High Performance Inverted Pyramidal Texture for Silicon Photovoltaics

Kitty Kumar

A thesis submitted in conformity with the requirements for the degree of Doctorate of Philosophy Department of Materials Science and Engineering University of Toronto

Kitty Kumar

Doctorate of Philosophy

Department of Materials Science and Engineering University of Toronto

2013

Abstract

An inverted pyramidal grating texture is known to reduce both surface reflection and to promote light trapping in crystalline silicon (c-Si) solar cells. However, these textures are not used in commercial solar cells mainly because of high fabrication costs, limited scalability of conventional fabrication techniques to thin wafers, and insufficient knowledge of the optimum grating parameters for silicon of different thicknesses. These issues become even more important as industry makes a transition to thinner Si wafers to reduce device cost.

The objective of this this thesis is to address all of these issues. Firstly, a new process for inverted pyramidal texturing of c-Si has been developed that is compatible with thin silicon wafers and foils. Secondly, a theoretical study of the optimum inverted pyramidal grating parameters has been done for a wide range of Si wafer thicknesses. Finally, the optical performance of the optimal textures has been experimentally verified.

The laser assisted texturing method produces high quality inverted pyramids in a non- cleanroom environment and is potentially scalable to mass production. Because of the contactless fabrication, the method can be used to texture fragile, ultra-thin Si foils. This ii approach also offers precise control of the patterned areas, which can exclude areas for front surface contacts on PV devices.

The wave-optical study of the size dependent performance of inverted pyramidal textures identifies a 1000 nm period as being universally optimal for silicon thicknesses ranging from 2 – 400 m.

As a point of comparison, inverted pyramidal textures were also fabricated by electron beam lithography. The measured reflectances show that textures with micron scale periodicity outperform a submicron periodic texture, in agreement with trends predicted by the wave-optical simulations.

Finally, a novel phenomenon of internal structuring within an optically transparent thin film is described that was discovered in the course of optimizing the laser processing parameters for making apertures in the hard mask layer for PV surface texturization. This phenomenon shows promise for varied applications such as marking of surfaces, and the production of buried channels within a thin film for lab on chip architectures.

iii

To my Parents

Dr. Jonas Salk said “Good parents give their children Roots and Wings. Roots to know where home is, Wings to fly away and exercise what's been taught them.”

I am lucky to have such great parents. Thank you Mom and Dad for Everything!

Acknowledgements

First and foremost, I would like to sincerely thank my supervisors, Professors Jun Nogami,

Nazir P. Kherani and Peter R. Herman, for giving me this wonderful opportunity to conduct my graduate studies in their world-class teams. I am grateful for their constant guidance, feedback and support throughout my doctoral study.

I would like to thank Kenneth K.C. Li for sharing his knowledge and experience in laser processing. Without his knowledge and assistance in laser processing this work would not have been possible. I would also like to express my gratitude towards Dr. Jianzhao Li for

ICCD experiments and Ali Khalatpour for simulation work.

I am also grateful to the University of Toronto, the Department of Materials Science and

Engineering, Hatch Ltd., Natural Sciences and Engineering Research Council of Canada, and the Ontario Research Fund – Research Excellence for their financial support.

Table of Contents

ABSTRACT ...... II ACKNOWLEDGEMENTS ...... V TABLE OF CONTENTS ...... VI LIST OF TABLES ...... VII LIST OF FIGURES ...... VIII LIST OF APPENDIX FIGURES ...... XIV CHAPTER 1 INTRODUCTION ...... 1

1.1 SOLAR ENERGY ...... 1 1.2 CRYSTALLINE SILICON SOLAR CELLS ...... 4 1.3 SURFACE TEXTURING ...... 5 1.4 GRATED SURFACE TEXTURES FOR LIGHT-TRAPPING ...... 7 1.5 RESEARCH OBJECTIVES ...... 9 1.6 THESIS OUTLINE ...... 10 CHAPTER 1 REFERENCES ...... 12 CHAPTER 2 QUANTIZED LASER PROCESSING OF THIN DIELECTRIC FILMS ON SILICON ... 16

2.1 INTRODUCTION ...... 16 2.1.1 Laser Processing of Transparent Dielectric Films ...... 17 2.2 FEMTOSECOND LASER-MATERIAL INTERACTION IN TRANSPARENT DIELECTRIC FILMS ...... 18 2.2.1 Theoretical Calculation of Electron Density inside the Film...... 19 2.3 QUANTIZED INTERNAL STRUCTURING AND EJECTION OF A TRANSPARENT FILM ...... 24 2.3.1 Film Morphology as a function of Laser Exposure ...... 25 2.4 VERIFICATION OF QUANTIZED FILM EJECTION WITH AN INTENSIFIED CCD CAMERA ...... 28 2.5 LARGE AREA PROCESSING OF FILMS ...... 31 2.6 SUMMARY ...... 34 2.7 DISCUSSION ...... 35 2.8 EXPERIMENTAL METHODS ...... 35 CHAPTER 2 REFERENCES ...... 38 CHAPTER 3 FEMTOSECOND LASER DIRECT HARD MASK WRITING FOR SELECTIVE FACILE MICRON-SCALE INVERTED-PYRAMID PATTERNING OF SILICON ...... 41

3.1 INTRODUCTION ...... 41 3.2 LASER PATTERNING OF A HARD MASK ...... 42 3.3 TEXTURING METHOD ...... 43 3.4 OPTIMIZATION OF SINGLE PULSE LASER EXPOSURE AND HARD MASK THICKNESS FOR HIGH RESOLUTION WRITING ...... 47 3.5 HARD MASK WRITING WITH MULTIPLE PULSE EXPOSURE ...... 51 3.6 FACILE PATTERNING FOR LARGE AREA INVERTED PYRAMIDAL TEXTURE, V-GROOVES AND RESERVOIRS53 3.7 DISCUSSION ...... 54 3.8 CONCLUSIONS AND OUTLOOK ...... 56 CHAPTER 3 REFERENCES ...... 57 CHAPTER 4 WAVE-OPTICAL STUDY OF INVERTED PYRAMIDAL TEXTURE ON THICK SILICON WAFERS AND ULTRA-THIN FOILS ...... 60

