen also on the growth surface. The 100
quantitative analysis, characterization, and control of the relative nanocrystalline and amorphous volume fractions within mixed-phase films is also a major challenge in Si:H manufacturing. Most often the nanocrystalline fraction is estimated from x-ray diffraction or Raman spectroscopy, which can yield values ranging an order of magnitude (Vetterl et al., 2000; Shah et al., 2003; Guha et al., 2011). Although these measurements are valuable, limitations exist. Typically ex situ x-ray diffraction measurements average information over the full depth of a thin film sample, and ex situ Raman spectroscopy averages information over a finite penetration depth into the sample that is dependent upon the wavelength of the probing laser, its power, and the absorption coefficient of the material. Si:H films may be inhomogeneous with thickness as crystallites nucleate from the amorphous phase and while the amorphous and nanocrystalline phases coexist.
Deconvolving gradients in crystallinity from ex situ x-ray diffraction and Raman spectroscopy measurements requires multiple samples, while in situ RTSE measurements applied during film deposition have been used to quantify structural gradients in crystallinity within a single film. The continuous deposition of microcrystalline silicon has been also monitored in situ with Raman spectroscopy to understand and control the deposition process (Muthmann et al., 2011)
With these motivations, an effort has been made to synthesize and analyze doped
(n-type) and undoped (intrinsic) PECVD nc-Si:H. The samples used were deposited on different substrate including that of n-i-p devices configuration. This chapter along with the analysis and development of high quality nc-Si:H materials also integrates some depositions in the device configuration used in the phase diagrams development from
Chapter 3 and 4. 101
5.2. Overview of Microstructural Evolution (a-Si:H to nc- Si:H)
Si:H films that evolve from the amorphous phase show two main transitions creating separate three different regimes in the phase evolution as discussed earlier in Chapters 3 and 4. The evolution of surface roughness layer thickness (ds) were obtained through analyses based on the global -minimization method with a two-layer optical model, as shown in Figure 5-1.
Figure 5-1. Results from RTSE analysis using a two-layer optical model for a Si:H film
deposition with the anc transitions. The ds and the σ of the mean square
deviation obtained in the LSR analysis results were plotted versus db. The vertical dash line separates three different regimes in the phase evolution of film growth i.e. amorphous, mixed, and nanocrystalline. 102
The evolution of ds versus the bulk layer thickness (db) and the evolution of the unweighted error function () obtained in the least-squares analysis are shown. It is clear that the increase in ds, reflecting the growth of the nc-Si:H phase over the a-Si:H phase, is associated with a significant increase in . This behavior illustrates that the nucleation of crystallites makes the simple two layer optical model insufficient to fully describe the optical and microstructural evolution of such films. The coalescence process is termed as
(a+nc)nc transition and occurs with an abrupt increase and subsequent decrease of ds.
For PECVD nc-Si:H films, various process parameters and the dilution of precursors, such as SiH4 and Si2H6 with H2, have a strong impact on the structure and morphology of the films. Some reports indicate that the nc-Si:H films deposited by PECVD do not show any straightforward correlation between the process parameters and the resulting film properties (Wang et al., 2003) due to the heterogeneity of films. On the other hand, others
(Shirai et al., 2002; Acciarri et al., 2005) have indicated that the crystallite size and densities can be controlled by deposition time, process pressure, plasma power, and substrate temperature. Therefore, more detail and careful investigations for the synthesis and characterization of PECVD nc-Si:H films are needed.
Virtual interface analysis (VIA) applied to RTSE data provides a means of tracking inhomogeneity in nc-Si:H layers to better identify structure and its variations.
Aspnes was the first to pioneer real time ellipsometry data analysis based on a three medium model (Aspnes 1993). He used an optical model consisting of an ambient/outer layer interface and an outer layer/pseudo substrate virtual interface (VI) instead of a multilayer stack for real time data analysis. The approach was adopted by Ferlauto et al.
(Ferlauto et al., 2000a) which has been used to solve this problem by modeling the film 103
as a series of stacked layers, including a pseudosubstrate, and to perform Σσ- minimization on each layer throughout the growth. The analysis determines spectra in of the material versus time in a stepwise manner (Ferlauto et al., 2000a). The VI optical model simplifies the management of a complicated graded layer by converting all the information of the underlying variations into a single pseudo-substrate. The need for a complicated multilayer stack is eliminated in VIA because only the top-most layers are analyzed. The discrete, stacked-layer method has its limitations like susceptibility to cumulative errors propagating from all underlying layers of the RTSE. In addition, due to the complexity of the stacked model, the surface roughness layer must still be described as a 0.5/0.5 mixture of a single material and void. This definition can lead to unrealistic surface roughness layer thicknesses due to optical property gradients occurring in the top- most material in the top layer of the stack. These type of disagreements were found as compared to AFM results noted by Fujiwara et al., (Fujiwara et al., 2001). The analysis method as reported by Aspnes (Aspnes, 1993) was later adjusted by Kim et al. including surface roughness on top of outer layer in the optical model (Kim and Collins 1996). The new model after modification is comprised of four media: (i) ambient, (ii) a surface roughness layer consisting of a mixture of outer layer and void, (iii) an outer layer with unique spectra in and (iv) the pseudo-substrate which contains the past history of the deposition.
Figure 5-2 shows a schematic of a four-medium optical model adapted from Stoke
(Stoke, 2008) justifying VIA and a four-medium optical model for films nucleating crystallites within.
104
Figure 5-2. a) Schematic of a four-medium optical model used in VIA of film nucleating crystallites. The use of this new optical model simplifies the management of a complicated graded layer by converting all the information of the underlying variations into a single pseudo-substrate analysis; the model consists of (i) the ambient, (ii) a surface roughness layer, (iii) an outer layer
with complex dielectric function spectra o, and (iv) a pseudosubstrate with complex dielectric function spectra <>. b) Schematic of a typical film with non-uniform optical response in the growth direction, whereby the nonuniformity is due to microstructural evolution from the pure amorphous phase through a mixed-phase regime to the pure nanocrystalline phase, (Stoke, 2008).
105
These model have been used in analysis of Si:H and amorphous silicon carbon alloys (a-
Si1–xCx:H) thin films (Kim et al., 1996; Fujiwara, et al., 1997; Ferlauto et al., 2004).
Additionally, the four medium optical model has been applied to silicon germanium alloys (Si1–xGex:H) thin films in order to determine the evolution of the volume fraction of crystallites, and in a-Si1–xGex:H films prepared with germanium content grading
(Podraza et al., 2008).
5.3. Experimental Details
The samples studied here tracking the evolution of crystallinity in detail were (i) undoped Si:H produced from SiH4 on native oxide covered c-Si, (ii) undoped Si:H produced from SiH4 + Si2H6 mixtures on native oxide covered c-Si (iii) n-type Si:H on thermal oxide coated c-Si, (iv) n-type Si:H on c-Si/SiO2/ZnO, (iv) n-type Si:H on
Ag/ZnO, and (v) undoped Si:H on Ag/ZnO/n-type a-Si:H. RTSE data acquisition was executed during each Si:H deposition using a rotating-compensator multichannel instrument (Lee et al., 1998) that can provide ellipsometric spectra from 0.75 to 5.88 eV as described in Chapter 2. Ellipsometric spectra were collected and averaged within a time of ~1.5 s. The angle of incidence for all depositions was fixed at a value within the range of 70.0o ± 0.5o. Before acquiring RTSE data during Si:H film deposition, single scan spectra were collected for well-known substrates at both room temperature and
200oC, to determine overlayer thicknesses and optical properties of each component. A four-medium optical model was employed in the analysis of materials having inhomogenous optical functions in the growth direction due to phase changes from a-Si:H to nc-Si:H on different substrates as initially detected from -minimization analysis. 106
The underlying principles and the optical model are adopted from Ferlauto et al. (Ferlauto et al., 2004). The Growth Rate and Optical Constant (GROC) model in J. A. Woollam
CompleteEASE version (v.5.08) contains a virtual interface. It assumes that the optical properties do not vary as a function of film thickness within the outer layer, and the growth rate is also constant.
5.4. Microstructural Evolution of nc-Si:H on c-Si
VIA was applied to determine the optical functions of nc-Si:H as well as the evolution of the nanocrystalline volume fraction (fnc) and void volume fraction (fvoid) on different substrates. The first step is to determine the spectra in that best represents the a-Si:H and nc-Si:H phases. This is achieved through a multi-step procedure that begins with identifying a time interval from which optical functions can be extracted. The a-Si:H spectra in is obtained from the global minimization procedure at ~200 Å of accumulated material thickness before the formation of nanocrystals from the amorphous phase as described in Chapter 2. The resulting a-Si:H spectra in is smoothened using an analytical formula consisting of a Cody-Lorentz oscillator. The bulk layer thickness and the surface roughness layer thickness evolutions are then determined from least squares regression analysis as described in Chapters 3 and 4. A distinct type of roughening transition is reported in which crystallites nucleate from the growing amorphous phase.
Because of the low crystallite nucleation density as observed by Ferlauto et al., and
Fujiwara et al., (Fujiwara et al., 2001, Ferlauto et al., 2004), the growth of crystalline protrusions produce a roughness layer that increases in thickness promptly with bulk layer thickness. Detailed microstructural evolution of Si:H films were previously
107
described in Section 3.2.2. The most reliable spectra in for nc-Si:H is then extracted at a bulk layer thickness ~200 Å after the coalescence thickness, the (a+nc)nc transition. In order to extract spectra in for nc-Si:H, the ellipsometric spectra at the corresponding time point are inverted assuming different surface roughness values by applying a two layer model. This model consists of a bulk nanocrystalline layer with a surface roughness layer on top. In this new analysis routine for extracting nc-Si:H , the surface roughness layer thickness ds, which occur at ~200 Å after the coalescence thickness is the only global free parameter. The optical response of a-Si:H, nc-Si:H, and void are used to track inhomegeitiy in the films.
For development of this analytical approach in determining structure and inhomogeneity for Si:H, a 1000 Å thick film deposited on native-oxide covered c-Si wafers was chosen as the test case in order to maintain the highest sensitivity to the evolution of surface roughness and bulk layer thickness with time. The film deposition
o parameters included a low substrate temperature of Ts = 200 C, a low total pressure of p
2 = 0.8 Torr, and a plasma power density of P = 0.04 W/cm with R = 30. This example is film from Chapter 3 synthesized with S = 0. The values for db as a function of time were deduced earlier in the simplified analysis using the two-layer model. The best fit growth rate obtained from the observed linear slope of db versus time is 0.53 Å/s and is fixed in the VIA. Spectra in that best represents the nanocrystalline phase is obtained through a multi-step procedure that begins with identifying a time interval which corresponds to
200 Å bulk layer thickness after crystallites coalescence. In this case, it is 30.56 minutes corresponding to 880 Å of total bulk layer thickness. The thickness should span a range
108
in the deposition process when the growing material is single phase and exhibits only
small surface roughness variation. Afterward, several sensible surface roughness values
over a range are identified such that each roughness values may be valid within the time
interval range of the analysis. Then, a series of nc-Si:H are generated from these trial
values of surface roughness. For example, the reasonable values for surface roughness for
a particular time point selected was from 10 to 50 Å, giving the grid spacing equal to 1-
2 Å as shown in Figure 5-3. The selection of ds values can have coarser grids and can be
used to identify a “general minimum.”
Figure 5-3. Spectrally averaged mean square error (MSE) versus assumed surface roughness layer thickness values using VIA and least squares regression for a Si:H film prepared at R = 30 on native oxide covered c-Si substrate. A three component EMA
(with volume fractions (fnc, fvoid) is used to represent the optical response of the outer layer of the film. The inset shows spectra in for nc-Si:H with the best fit
value of ds = 34.95 Å.
109
Once we have idea about the minima, the grid can be constructed then spectra in can be re-evaluated with a finer grid. Within this process, test values of ds are fixed and spectra in are extracted by numerical inversion. These spectra are parameterized with an analytical form consisting of two Lorentz oscillators with a common Tauc gap (Jellison and Modine, 1996a; Jellison and Modine, 1996b ). The Tauc-Lorentz oscillator is represented by:
2 AE0g E E 1 . E E 2 22g 2 2 2 EEE E, 0 (5.1) 0 E E g and
2 P2 d , (5.2) 1 22 Eg E where, A is the amplitude, E0 is the resonance energy, is the broadening parameter, and
Eg is band gap energy. These optical functions for a given value then serve as the “first guess” when selecting a new ds value higher or lower followed by numerical inversion.
This iteration is repeated until all spectra in have been generated using surface roughness values from ds,min = 10.95 Å to ds,max = 50.95 Å as illustrated in Figure 5-3.
These extracted optical functions are then used in a three component Bruggeman effective medium approximation (fa, fnc, fvoid) along with least-squares regression and VIA over the defined time range with ds and fnc and fvoid as free parameters. The time interval fit with this model corresponds to the point of nanocrystallite nucleation up to 200Å beyond nanocrystallite coalescence. The output of this analysis is an absolute MSE versus time. The time and spectrally averaged value of MSE which is performed with a different set of optical functions are then plotted as a function of ds. The clear minimum 110
of the plot is the surface roughness value corresponding to the best possible optical properties to represent the nc-Si:H component in this particular film. A plot resulting from such an analysis procedure are shown in Figures 5-3. If there is no clear minimum in MSE then the process is repeated over a different time range.
The results in Figure 5-4 demonstrates an example of the results of VIA applied to
RTSE data to obtain the surface roughness thickness, nanocrystallite fraction, void fraction, and average mean square error as defined in Equation (2.37) as functions of the bulk layer thickness. The VIA applied here utilizes spectra from 2.75 to 5.0 eV and for a-Si:H and nc-Si:H components. It is observed that the values for σ obtained in VIA are much lower (1.145×10-3) than those obtained in the standard two-layer model (33.44×10-
3). This reveals that fact that a much better description of the (a+nc)-Si:H and nc-Si:H growth regimes is obtained from VIA than global -minimization. The fnc determined in the VIA for the outer layer of thickness gives the average fractional area of the surface covered by nc-Si:H. The nanocrystallite fraction increases with bulk layer thickness, then converges to 1.0 (or 100%) around 600 Å of bulk layer thickness as expected for a nanocrystalline film. Voids initially appear with the nucleation of crystallites, which then subsequently decreases and stabilize to zero throughout the growth of this layer. For this particular case in Figure 5-4, it appears as though the grains are well passivated and dense. However, depending on the source of reference for nc-Si:H, a stable but nonzero value of fvoid can indicate that the grains under these conditions are not well passivated with a-Si:H as is desirable in nc-Si:H PV. The surface roughness evolution determined in the VIA displays the same behavior as that from the two-layer model, however the values
111
are larger for VIA. This behavior is expected as the value of the surface roughness layer thickness at coalescence of crystallites are found to be higher in the σ-minimization procedure used to obtain reference spectra in for nc-Si:H as part of the VIA procedure.
Figure 5-4. , ds, as well as fnc and fvoid in the top 10 Å of the bulk layer, plotted versus the accumulated bulk layer thickness for a R = 30 and S = 0 undoped Si:H film, as determined by VIA applied to RTSE data. The results of the VIA (open circles) are compared to results obtained in the standard RTSE analysis using the two-layer model and global -minimization (solid squares). 112
It has been reported by Ferlauto (Ferlauto, 2001) that the two-layer analysis delivers adequately precise results for estimation of the nucleation density and cone angle within the cone growth model. However, it steadily underestimates the surface roughness layer thickness at coalescence. Figure 5-5 shows the results of VIA applied to a R = 30, S
= 0.12 Si:H film. The sample has been taken to study the effect of disilane additives to the reactive gas SiH4. The a-Si:H spectra in was extracted before nucleation of crystallites at 100 Å whereas, nc-Si:H spectra in was extracted with the best fit value of ds = 50.37 Å around 1600 Å bulk thickness. The optical properties and best fit value of surface roughness were extracted in a same way as done for previous example R = 30 and
S = 0 undoped Si:H film deposited on native-oxide covered c-Si wafer. The deposition conditions for the Si:H film were reported in section 3.3 of Chapter 3. The film does not remain amorphous throughout growth but instead it evolves to the nanocrystalline phase.
The plot shows the comparison results of two and three component material effective medium approximation over-layers as determined by VIA. The combination of both voids and a-Si:H fractions leads to lower values of the error function, particularly for data at the onset of nanocrystallinity. This film shows an amorphous to mixed-phase
(aa+nc) transition at a bulk layer thickness 150 Å followed by a (mixed-phase) to single-phase nanocrystalline transition at approximately 1000 Å. Hence, comparing the S
= 0 and 0.12 films prepared at the lowest R where crystallites appear, the addition of
Si2H6 appears to cause initial crystalline nucleation at lower bulk layer thickness and suppresses earlier coalescence of crystallites. Also, these variations indicate that the presence of Si2H6 requires more H2-dilution to improve ordering in the amorphous film and nucleate crystallites. The initial appearance of crystallites may be the same, but the 113
delay in coalescence indicates a reduction in the crystallite nucleation density for S =
0.12. This behavior is expected as each Si2H6 molecule possesses double the silicon as
traditional SiH4, so the average silane radical to hydrogen ratio in the plasma is increased.
Spectra in used here in the analysis of the full data set were taken from the results
generating phase diagrams in Chapter 3 for fixed S = 0.12 and variable R.
Figure 5-5. , ds, as well as fnc and fvoid in the top 10 Å of the bulk layer, plotted versus the accumulated bulk layer thickness for a R = 30 and S = 0.12 undoped Si:H film.