4.1 INTRODUCTION ...... 60 4.1.1 Surface Textures for Enhanced Light Trapping ...... 60 4.1.2 Inverted Pyramidal Grating Texture...... 61 4.2 WAVE OPTICAL STUDY OF INVERTED PYRAMIDAL TEXTURE ON THICK C-SI WAFERS AND THIN FOILS ...... 62 4.2.1 Optical Loss in Bare 2 – -Si wafers ...... 64 4.2.2 Optimal Front Texture Parameters for Maximum Absorption in 2 – -Si wafers ...... 67 4.2.3 Optical Loss Due to Mesas ...... 70 4.2.4 Optimal Antireflective Coating (ARC) Parameters for the Texture with Minimal 100 nm Wide Mesas ...... 72 4.3 SUMMARY ...... 74 CHAPTER 4 REFERENCES ...... 76 CHAPTER 5 EXPERIMENTAL AND THEORETICAL STUDY OF REFLECTANCE FROM SUB- MICRON TO WAVELENGTH SCALE PERIODIC TEXTURES ...... 79

5.1 INTRODUCTION ...... 79 5.2 REFLECTANCE FROM SUB-MICRON TO WAVELENGTH SCALE PERIODIC TEXTURES ...... 81 5.3 INVERTED PYRAMIDAL TEXTURING OF C-SI WAFERS WITH ELECTRON BEAM LITHOGRAPHY ...... 85 5.3.1 Effect of E-Beam Dose on Etched Inverted Pyramidal Texture ...... 89 5.4 MEASURED AND SIMULATED REFLECTANCE FROM FABRICATED SAMPLES WITH AND WITHOUT ANTIREFLECTIVE COATING ...... 93 5.5 COMPARISON OF REFLECTANCE FROM SAMPLES FABRICATED BY LASER HARD MASK PATTERNING AND E- BEAM LITHOGRAPHY ...... 95 5.6 SUMMARY ...... 96 CHAPTER 5 REFERENCES ...... 98 CHAPTER 6 SUMMARY ...... 101 APPENDIX A EXTENSION OF QUANTIZED INTERNAL STRUCTURING OF TRANSPARENT THIN FILMS TO LARGER AREAS ...... 106 APPENDIX B ANISOTROPIC ETCHING OF SILICON THROUGH LASER WRITTEN DIELECTRIC HARD MASK ...... 112

APPENDIX B REFERENCES ...... 121 APPENDIX C WAVE OPTICAL SIMULATION METHOD ...... 122

APPENDIX C REFERENCES ...... 124

List of Tables

Table 5-1 The summary of inverted pyramid parameters and etched area for each EBL dose. .. 91

vii

List of Figures

Figure 1-1 Three year average of solar irradiance at different locations of the world, including nights and cloud coverage. TWe = Terawatt electrical power 2.The black dots indicate six potential sites for the installation of solar cells with only 8% efficiency that can collectively meet the current global energy demand of approximately 18 TWe...... 3 Figure 1-2 AM1.5 solar spectral irradiance6...... 3 Figure 1-3 Trend of thickness of as-cut silicon wafer thickness in mass production of solar cells9. Green, yellow and red represent the technology that is currently in production, exist in the market but is not currently in production, and for which the industrial solution is as yet unknown, respectively...... 5 Figure 2-1 A cartoon (a) depicting the division of the focal interaction volume of incident

522 nm laser light into thin interaction zones (orange discs) inside a 500 nm thick Si3N4 film as a result of the interference of Fresnel reflections of the incident light from the boundary interfaces. (b) The modulated intensity profile calculated for a Gaussian-shaped beam of 0.495 m (1/e2) radius (line) together with the electron density profile (shaded) expected inside the film at the threshold exposure of 9×1012 W/cm2 average incident

intensity. The electron density exceeds the critical plasma density (ncr) at the fringe maxima positions that manifests in the internal structuring of the film observed in cross- sectional and oblique top SEM views in (c). The threshold intensity results in the ejection of segments S1, S2 and blistering of fused segments S3 and S4 to form a nano-void at the fourth fringe maximum, and blistering of S5 overlying a second nanovoid. The positions of the cleavage planes align with the fringe maxima positions as depicted by red lines. The thickness of laser-plasma zones (FWHM of the electron density profile) at different times

during the () during the laser = 200 fs long pulse (d) shows the minimum width at

criticalfs that corresponds to the time at which the rate of impact ionization (wimp, green curve) starts diminishing at maxima positions relative to low intensity positions within the interference maxima while the multiphoton transition (wmpi) remains constant

viii

(blue curve) (e). The reduction in the rate of impact ionization occurs at critical when the effective collision frequency of electrons exceed the laser frequency (f)...... 23

Figure 2-2 The quantized ejection of multiple segments in a 945 nm Si3N4 film as observed with increasing fluence for a Gaussian-shaped beam to a maximum available fluence of 21.65 J/cm2, shown in top and cross-sectional SEMs (a) (i-viii). Slightly above threshold at 4.46 J/cm2 fluence (a (i)), 1st and 2nd segments are ejected and blistered, respectively,

together with damage at the Si3N4-silicon interface. The sequence of blistering, puncturing and ejection of each segment with increase in laser fluence is summarized graphically in (b) by the threshold fluences observed to form a solid blister (purple), a punctured blister (yellow), and an ejected blister (orange-red). Each segment was found to align between the cleavage planes (horizontal dashed lines) predicted by the positions of the Fabry-Perot intensity maxima shown graphically on the right...... 27 Figure 2-3 Time-resolved ICCD images (a) of ablation plumes recorded transversely from a 2 500 nm Si3N4 film exposed to 13.67 J/cm fluence and the top and cross-sectional SEMs (b) of the film after exposure. The plume emissions (a) appear in several clusters associated with the ejection of segments S1 and S2 (green) in the 3 - 9 ns window, partial ejection of S3 (orange) and S4 (red) in 173 - 193 ns and 233 - 393 ns window, respectively, and ejection of S5 (blue) in 393 - 1493 ns window. The observed position of clusters (c) from the surface as a function of time...... 30 Figure 2-4 A schematic (a) of a multifunctional device design consisting of combinations of optical, nanofluidic, and MEMs components over a large area together with SEM (gray) and optical (color) images of sample components constructed inside a film by interferometric laser processing. Uniform ejection of S1 (i), S2 (ii), or S3 (iii) segments over a large area by raster scanning a top hat laser beam that represent formation of single level reservoirs (R) and open serpentine channels (SC). The film color in the S1, S2 and S3 ejection zones was shifted from green (insets in (i) – (iii)). Different level ejections can also be combined to create multi-level reservoirs (R), mixing channels with pillars (v), and optical components such as a blazed grating (vi) and a Fresnel lens (vii). Nano-cavities at the 3rd Fabry Perot fringe position were stitched together (vii) to represent the writing of buried nanofluidic channels. A large area membrane structure (M) is anticipated with a large beam diameter. Optical image of a multi-component device (b) showing a Fresnel

lens, grating, single and multilevel reservoirs (R), and open serpentine (SC), crossed (CC) and mixing (Mixer) channels, fabricated with interferometric laser fabrication...... 33 Figure 3-1 A schematic illustrating (a) selective removal of thin dielectric film coated on crystalline silicon (001) by 522 nm femtosecond laser ejection to form a pattern of open apertures and (b) the resulting inverted-pyramid structure following anisotropic etching of exposed silicon. Variation in the laser written aperture pattern spacing, laser exposure and KOH etching time yield (c) V-channels, V-reservoirs, and arrays of inverted pyramids of diverse sizes as seen following HF removal of the dielectric mask...... 44 Figure 3-2 (a) SEM top views ((i) and (iii)) and ion-milled cross-section views ((ii) and (iv)) of the laser ejected mask aperture and ablation crater produced in 70 nm ((i) and (ii)) and -2 -2 20 nm ((iii) and (iv)) thick silicon nitride (SiNx) film at 0.52 Jcm and 0.45 Jcm fluence,

respectively. (b (i)) SEM images of laser modifications in a 20nm SiNx film on crystalline silicon (c-Si) with increasing laser fluence showing (left to right) blistering, blistering with a nanohole, collapsed blister and ejected blister together with the corresponding (b (ii)) atomic force micrographs and (b (iii)) line profiles...... 48

Figure 3-3 AFM micrographs of apertures produced in 100 nm thick (i) PECVD SiNx (ii)

of ~ 2 m in all three films. In (i) PECVD SiNx film remains partially attached to the aperture wall after ejection. Such partially opened apertures do not hinder anisotropic wet etching of the exposed silicon...... 49

Figure 3-4 The aperture diameter in 20 nm thick SiNx hard-mask and the size of inverted- pyramid together with respective error bars (rms values) observed after 2.5 minutes KOH etching are shown in (a) together with select SEM images as a function of fluence. The amount of mask undercutting after KOH etching decreases with the increase in laser fluence due to increase in laser induced c-Si substrate damage. Fluence zones for blistering

and high (~ 99.9%) and low (~33%) reproducibility ejection of SiNx are identified. (b) SEM views of (i) partially etched c-Si in KOH for 1 minute. The formation of fresh (111) planes is impeded by laser induced damage which is finally removed after 2 minutes of KOH etching to form (ii) clean 1.13 m inverted pyramid shown in (iii) finished with HF

removal of dielectric mask. The inverted pyramid size can be further reduced to 623 nm by lower the fluence (0.33 J cm-2) and compensating with the increase in number of pulses (10 pulses) as shown in b (iv)...... 52 Figure 3-5(a) SEM top views of a crystalline silicon wafer with a grid of laser ejected apertures at 1.5 μm periodic spacing and following (b) KOH etching, yielding an array of 1.3 μm inverted-pyramids. (c) Optical micrograph of University of Toronto crest pattern crest patterned with areas of high and low reflectivity corresponding to un-patterned and inverted-pyramid structure as seen in the SEM micrograph (outlined in red). (d) SEM images of 4x4 and 2x1 inverted pyramid arrays that have coalesced into a V-reservoir (left) and V-channel (right), respectively, as controlled by the laser pattern and KOH etching time...... 54 Figure 4-1 Schematic depiction of the solar cell architectures investigated in this work. A planar silicon slab in air (a) without and (b) with a perfect back surface reflector (BSR, orange), with front inverted pyramid texture (c) without mesa; filling fraction = 1 (d) with mesa; fill fraction <1 (e) with mesa and an antireflective coating (ARC, blue) and BSR with front side illumination...... 64 Figure 4-2 (a) The spectral reflectance (R), transmittance (T) and R+T loss in 400 m thick wafer. Spectral loss due to reflectance and transmittance in 2 - 400 m thick silicon (b) without and (c) with a perfect back surface reflector. (d) Absorption of AM1.5 spectrum in 2- 400 m thick silicon with a perfect reflector at rear surface...... 67 Figure 4-3 (a) Simulated photocurrent in silicon of thickness 2- 400 m with perfect back reflector and a front inverted pyramidal grating of periodicity ranging from 100 – 2000 nm. (b) Maximum photocurrent at grating period  ~ 1000 nm (wide solid bars) in 2- 400 m thick silicon and  = 850, 680 and 650 nm in 10, 5 and 2 m thick silicon, respectively, (narrow faded bars) corresponding to peak positions in (a). Solid lines above the bars represent Yablonovich limit for each silicon thickness...... 70 Figure 4-4 Simulated bsorption of AM1.5 solar spectrum in 400 m thick silicon above the band gap of silicon as a function of grating period () and filling fraction (inverted pyramid size/ grating periodicity) of the front inverted pyramidal grating texture...... 71 Figure 4-5 Degradation in photocurrent with the increase in mesa width in grating with optimum periodicity  ~ 1000 nm for various crystalline silicon thicknesses...... 72 xi

Figure 4-6 Simulated photocurrent in 2- 400 m thick silicon with 1000 nm periodic texture with 100 nm wide mesas coated for different thicknesses of antireflective coating (ARC) of refractive index 2.1 ...... 73 Figure 4-7 Simulated photocurrent current in 2 - 400 m thick silicon with optimum front texture with no mesas ( ) and with 100 nm mesa without ( ) and with 80 nm ARC ( ), and Lambertian front texture (×) ...... 74 Figure 5-1 1.5 m periodic array of 1.3 m inverted pyramids in top and side view, produced by laser patterning and chemical etching...... 81 Figure 5-2 Simulated reflectance from the  = 100 nm to  = 4000 nm periodic texture of inverted pyramids with no mesas. The simulated reflectance at each periodicity shows a sharp reduction in the reflectance at wavelengths around /p, where p is an integer, for

example, shown at 500 nm, 1000 and 1500 nm periodicity (dashed lines)...... 82 Figure 5-3 Simulated reflectance from the (a) 500 nm, (b) 1000 nm and (c) 1500 nm periodic texture of inverted pyramids with no mesas...... 84 Figure 5-4 SEM images of the sample after 9 minutes RIE and 45 seconds of KOH etching...... 87 Figure 5-5 (a) High and (b) low resolution SEM images of the new sample after 9 minutes of RIE, photoresist removal and 75 seconds of KOH etching...... 88 Figure 5-6 SEM images of a sample after 9 minutes of RIE, 75 seconds of KOH etching