114
The analysis employed not only compares the effect of adding disilane additives but also component of materials undertaken. The comparison of the σ, shows that three component effective medium approximation with amorphous, nanocrystalline, and void as the component material gives more accurate results of the desired sample. The increase in growth rate with S (0.69 compared to 0.53 Å/s) observed can be related with comparable or prompter nucleation of crystallites as compared to the S = 0 film prepared with only
SiH4.
5.5. Microstructural Evolution of n-type Si:H
The main limitation on single junction PV device performance is the restricted range of the solar spectrum that can be collected by the semiconductor and converted to electricity. By changing the deposition parameters or including another raw material such as germane gas in the PECVD process, the band gap of the intrinsic layer can be altered so that the layer can work in conjunction with other layers to maximally collect a wider range of the solar spectrum. Hence, Si:H solar cells constructed in multiple n-i-p junctions are beneficial if each intrinsic layer has the correct thickness for proper “current matching” among the junctions. The layer analysis involved here could be useful in bottom junctions of a thin Si:H-based triple junction solar cell, a-Si:H/nc-Si:H tandem devices, and single junction nc-Si:H cells.
Although it has several advantages over a-Si:H and a-SiGe:H, nc-Si:H solar cells also have challenges limiting volume manufacturing. The critical challenges are high rate deposition, large-area uniformity, and high performance as a component cell in multi- junction structures. As a start to addressing nc-Si:H growth in PV devices, we will study
115
variations in the bottom-most layer in n-i-p nc-Si:H devices. VIA was applied RTSE data for a 1000 Å thick n-type Si:H film deposited on a 6 inch diameter c-Si wafer coated with
o 1000 Å thick thermal oxide. Fixed deposition parameters include Ts = 200 C, p = 1.5
2 Torr, P = 0.032 W/cm , R = 100, and n-type doping gas ratio D = [PH3] / [SiH4] =
0.0125. Spectra of nanocrystalline component was obtained around 950 Å of bulk layer thickness as part of the VIA. We used the same method of doing VIA as done for previous samples. Different combinations of material component (a-Si:H, nc-Si:H, and void) with volume fractions fa, fnc and fvoid in an effective medium approximation is used to represent the optical response of the outer layer of the film. Among all combinations, the combination of both nc-Si:H and void fractions leads to lower values of the error function with sensible evolution of surface roughness throughout the film growth. The film nucleates as nc-Si:H and grows at a rate of 0.29 Å/s with no amorphous component material detected. The results from this analysis are shown in Figure 5-6. The nanocrystallite fraction in the outer layer for the film exhibits new features at 120 Å of bulk layer thickness evolution, as evidenced by the brief suppression of nanocrystallinity fnc 0.70. The decrease in ds within the first ~150 Å of material growth indicates that initial crystallites nucleate followed by the coalescence of initial clusters. After this coalescence the crystallite fraction increases then stabilizes near 0.7. Between ~150 and
300 Å, surface roughness increases slightly before decreasing. The nanocrystallite fraction increases after this point then converges to a completely nanocrystalline film after ~600 Å of accumulated material. From this analysis we suspect that randomly oriented crystallites nucleate on the substrate surface in the beginning. After ~150 Å of accumulated material, grains with more energetically favorable orientations evolve 116
preferentially over other grains. After ~600 Å of material accumulation, a densely packed nanocrystalline film is formed. The other samples are series of n-type nc-Si:H films deposited onto three inch diameter thermal oxide (~250 Å) covered c-Si substrates, initially pre-coated with varieties of ZnO.
Figure 5-6. Results of VIA applied to RTSE data for a 1000Å Si:H film deposited on a 6 inch diameter c-Si wafer coated with 1000 Å thermal oxide. Depth profiles in crystallite fraction within the bulk layer and surface roughness morphology evolution with bulk layer thickness are identified. No amorphous component material is observed in this film.
117
All samples were mounted in the deposition system at the same time so that nc-Si:H was co-deposited. Figure 5-7 shows a comparison of the experimental ellipsometric spectra collected for substrate c-Si/SiO2/ZnO in terms of Ν = Cos (2), Ϲ = Sin (2) Cos , Ѕ =
Sin (2) Sin and model fit. It should be noted that ZnO was produced through a collaboration, and details of the specific ZnO processing conditions will not be provided here. The samples are listed as “A,” “B,” and “C.” Samples A and B were processed at temperatures greater than that of the PECVD process, while sample C was not. The reference optical properties for c-Si and SiO2 were used from 0.75 to 5.88 eV (Herzinger et al., 1998), however the optical properties for ZnO were parameterized. The ZnO parameterization was done using optical properties reported in Section 4.3 of Chapter 4 as a starting material.
Figure 5-7. Comparison of the experimental ellipsometric spectra (Ν = Cos(2), Ϲ = Sin(2) Cos, Ѕ = Sin(2) Sin) in colored open circles and model fit as solid line for ZnO films on ~250 Å thermal oxide coated c-Si (sample A).
118
As expected, the amplitude of the CPPB oscillators used to model for ZnO were higher for the higher temperature processed samples. This variation can be attributed to formation of a denser material at higher temperature processing. All substrate are measured and mapped using ex situ spectroscopic ellipsometry techniques. The comparison of uniformity of ZnO thickness for each sample will be reported in the last part of this section.
After all pertinent information about these samples were obtained at room temperature, the samples were heated for two hours in H2 immediately prior to n-type
PECVD of R = 100 Si:H. The deposition conditions used for n-type Si:H were same as the deposition of Si:H on thermal oxide coated c-Si discussed in the beginning of this section and RTSE data was collected only for the film deposited on c-Si/~250 Å
SiO2/ZnO denoted as sample A. Figure 5-8 depicts the time dependence of bulk layer thickness of RTSE analysis performed on sample A showing continuous increase to a maximum thickness near 2100 Å. The analysis of RTSE data was divided into two ranges: nucleation and bulk evolution. In the nucleation stage, a single Bruggeman effective medium approximation layer consisting of void and nc-Si:H fractions is used to represent the not coalesced crystallites nucleating on the surface. The bulk growth regime is represented by a Bruggeman effective medium approximation consisting of variable void and nc-Si:H fractions and a surface roughness layer represented as a Bruggeman effective medium approximation of 0.5 bulk material and 0.5 void fractions. No a-Si:H component was detected for this film. The maximum value of nanocrystallite fraction compared to voids in the initial nucleation stage corresponds to almost same percentage as that of relative nanocrystallite fraction compared to voids within the bulk film layer 119
with overlayer surface roughness. Surface roughness thickness reaches a maximum, decreases, and then stabilizes indicating that crystallites have coalesced. The comparison of evolution from VIA using different components like a-Si:H, nc-Si:H, and void are shown in Figure 5-9.
Figure 5-8. Time dependence of bulk layer thickness, surface roughness, relative nanocrystallite fraction compared to voids within the bulk film layer and nanocrystallite fraction compared to voids in the initial nucleation stage with two component model for ZnO films on ~250 Å thermal oxide coated c-Si (sample A).
120
The a-Si:H component used here is taken from R = 50 n-type Si:H used in phase diagram development in Chapter 4 and nc-Si:H was obtained at 2000 Å of bulk layer growth for this sample. Regardless of the model applied to describe growth evolution, crystallite coalescence occurs near 100 Å of bulk layer thickness with a densely packed nc-Si:H film developing by 1000 Å of accumulated material.
Figure 5-9. , ds, fnc and fvoid in the top 10 Å of the bulk layer, plotted versus the accumulated bulk layer thickness for a R = 100 and n-type Si:H film. The plot shows the results of two component and 3 component effective medium layers. 121
The reduction in value of the error decreases significantly in the beginning of deposition by introducing an a-Si:H component. Earlier for R = 100 Si:H on thermal oxide, it was implied that there is not an a-Si:H component developing. The reduction in the error for
(nc-Si:H + a-Si:H) and (nc-Si:H + a-Si:H + void) however, reveals the presence of amorphous component. These models suggest the nucleation of amorphous material at least for the bulk layer thickness below 50 Å. But they shows failure around 350 Å of thickness as the amorphous component fraction after the crystallite coalescence increases around that thickness. It can be possible that grains with more energetically favorable orientations evolve preferentially right after the a-Si:H nucleation, coupled with comparable or prompter nucleation of crystallites.
VIA is next applied to a 1300 Å thick n-type Si:H deposited on a Ag/ZnO back
o reflector. Fixed deposition parameters included Ts = 200 C, p = 1.5 Torr, P = 0.032
2 W/cm , R = [H2] / [SiH4] = 80 and n-type doping gas ratio D = 0.0125. Spectra in in crystalline component around 1200 Å of bulk layer thickness along with that for void, were then used in a two component effective medium approximation layer. The least squares regression within the VIA was employed with ds and the relative nanocrystallite
(fnc) and void (fvoid) fractions as free parameters. The lowest average error was obtained for two component analysis with nc-Si:H and void for this example. The outcomes of analysis are shown in Figure 5-10. The film nucleates as nc-Si:H with 0.76 nanocrystalline volume fraction and growth rate of 0.38 Å/s. In these previous studies, it was shown that when crystalline nuclei develop on an a-Si:H thin film surface, they protrude outward. This gives rise to a rapid increase in surface roughness as the bulk layer thickness increases, due to the preferential growth of the nanocrystalline phase. 122
Thus, a sharp increase in surface roughness on a plot of db versus ds marks the transition from amorphous material to a mixed-phase material of amorphous and nc-Si:H. For this sample, however, surface roughness with respect to bulk layer thickness is initially stable and steadily decreases with increasing nc-Si:H fraction and decreasing void fraction.
Figure 5-10. Results of VIA applied to RTSE data for a 1300 Å thick R = 80 n-type Si:H film deposited on Ag/ZnO back reflector coated 6” x 6” borosilicate
glass. Depth profiles in fnc within the bulk layer, ds morphology and fvoid evolution with bulk layer thickness are identified. No amorphous component material is observed in this film.
123
Figure 5-11 shows 2 comparison of n-layer Si:H on three different substrate namely
Ag/ZnO, c-Si/SiO2 and c-Si/SiO2/ZnO. Spectra in for nc-Si:H film shows dampening of the amplitude and increased broadening of the critical point features relative to c-Si. The location of these critical points in for nc-Si:H are still preserved, unlike in a-Si:H.
Spectra in ε2 for the R = 80 n-layer Si:H deposited on back reflector had significantly lower amplitude. The nanocrystallite fraction in the outer layer for the film reaches fnc
0.85 near 100 Å of accumulated material and remains below ~0.90 for several hundred angstroms afterward. The surface roughness versus bulk layer thickness is also stable in this region of material growth then decreases continuously until the end of the deposition.
Figure 5-11. Spectra in (2) used in VIA applied to RTSE data for a R = 80 n-type Si:H film deposited on Ag/ZnO back reflector, R = 100 n-type Si:H film
deposited on c-Si/SiO2 and R = 100 n-type Si:H film deposited on c-
Si/SiO2/ZnO. 124
This decrease in ds indicates that crystallite nucleation on the substrate is followed by coalesce of the initial clusters. The surface roughness shows almost same value until 100
Å of accumulated material and then decreases to a value of 50 Å. The nanocrystallite fraction converges to a completely nanocrystalline film after ~800 Å of accumulated material when more energetically favorable orientations evolve preferentially over initial crystallites.
Comparison between c-Si/1000 Å SiO2/R=100 n-type Si:H, c-Si/250 Å
SiO2/ZnO/R=80 n-type Si:H and Ag/ZnO/R=80 n-type Si:H shows that all of them nucleate nanocrystallites during their growth. It can be seen from the evolution that the coalescence of the nanocrystallites occurs way sooner for the film on c-Si/250 Å SiO2/ZnO substrate. The other two indicates the shift in the development and coalescence of higher
bulk layer thicknesses, db, coal = 600 Å despite of their different R. This behavior results in occurrence of 0.90 nc-Si:H fraction occurring over a longer period of bulk layer thickness in case of nc-Si:H film on BR. The nanocrystallite fraction in the outerlayer for the film exhibits new features during its evolution, as evidenced by the brief suppression of nanocrystallinity in c-Si/1000 Å SiO2/ R=100 n-type Si:H and Ag/ZnO/R=80 n-type Si:H samples, showing the preferential growth of more energetically favorable crystallites within. The features exhibited by the surface roughness evolution were highest for the nc-
Si:H films on Ag/ZnO. The surface roughness of the back reflector substrate could be the reason for rougher film. The growth rate are 0.29 Å/s, 0.43 Å/s and 0.38 Å/s for c-Si/1000 Å
SiO2/R=100 n-type Si:H, c-Si/250 Å SiO2/ZnO/R=80 n-type Si:H and Ag/ZnO/R=80 n- type Si:H respectively. The evolution of void fraction and the noteworthy increase in
125
amplitude of spectra in ε2 for the R = 100 n-layer Si:H deposited on c-Si/250 Å SiO2/ZnO indicates that the film is densely packed among all.
5.6. Microstructural Evolution of Si:H on (n-type a-Si:H)/BR
The role of hydrogen dilution on the relative a-Si:H and nc-Si:H volume fractions in a-Si::H i-layer is studied here. Microstructural evolutions of intrinsic layers on the basis of two layer models were already presented in Chapter 4 with respect to phase diagram development. Depositions were performed at T = 200oC, P = 0.04 W/cm2, and p
= 0.8 Torr. The VIA applied here utilizes spectra from 2.75 to 5.0 eV and for a-Si:H and nc-Si:H components as shown in Figure 5-12. Spectra in for nc-Si:H was obtained from the end of the respective deposition, ~1150 Å of a 1300 Å thick film using the same optical model as was used for the i-layer growth evolution diagram. In this model the free parameters are db and ds. Spectra in for the amorphous phase was taken from the analysis of a R = 15 deposition corresponding to a time within the first ~200 Å of bulk material prior to the nucleation of nanocrystallites. There is strong optical contrast between the two sets of for Si:H, in that the amorphous phase has only a single broad resonance while that of nanocrystallite material has two features representative of dampened and broadened critical point features found in single c-Si. These reference spectra in for a-Si:H and nc-Si:H, along with that for void where = 1, were then used in a three component Bruggeman effective medium approximation (Aspnes, 1982;
Fujiwara, 2000) layer and a least-squares regression within the VIA with ds and the relative nanocrystallite (fnc) and void (fvoid) fractions as free parameters.
126
Figure 5-13 shows an example of the results of VIA applied to RTSE data to obtain the surface roughness thickness, nanocrystallite fraction, void fraction, and average mean square error as functions of the bulk layer thickness for a R = 50 i-layer on a BR over-coated with a 200 Å R = 50 n-layer. Results of VIA show an increase in surface roughness followed by a decrease within the first ~300 Å of material accumulation, indicating crystallite nucleation on the substrate followed by coalesce of the clusters.
Figure 5-12. Spectra in of a-Si:H and nc-Si:H reference material used in VIA as functions of photon energy extracted over a spectral range from 2.75 to 5.0 eV. Both for a-Si:H and nc-Si:H were obtained by numerical inversion at a bulk layer thickness of 200 Å and 1150 Å, respectively.
127
The nanocrystallite fraction increases with bulk layer thickness, then converges to 1.0 as expected for a nanocrystalline film. Voids initially appear with the nucleation of crystallites, which then subsequently decrease and stabilize near 0.04 throughout the growth of this layer. Spectrally averaged mean error for fvoid, fnc, and ds are 0.003, 0.024, and 0.8 Å respectively.
Figure 5-13. MSE, fvoid, fnc, and ds in the top 10 Å of the bulk layer, plotted versus the accumulated bulk layer thickness for a R = 50 Si:H film deposited on R = 50 n-type a-Si:H/Ag/ZnO, as determined by VIA applied to RTSE data.
Spectrally averaged mean error for fvoid, fnc, and ds are 0.003, 0.024, and 0.8 Å respectively. 128
Depending on the source of reference spectra in for nc-Si:H, this behavior could indicate that the grains under these conditions were not well passivated with a-Si:H as is desirable in nc-Si:H PV (Shah et al., 2003; Kroll et al., 1996). Optimized nanocrystalline
/microcrystalline PV devices often incorporate layers prepared at lowest hydrogen dilution where crystallite growth can occur, and nc-Si:H layers are often fabricated using hydrogen dilution grading approaches to manipulate the degree of crystallinity with the layer. For very high values of hydrogen dilution, such as R = 50 in this example, the material is likely not optimized for solar cells, because cracks related to voids can prompt shunts in the cells and channels by which contamination (e.g. oxygen) can enter into the layer (Stoke et al., 2008b; Vetterl et al., 2000; Funde et al., 2008).
5.7. Summary
RTSE and VIA have been used to track the amorphous-to-nanocrystalline transition in doped and undoped Si:H films and identify how both the microstructure and the optical properties change continuously and significantly with thickness. VIA has been successfully applied to report the evolution of nc-Si:H on different substrates. Phase transitions from amorphous to mixed-phase and from mixed-phase to single-phase nc-
Si:H are identified in case of Si:H deposited with and without Si2H6 gas added to SiH4. A much better description of the (a+nc)-Si:H and nc-Si:H growth regimes has been employed obtained from the VIA compared to Σ-minimization alone. Distinct features in the nanocrystallite fraction evolution for Si:H, have been observed. Further analysis
129
methods such as XTEM and AFM will be necessary to properly identify the nature of these features.