and SiNx layer removal with BOE at (a) high and (b) low resolution...... 88

Figure 5-7 High (left) and low (right) resolution SEM images of SiNx sample prepared with electron beam doses of (a) 180 C/cm2 (b) 320C/cm2 and (c) 480 C/cm2 followed by KOH etching...... 90 Figure 5-8 Average length-to-width ratio of fabricated inverted pyramids as a function of EBL dose ...... 92 Figure 5-9 Average percentage of etched area as a function of EBL dose...... 92 Figure 5-10 Simulated (dashed lines) and measured (solid lines) total reflectance for three samples at the three different periodicities () and feature sizes (s) with calculated Solar weighted reflectance (SWR) from simulated and experimental reflectance spectra in inset...... 94

xii

Figure 5-11 Simulated (dashed lines) and measured (solid lines) total reflectance for three

samples at the three different periodicities and feature sizes coated with 70 nm SiNx film used as an antireflective coating...... 95 Figure 5-12 Spectral total reflectance measured from  = 1500 nm periodic textures fabricated by laser and electron beam lithography with feature size 1300 nm and 1400 nm, respectively, for the case with and without and antireflective coating...... 96

xiii

List of Appendix Figures

Video A1 depicts the interferometric internal structuring of a 500 nm thin transparent film on high index substrate with laser exposure greater than the threshold intensity. The exposure results in sequential ejection of three segments from the top and blistering of a fourth

segment close to the Si3N4-silicon interface, all separated by the /2nfilm Fabry-Perot fringe spacing...... 107

Figure A2 shows cross-sectional SEM views of a 500 nm thick Si3N4 film exposed to a top- hat beam profile for fluences of (i) 93.5 mJ/cm2, (ii) 140.2 mJ/cm2, (iii) 303.8 mJ/cm2, and (iv) 436.2 mJ/cm2. The threshold fluence of 93.5 mJ/cm2 shows (i) the onset of blistering for the first ~ 29 nm thick segment of the film, which is seen ejected at the higher fluence exposure in (ii). Segment 1 and 2 are both removed at the higher fluences as shown in (iii) and (iv), yielding a more uniform morphology in contrast with the case of Gaussian beam exposure shown in Figure2-1 - 3...... 108

Figure A3 shows SEM images of 500 nm thick Si3N4 film exposed with uniform-square beam profile of 396 mJ/cm2 (a) and 339 mJ/cm2 (b) fluence on hexagonal patterns varying with spot-to-spot offsets of 0.64 - 0.8 m ((i) - (v))...... 109

Figure A4 Spectral reflectance calculated for as a function of thin Si3N4 film thickness. Vertical cyan dashed lines highlight the reflected wavelength at 500 nm (full film thickness) and 471 nm (film thickness after first ejection of the first segment) which match well to the observed colors, green and red, respectively, under the optical microscope (inset) and Figure 4-4 (a) (i)...... 111

Figure B1 shows cross-section of typical etch profiles in (110) and (100) wafers after (a) isotropic etching and (b) anisotropic etching...... 113 Figure B2 Scanning electron micrograph (SEM) of 22.5 m x 22.5 m laser patterned sample after continuous etching in 30 wt% KOH maintained at 60ᵒC for 2.5 minutes. The sample etched non-uniformly in to regions of clean inverted pyramids, smooth inverted pyramids, and incomplete inverted pyramids and over etched regions, e.g. enclosed by dashed square, polygon and rectangle, respectively. The trench outlining the patterned

region is the etched marker made by removing SiNx mask at higher laser fluence...... 114

xiv

Figure B3 Shows the result of etching a laser patterned sample in two steps in 30 wt% KOH at 60ᵒC. The exposed silicon craters were etched in to complete inverted pyramids with smooth (111) planes, with a few sites consisting of fused neighbouring inverted pyramids (enclosed by dashed rectangles). Inset shows magnified view of inverted pyramids formed

in silicon under a 20 nm thick SiNx hard mask (outlined by dashed circle) used for selective etching...... 115 Figure B4 Shows the result of etching a sample in two steps in 30 wt% KOH at 75ᵒC. The etching resulted in a pattern of incomplete inverted pyramids. Fresh (111) planes were formed closer to the top of silicon crater, i.e. in the region with minimum laser damage, however, their growth was inhibited by high laser induced damage and debris inside the crater...... 116 Figure B5 Shows the result of etching a sample in two steps in 30 wt% TMAH at 60ᵒC. The exposed silicon craters were etched into incomplete inverted pyramids and fused over etched region. Inset shows magnified image of the incomplete inverted pyramids with flat base in a pattern with ~ 170 nm wide mesa...... 117 Figure B6 Schematic of wet etching process flow to achieve high density pattern of high fidelity inverted pyramids from a pattern of laser ablated silicon craters...... 119 Figure B7 Large area optical images of etched samples with inverted pyramids textures with (a) 300 nm (b) 200 nm mesa width. Formation of clusters (inside dashed circles) in (b) suggests the fusion of neighbouring inverted pyramids...... 120

Figure C1 The scattering matrix method approximate real grating as a stack of rectangular gratings (a). Each layer is treated as a separate grating (periodic in one or two dimensions) with its own effect on the incoming and outgoing fields (b). The grating is expressed in terms of periodic material properties such as refractive index and permittivity...... 123 Figure C2 Reflected AM1.5 power at 450 nm wavelength from 1000 nm periodic inverted pyramidal texture on 400 micron wafer as a function of number of modes...... 123

Chapter 1 Introduction

There is an inevitable shift towards renewable energy in current times. According to an

International Energy Agency (IEA) World Energy Outlook report1, 81% of global energy consumption in 2011 was generated from fossil fuel sources. It is projected that global energy consumption will increase by 33% by the year 2035 in comparison with 2011, resulting in a 3.6 ᵒC rise in long-term global average temperature because of carbon dioxide emissions from fossil fuels. Emerging economies like China, India, and the

Middle East contribute 60% towards the total rise in energy demand with the improving living standards in these countries.

1.1 Solar Energy

Solar energy is by far the most available permanent energy source on Earth. The Earth intercepts about 173000 TW power from the sun, of which ~ 50% (86000 TW) reaches the surface of our planet. It is estimated that only 8% photovoltaic (PV) conversion efficiency solar cells installed at six locations as indicated by black dots on the world map shown in Figure 1-12, would generate enough energy to power the whole world.

With current advances in technology and knowledge, it is thus only natural that PV materials are further explored to harness the sun’s energy that ultimately provides the energy for all life on our planet. It will still be some time before solar energy will reach

1 grid parity1 everywhere, as the price of solar power strongly depends on climate as well as availability and cost of solar technology (cost per Watt). However, due to the ruggedness of solar panels as well as the stand alone practicality of such systems, rural places that do not readily have access to electricity can benefit the most. Developing countries stand to benefit the most and hence scientific and technological advances in PV knowledge can be a huge turning point for such places in the long run.

Over the last decade, the solar industry is growing at an average annual rate of

30% underpinned by progress in technology, continuous increases in cell efficiency, lowering of module costs, increasing government subsidies and growing prices of grid electricity, with the highest growth of 66% in developing Asian markets in 20123, 4. Many

PV materials have been studied for effective conversion of solar into electrical energy.

Of the various PV materials, silicon is one of the most widely used and investigated for this purpose. Silicon has a band gap of 1.1 eV, which results in the ability to absorb solar energy below a wavelength of 1127 nm. Figure 1-2 shows the solar radiation spectrum which has a maximum energy output in the visible range of the spectrum, hence the reason for silicon’s viability as the material of choice in photovoltaic cells. In 1961,

Shockley and Queisser predicted that a maximum 30% efficiency can be achieved in a single p-n junction crystalline silicon solar cell limited by its band gap5.

1 Grid parity is the ratio of price to power at which energy produced with renewable means is equal to or lower than the price of power from the electricity grid. Standard grid parity target is installed cost of $1/ Wp, where Wp, watt-peak, is the nominal power of a solar device measured under standard illumination conditions. 2

Figure 1-2 AM1.5 solar spectral irradiance6.

1.2 Crystalline Silicon Solar Cells

Since the demonstration of p-n homojunction based crystalline silicon PVs by Chapin,

Fuller and Pearson in 19547, remarkable technological progress has been made by the solar industry, to the extent that the production of crystalline silicon based solar cells of

150 to 200 m thickness with efficiencies exceeding 16 to 20% is routine and that the cost of the generated electricity is approaching or below time-of-day retail pricing in many regions of the world. Today wafer based crystalline silicon (c-Si) based technology holds 85% of the PV market share and rest is shared by new thin-film based technologies8. Notwithstanding these advances, industry continues to demand even lower cost per watt solar cells so that PV electricity can ultimately compete with conventional sources of energy. One viable approach in attaining this goal is to markedly reduce the wafer thickness while maintaining the energy conversion efficiency of the device9.

Figure 1-3 shows the projected thickness of silicon wafers up to year 2020. The advantage of thinner wafers is that they suffer lower bulk recombination losses and hence offer opportunity to replace high quality expensive silicon wafers by low grade crystalline silicon with short carrier diffusion lengths, which will further reduce the material cost.

In this direction, a number of processes have also been developed to effectively produce ultra-thin silicon foils at thicknesses of the order of few to tens of m10, 11. In contrast with thin wafers, ultra-thin silicon foils offer high flexibility and expected to reduce the cell cost further by reducing breakage loss during cell fabrication. Further,

4 solar cells can be fabricated on the flexible foils with roll-to-roll processes that may significantly cut down the manufacturing costs.

Figure 1-3 Trend of thickness of as-cut silicon wafer thickness in mass production of solar cells9. Green, yellow and red represent the technology that is currently in production, exist in the market but is not currently in production, and for which the industrial solution is as yet unknown, respectively.

However, thin Si wafers and ultra-thin foils do not readily lend themselves to high-efficiency PV devices owing to the large penetration depth of infrared wavelengths.

The corresponding optical loss due to inadequate absorption of longer wavelength light requires the integration of an effective light-trapping scheme in addition to the need for a broad-band anti-reflective surface.

1.3 Surface Texturing

Current methods of reducing the surface reflection of silicon employ processes such as wet chemical etching, dry etching and electrochemical etching12-18 to increase the roughness of the surface in order to provide a gradient between the index of refraction of air and c-Si. Much interest has been directed toward anisotropic etching of crystalline

5 silicon with potassium hydroxide (KOH) as this approach allows research advances in the area of selected surface texturization to form desired surface morphologies at low cost.

An anisotropic wet etch on silicon wafer creates a cavity with a trapezoidal cross section, which results in an inverted pyramid with an angle of 54.7  from the <100> surface13, 14,

19, 20. In present commercial silicon solar cells, 180-200 m thick saw-cut c-Si wafers are chemically etched in hot KOH /isopropyl alcohol (IPA) solution to form a high density texture of random pyramids having size distribution of a few to tens of microns on the wafer surface that serve to reduce the surface refection by promoting multiple bounces of the incident light. Such texture reduces the surface reflection by 11-14 % compared to

~35% for bare silicon17, 21, 22. The textured surface is then coated with multiple layers of optically thin films to reduce the surface reflection to ~ 2%23. The solar cells are then encapsulated in thick cover glass that further cuts down the reflection losses24.

In addition to decreasing surface reflection, a textured front surface also lengthens the path travelled by light inside the absorbing layer and enhances absorption of weakly absorbing wavelengths in the cell. Light trapping properties of the random pyramidal textures on silicon with depths greater than the wavelength of light has been studied using geometrical optics 25. For pyramid size and absorbing layer depth greater than the wavelength of light, the information on the wavelength and phase of electromagnetic waves is not important and the optical flow of energy can be modeled by using geometrical ray optics25-27. With this approach, one finds that the light can be most effectively trapped by incorporating an ideal rough surface for Lambertian scattering at the front or both surfaces bounding the absorbing layer. With a Lambertian surface at the front and a lossless reflector at the back surface, the optical path length near the

6 absorption band edge (i.e., in the weakly absorbing region) can be increased by a maximum factor of 4n2 over that of a single pass, where n is the refractive index of the absorbing medium. This 4n2 enhancement limit is also known as the Lambertian or

Yablonovitch limit26, 28, 29, and represents the maximum enhancement that can be achieved in geometrical optics regime by any front texture with the acceptance cone traversing a full hemisphere. This suggests the possibility of a ~50 fold increase in the path length in silicon (n ~ 3.5) and, if achieved, the prospect of a few micron thick c-Si solar cell with similar efficiencies to current 180 m thick c-Si PV cells.

The Yablonovitch limit is derived using geometrical optics and is not strictly applicable when the wave effects of light become important. This occurs for the following cases30-33: (i) when the thickness of the absorbing layer is comparable or smaller than the wavelength of light, and (ii) for ordered or grating textures with periodicity on the order of incident wavelengths. The grating period approximating the wavelength increases the optical thickness of the cell by coupling the light into diffractive modes that propagate at large off-normal angles within the silicon wafer. The wave effect of light becomes important and the absorption can surpass the Lambertian limit in a limited spectral range at normal incidence at the cost of absorption at other incident angles, keeping the angle-averaged absorption below the Lambertian limit30-33.