For Si:H films on native oxide covered c-Si, adding of Si2H6 requires more H2- dilution or higher dilution to improve ordering in the nucleated crystallites. This behavior is expected as each Si2H6 molecule possesses double the silicon as traditional SiH4, so the average silane radical to hydrogen ratio in the plasma is increased. Both n-type Si:H deposited on 1000 Å thermal oxide covered c-Si and on sputtered Ag/ZnO back reflector revealed a short-term suppression of nanocrystallinity which could be a result of formation of randomly oriented crystallites on the substrate surface or less energetically favorable orientations evolution with bulk. No a-Si:H component is observed for these samples. Whereas, the R = 100 n-type Si:H deposited on high temperature processed c-
Si/SiO2/ZnO substrate does not show the same nc-Si:H suppression, but requires an a-
Si:H component to fit ellipsometric spectra in the early stages of growth. The undoped
Si:H deposited on n-type a-Si:H coated BR shows the dependence of microstructural evolution for intrinsic layer growth in a PV device configuration. The stabilization of voids to about 4% after the coalescence of crystallites until the end of the deposition could indicate that the grains under these conditions were not well passivated.
130
Chapter 6
Infrared Extended Spectroscopic Ellipsometry Applied to Characterization of Thin Films and PV Device Structures
6.1. Overview
Electrical and optical measurements of amorphous thin films (a-Si:H) have demonstrated that a variety of industrial applications including PV may be achieved depending on the deposition technique and conditions (Brodsky et al., 1977; Ferlauto et al., 2002; Collins et al., 2003; Saint John et al., 2011). For any type of film, variations in reactor design often modify the local growth evolution of each film such that comparable processing conditions may not guarantee the same film growth, though correlations will exist. NIR to UV wavelength range spectroscopic ellipsometry is often employed for both in situ (Collins et al., 2003; Podraza et al., 2009; Dahal et al., 2010) and ex situ (Ferlauto et al., 2000; Attygalle et al., 2010; Saint John et al., 2011) metrology of PECVD thin films. Structure and optical properties gained over this range can provide some useful information in terms of the relative density and band gap of the amorphous material, however the change in measured electrical behavior and chemistry of the samples may only be weakly correlated to features in the optical response over this spectral range.
131
TCO’s are wide bandgap semiconductors intentionally doped to enhance their electrical conductivity. As the TCO layer is meant to serve as one of the contacts in the solar cell device by carrying current to the external circuit, the layer must exhibit a high electrical conductivity. An increase in the free electron concentration in the TCO layer by doping has the desired effect of increasing the conductivity of the layer but also the undesired effect of reducing the optical transmission in the near infrared (NIR) region of the solar spectrum due to absorption by the free electrons. Additionally, Si:H may be deposited on the TCO in the p-i-n superstrate device. There is a contrast in optical properties between substrate and semiconductor Si:H layers as the TCO’s are strongly absorbing in infrared region. These differences allows relatively easy determination of
TCO layer’s modification by depositing doped and undoped Si:H on top of them.
The IR spectral region of is host to the vibrational bonding and phonon modes present in these materials. This spectral region has been explored via infrared absorption measurements to quantify the silicon-nitrogen vibrational modes and the hydrogen content and bonding character of a-Si:H films (Brodsky et al., 1977; Langford, et al.,
1992; Smets et al., 2008; Attygalle et al., 2010). The availability of reflection mode infrared extended spectroscopic ellipsometry (IR-SE) data has enabled the measurement and modeling of complex over a range sensitive to the bonding character in a-Si:H films represented by the SiHn absorption features in device configurations. These tools allow the bonding characteristics to be probed non-invasively for thinner films than those used in earlier studies as well as for films in the respective complete PV device configuration. The results of the analysis of IR-SE data were used to study absorption in the BR and TCO components, which can be used to extract electrical transport properties 132
and phonon modes of these component materials. Comparison of optical absorption features afford a method of assessing film character, which then suggests ways to improve material quality and potentially device performance. It also contributes to the development of realistic optical models for in situ and ex situ optical spectra analysis.
6.2. Experimental Details
Ellipsometric spectra [in Ν = Cos(2), Ϲ = Sin(2)Cos, Ѕ = Sin(2)Sin] were collected for each sample at an angle of incidence = 70° using a variable-angle rotating compensator multichannel ellipsometer initially over a spectral range from 0.734 eV to
5.887 eV. A rotating compensator Fourier transform infrared (FTIR) ellipsometer operating from 0.033 to 0.742 eV was used to acquire ellipsometric data for all samples at = 70°. Comparison of the experimental and model fit ellipsometric spectra (Ν, Ϲ, Ѕ) for BR/n-type Si:H/intrinsic Si:H sample is shown in Figure 6-1 as an example. Multiple samples consisting of Ag/ZnO back reflector, Si:H films on an Ag/ZnO BR over-coated with a R = 50 a-Si:H n-layer, TCO coated glass (TECTM-15 substrate consisting of soda lime glass/SnO2/SiO2/SnO2:F provided by NSG Pilkington, USA) and Si:H film on
TECTM-15 substrate were studied here. These samples were chosen to represent the actual n-i-p substrate and p-i-n superstrate configurations of a-Si:H PV devices, although the final contact layers were not deposited here. Simultaneous analysis of ellipsometric spectra collected over the mid-IR to near UV range from multiple samples was used to yield a common for each layer while structural parameters such as the bulk layer thickness (db) and surface roughness (ds) may be varied separately as in RTSE data
133
analysis discussed in Chapter 2, 3, and 4 and similar in methodology to the divided spectral range approach (Gautam et al., 2014). The results of BR, TCO, and other underlying layer analysis obtained prior to semiconductor deposition are compared with the results obtained after film deposition. The appropriate structural and optical models were developed for each sample and the experimental ellipsometric spectra fit using least squares regression analysis as described in Chapter 2.
Figure 6-1. Comparison of the experimental ellipsometric spectra [Ν = Cos(2), Ϲ = Sin(2)Cos, Ѕ = Sin(2)Sin] in colored open circles and model fit as solid line for R = 10 a-Si:H films on BR over-coated with a R = 50 a-Si:H n- layer.
134
6.3. Data Analysis and Results
6.3.1. Si:H Layers on Ag/ZnO Back-Reflectors
Ellipsometric spectra from 0.04 to 5.0 eV were collected experimentally and analyzed for Ag/ZnO BR and BR/n-type/intrinsic a-Si:H structures. The models and thicknesses described correspond to two different ZnO coated Ag BR samples: one with and one without over-deposited n- and i-type a-Si:H layers. All layers of two samples as mentioned earlier were deposited without vacuum break as given in Table 4.1 of Chapter
4. In situ spectroscopic ellipsometry data from 0.73 eV to 5.88 eV was collected for each deposited layer and the model generated was used for extended IR-SE analysis. Data collected for semi-infinite Ag substrate were taken before ZnO layer deposition at room temperature and were analyzed in the energy range 0.734 to 5.88 eV. All parameters describing Ag and its surface roughness are listed in Table 4.2. The structural and parametric model for the Ag/ZnO BR in the energy range 0.73 to 5 eV consisted of Ag,
ZnO, and the Ag + ZnO interface and are listed in Tables 4.2 and 4.3. The analysis was extended to the IR by fitting parameters defining for ZnO only and fixing those defining for Ag and the Ag + ZnO interface as well as the interface layer thickness as shown in Table 6.1. The analysis approach was chosen as free carrier absorption represented by the Drude feature dominates the IR response of Ag and the Ag + ZnO interface layers and is already established from near IR to UV spectral range analysis. A common parameterization of was applied for the data collected from the two instruments with spectral ranges from 0.04-0.73 eV and 0.73-5.0 eV, respectively.
135
Table 6.1. Parameters describing and structure for a ZnO film deposited on Ag/ZnO BR. Experimental ellipsometric spectra were collected ex situ using near infrared to ultraviolet (0.734 to 5.0 eV) and infrared (0.04 to 0.734 eV) spectral range instruments and fit jointly using least squares regression analysis with an unweighted estimator error function, = 8 x 10-3. Parameters describing for Ag and the Ag + ZnO interface were fixed from Tables 4.2 and 4.3, respectively. The ZnO bulk layer thickness was allowed to vary separately for each set of ellipsometric spectra; all other parameters are common to both analyses. For ZnO, the parameterization of consisted of two CPPB oscillators, three Lorentz oscillators, and .
Layer Oscillators
CPPB (µ = 0.5) = 2.43 ± 0.01
ZnO A (eV) En (eV) Ө (degrees) 2.82 ± 0.02 0.209 ± 0.002 3.364 ± 0.001 20.8 ± 0.4 db (NIR to UV) 1.23 ± 0.02 3.95 ± 0.03 3.94 ± 0.02 0
= 2996 ± 2 Å A (Unitless) (eV) E0 (eV) Lorentz 0.75 ± 0.05 0.196 ± 0.005 0.264 ± 0.002 db (IR) Lorentz 3.17 ± 0.03 0.169 ± 0.007 0.134 ± 0.001 = 3025 ± 2 Å Lorentz 46 ± 2 0.0093 ± 0.0004 0.0501 ± 0.0002 ds = 84 ± 1 Å
The bulk ZnO layer thickness was allowed to vary to account for measurement on different spots over the sample surface. A common surface roughness thickness between the two sets of measured spectra was obtained, as this effect will vary less with non- uniformity than the overall bulk layer thickness. Figure 6-2 shows parameterized for
ZnO. The spectra in is represented by a combination of CPPB oscillators for electronic transitions, Lorentz oscillators representing IR phonon modes, and a constant real additive term to account for dispersion from absorption features outside the measured 136
spectral range from 0.04 to 5 eV with parameters given in Table 6.1. The near IR to near
UV range shows only small absorption below the lowest direct transition at 3.364 eV as expected for direct band gap ZnO (Jellison and Boatner, 1998). Phonon modes for wurtzite ZnO are opt = 1A1 + 2B1 + 1E1 + 2E2, with A1 and E1 modes IR-active. Only one characteristic transverse optical (TO) mode for ZnO with E1 symmetry at 0.0501 eV
(404.08 cm-1) is resolved for this sample (Tzolov et al., 2000; Ashkenov et al., 2003;
Bundesmann et al., 2004).
Figure 6-2. Spectra in as a function photon energy extracted over a spectral range from 0.04 to 5 eV for 3010 2 Å ZnO films on Ag. The inset shows high energy features of ZnO based on parameterization assuming critical point parabolic bands. 137
Weak absorption bands in the spectral region from 0.134 to 0.264 eV (1080 to 2130 cm-1) have been observed and are often associated with hydrogen-associated bending modes; stretching modes of hydrogen bonded to heavier elements like zinc; and various carbon, oxygen, and nitrogen-related stretching modes not involving hydrogen (Keyes et al.,
2005).
A similar sample was overcoated with a-Si:H to determine for a-Si:H over the
0.04 to 5.0 eV range as well as modifications to the underlying ZnO due to this over deposition. The structural model for the a-Si:H coated Ag/ZnO BR consisted of a semi- infinite opaque Ag metal layer, a 108 Å Ag + ZnO interfacial layer with fixed thickness from previous analysis in Table 4.2 and 4.3, an average 2751 5 Å bulk ZnO layer, a 84
1 Å 0.5 ZnO/0.5 n-type a-Si:H Bruggeman effective medium approximation interfacial layer, a 278 1 Å a-Si:H n-layer, a 30 1 Å 0.5 n-type a-Si:H/0.5 intrinsic Bruggeman effective medium approximation interfacial layer, a 3621 2 Å bulk intrinsic a-Si:H layer, and a 29 1 Å surface roughness represented using Bruggeman effective medium approximation of 0.5 intrinsic a-Si:H/0.5 void volume fractions, and an air ambient.
Figure 6-3 represents a schematic diagram of this structure. The optical model has become quite complicated with the addition of subsequent layers and interfaces. The
ZnO/n-layer interface, n-layer bulk layer, and n-layer surface roughness thicknesses are obtained from in situ RTSE measurements and analysis prior to intrinsic a-Si:H deposition. The n-type/intrinsic a-Si:H interface thickness is set at the same value as the n-layer surface roughness assuming that overdeposited intrinsic a-Si:H fill the voids in the n-layer surface. Parameters describing for ZnO and a-Si:H are listed in Table 6.2.
138
As with the IR extended analysis of the Ag/ZnO sample, a common parameterization of for the materials over the full spectral range was applied, the bulk layer thicknesses for the ZnO and intrinsic a-Si:H layers were fit independently for spectra collected from each instrument, and all other layer thicknesses were either fixed from prior analyses or kept common between the two sets of spectra. For the i-layer, the nominal substrate temperature and hydrogen dilution ratio were 200oC and R = 10, respectively. The optimized n-type a-Si:H thickness was fixed at 278 Å for R = 50 as found by RTSE growth evolution studies. As film growth is initiated, the depositing material starts filling the void space in the surface roughness layer of the underlying film.
29 ± 1 Å
i-Si:H 3671 ± 2 Å
n-Si:H + i-Si:H 30 ± 1 Å
n-Si:H 278 ± 1 Å ZnO + n-Si:H 84 ± 1 Å
ZnO 2751 ± 5 Å
Ag + ZnO 108 ± 11 Å
Ag Opaque
Figure 6-3. Optical model used in the analysis of the SE data collected on the a-Si:H based n-i-p substrate solar cell. The parameters that were fixed, fit, or coupled in the analysis are also shown.
139
At the same time, the protrusions in the surface roughness of the substrate film are coated with the depositing material, generating a roughness layer associated with the growing film. Bruggeman effective medium approximation defines spectra in for these interfaces.
Differences in for the ZnO are expected when overcoated with a-Si:H. PECVD of a-Si:H raises the temperature of ZnO to 200oC and exposes it to hydrogen in the plasma. There are many studies on the growth and various effects of annealing on the optical and structural properties of ZnO layers (Puchert et al., 1996; Kang et al., 2004;
Jung et al., 2005; Rolo et al., 2007). It is well identified that the properties of ZnO layers are strongly affected by not only the deposition conditions but also the post deposition annealing conditions or temperature treatment. Annealing has a large effect on the crystallinity of the layers in terms of grain size, residual strain, and the defect density as compared to as-deposited films. As noted in Table 6.2, the amplitude of the CPPB oscillators of ZnO overcoated with a-Si:H were fitted to account for changes in occurring during PECVD of a-Si:H. The increase in amplitude for higher energy absorption features in and the decrease in film thickness compared to the sample
o without a-Si:H coating at T = 200 C indicate that the as-deposited ZnO material densifies after annealing at the a-Si:H deposition temperature. The comparison of different phonon modes in ZnO with and without a-Si:H coating is shown in Figure 6-4. The characteristic
-1 TO modes with A1 and E1 symmetry at 0.0467 eV (376.66 cm ) and 0.0506 eV (408.12 cm-1) are present and able to be resolved (Tzolov et al., 2000; Ashkenov et al., 2003;
Bundesmann et al., 2004).
140
Table 6.2. Parameters describing and structure for a Ag/ZnO BR coated with n-type and intrinsic a-Si:H in 0.04 to 5.0 eV spectral range using least squares regression analysis with an unweighted estimator error function, = 11 x 10-3. Parameters describing for Ag and the Ag + ZnO interface were fixed from Tables 4.2 and 4.3, respectively. Parameters describing for the n-layer were determined from 200oC RTSE analysis of data and then parameter values extrapolated to RT. The db of ZnO and intrinsic a-Si:H were allowed to vary separately for each set of spectra; all other parameters are common to both analyses. For ZnO, the parameterization of consisted of two CPPB, four Lorentz oscillators, and . A Cody-Lorentz, a Sellmeier, three Gaussian oscillators and were added to the parameterization of for intrinsic a-Si:H.
Layer Oscillators
i-type a-Si:H Cody-Lorentz Eg (T&R) = 1.780 ± 0.001 = 1.50 ± 0.01 A (unitless) (eV) E0 (eV) db (Near IR to UV) = Gaussian 1.732 ± 0.06 0.013 ± 0.001 0.079 ± 0.001 3623 ± 1 Å Gaussian 0.28 ± 0.01 0.010 ± 0.001 0.249 ± 0.001 Gaussian 0.41 ± 0.04 0.016 ± 0.002 0.106 ± 0.001 d (IR) = 3619 ± 2 Å b Sellmeier 0.0050 ± 0.0002 eV2 0
ds = 29.0 ± 0.3 Å
n-type a-Si:H/i- a-Si:H = 1 Interface = 30 ± 1 Å A (eV) (eV) E0 (eV) Eg Ep (eV) (eV) n-layer (db) = 278 ± 1 Å Cody- 62 2.01 3.99 1.65 1.05 Lorentz ZnO/n-type a-Si:H Interface = 84 Å
ZnO = 1.91 ± 0.02 µ = 0.5 A (eV) En (eV) Ө d (Near IR to UV) = b CPPB 4.04 ± 0.05 0.209 3.364 20.8 2763 ± 3 Å CPPB 1.31 ± 0.02 3.95 3.94
db IR) = 2738 ± 5 Å A (Unitless) (eV) E0 (eV) Lorentz 3.89 ± 0.1 0.233 ± 0.001 0.162 ± 0.002 Lorentz 82 ± 4 0.0030±0.0003 0.0506 ± 0.0001 Lorentz 16.4 ± 0.4 0.039 ± 0.002 0.085 ± 0.001 Lorentz 13 ± 3 0.004 ± 0.002 0.047 ± 0.001
141
The vibrational mode at 0.0847 eV (683.15 cm-1) can be attributed to longitudinal optical
(LO) mode with A1 symmetry (Tzolov et al., 2000; Özgür et al., 2005). The orientation of grains in film could be a reason for shifting of modes to slightly higher or lower wavenumbers. The splitting of the peak observed at 404 cm-1 for the uncoated ZnO into two expected peaks at 377 and 408 cm-1 for ZnO overcoated with a-Si:H and appearance of the 683 cm-1 mode is likely due to grain structure refinement from annealing at at
o -1 200 C (Rolo et al., 2007). The increase in amplitude for 2 of phonon mode at 408 cm also supports the idea that grain restructuring and material densification occurs.