1.4 Grated Surface Textures for Light-Trapping

A periodically structured surface or grating was first proposed by Sheng et al. in 1983 for amorphous-silicon thin film cells31 and later studied by Heine and Morf for wafer-based high efficiency c-Si cells34. Prior studies have shown that a high performance grating

7 structure consists of tapered features highly packed in a two dimensional array with grating period slightly smaller than the wavelength range of interest30, 32, 35. The tapered features reduce reflection losses at the front surface by producing a refractive index profile that gradually increases from n = 1 from air to n = 3.5 in the medium (silicon) directly above the wafer and the grating periodicity facilitates diffractive light trapping.

Grating structures with hemispherical35, triangular36, conical37, pyramidal30, 38, inverted pyramidal39, nano-holes40 and rod-like40 features have been previously studied, with the objective of maximizing optical absorption within a cell. By comparison, an inverted pyramid grating offers good light trapping and also minimally increases the surface area upon texturing which is required to achieve low surface recombination losses. The inverted pyramidal features decrease surface reflection by providing a graded refractive index from air to silicon as the width of the inverted pyramid reduces from its base width to zero inside the silicon slab and are reported to perform better than random and ordered upright pyramids41. Inverted-pyramidal grating textures have been applied on both thick and thin crystalline silicon using various fabrication methods such as photolithography, nanoimprint, colloidal lithography, laser equipment and electron beam lithography 39, 42-44 to demonstrate high efficiency PV cells 39, 42, 45. However, this structure has yet to break through in commercial production of silicon solar cells because the fabrication methods are not readily scalable to industrial demands and involve high costs due to multi-step clean-room processing. A new fabrication technique is thus required to make such structures for high-throughput large area applications with minimum number of fabrication steps in non-cleanroom environments. Any new

8 technique should be versatile to create high efficiency structures suitable for both thick and thin silicon wafers as the industry makes a shift to thinner wafers.

In prior theoretical and experimental studies of optical efficiency, inverted pyramid grating texture of specific grating periodicities and feature size have been applied on very thick (~ 400 m)43, 44, 46 and very thin silicon membranes (< 50 m)39, 45 commensurate with the capability of the fabrication methods. In 2009, 20.7% efficient solar cells were demonstrated by Kray et al. on 40 m thick high quality c-Si with an inverted pyramid grating of period, ~2 m and recently, Mavrokefalos et al. employed a 700 nm period inverted pyramidal grating aiming to achieve ~25% efficient cells on 10

m thick c-Si foils. However, heretofore the understanding of the precise interplay between grating periodicity, wafer thickness, and absorption enhancement is still unclear.

It is not clear if the optimal grating parameters (periodicity and feature size) for effective photoabsorption in the cell will shift with the wafer thickness and will necessitate new fabrication methods for different wafer thicknesses.

1.5 Research Objectives

The aim of this doctoral thesis is to investigate and optimize inverted pyramidal textures for maximum absorption of AM1.5 solar spectrum in the solar cells built on silicon of thicknesses ranging from 2 to 400 m and to develop a fabrication technique compatible with thin to thick silicon wafers and foils. In particular, the objectives are to:

(i) study inverted pyramidal texture of different periodicities on 2 - 400 m thick

silicon and identify a high performance inverted-pyramidal texture design using

wave-optical simulations for various c-Si thicknesses,

(ii) design a new method to fabricate optimized texture on thick to thin wafers as an

alternative to other available lithographic methods with minimum number of

fabrication steps and least silicon damage and removal,

(iii) to measure the optical efficiency of the fabricated textures and compare the

performance with high resolution samples fabricated with electron beam

lithographic tool and wave-optical simulation results.

In this research a new method is developed for the fabrication of inverted pyramidal texture on silicon wafer surfaces. Specifically, this is a non-conventional lithographic technique which involves pattering of a thin dielectric film on silicon by femtosecond laser pulses, followed by selective anisotropic etching of silicon in KOH, resulting in a high resolution inverted pyramidal texture. Within this framework, laser processing of dielectric films is comprehensively studied as a part of the thesis project.

1.6 Thesis Outline

We now give an outline of the thesis. Chapters 2 and 3 contain details on the processing of thin dielectric films on silicon with  = 522 nm wavelength femtosecond laser pulses.

Chapter 2 focuses on the processing of films of thickness > 4nfilm, where nfilm is the refractive index of the deposited film. A novel interaction controlled by Fabry Perot interference patterns leads to highly resolved axial processing within a thin transparent film on silicon. Chapter 3 concentrates on the delamination of films of thickness <

4nfilm, driven by confinement of laser light at the film-silicon interface. The optimization of laser exposure conditions and dielectric film (hard-mask) thicknesses is done to create small mask apertures using a single laser pulse for high-resolution

10 patterning, to facilitate the subsequent etching of c-Si into an inverted-pyramidal texture.

The optimization of KOH etching procedures for selectively etching silicon through the patterned hard mask is given in Appendix A.

The results of wave optical studies of inverted pyramidal textures on thick silicon are discussed in Chapter 4. Inverted pyramidal textures of periodicities ranging from

100-2000 nm are studied on 2 – 400 m thick crystalline silicon. The optimal texture design parameters are identified to maximize photoabsorption in cells of various thicknesses with a perfect back reflector under normal incidence. We report a universal texture parameter that will maximize photoabsoption in silicon, irrespective of its thickness. Further, the optical loss is studied for the role of the flat ridges/mesas between the inverted pyramids in the texture. The photoabsoption is also studied with the addition of an antireflective coating and optimized for highest short-circuit current.

Chapter 5 details the fabrication of high-resolution inverted pyramidal textures by electron beam lithography. The measured spectral reflectances from the samples fabricated by laser hard mask writing technique and high precision electron beam lithography are compared to examine the quality of laser processed samples. The measured spectral reflectance is further compared with simulated reflectance.

Chapter 6 summarizes the major findings of this work and presents possible future research directions.

Chapter 1 References

1. International Energy Agency (IEA). World energy outlook 2012 (2012).

2. http://www.ez2c.de/ml/solar_land_area/index.html.

3. Masson, G. PVPS Report - A Snapshot of Global PV - 1992-2012 (2013).

4. http://www.epia.org/news/publications/global-market-outlook-for-photovoltaics-2013-2017/.

5. Shockley, W. & Queisser, H. J. Detailed Balance Limit of Efficiency of p-n Junction Solar Cells. J. Appl. Phys. 32, 510 (1961).