Figure 6-4. Comparison of 2 between the Ag/ZnO back reflector (BR) samples with and without a-Si:H over-deposition. Each BR stack was fabricated under same deposition conditions as given in Table 4.1 of Chapter 4.
142
The presence of additional absorption mode at 0.162 eV (1306.62 cm-1) can be associated with oxygen-hydrogen (O–H) bonds in the thin film, such as the formation of zinc hydroxide or absorbed water or stretching modes of hydrogen bonded to heavier elements like zinc (Keyes et al., 2005). The significant increase in broadening of this absorption peak could be due to the modification or damage to the ZnO as a result of exposure to hydrogen during PECVD.
After ZnO deposition, a 278 Å thick n-layer was deposited onto a Ag/ZnO coated substrate with deposition conditions given in Table 4.1 of Chapter 4. The n-layer optical properties, as well as its db and ds, were obtained using RTSE analysis. The final numerically inverted spectra in for the n-layer were fit to a Cody-Lorentz oscillator
(Equation 2.38 and 2.39). Parameters describing for the n-layer at the deposition
o temperature T = 200 C are A = 59 ± 2 eV, = 2.12 ± 0.02 eV, E0 = 3.99 ± 0.01 eV, Eg =
1.58 ± 0.04 eV, and Ep = 0.96 ± 0.09 eV. Figure 6-5 shows spectra in for the R = 10 a-
Si:H intrinsic layer parameterized using a Cody-Lorentz oscillator at high energies and
Gaussian oscillators to represent the IR vibrational modes. Each Gaussian oscillator is described by:
EEEE nn (E) Ae Ae (6.1)
(6.2) 2 ln 2 where A, , and En represent amplitude, broadening, and resonance energy respectively.
The 1 2 ln 2 factor defines broadening approximately equals to full width at half
143
maximum (FWHM). The Cody-Lorentz oscillator parameters for intrinsic a-Si:H were linked to a single fit parameter, Eg, by linear relationships previously determined for PV device quality a-Si:H (Section 2.2.1.3 of Chapter 2). This technique minimizes the number of fit parameters allowing for physically realistic ε to be determined. Parameters describing spectra in for the underlying n-layer were extrapolated based on previously observed trends in the Cody-Lorentz oscillator parameters with temperature (Podraza et al., 2006a).
Figure 6-5. Spectra in extracted over a spectral range from 0.04 to 5 eV for 3621 2 Å R = 10 a-Si:H films on BR over-coated with a R = 50 a-Si:H n-layer. The inset shows lower energy features in for a-Si:H as function of photon energy in the 0.04 to 0.73 eV spectral range.
144
IR vibrational studies of a-Si:H have been useful in understanding the role of Si-H bonding in determining a-Si:H properties. High mobility and reactivity of hydrogen enables passivation of the electronic defect states in a-Si:H and relaxes the a-Si:H network to improve electronic and structural properties. IR-absorption studies have shown that hydrogen in a-Si:H is bonded as Si-Hn , with n =1, 2, and 3 (Brodsky et al.,
1977; Langford et al., 1992). IR features in for the intrinsic a-Si:H film are highlighted in the inset of Figure 6-5. Spectra in for a-Si:H in the n-i-p device configuration exhibited bending modes near 0.079 eV (635.6 cm-1) and a stretching monohydride (Si-
H) mode around 0.249 eV (2008.3 cm-1). In addition to the expected SiH modes, this a-
Si:H sample exhibited an absorption mode centered around 0.106 eV (854.9 cm-1), which can be attributed with the bending or scissors mode of SiH2 dihydride. The peak
-1 centered ~2100 cm assigned to the dihydride (SiH2) or clustered hydrogen is not observed. The absence of that peak usually confirms the presence of ordered dense Si:H material (Smets et al., 2003; Müllerová et al., 2006). Inferences about film density can be made from the amplitude of 1 and the relatively high amplitude of the near IR to UV absorption feature in 2 which is characteristic of device quality a-Si:H.
6.3.2. Si:H Layers on TCOs
In the superstrate configuration, light enters into the device from glass side and any optical absorption in the component layers of the device before the light reaches the semiconductor absorber does not generate photocurrent. Hence, there is often a trade-off between high electrical conductivity and high optical transparency of the TCO layer
(Gordon, 2000). In order to reach the highest overall performance of solar cell device, the 145
TCO layer must be optimized with respect to its electrical conductivity and optical transparency. TECTM glasses are a product of NSG Pilkington, USA, and are used for a variety of applications. Thin film a-Si:H solar cells in the superstrate configuration are fabricated on TECTM glasses. Soda lime glass (SLG) including layers of undoped tin oxide (SnO2), SiO2, and fluorine doped tin oxide (SnO2:F) constitute the TEC™-15 coated glass product as shown in Figure 6-6. In order to analyze the ellipsometric spectra acquired on a sample with a relatively complicated optical structure, a stepwise analysis approach is considered to be effective (Chen et al., 2009a).
EMA 4 (258 ± 1 Å, void 52%)
EMA 3 (422 ± 2 Å, void 15%)
EMA 2 (598 ± 2 Å, void 8%)
EMA 1 (552 ± 2 Å, void 4%)
5 SnO : F (870 ± 4 Å) 2
4 SnO2: F (591 ± 3 Å)
3 SnO2: F (28 ± 2 Å)
2 SiO2 (186 ± 2 Å)
1 SnO (316 ± 3 Å) 2 Soda lime glass substrate
Figure 6-6. Optical model showing different layers for the glass/SnO2/SiO2/SnO2:F (TEC™-15) substrate used in analysis of SE data (0.04 to 5 eV) collected at RT. Structural parameter values for the components with their confidence -3 limits were also depicted. The σ = 1.65 x 10 was obtained in this analysis. The numeric shows the layer numbers used for the optical model.
146
To achieve the goal, ellipsometric spectra at room temperature from 0.734 to 5.0 eV and
IR-SE data from 0.04 to 0.73 eV were analyzed simultaneously using a least squares regression analysis to an optical model shown in Figure 6-6. A spectrally averaged unweighted error function () was used to fit the experimental ellipsometric spectra to an optical model. Spectra in for glass, undoped SnO2 and SiO2 were used from a database that has been established previously (Chen, 2010) and were plotted in Figure 6-7.
Figure 6-7. Spectra in for Soda lime glass, undoped SnO2, and SiO2 layers adapted from Chen, 2010) for the spectroscopic ellipsometry analysis of TEC™-15 substrates based on the optical model shown in Figure 6-6.
147
The presence of relatively thick surface roughness layer on the SnO2:F makes a single layer inadequate to describe its optical effects. A second possible deviation from an ideal model is non-uniformity throughout the thickness of the SnO2:F layer due to an evolution in the grain structure or dopant atom incorporation during the growth process. Beginning from a simplified structure of a single SnO2:F layer and single surface roughness layer, the model was modified to better match experimental ellipsometric spectra. The bulk nominally 3000 Å thick SnO2:F in the range 0.04 to 0.73 eV was described by three distinct sublayers as plotted in Figure 6-8.
(Layer 5)
(Layer 4)
(Layer 3)
1 2 3 4 5
Photon energy (eV)
Figure 6-8. Spectra in in the high energy spectral range 0.734 to 5.0 eV for three
sublayers for thick 3000 Å SnO2:F layer of substrate TEC™-15 based on the optical model shown in Figure 6-6. 148
with the underlying layers SLG, undoped SnO2 and SiO2 extrapolated from database
(Chen, 2010). Four sublayers were used to simulate the surface roughness, such that the spectra in of each layer is simulated using the Bruggeman effective medium approximation of the top-most SnO2:F material (layer 5) and void. The topmost 870 Å thick SnO2:F layer (layer 5) shows higher energy electronic transition modeled using
Lorentz oscillator (Equation 4.3) at 4.866 eV, confirming it as a wide band gap TCO. The
IR spectra in of different layers in TEC™-15 glass are shown in Figure 6-9. The final deduced optical and structural parameters were presented in Table 6.3.
(Layer 5)
(Layer 4)
(Layer 3)
Figure 6-9. The IR spectra in for Layer 3, 4, and 5 of SnO2:F of the substrate TEC™- 15 in the energy range 0.04 to 0.73 eV at room temperature based on the optical model shown in Figure 6-6. 149
Table 6.3. Optical parameters are tabulated for the three incorporated SnO2:F layers for TEC™-15 substrate as determined in an analysis of ellipsometric spectra data 0.04 to 5 eV collected at room temperature using least squares regression analysis with an unweighted estimator error function, = 1.65 x 10-3. The variable parameters associated with these contributions are listed along with their confidence limits. Oscillator parameters without confidence limits are fixed in the analysis. of the SLG, SnO2 and
SiO2 layer was adopted from a database presented in previous work (Chen, 2010).
Layer Oscillator Oscillator’s parameters
A (unitless) (eV) Eo (eV) Lorentz 1.40 ± 0.02 0.725 ± 0.008 4.866 ± 0.003 5 Gaussian 0.81 ± 0.05 0.1770 ± 0.522 ± 0.002 0.0004 (SnO2:F) Gaussian 0.377 ± 0.009 0.257 ± 0.004 0.703 ± 0.003 Gaussian 29 ± 3 0.073 ± 0.005 0.073 ± 0.004 Gaussian 2.9 ± 0.2 0.010 ± 0.001 0.155 ± 0.009 Gaussian 84 ± 5 0.026 ± 0.001 0.051 ± 0.001 Sellmeier 183 ± 1 eV2 7.59 ± 0.02 Drude (x 10-4 cm) (fs) 2.70 ± 0.04 14.4 ± 0.3 A (unitless) (eV) Eo (eV) Gaussian 218 ± 12 0.01 ± 0.02 0.0790 ± 4 0.0002 Sellmeier 450 ± 20 eV2 11.3 ± 0.3 (SnO2:F) Drude (x 10-3 cm) (fs) 1.41 ± 0.02 1.42 ± 0.02
3 Lorentz A (unitless) (eV) Eo (eV) 4.2 ± 0.3 0.08 ± 0.02 3.992 ± 0.008 (SnO2:F)
The various Gaussian oscillators (Equation 6.1 and 6.2) used in the infrared range 0.03 to
0.09 eV (241.96 to 725.89 cm-1) can be ascribed to SnO and SnOSn vibrational modes (Banerjee et al., 2003; Du et al., 2005)The vibrational modes at 1250.16 cm-1 and
4210.20 cm-1 can be attributed to SnOH due to atmospheric moisture. We used the
150
model generated from IR-SE extended data of TEC™-15 in RTSE analysis to study the role of H2-dilution of a-Si:H in modifying the underlying transparent conducting oxide and to develop a model to apply it in ex situ IR-SE studies of the intrinsic a-Si:H on
TEC™-15 glass substrate. The model developed for characterization of TEC™-15 from
0.04 to 5.0 eV was applied in the analysis of RTSE data collected to study the role of modification of the TCO layer by plasma in the plasma enhanced chemical vapor deposition process and to determine the infrared optical response of a-Si:H in this device configuration. The R = 10 intrinsic a-Si:H layer has been fabricated using rf PECVD at
200oC onto 6 × 6 square inch TEC™-15 glass from the previous analysis in Table 6.3 and
Figure 6-6. Total process gas pressure of 0.45 Torr and a plasma power density of 0.017
W/cm2 was applied. The growth of ~3000 Å undoped a-Si:H layer has been monitored with real time spectroscopic ellipsometry using a multichannel rotating compensator instrument over a spectral range of 0.734 to 5.88 eV. In order to extract the temperature dependence of for the SnO2:F layer, the spectroscopic ellipsometric data measured at
o 200 C were analyzed. The thicknesses of the component layers, the SnO2:F/void percentages in the effective medium approximation layer were fitted. In addition, the amplitude of Lorentz oscillators in near infrared to ultraviolet range were fitted. For the analysis of the SnO2:F, the optical properties of the glass substrate and the SnO2 and SiO2 layer as used in IR-SE at the room temperature were employed. The effect of temperature variation is expected to be very weak and so is neglected in this study as SiO2 is large bandgap material (~9 eV) and is transparent over the spectral range of 0.6 to 6 eV. Table
6.4 shows parameters influencing spectra in over this spectral range that were fitted for data at 200oC. 151
Table 6.4. Optical property parameters for the four Bruggeman EMA and three incorporated SnO2:F layers for TEC™-15 as determined in an analysis of spectroscopic ellipsometry data 0.734 to 5 eV collected at 200oC using least squares regression analysis with = 10 x 10-3. The variable parameters associated with these contributions are listed along with their confidence limits. Oscillator parameters without confidence limits are fixed in the analysis.
Layer Structural parameters EMA 4 267.18 ± 1.50 Å, void 52.5% EMA 3 541.01 ± 6.40 Å, void 15.8% EMA 2 532.20 ± 6.46 Å, void 9.4% EMA 1 536.34 ± 7.3 Å, void 5.0% Oscillator Oscillator’s parameters
SnO2:F (Layer 5) A (unitless) (eV) Eo (eV) Lorentz 1.032 ± 0.06 0.725 4.866 db = 945.9 ± 10.3 Å Gaussian 0.81 0.1770 0.522 Gaussian 0.377 0.257 0.703 Gaussian 29 0.073 0.073 Gaussian 2.9 0.010 0.155 Gaussian 84 0.026 0.051 Sellmeier 184.01 ± 0.90 eV2 7.59 Drude (x 10-4 cm) (fs) 4.44 ± 0.13 9.2 ± 0.2 SnO2:F (Layer 4) A (unitless) (eV) Eo (eV) Gaussian 218 0.01 0.0790 db = 526.6 ± 6.4 Å Sellmeier 422.20 ± 2.91 eV2 11.3 Drude (x 10-3 cm) (fs) 1.97 ± 0.07 1.17 ± 0.03 SnO2:F (Layer 3) A (unitless) (eV) Eo (eV) Lorentz 1.25 ± 0.2 0.50 ± 0.02 3.992 db = 27.11 ± 5.4 Å
The room temperature spectra in of the SLG, SnO2 and SiO2 layer was adopted from a database presented in previous work (Chen, 2010). The layer thicknesses were allowed to vary to account for measurement on different spots over the sample surface in RTSE measurement for a similar 6 × 6 square inch TEC™-15 glass from previous analysis in
Table 6.3 and Figure 6-6.
152
Global -minimization analysis was applied to RTSE data after ~250 Å of accumulated bulk layer material to obtain a single set of ε for a plasma modified SnO2:F material and the deposited a-Si:H as shown in Figure 6-10. After obtaining spectra in , db and ds for intrinsic layer for each individual time points were allowed to vary independently. The optical parameters for plasma modified SnO2:F and coupled band gap
(Eg) of Cody-Lorentz oscillator representing a-Si:H were hold common over time.
Spectra in for the a-Si:H layer is linked to known relationships between Cody-Lorentz oscillator parameters, as discussed in Chapter 2, to reduce the number of fit parameters in the analysis.
Figure 6-10. Spectra in showing a plasma modified SnO2:F component of TEC™-15. RTSE data collected at 200oC temperature reveals that the outermost ∼500
Å of SnO2:F material in TCO sample was damaged by plasma exposure.
153
These optical constants were then used to extract the microstructural evolution of Si:H film with time. When a-Si:H film was deposited on the TCO, the optical properties of the outermost ~500 Å of SnO2:F layer of TCO was observed to be different from ε determined before the deposition. Specifically, ε2 featured a significant absorption peak shifted in the middle of the measured visible spectral range. The prolonged exposure to the hydrogen plasma could result in deteriorations of electrical and optical properties of the film such as increased absorption. X-ray photoelectron spectroscopy (XPS) analysis and Field Emission Scanning Electron Microscope (FESEM) observation of FTO surfaces, as reported by Choi et al. (Choi et al., 2013) show that the chemical reduction of SnO2 to Sn metallic state occurs on the surface region and is accompanied by etching.
The work proposes that a conformal over coating of a few nanometers thick highly durable plasma buffer layer on TCO may effectively improve the performance of a-Si:H
PV without changing the structure or process conditions for the cell fabrication. To avoid reoptimizing p-i-n a-Si:H PV processing, variations to the TCO can be employed to enhance PV device performance. Here, we have demonstrated that changes in optical properties of TCO by annealing and plasma depositions as shown in Table 6.5. These analysis techniques and observations can be used as a simple and fast method for identifying the properties of films and the modification of underlying layers in the device configuration. Coupled with IR-extended ellipsometric measurements, these tools potentially allow the plasma modification and bonding characteristics to be probed non- invasively for the intrinsic films in the respective complete PV device configuration. The
IR spectra shows stretching, bending, and wagging vibrations of Si:H and their complexes due to one or more H atom attached to an atom of Si. 154
TM Table 6.5. Drude parameters of the SnO2:F TCO layers in the TEC -15 substrates are tabulated as measured by spectroscopic ellipsometry (0.073 to 5 eV) at RT, 200oC, and plasma modified.
SnO2:F parameter Magnitude RT 200oC Plasma modified Resistivity () (104 cm) 2.70 ± 0.02 4.44 ± 0.13 3.8 ± 0.2 Scattering time 14.4 ± 0.3 9.2 ± 0.2 10.8 ± 0.4 Electron Conc. (N) (cm3) 3.48 x 1020 3.31 x 1020 3.29 x 1020 Mobility () 66.55 42.52 49.92 () (fs)
2 These type (cm of features/V s) are useful in the identification of confined environment of the bonded H atom. Room temperature IR-SE data for intrinsic R = 10 a-Si:H layer on a
TEC™-15 substrate were analyzed from 0.04 to 5 eV as explained in the Experimental section of this chapter. Table 6.6 shows the structural model used for analysis of a-Si:H layer on TECTM-15 glass. This model is adapted from RTSE study of the same sample to incorporate damage to the TCO from hydrogen and the PECVD process. The absorption peak arising from plasma modification and modeled by a Lorentz oscillator (Ferlauto et al., 2005) and other parameters were kept fixed for the TECTM-15. The top two interface layers with different sets of fitted volume fractions for the plasma modified SnO2:F/a-
Si:H layer/void components were modeled using Bruggeman effective medium approximation. Fit parameters consist of the bulk layer thickness of a-Si:H and all four interface layer thicknesses below it. Cody-Lorentz oscillator (Ferlauto et al., 2002) parameters for intrinsic a-Si:H were linked Eg, by linear relationships previously determined to minimize the number of fit parameters. The single fit bandgap parameter corresponds to 1.670 ± 0.004 eV in the analysis performed here.
155
Table 6.6. Structural model along with the values of the fixed parameters and the best fit parameters (the latter with confidence limits) obtained from the SE analysis (0.04 eV to 5 eV) of a R = 10 a-Si:H layer on TEC™-15 substrate. The substrate was fixed to a parameterized model discussed in Table 6.4 with plasmon feature in modified SnO2:F from RTSE analysis. Fit parameters consist of the bulk layer thickness of a-Si:H and all four interface layer thicknesses with their component material fraction below it. σ = 25.64 x 10-3 was obtained in this analysis.
a-Si:H layer surface roughness layer ds = 16 ± 1 Å; a-Si:H / void = 0.5/ 0.5 (EMA)
Voidy i-layer (coupled to a-Si:H layer) db = 172 ± 6 Å; (fvoid) = 0.110 ± 0.004
a-Si:H layer db = 3950 ± 13
C-L coupled bandgap 1.672 ± 0.004 eV
Gaussian oscillator A=1.4 ± 0.1; =0.011 ± 0.001 eV; Eo=0.079 ± 0.001 eV
Gaussian oscillator A=0.33 ± 0.02; =0.015 ± 0.001 eV; Eo=0.250 ± 0.001 eV
Gaussian oscillator A=0.426 ± 0.030; =0.199 ± 0.008 eV; Eo=0.377 ± 0.003 eV
Gaussian oscillator A=0.15 ± 0.02; =0.019 ± 0.002 eV; Eo=0.166 ± 0.001 eV
Gaussian oscillator A=0.12 ± 0.04; =0.066 ± 0.033 eV; Eo=0.120 ± 0.012 eV
Plasma modified SnO2:F/a-Si:H SnO2:F fraction = 0; i-layer fraction = 0.720 ± 0.001; void
layer/void interface di = 270 ± 6 Å fraction = 0.280 ± 0.001
Plasma modified SnO2:F/a-Si:H SnO2:F fraction 0.950 ± 0.001; i-layer fraction = 0.050 ± 0.001;
layer/void interface di = 575 ± 41 Å; void fraction = 0.000
SnO2:F/a-Si:H layer/void SnO2:F fraction = 0.920 ± 0.002; i-layer fraction = 0.000; void
interface di = 495 ± 52 Å; = 0.080 ± 0.001
SnO2:F/a-Si:H layer/void SnO2:F fraction = 0.960 ± 0.001; i-layer fraction = 0.000; void
interface di=540 ± 66 Å; fraction = 0.040 ± 0.001
Plasma modified SnO2:F layer 0 Å ( generated used in interfacial regions)
Drude = (3.8 ± 0.2) x 10-4 cm; = 10.8 ± 0.4 fs
Lorentz A = 1.182 ± 0.14 eV; = 0.647 ± 0.05 eV; E0 = 2.64 ± 0.01 eV
156
In order to improve the fit to the experimental data, a void-rich i-layer represented by a two component Bruggeman effective medium approximation was added between the surface roughness and 4000 Å thick i-layer with variable void fraction. The simultaneous analysis of ellipsometric spectra collected from the two instruments yielded db = 4101 ± 14 Å and db = 3810 ± 14 Å for the a-Si:H layer.
IR vibrational studies of a-Si:H on BR have remained useful in understanding the role of Si-H bonding in determining a-Si:H properties. IR-absorption studies have shown that hydrogen in a-Si:H is bonded as SiHn , with n = 1, 2, and 3. Lucovsky et al., reports that the methyl (CH3) and hydroxyl (OH) groups are common contaminants of air exposed metal surfaces (Lucovsky 1979). They also reports the shifts in the SiHn vibrational frequencies due to change in local chemical environment with respect to the bulk material formed at the film-substrate interface on top of a glass substrate. Figure 6-
11 shows spectra in for R = 10 a-Si:H deposited onto TEC™-15. Parameterization of was done using a Cody-Lorentz oscillator (Ferlauto et al., 2002) at high energies and
Gaussian oscillators (De et al., 2006) to represent the IR vibrational modes. Spectra in for a-Si:H on TEC™-15, a starting point for evaluating material in the p-i-n device configuration, exhibited bending modes near 0.079 eV (641 cm-1) and a stretching monohydride (SiH) mode around 0.250 eV (2016 cm-1). These modes are nearly the same as seen in a-Si:H on BR/n-type a-Si:H substrate. The feature exhibited by bending modes can be integrated and associated with the total hydrogen content, whereas the stretching mode region (around 2000 -2100 cm-1) can be correlated with the relative
-1 -1 amount of monohydride SiH (2000 cm ) and dihydride SiH2 (2100 cm ) bonding. In
157
addition to the expected SiH modes, the a-Si:H sample exhibited an absorption mode centered around 0.120 eV (968 cm-1), which relates to absorption reported in a-Si:H with small amounts of oxygen present (Freeman et al., 1978; Pollard et al., 1981).The corresponding peak observed here can be related to asymmetric SiOSi stretching or
SiOx due to presence of interstitial oxygen (Müllerová et al., 2006). The vibrational modes at 0.166 eV (1339 cm-1) and 0.377 eV (3041 cm-1) can be assigned to stretching modes of silanol (SiOH) group and are results of interaction with adsorbed water in the sample.
Figure 6-11. Spectra in for 3950 13 Å thick R = 10 a-Si:H films on TEC™-15 as a part of IR-extended SE analysis. The plot shows lower energy features in as function of photon energy in the 0.04 to 0.73 eV spectral range.
158
It is possible that some absorption in this region may also originate from SiF bonding
(Lucovsky 1979; Street 2005). Exact comparable features, however, are seen in sputtered a-Si:H films (Freeman et al.,1978, Shimada et al., 1980). The analysis performed for
Si:H layers on TCO does not show peaks centered at 845 to 910 cm-1 and ~2100 cm-1 assigned to the dihydride (SiH2) or clustered hydrogen. The presence of the peaks relating to high proportion of SiH2 bonds would, otherwise, result in poor quality of material.
6.4. Summary
Room temperature ex situ SE, IR-SE, over a wide spectral range (0.04 - 5 eV) spanning the mid-IR to near-UV was used to determine for the Ag/ZnO BR and
TEC™-15 glass components, and a-Si:H layers. Simultaneous analysis of ellipsometric spectra collected from multiple instruments was used to yield a common for each layer while structural parameters such as db and ds may be varied separately. By using various data analysis techniques, especially the coupling of all fit parameters to a single-bandgap value in a-Si:H, reduces the number of fitted parameters. A parametric approach combining experimental ellipsometric spectra ranges corresponding to “electronic transitions” and “vibrational transitions” behavior in the materials of interest been applied. Structural parameters extracted were common to both ranges but relied upon different, physically valid parameterizations of within each respective range.
Specific outcomes are identified with respect to each material system. The results of the analysis of IR-SE data were used to study the absorption in BR and TCO components, which can be used to extract electrical transport properties and phonon 159
modes. In addition, for intrinsic a-Si:H extracted from IR-SE data are sensitive to the
Si-H bonding. Comparison of optical absorption features affords a method of assessing film character, which then suggests ways to improve material quality and potentially device performance. In addition to being sensitive to the SiH related SiH2 absorption modes, IR-SE analysis is sensitive to absorption caused by more disordered or SiH2 modes, SiOSi and SiOH features resulting from unwanted environmental H2O, oxygen contamination, or even from plasma interactions with SnO2:F.
160
Chapter 7
Summary and Future Research
7.1 Summary
Thin film Si:H layers in solar cell configurations have been studied in detail with the aim to further optimize the a-Si:H material system in devices. This study makes the use of spectroscopic ellipsometry as the primary tool to characterize individual layer structure and optical response of different layers used in solar cells, especially in n-i-p configuration devices.
Although Si:H based multijunction PV devices can have high efficiencies, there is the economical need for faster and cheaper fabrication. We seek to increase the growth rate of Si:H using different reactive gas mixtures to simultaneously manipulate film structure and growth rate. In Chapter 3, we studied the growth evolution of Si:H films prepared by PECVD on smooth native oxide covered c-Si wafers where Si2H6 is added to the standard silane SiH4 gas mixture. RTSE has been used track the impact of using Si2H6 on the evolution of crystallinity, material deposition rate, and electronic quality of the material and to produce the growth evolution diagrams of Si:H films prepared under variable H2-dilution and silane-to-disilane ratio deposition conditions. Films with pure
161
SiH4 (S = 0) and SiH4 replaced with in part (S = 0.12) or completely (S = 1) by Si2H6 were deposited. The phase transitions aa, a(a+nc) and (a+nc)nc were observed with addition of Si2H6 indicating that the fundamental growth evolution is not affected.
The shifting of aa transition to a lower bulk layer thickness were observed for a range of R where material remains amorphous.
Optimized a-Si:H quality degrades with increasing S. On the other hand, R can be increased to maintain film quality. Increased rate is expected as more atomic silicon is present with Si2H6, however the film growth rate of optimized a-Si:H material appears to stabilize regardless of S. Small decreases in rate with addition of Si2H6 for optimized-
Si:H under these conditions may be due to etching of weakly bonded materials by increased atomic H in the plasma and could indicate inferior a-Si:H quality. Higher aa transition thicknesses at R below the optimum values for film prepared with S 0 indicates that some improvement in both a-Si:H film quality and growth rate can be achieved for non-optimum conditions, potentially widening the process window.
Nanocrystallite nucleation occurs more promptly at minimum values of R with addition of Si2H6 as compared to pure SiH4 films. Furthermore, comparing the structurally similar nc-Si:H material for pure Si2H6 R = 107 and pure SiH4 R = 40 shows that the film with
Si2H6 has 25% higher growth rate. The nc-Si:H films prepared at higher R and S can be deposited at higher rates and with a greater degree of crystallinity. This behavior indicates that the addition of Si2H6 to PECVD of Si:H can be used to potentially decrease the fabrication time and improve the material quality in nc-Si:H layers. These studies
162
have led to the development of deposition phase diagrams for fixed values of S with variable R as discussed in Chapter 3.
In Chapter 4, in situ RTSE was applied to generate deposition phase diagrams of the n-, i-layer, and p-layer on Ag/ZnO back reflectors or structures otherwise mimicking the n-i-p solar cell configuration. From these diagrams, the most ordered protocrystalline phases of all doped and undoped layers of single junction n-i-p could be confirmed. For example, the p-layer deposition process is critical as it controls key device performance parameters, but its optimized thickness is only ~120 Å. This implies a considerable optimization challenge, one that can be uniquely met by ellipsometry. Also for back reflectors, spectroscopic ellipsometry was used to study the metal, metal/TCO interfaces, and TCO materials in efforts to better understand the optical response of back-reflector components in thin film Si:H solar cells. The phase diagrams constructed in Chapter 4 include information on the phase transitions of doped and undoped Si:H layers in the device configuration from amorphous to mixed-phase and from mixed-phase to single- phase nc-Si:H as a function of accumulated thickness plotted as a function of R. The protocrystalline n-, i- and p-layers prepared with maximum H2 dilution prior to crystallite nucleation have been identified as the best such layers for optimum stabilized thin film a-
Si:H n-i-p solar cell with deposition conditions given in Table 4.1 and 4.2. For n- i- p- layers hydrogen dilution ratio of R = 50, 10 and 150 were identified for intended thicknesses ~ 200 Å, 3000 Å and 120 Å bulk thickness, respectively.
The trends in Cody gap with respect to R have been plotted for a-Si:H n-, i-, p- layers used in phase diagrams. The increase in n-type a-Si:H band gap with the increase of R is mostly due to the alloying effect of the amorphous Si matrix by hydrogen atoms. 163
Higher R suggests that more hydrogen atoms are incorporated that relax the compressive strain in the network. The band gaps of the i- layers deposited at R = 10 and 15, show a linearly decreasing trend with R. The lower value of band gap of R = 15 could be a result of enhanced density expected from extra H-etching or could be a stronger absorption near the band edge of a-Si:H due to nanocrystals appearance. At R > 15, the i-layer transitions to mixed-phase Si:H at a very early stage of the deposition and nc-Si:H for R > 15 has been obtained from VIA. These optical properties were valid only to 2.75 eV which is above the optical band gap of the bulk material of the intrinsic layers and would not necessarily be representative of the full spectral. The high band gap is always desirable for the p-type a-Si:H layer in device configuration. The depositions grown with R = 50 to
120 show a linear increase in band gap with increasing R. At R = 130 and 150, however, the trend decreases and becomes flat with very low rise in band gap energy with the increase in R. This consequence may be due to an increasing fraction of the nanocrystallites at even low volume fractions within the top-most portion of the first 200
Å deposited bulk material. The result could also be due to increased film density.
In Chapter 5, VIA has been applied to doped and undoped Si:H. The nc-Si:H optical properties as well as the evolution of fnc and fvoid on different substrates were extracted successfully. For Si:H films on native oxide covered c-Si, the film deposited with Si2H6 shows that the adding of Si2H6 requires more higher R to improve ordering.
This behavior is expected as each Si2H6 molecule possesses double the silicon as traditional SiH4, so the average silane radical to hydrogen ratio in the plasma is increased.
A comparison between n-type Si:H layer deposited on 1000 Å thick thermal oxide covered c-Si and on sputtered Ag/ZnO back reflectors revealed a short-term suppression 164
of nanocrystallinity which could be a result of formation of randomly oriented crystallites on the substrate surface or less energetically favorable orientations which subsequently are suppressed. R = 100 n-type Si:H deposited on high temperature processed ZnO does not show this suppression but inclusion of a-Si:H at the beginning of growth. The undoped Si:H deposited on n-type a-Si:H coated BR shows the substrate dependence of microstructural evolution. The stabilization of voids to about 5% after the coalescence of crystallites to end of the intrinsic Si:H growth could indicate that the grains under these conditions were not well passivated. This capability allows for better quantifying the nc-
Si:H structure which is critical for optimizing associated PV devices—namely well- passivated nanocrystallites with minimal voids are desired.
A detailed IR-SE extended analysis has been performed and presented in
Chapter 6 for Ag/ZnO back reflectors, a-Si:H on Ag/ZnO, TEC™-15, and a-Si:H on
TEC™-15. The effects of Si:H deposition on the underlying Ag/ZnO back reflector and
SnO2:F top layer of the TEC™-15 substrates have been explored. Throughout this
study, the substrate dependence of the a-Si:H properties has been explored in both in n-
i-p and p-i-n configuration of solar cells. The component layers of stack including Ag,
ZnO, SnO2:F exhibit concentrations of free electrons whose characteristics of
resistivity, scattering time, and plasma energy can be determined from IR-SE analysis.
Common features are discovered near the visible spectral range and higher energy
absorption due to electronic transitions in and R = 10 a-Si:H deposited on n-type coated
BR and TEC™-15 glass. The effect of plasma or plasma damage on Si:H overcoated
TCO substrate substrates have been explored too. Spectra in for a-Si:H in the n-i-p
device configuration exhibited modes near 0.079 eV (637.2 cm-1) and 0.249 eV (2008.3 165
cm-1). In addition to the expected SiH modes, this a-Si:H sample exhibited the bending
-1 or scissors mode of SiH2 dihydride centered around 0.106 eV (854.9 cm ).
Free carrier properties of the back reflector layers of n-i-p configuration solar cell have been extracted. A resistivity of 3.02 0.03 x 10-6 cm and a scattering time of 16.7
0.1 fs were determined from the Drude oscillator (Tiwald et al., 1998) parameters of
Ag film. The interface layer in between Ag and ZnO showed a low resistivity of 3.7 ± 0.5 x 10-5 cm and a scattering time of 2.7 ± 0.3 fs. These values indicate that the interface is less conductive than the bulk material as expected due to the high resistivity and disordered ZnO material on top. Combination of CPPB oscillators (Aspnes, 1980) for electronic transitions, Lorentz oscillators (Ferlauto et al., 2005) representing IR optical phonon modes, and a constant real additive term to account for dispersion from absorption features outside the measured spectral range have been used to represent for
-1 ZnO. The characteristic TO modes with A1 and E1 symmetry at 0.0468 eV (377.47 cm ) and 0.0506 eV (408.12.3 cm-1) were detected for ZnO in the sample coated with a-Si:H
-1 whereas only TO (E1) mode at 405 cm was present for the as-deposited BR. The refinement of grains after heating to 200oC could explain the resolution of these features.
The vibrational mode at 0.085 eV (683.15 cm-1) in the sample over coated with a-Si:H was identified as a LO mode with A1 symmetry.