6. http://rredc.nrel.gov/solar/spectra/am1.5/.

7. Chapin, D. M., Fuller, C. S. & Pearson, G. L. A New Silicon p-n Junction Photocell for Converting Solar Radiation into Electrical Power. J. Appl. Phys. 25, 676-677 (1954).

8. Jäger-Waldau, A. JRC Scientific and Policy Reports- PV Status Report 2012 (2012).

9. Group, C. T. M. International Technology Roadmap for Photovoltaics- Results 2012. ITRPEV. net 2 (2012).

10. Janssen, E. &Kleiman, R. Novel process flow and cell architecture for 10 µm thick membrane single-crystalline silicon solar cell (38th IEEE Photovoltaic Specialists Conference (PVSC), 2012 , IEEE, 2012).

11. Morenoa, M. & Roca i Cabarrocas, P. Ultra-thin crystalline silicon films produced by plasma assisted epitaxial growth on silicon wafers and their transfer to foreign substrates. EPJ Photovolt. 1, 10301-10306 (2010).

12. Franssila, S. in Introduction to microfabrication (2nd Edition) 534 (John Wiley & Sons, Chichester, West Sussex [England], 2010).

13. Gad-el-Hak, M. in MEMS : design and fabrication 664 (CRC/Taylor & Francis Group, Boca Raton, FL, 2006).

14. Hull, R. in Properties of crystalline silicon 1016 (INSPEC, London, 1999).

15. Madou, M. J. in Fundamentals of microfabrication and nanotechnology 656 (CRC Press, Boca Raton, FL, 2012).

16. Abbott, M. & Cotter, J. Optical and electrical properties of laser texturing for high-efficiency solar cells. Prog. Photovoltaics 14, 225-235 (2006).

17. Sparber, W. et al. Comparison of texturing methods for monocrystalline silicon solar cells using KOH and Na2CO3 (Proceedings of 3rd World Conference on Photovoltaic Energy Conversion Ser. 2, IEEE, 2003).

18. Shikida, M., Sato, K., Tokoro, K. & Uchikawa, D. Differences in anisotropic etching properties of KOH and TMAH solutions. Sensors & Actuators: A.Physical 80, 179-188 (2000).

19. Seidel, H. Anisotropic Etching of Crystalline Silicon in Alkaline Solutions. J. Electrochem. Soc. 137, 3626-3632 (1990).

20. Wind, R. A., Jones, H., Little, M. J. & Hines, M. A. Orientation-Resolved Chemical Kinetics: Using Microfabrication to Unravel the Complicated Chemistry of KOH/Si Etching. J. Phys. Chem. B 106, 1557-1569 (2002).

21. Birmann, K., Matthias, D. & Rein, H. Optical characterization of Random Pyramid Texturization. Small 2, 11-13 (2011).

22. Munoz, D. et al. Optimization of KOH etching process to obtain textured substrates suitable for heterojunction solar cells fabricated by HWCVD. Thin Solid Films 517, 3578-3580 (2009).

23. Gee, J. M., Gordon, R. & Liang, H. Optimization of textured-dielectric coatings for crystalline-silicon solar cells (Proceedings of 25th IEEE Photovoltaic Specialists Conference, IEEE, New York, 1996).

24. Landis, G. A. A light-trapping solar cell coverglass (Proceedings 21st IEEE Photovoltaic Specialists Conference, IEEE, New York, 1990).

25. Green, M. A. &Campbell, P. Light trapping properties of pyramidally textured and grooved surfaces ((in amorphous and crystalline silicon solar cells)) (19th IEEE Photovoltaic Specialists Conference, IEEE, New York, 1987).

26. Yablonovitch, E. & Cody, G. D. Intensity enhancement in textured optical sheets for solar cells. IEEE Trans. Electron Devices 29, 300-305 (1982).

27. Campbell, P. & Green, M. A. Light trapping properties of pyramidally textured surfaces. J. Appl. Phys. 62, 243-249 (1987).

28. Deckman, H. W., Roxlo, C. B. & Yablonovitch, E. Maximum statistical increase of optical absorption in textured semiconductor films. Opt. Lett. 8, 491-493 (1983).

29. Campbell, P. R. & Green, M. A. On "Intensity enhancement in textured optical sheets for solar cells". IEEE Trans. Electron Devices 33, 1834-1835 (1986).

30. Han, S. E. & Chen, G. Toward the Lambertian limit of light trapping in thin nanostructured silicon solar cells. Nano Lett. 10, 4692-4696 (2010).

31. Sheng, P., Bloch, A. N. & Stepleman, R. S. Wavelength-selective absorption enhancement in thin-film solar cells. Appl. Phys. Lett. 43, 579-581 (1983).

32. Yu, Z., Raman, A. & Shanhui, F. Fundamental limit of light trapping in grating structures. Opt. Express 18, A366-A380 (2010).

33. Callahan, D. M., Munday, J. N. & Atwater, H. A. Solar Cell light trapping beyond the ray optic limit. Nano Lett. 12, 214-218 (2012).

34. Heine, C. & Morf, R. H. Submicrometer gratings for solar energy applications. Appl. Opt. 34, 2476-2482 (1995).

35. Song, Y. M., Yu, J. S. & Lee, Y. T. Antireflective submicrometer gratings on thin-film silicon solar cells for light-absorption enhancement. Opt. Lett. 35, 276-278 (2010).

36. Dewan, R. et al. Light trapping in thin-film silicon solar cells with submicron surface texture. Opt. Express 17, 23058-23065 (2009).

37. Wang, K. X., Yu, Z., Liu, V., Cui, Y. & Fan, S. Absorption enhancement in ultrathin crystalline silicon solar cells with antireflection and light-trapping nanocone gratings. Nano Lett. 12, 1616-1619 (2012).

38. Chutinan, A., Li, C. W. W., Kherani, N. P. & Zukotynski, S. Wave-optical studies of light trapping in submicrometre-textured ultra-thin crystalline silicon solar cells. J. Phys. D 44, 262001-1-262001-4 (2011).

39. Mavrokefalos, A., Han, S. E., Yerci, S., Branham, M. S. & Chen, G. Efficient Light Trapping in Inverted Nanopyramid Thin Crystalline Silicon Membranes for Solar Cell Applications. Nano Lett. 12, 2792-2796 (2012).

40. Han, S. E. & Chen, G. Optical absorption enhancement in silicon nanohole arrays for solar photovoltaics. Nano Lett. 10, 1012-1015 (2010).