When exposed to H2, heated, and coated with a-Si:H the topmost SnO2:F (Layer
5) of TEC™-15 shows plasmonic absorption modeled by a Lorentz oscillator (Ferlauto et al., 2005) at 4.866 eV and various SnO and SnOSn vibrational modes in the IR range from 0.03 to 0.09 eV (241.96 to 725.89 cm-1). The model of TECTM-15 was
166
improved by incorporating four Bruggeman effective medium approximation layers to simulate the surface roughness. Following the Si:H layer deposition, ε of this outer TCO material was observed to be showing a significant absorption peak in the middle of the measured spectral range at 2.64 eV as a source of the modification / damage to the TCO.
The plasma damage may reduce some of the SnO2:F, resulting in metallic tin-rich clusters at the interface that could enhance parasitic absorption and decrease device performance.
Spectra in for a-Si:H in the p-i-n device configuration exhibited bending modes near
0.0795 eV (641.2 cm-1) and a stretching monohydride (SiH) mode around 0.250 eV
(2016.4 cm-1), and an oxygen related absorption mode centered around 0.120 eV (967.9 cm-1).
7.2 Future Work
This thesis work describes comprehensive in situ real time and ex situ spectroscopic ellipsometry analyses of the component layers of thin film Si:H solar cells.
A robust optical model for the device structure and a database of spectra in at working temperature will help to guarantee the extraction of reliable physical parameters in subsequent analyses. Possible pathways to improve Si:H based PV performance related to this work are:
(1) The growth evolution of Si:H films deposited using Si2H6 additives has been reported in this dissertation. Phase diagrams for undoped Si:H were developed only for hydrogen to reactive gas ratio (R) with fixed SiH4 to Si2H6 ratio (S). The optimization of growth kinetics should be developed for the different deposition parameters of rf power density and total gas pressure. It is well known that the lowest rf power density value 167
possible for a stable plasma yields highest performance devices. Both lowest and highest gas pressure responsible for short-lived SiH2 radicals and polysilane radicals generated in the plasma may produce lower quality material. Hence, in addition to R, additional deposition parameter values should be selected so as to obtain the highest performance for the desired protocrystalline layers in solar cells. Films deposited under the conditions that yield the best solar cells should also be analyzed by a direct structural probe such as cross-sectional transmission electron microscopy (XTEM) to determine if the i-layer film incorporates any nanocrystalline component near its top surface. Single junction devices should be fabricated incorporating these layers.
(2) Introduction of rough silver in BR increase the short circuit current Jsc without degradation of the other performance parameters in the solar cells by maintaining the overall cell performance. For that purpose, Ag should be roughened in a controlled manner by increasing the deposition temperature above room temperature in the sputtering process to confirm a large grain size, which serves to reduce free carrier absorption. Depositing the initial ZnO layer on Ag at an elevated Ar pressure is expected to reduce the ZnO index of refraction to ~1.5 in the range of 700-800 nm. Reduction in the screening and shift in the Ag/ZnO interface absorption peak towards high energy
(blue shifted) were studied by (Sainju, 2015). The roughness of the Ag layer should be
~250 Å, as measured by RTSE to enhance scattering of light in back reflector for maximization of Jsc. In addition optimization of Si:H layers in a textured device configuration should be studied.
(3) Potential optimized conditions have been identified for n-i-p a-Si:H PV. The use of these optimized layers in a tandem or even a triple-junction thin film Si:H-based 168
solar cell is required to optimize PV technology. A controllable band gap engineering approach must developed to achieve a desirable band gap for the and middle i-layers in triple junction thin film Si:H-based solar cells. The nc-Si:H with a band gap around 1.1 eV as low as that of single crystal Si can serve as an improved bottom cell material.
The thickness of these individual stacked cells must be selected such that the same amount of current is produced in each junction. The deposition of the nanocrystalline i- layer for the bottom cell is facilitated by the desired nanocrystalline Si:H n-layer at the very bottom of the semiconductor stack. In fact, in the p-i-n superstrate configuration, an optimum solar cell fill-factor is obtained when the n-layer reaches a ~ 58 vol.% - 42 vol.% mixture of amorphous and nanocrystalline phases at its top surface (Huang et al.,
2013). The VIA routine developed here thus provides a powerful technique to accomplish this task.
(4) Simpler optimization strategies of device grade intrinsic a-Si:H material in n-i- p and p-i-n solar devices are highly desirable to reduce the time consuming optimization process. Here, we have demonstrated that changes in optical properties of transparent conducting oxides (TCOs) by heating in hydrogen or plasma depositions can be easily detected using RTSE analysis. These analysis techniques and observation can be used as a simple and fast optimization for the properties of high-rate deposited film in the device configuration. These tools would allow the plasma modification and bonding characteristics to be probed non-invasively for films in the respective complete photovoltaic device configuration over wide photon energy range and for deposition parameter variations. Comparison between the electrical performances and quantum efficiency (QE) of the devices studying the effect of plasma damage should be made for 169
better device performances. Moreover measurement of the hydrogen vibrational modes for a-Si:H materials in complete high (and low) efficiency PV device structures should also be performed.
170
References
Abelson, 1993 Abelson, J. R., “Plasma deposition of hydrogenated amorphous silicon: Studies of the growth surface.” Appl. Phys. A 56, 493 (1993).
Aberle et al., 2001 Aberle, A.G., Harder, N.-P., Oelting, S., “Formation of large-grained uniform poly-Si films on glass at low temperature”. Journal of crystal growth, 226, 209- 214 (2001).
Acciarri et al., 2005 M. Acciarri, S. Binetti, M. Bollani, A. Comotti, L. Fumagalli, S. Pizzini, H. Von Känel, “Nanocrystalline silicon film grown by LEPECVD for photovoltaic applications.” Solar energy materials and solar cells, 87, 11-24 (2005).
Ahn et al., 2012 Ahn, S., Lee, S., Lee, H., “Toward commercialization of triple-junction thin-film silicon solar panel with >12% efficiency.” 27th European Photovoltaic Solar Energy Conference, (2012).
Amor et al., 2014 Amor, S. B., M. Atyaoui, R. Bousbih, I. Haddadi, W. Dimassi and H. Ezzaouia. "Effect of substrate temperature on microstructure and optical properties of hydrogenated nanocrystalline Si thin films grown by plasma enhanced chemical vapor deposition." Solar Energy 108: p. 126 (2014).
An et al., 1990 An, I., Nguyen, H. V., Nguyen, N. V., and Collins, R. W., "Microstructural evolution of ultrathin amorphous-silicon films by real-time spectroscopic ellipsometry." Phys. Rev. Lett. 65, p. 2274 (1990).
An et al., 1991 An, I., Cong, Y., Nguyen, N. V., Pudliner, B. S., and Collins, R. W., "Instrumentation considerations in multichannel ellipsometry for real-time spectroscopy." Thin Solid Films 206, p. 300 (1991).
171
Aryal et al., 2011 Aryal, P., Chen, J., Huang, Z., Dahal, L. R., Sestak, M. N., Attygalle, D., Jacobs, R., Ranjan, V., Marsillac, S., and Collins, R. W., "Quantum efficiency simulations from on-line compatible mapping of thin-film solar cells," Proc. 37th IEEE Photovoltaics Specialists Conference proceedings, June 19-24, Seattle, WA, p. 2241 (2011).
Ashkenov et al., 2003 Ashkenov, N., B. N. Mbenkum, C. Bundesmann, V. Riede, M. Lorenz, D. Spemann, E. M. Kaidashev, A. Kasic, M. Schubert, M. Grundmann, G. Wagner, H. Neumann, V. Darakchieva, H. Arwin and B. Monemar. "Infrared complex dielectric function spectra and phonon modes of high-quality ZnO films." Journal of Applied Physics 93(1), p.126 (2003).
Aspnes, 1980 Aspnes, D. E., in: A. M. Balkanski (North-Holland, 1980) (Ed.) Handbook on semiconductors, pp. 125-127. “Optical properties of thin films.” Thin Solid Films 89, 249 (1982).
Aspnes, 1982 Aspnes, D. E., "Optical properties of thin films." Thin Solid Films 89, 249 (1982).
Aspnes, (1993) Aspnes, D., "Minimal‐data approaches for determining outer‐layer dielectric responses of films from kinetic reflectometric and ellipsometric measurements." Applied physics letters 62(4), 343 (1993).
Attygalle et al., (2010) Attygalle, O., Z. Huang, P. Koirala, P. Aryal, M. N. Sestak, L. R. Dahal, M. R. Mapes, C. Salupo and R. Collins., "Optical metrology of thin film solar cells from 0.2 to 30 µm." Proc. 35th IEEE Photovoltaics Specialists Conference proceedings, June 20-25, Honolulu, HI, (IEEE, Piscataway NJ) p. 631 (2010).
Azzam and Bashara, 1977 Azzam, R. M. A., and Bashara, N. M., "Ellipsometry and Polarized Light." Amsterdam, North Holland, 1977.
Bales et al., 1989 Bales, G. S., Redfield, A. C., and Zangwill, A., "Growth dynamics of chemical vapor-deposition." Phys. Rev. Lett. 62, 776 (1989).
Banerjee et al., (2003 Banerjee, A., S. Kundoo, P. Saha and K. Chattopadhyay, "Synthesis and characterization of nano-crystalline fluorine-type tin oxide thin films by sol-gel method." Journal of sol-gel science and technology 28(1), p.105 (2003). 172
Becquerel et al., 1839 Becquerel, A., "On electric effects under the influence of solar radiation." CR Acad. Sci 9: p. 711 (1839).
Beshkov et al., 2002 Beshkov, G., S. Lei, V. Lazarova, N. Nedev and S. Georgiev., "IR and Raman absorption spectroscopic studies of APCVD, LPCVD and PECVD thin SiN films." Vacuum 69(1): 301 (2002).
Beyer et al., 2007 Beyer, W., J. Hüpkes and H. Stiebig.,"Transparent conducting oxide films for thin film silicon photovoltaics." Thin Solid Films 516: 147 (2007).
Boyle, 2004 Boyle, G. Renewable energy, OXFORD university press. (2004)
Brewer et al., 2002 Brewer, S. H. and S. Franzen., "Optical properties of indium tin oxide and fluorine-type tin oxide surfaces: correlation of reflectivity, skin depth, and plasmon frequency with conductivity." Journal of Alloys and Compounds, 338: 73 (2002)
Brodsky et al., 1977 Brodsky, M., M. Cardona and J. Cuomo. "Infrared and Raman spectra of the silicon-hydrogen bonds in amorphous silicon prepared by glow discharge and sputtering." Physical Review B, 16(8): 3556 (1977).
Bundesmann et al., 2004 Bundesmann, C., N. Ashkenov, M. Schubert, A. Rahm, H. Wenckstern, E. Kaidashev, M. Lorenz and M. Grundmann. "Infrared complex dielectric function spectra and crystal orientation of a-plane ZnO thin films on r-plane sapphire determined by generalized ellipsometry." Thin Solid Films 455: 161 (2004).
Cao et al., 2008 Cao, X., J. A. Stoke, J. Li, N. J. Podraza, W. Du, X. Yang, D. Attygalle, X. Liao, R. W. Collins and X. Deng., "Fabrication and optimization of single-junction nc- Si: H n–i–p solar cells using Si: H phase diagram concepts developed by real time spectroscopic ellipsometry." Journal of Non-Crystalline Solids 354(19): 2397 (2008)
Capezzuto et al., 1995 Capezzuto, P. and A. Madan., "Plasma deposition of amorphous silicon-based materials." Academic Press, (1995)
173
Carlson and Wronski, 1976 Carlson, D. E., and Wronski, C. R., "Amorphous silicon solar cells." Appl. Phys. Lett. 28, 671 (1976).
Chapman, 1980 Chapman, Brian, Glow Discharge Processes, John Wiley & Sons, New York, 1980.
Chen et al., 2004 Chen, L.Y., W.H. Chen, J.J. Wang, F. C.N. Hong and Y.K. Su., "Hydrogen-type high conductivity ZnO films deposited by radio-frequency magnetron sputtering." Applied Physics Letters 85, 5628 (2004).
Chen et al., 2009a Chen, J., Li, J., Thornberry, C., Sestak, M.N., Collins, R.W., Walker, J.D., Marsillac, S., Aquino, A.R., and Rockett, A., “Through-the-glass spectroscopic ellipsometry of CdTe solar cells”, Conf. Record 34th IEEE PV Spec. Conf., Philadelphia, PA, 7-12 June 2009 (IEEE, New York NY, 2009) p. 1748.
Chen, 2010 Chen, J., "Spectroscopic ellipsometry studies of II-VI semiconductor materials and solar cells." Ph.D. Thesis, The University of Toledo, Toledo, 2010.
Cheng et al., 2000 Cheng, H.-C., Huang, C.-Y., Wang, F.-S., Lin, K.-H., Tarntair, F.-G., "Thin-film transistors with polycrystalline silicon films prepared by two-step rapid thermal annealing". Japanese Journal of Applied Physics, 39, L19 (2000).
Chittick et al., 1969 Chittick, R. C., Alexander, J. H., and Sterling, H. F., "The preparation and properties of amorphous silicon." J. Electrochem. Soc. 116, 77 (1969).
Choi et al., 2013 Choi, M.-S., Y.-J. Lee, J.-D. Kwon, Y. Jeong, J.-Y. Park, Y.-C. Kang, P.-K. Song and D.-H. Kim "Effects of hydrogen plasma treatment on SnO 2: F substrates for amorphous Si thin film solar cells." Current Applied Physics 13(8): 1589 (2013).
Collins and Yang, 1989 Collins, R. W., and Yang, B. Y., "In situ ellipsometry of thin-film deposition - implications for amorphous and microcrystalline Si growth." J. Vac. Sci. Technol. B 7, 1155 (1989).
Collins, 1990 Collins, R. W., "Automatic rotating element ellipsometer - calibration, operation, and real-time applications." Rev. Sci. Instrum. 61, 2029 (1990). 174
Collins, 1998 Collins, R. W., "Advances in multichannel spectroscopic ellipsometry." Thin Solid Films 313:18 (1998).
Collins et al., 2000 Collins, R. W. ; Koh, J.; Fujiwara, H.; Rovira, P. I. ; Ferlauto, A. S.; Zapien, J. A. ; Wronski, C.R.; and Messier, R., "Recent Progress in Thin Film Growth Analysis by Multichannel Spectroscopic Ellipsometry." App. Surf. Sci. 154, 217 (2000).
Collins et al., 2001 Collins, R. W., An, I., Lee, J., and Zapien, J. A., “Multichannel ellipsometry.” In: Tompkins, H. G. and Irene, E. A. (Eds.), Handbook of Ellipsometry. New York, Noyes, 2001.
Collins et al., 2003 Collins, R. W., Ferlauto, A. S., Ferreira, G. M., Chen, C., Koh, J., Koval, R., Lee, Y., Pearce, J. M., and Wronski, C. R., "Evolution of microstructure and phase in amorphous, protocrystalline, and microcrystalline silicon studied by real time spectroscopic ellipsometry." Solar Energy Mater. Solar Cells 78, 143 (2003).
Collins and Ferlauto, 2005 Collins, R. W., A. S. F.,"Handbook of ellipsometry." H. Tompkins and E. A. Irene, William Andrew, Norwich NY, pp. 159 / 125-129 (2005).
Dahal et al., 2009 Dahal, L. R., Sainju, D., Li, J., Podraza, N. J., Sestak, M. N., and Collins, R. W., "Comparison of Al/ZnO and Ag/ZnO Interfaces of Back-reflectors for Thin Film Si Photovoltaics." 34th IEEE Photovoltaics Specialists Conference, June 7-12, 2009, Philadelphia, PA (IEEE, Piscataway NJ, 2009), p. 1702.
Dahal et al., 2010 Dahal, L. R., Huang, Z., Attygalle, D., Sestak, M. N., Salupo, C., Marsillac, S., and Collins, R. W., “Application of real time spectroscopic ellipsometry for analysis of roll-to-roll fabrication of Si:H solar cells on polymer substrates.” 35th IEEE Photovoltaics Specialists Conference, June 20-25, 2010, Honolulu, HI, (IEEE, Piscataway NJ) p. 631 (2010).
Dahal, 2013 Dahal, Lila Raj, "Spectroscopic Ellipsometry Studies of Thin Film a-Si:H Solar Cell Fabrication by Multichamber Deposition in the n-i-p Substrate Configuration, Ph.D. Thesis, The University of Toledo, 2013.
Dahal et al., 2014 Dahal, L. R., J. Li, J. A. Stoke, Z. Huang, A. Shan, A. S. Ferlauto, C. R. Wronski, R. W. Collins and N. J. Podraza., "Applications of real-time and mapping 175
spectroscopic ellipsometry for process development and optimization in hydrogenated silicon thin-film photovoltaics technology." Solar Energy Materials and Solar Cells 129, 32 (2014).
De Sousa et al., 2006 De Sousa Meneses, D., M. Malki and P. Echegut., "Structure and lattice dynamics of binary lead silicate glasses investigated by infrared spectroscopy." Journal of Non-Crystalline Solids 352, 769, (2006).
Deng et al., 2003 Deng, X., and Schiff, E. A., “Amorphous silicon-based solar cells.” In: Luque, A. and Hegedus, S. (Ed.), Handbook of Photovoltaic Science and Engineering (pp. 505), New York, John Wiley and Sons, 2003.
Doshi et al., 1997 Doshi, P., G. E. Jellison and A. Rohatgi (1997). "Characterization and optimization of absorbing plasma enhanced chemical vapor deposited antireflection coatings for silicon photovoltaics." Applied optics 36, 7826 (1997)
Doughty et al., 1990 Doughty, D. A., Doyle, J. R., Lin, G. H., and Gallagher, A., "Surface-reaction probability of film-producing radicals in silane glow-discharges." J. Appl. Phys. 67, 6220 (1990).
Doyle et al., 1992 Doyle, J. R., Doughty, D. A., and Gallagher, A., "Plasma chemistry in disilane discharges ." J. Appl. Phys. 71, 4771-4780 (1992).
Du et al., 2005 Du, X., Y. Du and S. Georgea., "In situ examination of tin oxide atomic layer deposition using quartz crystal microbalance and Fourier transform infrared techniques." molecules 5, 581 (2005)
Dyer et al., 1993 Dyer, T., Marshall, J., Pickin, W., Hepburn, A., Davies, J., "Polysilicon produced by excimer (ArF) laser crystallisation and low-temperature (600°C) furnace crystallisation of hydrogenated amorphous silicon (a-Si: H)". Journal of non- crystalline solids, 164 1001-1004 (1993)
El Manouni et al., 2005 El Manouni, A., F. Manjón, M. Mollar, B. Marí, R. Gómez, M. López and J. Ramos-Barrado "Effect of aluminium doping on zinc oxide thin films grown by spray pyrolysis." Superlattices and Microstructures 39, 185 (2006). 176
Ferlauto et al., 2000 Ferlauto, A. S., Koh, J., Rovira, P. I., Wronski, C. R., Collins, R. W., and Ganguly, G., "Modeling the dielectric functions of silicon-based films in the amorphous, nanocrystalline and microcrystalline regimes." J. Non-Cryst. Solids 266-269, 269 (2000).
Ferlauto et al., 2000a Ferlauto, A. S., Rovira, P. I., Koval, R. J., Wronski, C. R., and Collins, R. W., "Study of the amorphous-to-microcrystalline transition during silicon film growth at increased rates: Extensions of the evolutionary phase diagram." Mater. Res. Soc. Symp. Proc. 609, A2.2 (2000).
Ferlauto, 2001 Ferlauto, A. S., "Phase Diagrams for Guiding Silicon Thin Film Deposition in Photovoltaics Applications as Derived by Spectroscopic Ellipsometry". Ph.D. Thesis, The Pennsylvania State University, University Park, 2001.
Ferlauto et al., 2002 Ferlauto, A. S., Ferreira, G. M., Pearce, J. M., Wronski, C. R., Collins, R. W., Deng, X., and Ganguly, G., “Analytical model for the optical functions of amorphous semiconductors from the near-infrared to ultraviolet: applications in thin film photovoltaics.” J. Appl. Phys. 92, 2424 (2002).
Ferlauto et al., 2004 Ferlauto, A. S., Ferreira, G. M., Koval, R. J., Pearce, J. M., Wronski, C. R., Collins, R. W., Al-Jassim, M. M., and Jones, K. M., “Evaluation of compositional depth profiles in mixed-phase (amorphous + crystalline) silicon films from real time spectroscopic ellipsometry.” Thin Solid Films 455-456, 665 (2004).
Ferreira et al., 2004 Ferreira, G. M., Ferlauto, A. S., Pearce, J. M., Wronski, C. R., Ross, C., and Collins, R. W., “Comparison of phase diagrams for vhf and rf plasma-enhanced chemical vapor deposition of Si:H films.” Mater. Res. Soc. Symp. Proc. 808, 215 (2004).
Fonrodona et al., 2006 Fonrodona, M., D. Soler, J. Escarré, F. Villar, J. Bertomeu, J. Andreu, A. Saboundji, N. Coulon, T. Mohammed-Brahim, "Low temperature amorphous and nanocrystalline silicon thin film transistors deposited by Hot-Wire CVD on glass substrate". Thin Solid Films, 501, 303-306 (2006).
Freeman and Paul, 1978 Freeman, E. and W. Paul., "Infrared vibrational spectra of rf-sputtered hydrogenated amorphous silicon." Physical Review B 18, 4288 (1978).
177
Fritzsche, 2001 Fritzsche, H. "Development in understanding and controlling the Staebler- Wronski effect in a-Si: H." Annual Review of Materials Research 31, 47 (2001).
Fujiwara et al., 1997 Fujiwara, H., J. Koh, C. Wronski and R. Collins "Application of real time spectroscopic ellipsometry for high resolution depth profiling of compositionally graded amorphous silicon alloy thin films." Applied physics letters 70, 2150, (1997).
Fujiwara et al., 1998 Fujiwara, H., J. Koh, C. Wronski, R. Collins and J. Burnham., "Parameterization of the optical functions of a-Si1-xCx:H: applications to C depth-profiling and surface temperature monitoring in solar cell preparation." Journal of non- crystalline solids 227, 460 (1998).
Fujiwara et al., 2000 Fujiwara, H., J. Koh, P. Rovira and R. Collins "Assessment of effective-medium theories in the analysis of nucleation and microscopic surface roughness evolution for semiconductor thin films." Physical Review B 61, 10832 (2000).
Fujiwara et al., 2001 Fujiwara, H., M. Kondo and A. Matsuda., "Real-time spectroscopic ellipsometry studies of the nucleation and grain growth processes in microcrystalline silicon thin films." Physical Review B 63,115306 (2001).
Fujiwara, 2007 Fujiwara, H., "Spectroscopic ellipsometry: principles and applications, John Wiley & Sons." (2007).
Funde et al., 2008 Funde, A. M., N. A. Bakr, D. K. Kamble, R. R. Hawaldar, D. P. Amalnerkar and S. R. Jadkar., "Influence of hydrogen dilution on structural, electrical and optical properties of hydrogenated nanocrystalline silicon (nc-Si:H) thin films prepared by plasma enhanced chemical vapour deposition (PE-CVD)." Solar Energy Materials and Solar Cells 92 1217 (2008).
Gallagher, 1988 Gallagher, A., "Neutral radical deposition from silane discharges." Journal of applied physics 63, 2406 (1988).
Ganguly and Matsuda, 1993 Ganguly, G. and A. Matsuda.,"Defect formation during growth of hydrogenated amorphous silicon." Physical Review B 47 3661 (1993)
178
Gautam et al., 2014 Gautam, L.K., Haneef, H., Junda, M., Saint John, D., Podraza, N., "Approach for extracting complex dielectric function spectra in weakly-absorbing regions." Thin Solid Films, 571 548-553 (2014).
Gevelber and Stephanopoulos, 1987 Gevelber, M. A. and G. Stephanopoulos., "Dynamics and control of the Czochralski process: I. Modelling and dynamic characterization." Journal of crystal growth 84 647 (1987).
Gordon, 2000 Gordon, R. G., "Criteria for choosing transparent conductors." MRS bulletin 25, 52 (2000).
Green et al., 2014 Green, M. A., Emery, K., Hishikawa, Y., Warta, W., and Dunlop, E. D., "Solar cell efficiency tables (version 43)," Prog. Photovolt.: Res. Appl. 22, p.1 (2014).
Green and Emery, 1993 Green, M. A. and K. Emery., "Solar cell efficiency tables." Progress in Photovoltaics: Research and Applications 1 25 (1993).
Guha et al., 2011 Guha, S., D. Cohen, E. Schiff, P. Stradins, P. Taylor, J. Yang, "Industry-academia partnership helps drive commercialization of new thin-film silicon technology"., Photovoltaic International, 134 (2011).
Herzinger et al., 1998 Herzinger, C., B. Johs, W. McGahan, J. Woollam and W. Paulson "Ellipsometric determination of optical constants for silicon and thermally grown silicon dioxide via a multi-sample, multi-wavelength, multi-angle investigation." Journal of Applied Physics 83 3323 (1998).
Honsberg and Bowden, 2010 Honsberg, C., and Bowden, S., www.pveducation.org/pvcdrom (2010).
Hsiao et al., 1999 Hsiao, H., H. Hwang, A. Yang, L. Chen, T. Yew, "Study on low temperature facetting growth of polycrystalline silicon thin films by ECR downstream plasma CVD with different hydrogen dilution". Applied surface science, 142, 316-321 (1999).
Huang et al., 2013 Huang, Z., Dahal, L.R., Salupo, C., Ferlauto, A.S., N.J. Podraza, R.W. Collins, Optimization of a-Si:H p-i-n solar cells through development of n-layer growth 179
evolution diagram and large area mapping, Photovoltaic Specialists Conference (PVSC), 2013 IEEE 39th, 2013, pp. 1788-1793.
Jackson, 1998 Jackson, J. D., "Classical Electrodynamics." New York, John Wiley & Sons, 1998.
James., 2010 James, F. A., "Michael Faraday: a very short introduction," Oxford University Press. (2010).
Jellison and Modine, 1996a Jellison, G. E. Jr., and Modine, F. A., “Parameterization of the optical functions of amorphous materials in the interband region.” Appl. Phys. Lett. 69, 371 (1996).
Jellison and Modine, 1996b Jellison, G. E. Jr., and Modine, F. A., Erratum: ‘‘Parameterization of the optical functions of amorphous materials in the interband region’’ Applied Physics Letters, 69 2137-2137 (1996).
Jellison, 1998 Jellison, G. E., "Spectroscopic ellipsometry data analysis: Measured versus calculated quantities." Thin Solid Films 313, 33 (1998).
Jellison and Boatner, 1998 Jellison, G.E., Boatner, L.A., "Optical functions of uniaxial ZnO determined by generalized ellipsometry." Physical Review B, 58, 3586-3589 (1998).
Johs et. al., 1999 Johs, B., Woollam, J. A., Herzinger, C., Hilfiker, J., Synowicki, R., and Bungay, C., “Overview of variable angle spectroscopic ellipsometry (VASE), Part II: Advanced applications.” Proc. Soc. Photo-Opt. Instrum. Eng. CR-72, 29, 1999.
Jung et al., 2005 Jung, M., J. Lee, S. Park, H. Kim and J. Chang., "Investigation of the annealing effects on the structural and optical properties of sputtered ZnO thin films." Journal of crystal growth 283, 384 (2005).
Kajzar et al., 1986 Kajzar, F., J. Messier and C. Rosilio., "Nonlinear optical properties of thin films of polysilane." Journal of applied physics 60, 3040 (1986).
180
Kang et al., 2004 Kang, H. S., J. S. Kang, J. W. Kim and S. Y. Lee.," Annealing effect on the property of ultraviolet and green emissions of ZnO thin films." Journal of Applied Physics 95, 1246 (2004).
Keyes et al., 2005 Keyes, B., L. Gedvilas, X. Li and T. Coutts., "Infrared spectroscopy of polycrystalline ZnO and ZnO: N thin films." Journal of crystal growth 281, 297 (2005).
Kim and Collins, 1995 Kim, S. and R. Collins., "Optical characterization of continuous compositional gradients in thin films by real time spectroscopic ellipsometry." Applied physics letters 67, 3010 (1995).
Kim et al., 1996 Kim, S., J. Burnham, J. Koh, L. Jiao, C. Wronski and R. Collins., "Real time spectroscopic ellipsometry characterization of optical gap profiles in compositionally‐graded semiconductor structures: Applications to bandgap engineering in amorphous silicon‐carbon alloy solar cells." Journal of applied physics 80, 2420 (1996).
Kim et al., 2010 Kim, D.H., S.H. Lee, G.H. Lee, H.B. Kim, K. H. Kim, Y.G. Lee and T.H. Yu., "Effects of deposition temperature on the effectiveness of hydrogen doping in Ga- type ZnO thin films." Journal of Applied Physics 108, 023520 (2010).
Kitao et al., 2001 Kitao, J., H. Harada, N. Yoshida, Y. Kasuya, M. Nishio, T. Sakamoto, T. Itoh, S. Nonomura and S. Nitta., "Absorption coefficient spectra of μc-Si in the low- energy region 0.4–1.2 eV." Solar energy materials and solar cells 66, 245 (2001).
Klein et al., 2002 Klein, S., Finger, F., Carius, R., Rech, B. Houben, L., Luysberg M., and Stutzmann, M., "High efficiency thin film solar cells with intrinsic microcrystalline silicon prepared by hot wire CVD." MRS Proceedings, Cambridge Univ Press 715 (2002).
Koh et al., 1995 Koh, J., Y. Lu, S. Kim, J. Burnham, C. Wronski and R. Collins., "Real time spectroscopic ellipsometry study of hydrogenated amorphous silicon p‐i‐n solar cells: Characterization of microstructural evolution and optical gaps." Applied physics letters 67, 2669 (1995).
181
Koh et al., 1999 Koh, J., A. Ferlauto, P. Rovira, C. Wronski and R. Collins. "Evolutionary phase diagrams for plasma enhanced chemical vapor deposition of silicon thin films from hydrogen-diluted silane." Applied physics letters 75, 2286 (1999)
Kojima et al., 2009 Kojima, A., Teshima, K., Shirai, Y., Miyasaka, T.," Organometal halide perovskites as visible-light sensitizers for photovoltaic cells." J. Am. Chem. Soc. 131, 6050-6051 (2009).
Koval et al., 2002 Koval, R., C. Chen, G. Ferreira, A. Ferlauto, J. Pearce, P. Rovira, C. Wronski and R. Collins "Maximization of the open circuit voltage for hydrogenated amorphous silicon nip solar cells by incorporation of protocrystalline silicon p- type layers." Applied physics letters 81, 1258 (2002).
Kroll et al., 2009 Kroll, U., Meier, J., Shah, A., Mikhailov, S., Weber, J., "Hydrogen in amorphous and microcrystalline silicon films prepared by hydrogen dilution". Journal of Applied Physics, 80 4971-4975 (1996).
Kryukov et al., 2009 Kryukov, Y., N. Podraza, R. Collins and J. Amar., "Experimental and theoretical study of the evolution of surface roughness in amorphous silicon films grown by low-temperature plasma-enhanced chemical vapor deposition." Physical Review B 80, 085403 (2009).
Langford et al., 1992 Langford, A., M. Fleet, B. Nelson, W. Lanford and N. Maley., "Infrared absorption strength and hydrogen content of hydrogenated amorphous silicon." Physical Review B 45, 13367 (1992).
Lautenschlager et al., 2004 Lautenschlager, P., M. Garriga, L. Vina and M. Cardona., "Temperature dependence of the complex dielectric function spectra and interband critical points in silicon." Physical Review B 36, 4821 (1987).
Leblanc et al., 1994 Leblanc, F., J. Perrin and J. Schmitt., "Numerical modeling of the optical properties of hydrogenated amorphous‐silicon‐based p‐i‐n solar cells deposited on rough transparent conducting oxide substrates." Journal of Applied Physics 75, 1074 (1994)
182
Lee et al., 1998 Lee, J., P. Rovira, I. An, Collins R.W., "Rotating-compensator multichannel ellipsometry: Applications for real time Stokes vector spectroscopy of thin film growth." Review of scientific instruments 69, 1800 (1998)
Li et al., 1992 Li, Y. M., An, I., Nguyen, H. V., Wronski, C. R., and Collins, R. W., "Thin-film coalescence in hydrogenated amorphous-silicon probed by spectroscopic ellipsometry with millisecond-scale resolution." Phys. Rev. Lett. 68, 2814 (1992).
Lu et al., 1991 Lu, G. Q., E. Nygren and M. J. Aziz., "Pressure‐enhanced crystallization kinetics of amorphous Si and Ge: Implications for point‐defect mechanisms." Journal of applied physics 70, 5323 (1991). Lu et al., 1994 Lu, Y., S. Kim, M. Gunes, Y. Lee, C. Wronski and R. Collins., "Process-property relationships for a-Si 1-x C x: H deposition: excursions in parameter space guided by real time spectroellipsometry." MRS Proceedings, Cambridge Univ Press (1994).
Lucovsky, 1979 Lucovsky, G., "Chemical effects on the frequencies of Si-H vibrations in amorphous solids." Solid State Communications 29, 571 (1979)
Luft and Tsuo., 1993 Luft, W. and Y. S. Tsuo., "Hydrogenated amorphous silicon alloy deposition processes." (1993).
Luque and Hegedus, 2011 Luque, A. and S. Hegedus., "Handbook of photovoltaic science and engineering." John Wiley & Sons (2011).
Mack, 2011 Mack, C. "Fifty years of Moore's law." Semiconductor Manufacturing, IEEE Transactions on 24, 202 (2011).
Major et al., 1986 Major, S., S. Kumar, M. Bhatnagar and K. Chopra., "Effect of hydrogen plasma treatment on transparent conducting oxides." Applied Physics Letters 49, 394 (1986).
Mazor et al., 1988 Mazor, A., Srolovitz, D. J., Hagan, P. S., and Bukiet, B. G., "Columnar growth in thin-films." Phys. Rev. Lett. 60, 424 (1988).
183
Matsuda, 1998 Matsuda, A., "Plasma and surface reactions for obtaining low defect density amorphous silicon at high growth rates." Journal of Vacuum Science & Technology A 16, 365 (1998).
Matsuda et al., 1990 Matsuda, A., K. Nomoto, Y. Takeuchi, A. Suzuki, A. Yuuki and J. Perrin., "Temperature dependence of the sticking and loss probabilities of silyl radicals on hydrogenated amorphous silicon." Surface Science 227, 50 (1990).
Matsuda and Tanaka, 1986 Matsuda, A. and K. Tanaka., "Investigation of the growth kinetics of glow‐discharge hydrogenated amorphous silicon using a radical separation technique." Journal of applied physics 60, 2351 (1986).
Matsui et al., 2013 Matsui, T., H. Sai, K. Saito, M. Kondo, "High‐efficiency thin‐film silicon solar cells with improved light‐soaking stability". Progress in Photovoltaics: Research and Applications, 21, 1363-1369 (2013).
Meier et al., 1994 Meier, J., R. Flückiger, H. Keppner and A. Shah., "Complete microcrystalline p‐i‐n solar cell—Crystalline or amorphous cell behavior?" Applied Physics Letters 65, 860 (1994).
Meier et al., 1996. Meier, J., P. Torres, R. Platz, S. Dubail, U. Kroll, J. A. Selvan, N. P. Vaucher, C. Hof, D. Fischer and H. Keppner, "On the way towards high efficiency thin film silicon solar cells by the “Micromorph” concept". MRS Proceedings, Cambridge Univ Press. (1996).
Müllerová et al., 2006 Müllerová, J., S. Jurečka and P. Šutta (2006). "Optical characterization of polysilicon thin films for solar applications." Solar Energy 80, 667 (2006).
Muthmann et al., 2011 Muthmann, S., Köhler, F., Meier, M., Hülsbeck, M., Carius, R., Gordijn, A., "Monitoring of the growth of microcrystalline silicon by plasma-enhanced chemical vapor deposition using in-situ Raman spectroscopy". physica status solidi (RRL) – Rapid Research Letters, 5, 144-146 (2011).
Nasuno et al., 2002 Nasuno, Y., M. Kondo and A. Matsuda "Key issue for the fabrication of high- efficiency microcrystalline silicon thin-film solar cells at low temperatures." Japanese journal of applied physics 41, 5912 (2002). 184
Nelson, 2003 Nelson, J., "The physics of solar cells." World Scientific. (2003)
O'Dowd, 1987 O'Dowd, J. G., "A model for light scattering by rough tin oxide." 30th Annual Technical Symposium, International Society for Optics and Photonics. (1987).
Oldham, 1969 W. Oldham, "Numerical techniques for the analysis of lossy films". Surface Science, 16, 97-103 (1969).
Özgür et al., 2005 Özgür, Ü., Alivov, Y.I., Liu, C., Teke, A., Reshchikov, M., Doğan, S., Avrutin, V., Cho, S.-J., Morkoc, H., "A comprehensive review of ZnO materials and devices," Journal of applied physics, 98, 041301(2005).
Palmer and Gordon, 1988 Palmer, B. J., and Gordon, R. G., "Local equilibrium-model of morphological instabilities in chemical vapor-deposition." Thin Solid Films 158, 313 (1988).
Park, 2013 Park, N. G., "Organometal perovskite light absorbers toward a 20% efficiency low-cost solid-state mesoscopic solar cell." J. Phys. Chem. Lett. 4, 2423– 2429 (2013).
Pearce et al., 2007 Pearce, J. M., Podraza, N. J., Collins, R. W., Al-Jassim, M. M., Jones, K. M., Deng, J. and Wronski, C. R., “Optimization of open circuit voltage in amorphous silicon solar cells with mixed-phase (amorphous+nanocrystalline) p-type contacts of low nanocrystalline content”, J. Appl. Phys., 101, 114301 (2007).
Perrin et al., 1989 Perrin, J., Y. Takeda, N. Hirano, Y. Takeuchi and A. Matsuda., "Sticking and recombination of the SiH3 radical on hydrogenated amorphous silicon: The catalytic effect of diborane." Surface Science 210, 114 (1989).
Perrin, 1995 Perrin, J.,"4 Reactor Design for a-Si: H Deposition." Plasma Deposition of Amorphous Silicon-Based Materials." 177 (1995).
Perrin et al., 1998 Perrin, J., M. Shiratani, P. Kae-Nune, H. Videlot, J. Jolly and J. Guillon., "Surface reaction probabilities and kinetics of H, SiH3, Si2H5, CH3, and C2H5 during
185
deposition of a-Si: H and aC: H from H2, SiH4, and CH4 discharges." Journal of Vacuum Science & Technology A 16, 278 (1998).
Philipp, 1998 Philipp, H. R.,"Handbook of Optical Constants of Solids." E. D. Palik, Academic Press, San Diego CA, 1, 774 (1998).
Podraza et al., 2006 Podraza, N., C. Wronski and R. Collins., "Model for the amorphous roughening transition in amorphous semiconductor deposition." Journal of non-crystalline solids 352, 950 (2006).
Podraza et al., 2006a Podraza, N. J., Wronski, C. R., Horn, M. W., and Collins, R. W.; “Complex dielectric function spectra of a-Si1-xGex:H versus Ge content, temperature, and processing: advances in optical function parameterization.” Materials Research Society Symposium Proceedings: Amorphous and Polycrystalline Thin-Film Silicon Science and Technology - 2006, (MRS, Warrendale, 910, p. 259 (2006)
Podraza et al., 2008 Podraza, N., J. Li, C. Wronski, E. Dickey, M. Horn and R. Collins., "Analysis of Si1-xGex:H thin films with graded composition and structure by real time spectroscopic ellipsometry." physica status solidi (a) 205, 892 (2008).
Podraza et al., 2009 Podraza, N., J. Li, C. Wronski, E. Dickey and R. Collins., "Analysis of controlled mixed-phase (amorphous+microcrystalline) silicon thin films by real time spectroscopic ellipsometry." Journal of Vacuum Science & Technology A 27, 1255 (2009).
Pollard and Joannopoulos, 1981 Pollard, W. and J. Joannopoulos., "Vibrational properties of amorphous silicon alloys." Physical Review B 23, 5263 (1981).
Prasad et al., 1991 Prasad, K., F. Finger, S. Dubail, A. Shah and M. Schubert., "Deposition of phosphorus doped microcrystalline silicon below 70 C at 70 MHz." Journal of non-crystalline solids 137, 681(1991).
Puchert et al., 1996 Puchert, M., P. Timbrell and R. Lamb., "Postdeposition annealing of radio frequency magnetron sputtered ZnO films." Journal of Vacuum Science & Technology A 14, 2220 (1996).
186
Robertson, 2000 Robertson, J., "Deposition mechanism of hydrogenated amorphous silicon." Journal of Applied Physics 87 2608 (2000).
Robertson et al., 2000a Robertson, J., "Growth mechanism of hydrogenated amorphous silicon." Journal of non-crystalline solids 266: 79-83 (2000).
Rolo et al., 2007 Rolo, A. G., J. A. de Campos, T. Viseu, T. de Lacerda-Arôso and M. Cerqueira.,"The annealing effect on structural and optical properties of ZnO thin films produced by rf sputtering." Superlattices and Microstructures 42, 265 (2007).
Rossnagel, et al., 1990 Rossnagel, M. S., Cuomo, J. J., and Westwood, W. D., Handbook of Plasma Processing Technology, Noyes Publication, Park Ridge, New Jersey, 1990.
Rovira and Collins, 1999 Rovira, P. and R. Collins., "Analysis of specular and textured SnO2:F films by high speed four-parameter Stokes vector spectroscopy." Journal of applied physics 85 2015 (1999).
Rovira, 2000 Rovira, P. I., "Real Time Stokes Vector Spectroscopy and its Application to the Characterization of Inhomogeneous Thin Films." Ph.D. Thesis, The Pennsylvania State University, University Park, 2000.
Rovira et al., 2000 Rovira, P.I., Ferlauto, A.S., Koval, R.J., Wronski, C.R., Collins, R.W., and Ganguly, G., “Real time optics of p-type microcrystalline silicon deposition on specular and textured ZnO-coated glass”, Mater. Res. Soc. Symp. Proc. 609, A.19.6.1 (2000).
Sainju et al., 2006 Sainju, D., Oever, P. J., Podraza, N. J., Syed, M., Stoke, J. A., Chen, J., Yang, X., Deng, X., and Collins, R. W., “Origin of optical losses in Ag/ZnO back-reflectors for thin film Si photovoltaics.” Proc. 4th World Conf. PV Energy Conversion, (IEEE, Piscataway NJ, 2006), p. 1732.
Sainju, 2015 Sainju, D., "Spectroscopic Ellipsometry Studies of ZnO Layers and Interfaces for Thin Film Photovoltaics". Ph.D. Thesis in progress, The University of Toledo, Toledo Ohio, 2015.
187
Saint John et al., 2011 Saint John, D., H.B. Shin, M.Y. Lee, S. Ajmera, A. Syllaios, E. Dickey, T. Jackson and N. Podraza., "Influence of microstructure and composition on hydrogenated silicon thin film properties for uncooled microbolometer applications." Journal of Applied Physics 110, 033714 (2011).
Sai et al., 2014 Sai, H. T. Matsui, K. Matsubara, M. Kondo, I. Yoshida, "11.0%-Efficient Thin- Film Microcrystalline Silicon Solar Cells with Honeycomb Textured Substrates". Photovoltaics, IEEE Journal, 4 1349-1353 (2014).
Sakata et al., 1993 Sakata, I., M. Yamanaka and Y. Hayashi., "Properties of hydrogenated amorphous silicon prepared by alternatively repeating chemical‐vapor deposition from disilane and hydrogen plasma treatment." Journal of applied physics 74, 2543 (1993).
Saleh and Nickel, 2003 Saleh, R. and N. Nickel., "Raman spectroscopy of B-type microcrystalline silicon films." Thin Solid Films 427, 266 (2003).
Schropp et al., 2002 Schropp, R., Y. Xu, E. Iwaniczko, G. Zaharias and A. Mahan., "Microcrystalline silicon for solar cells at high deposition rates by hot wire CVD." MRS Proceedings, Cambridge Univ Press (2002).
Scott et al., 1982 Scott, B., J. Reimer, R. Plecenik, E. Simonyi, W. Reuter, "Low defect density amorphous hydrogenated silicon prepared by homogeneous chemical vapor deposition". Applied Physics Letters, 40, 973-975 (1982).
Shah et al., 2002 Shah, A., J. Meier, E. Vallat-Sauvain, C. Droz, U. Kroll, N. Wyrsch, J. Guillet and U. Graf., "Microcrystalline silicon and ‘micromorph’ tandem solar cells." Thin Solid Films 403, 179 (2002).
Shah et al., 2003 Shah, A., J. Meier, E. Vallat-Sauvain, N. Wyrsch, U. Kroll, C. Droz and U. Graf., "Material and solar cell research in microcrystalline silicon." Solar Energy Materials and Solar Cells 78, 469 (2003).
Shah et al., 2004 Shah, A., H. Schade, M. Vanecek, J. Meier, E. Vallat‐Sauvain, N. Wyrsch, U. Kroll, C. Droz and J. Bailat., "Thin‐film silicon solar cell technology." Progress in photovoltaics: Research and applications 12, 113 (2004). 188
Shimada et al., 1980 Shimada, T., Y. Katayama and S. Horigome., "Infrared Spectra of Amorphous Silicon-Fluorine Alloys Prepared by Sputtering in Fluorosilane-Argon Gas Mixture." Japanese Journal of Applied Physics 19, L265 (1980).
Shirai et al., 1991 Shirai, H., J.i. Hanna and I. Shimizu., "Role of atomic hydrogen during growth of hydrogenated amorphous silicon in the “chemical annealing”." Japanese journal of applied physics 30, L679 (1991).
Shirai et al., 2002 Shirai, H., Y. Fujimura, S. Jung, "Formation of nanocrystalline silicon dots from chlorinated materials by RF plasma-enhanced chemical vapor deposition." Thin Solid Films, 407, 12-17 (2002).
Smets et al., 2003 Smets, A., W. Kessels and M. Van de Sanden.,"Vacancies and voids in hydrogenated amorphous silicon." Applied physics letters 82, 1547 (2003)
Smets et al., 2008 Smets, A., T. Matsui and M. Kondo., "Infrared analysis of the bulk silicon- hydrogen bonds as an optimization tool for high-rate deposition of microcrystalline silicon solar cells." Applied Physics Letters 92, 33506 (2008).
Spear and Comber, 1975 Spear, W. and P. Le Comber., "Substitutional doping of amorphous silicon." Solid State Communications 17 1193 (1975).
Staebler and Wronski, 1977 Staebler, D. and C. Wronski., "Reversible conductivity changes in discharge‐produced amorphous Si." Applied Physics Letters 31 292 (1977).
Stoke et al., 2008 Stoke, J. A., N. J. Podraza, J. Li, X. Cao, X. Deng and R. W. Collins "Advanced deposition phase diagrams for guiding Si: H-based multijunction solar cells." Journal of Non-Crystalline Solids 354, 2435 (2008).
Stoke et al., 2008b Stoke, J. A., Dahal, L., J. Li, N. J. Podraza, X. Cao, X. Deng, R. Collins, "Optimization of Si: H multijunction nip solar cells through the development of deposition phase diagrams". Photovoltaic Specialists Conference, 2008. PVSC'08. 33rd IEEE, IEEE, pp. 1-6 (2008). 189
Stoke, 2008c Stoke, J. A., "Spectroscopic ellipsometry analysis of the component layers of hydrogenated amorphous silicon triple junction solar cells". M.S. Thesis, The University of Toledo, Toledo OH, 2008.
Street, 2005 Street, R. A., "Hydrogenated amorphous silicon." Cambridge University Press.
Sze and Kwok, 2006 Sze, S. M. and K. K. Ng., "Physics of semiconductor devices." Wiley. com (2006)
Takai et al., 2000 Takai, M., Nishimoto, T., Takagi, T., Kondo, M., and Matsuda, A., "Guiding principles for obtaining stabilized amorphous silicon at larger growth rates." J. Non-Cryst. Solids 266, 90 (2000).
Tanaka, 1989 Tanaka, K. (Ed.), Glow-discharge Hydrogenated Amorphous Silicon. Boston, Kluwer, 1989.
Temple-Boyer et al., 2010 Temple-Boyer, P., Rousset, B., Scheid, E., "Influences of deposition and crystallization kinetics on the properties of silicon films deposited by low- pressure chemical vapour deposition from silane and disilane". Thin Solid Films, 518, 6897-6903 (2010).
Thiart et al., 2000 Thiart, J. J., Hlavacek, V., and Viljoen, H. J., "Chemical vapor deposition and morphology problems." Thin Solid Films 365, 275 (2000).
Tiwald et al., 1998 T.E. Tiwald, D.W. Thompson, J.A. Woollam, W. Paulson, R. Hance, "Application of IR variable angle spectroscopic ellipsometry to the determination of free carrier concentration depth profiles". Thin Solid Films, 313, 661-666 (1998).
Toyoshima et al., 1991 Toyoshima, Y., Arai, K., Matsuda, A., and Tanaka, K., "In situ characterization of the growing a-Si:H surface by IR spectroscopy." J. Non-Cryst. Solids 137, 765 (1991). Toyoshima, Y., Arai, K., Matsuda, A., and Tanaka, K., "In situ characterization of the growing a-Si:H surface by IR spectroscopy." J. Non-Cryst. Solids 137, 765 (1991).
190
Tsai et al., 1986 Tsai, C. C., Knights, J. C., Chang, G., and Wacker, B., "Film formation mechanisms in the plasma deposition of hydrogenated amorphous-silicon." J. Appl. Phys. 59, 2998 (1986).
Tsu et al., 1997 Tsu, D. V., Chao, B. S., Ovshinsky, S. R., Guha, S., and Yang, J., "Effect of hydrogen dilution on the structure of amorphous silicon alloys." Appl. Phys. Lett. 71, 1317 (1997).
Tzolov et al., 2000 Tzolov, M., N. Tzenov, D. Dimova-Malinovska, M. Kalitzova, C. Pizzuto, G. Vitali, G. Zollo and I. Ivanov., "Vibrational properties and structure of undoped and Al-type ZnO films deposited by RF magnetron sputtering." Thin Solid Films 379, 28 (2000).
USEIA, 2014 U.S. Energy Information Administration, 2014, Annual Energy Outlook 2014 with projections to 2040, (2014).
Vetterl et al., 2000 Vetterl, O., Finger, F., Carius, R., Hapke, P., Houben, L., Kluth, O., Lambertz, A., Muck, A., Rech, B., and Wagner, H., “Intrinsic microcrystalline silicon: A new material for photovoltaics”. Solar Ener. Mater. Solar Cells, 62, pp. 97-108 (2000).
Wang et al., 2003 Wang, Y., J. Lin and C. Huan. "Structural and optical properties of a-Si: H/nc-Si: H thin films grown from Ar–H2–SiH4 mixture by plasma-enhanced chemical vapor deposition." Materials Science and Engineering: B 104, 80 (2003)
Wan et al., 2011 Wan Z., Huang, S., Green, M.A., Conibeer, G., "Rapid thermal annealing and crystallization mechanisms study of silicon nanocrystal in silicon carbide matrix". Nano Res Lett, 6 129 (2011).
Wasa and Hayakawa, 1992 Wasa, K. and S. Hayakawa., "Handbook of Sputter Deposition Technology: Principles." Technology, and Applications (Park Ridge: Noyes) (1992).
Wotten, 2013 Wooten, F., "Optical properties of solids, Academic press (2013).
191
Wronski et al., 2004 Wronski, C. R. and Collins, R. W., “Phase engineering of a-Si:H solar cells for optimized performance.” Solar Energy 77, 877 (2004).
Yan et al., 2011 Yan, B., Yue, G., Sivec, L., Yang, J., Guha, S., Jiang, C.-S. "Innovative dual function nc-SiOx:H layer leading to a >16% efficient multi-junction thin-film silicon solar cell". Applied Physics Letters, 99, 113512 (2011).
Yang, et al., 1997 Yang, J., Banerjee, A., and Guha, S., "Triple-junction amorphous silicon alloy solar cell with 14.6% initial and 13.0% stable conversion efficiencies." Appl. Phys. Lett. 70, 2975 (1997).
Yang et al., 2003 Yang, J., Benerjee, A., and Guha, S., “Amorphous silicon based photovoltaics, - from earth to the final frontier.” Solar Energy Material & Solar Cell 78, 597-612 (2003).
192