41. Baker-Finch, S. C. & McIntosh, K. R. Reflection of normally incident light from silicon solar cells with pyramidal texture. Prog Photovoltaics Res Appl 19, 406-416 (2011).

42. Zhao, J., Wang, A. H. & Green, M. A. 24.5% efficiency PERT silicon solar cells on SEH MCZ substrates and cell performance on other SEH CZ and FZ substrates. Sol. Energ. Mat. Sol. C. 66, 27-36 (2001).

43. Sun, C. H., Min, W. L., Linn, N. C., Jiang, P. & Jiang, B. Templated fabrication of large area subwavelength antireflection gratings on silicon. Appl. Phys. Lett. 91, 231105-1-231105-3 (2007).

44. Chaturvedi, N., Hsiao, E., Velegol, D. & Kim, S. H. Maskless Fabrication of Nanowells Using Chemically Reactive Colloids. Nano Lett. 11, 672-676 (2011).

45. Kray, D. & McIntosh, K. R. Analysis of ultrathin high-efficiency silicon solar cells. Phys. Status Solidi (a) 206, 1647-1654 (2009).

46. Parretta, A. et al. Angle-dependent reflectance measurements on photovoltaic materials and solar cells. Opt. Commun. 172, 139-151 (1999).

Chapter 2 Quantized Laser Processing of Thin Dielectric Films on Silicon

2.1 Introduction

The advent of ultrashort-pulsed lasers has dramatically improved the precision of light- matter interactions owing to greatly reduced heat-affected zone and strong nonlinear optical absorption. Inside a transparent medium, femtosecond laser light can be tailored to drive strong nonlinear absorption confined to a small focal volume created by high numerical aperture lenses. Multi-photon fluorescence can be locally excited to enable high resolution three dimensional (3D) microscopy of living cells1 while higher exposure can induce gentle refractive index changes for writing into 3D optical circuits 2 or driving micro-explosions for 3D memory or marking3, 4. Femtosecond laser light has also been transmitted through a thin transparent film and confined to interact within the thin penetration depth of an underlying silicon substrate5, 6. In this way, the laser-plasma interaction at the buried interface has enabled the formation of thin-film blisters and nano-fluidic networks7 at low exposure or the precise ejection of the film at higher exposure. These ejection principles are promising for patterning and repair in microelectronic circuits, photovoltaic cells8 and glass display manufacturing and further underlie the driving mechanisms in laser induced forward transfer (LIFT)9 for printing or additive manufacturing, and cell ejection by laser pressure catapulting1, 10.

2.1.1 Laser Processing of Transparent Dielectric Films

On the other hand, the concept of processing directly within such films is an unexplored area that could significantly improve the functionality of CMOS, flexible electronic, photovoltaics, MEMS, LED, lab-on-a-chip devices where thin films are widely deployed during their manufacture. In this direction, multi-surface Fresnel reflections of laser light are well known to interfere and create a standing wave interference pattern with fringe maxima spaced by /2nfilm inside a transparent film of thickness, d > /2nfilm, where nfilm is the refractive index of the film. Laser interaction models have predicted enhanced laser dissipation at these intensity maxima that in experiments was associated with a lower breakdown threshold for damage in thick single11, 12 or multilayer13 dielectric films. However, such spatially localized laser interaction has not been directly observed inside the film, and any associated material modification was concentrated either at an interface or diffused over the bulk volume of the film13.

We harness nonlinear absorption of femtosecond laser light to overcome thermal transport and enable a strongly localized laser-plasma formation within zones narrower than the Fabry-Perot interference fringes. These thin plasma disks facilitate material modification inside the film volume at a length scale much smaller than the focal

Rayleigh range for the first time. The laser interaction volume is found to be divided into an array of thin axial planes that align with the Fabry-Perot fringe maxima inside the dielectric film. We show that this novel localized interaction can be controlled by the laser exposure to modify the film interior periodically to either open internal nano-voids or for quantized ejection of film disks in multiples of /2nfilm thickness. The temporal 17

ejection of multiple disks by a single laser pulse was verified by time-resolved imaging of the ablation plume with an intensified CCD camera. These varying interactions enable the fabrication of 3D nanofluidic structures inside the thin film while the quantized surface structuring defines a new approach for film coloring and labeling or multilevel surface structuring. Together the nano-voids and quantized ejection are attractive for structuring thin films such as widely used in CMOS and many other manufacturing processes that promise to improve the functionality of microelectronic, photonic, MEMS, optofluidic and sensor devices as well as opening new directions for developing flexible electronic or display films or new lab-in-a-film concepts.

2.2 Femtosecond Laser-Material Interaction in Transparent Dielectric Films

Laser light entering a thin transparent film of thickness greater than /2nfilm will lead to formation of a Fabry-Perot interference pattern owing to Fresnel reflections and transmissions at the air-film and the film-substrate interface. The resulting fringe contrast can be controlled by tailoring the refractive index of the film and substrate materials. For the case of silicon nitride (Si3N4; nfilm = 1.98) film of d = 500 nm thickness on a silicon substrate (nsi = 4.192 and si = 0.036), a 522 nm wavelength laser beam incident from the top in Figure 2-1(a) will generate four fringe maxima (m ≈ d/(/2nfilm)) on /2nfilm =

131.8 nm period with a fringe visibility of 0.63 as shown in Figure 2-1(b). A node is positioned near the Si3N4-silicon interface due to the high index contrast, therefore locking the fringe pattern with the last fringe maximum positioned at /4nfilm = 65.9 nm from the bottom interface. As a consequence, the position of the first fringe maximum

from the top surface will vary with the film thickness, shown at d - m/2nfilm = 38.7 nm from the air-Si3N4 interface for the case in Figure 2-1(b).

At moderately low laser intensity, stronger linear interactions in the silicon substrate dominate over the nonlinear plasma excitation in the transparent film to drive laser heating only to a penetration depth of 1/si = 28 nm in the silicon. The machining at the film-silicon interface due to this thin heating zone underpins the physics for blistering and ejection of films as reported in references [5, 14-18] over varying film thickness without evidence of internal structuring of the transparent film. However, such interface machining was found together with the first evidence of internal structuring of the film as shown in Figure 2-1(c) by the scanning electron micrograph (SEM) images for the

12 present 500 nm thick film exposed at a threshold incident intensity of Iavg = 9×10

W/cm2. Laser induced cleavage planes are seen to have ejected thin disk segments or formed nano-voids at positions found to align with the calculated positions of the interference maxima as indicated by the red connecting lines to Figure 2-1(b). A radius of 0.53 m is observed for the fully ejected second disk in Figure 2-1(c), which is

2 commensurate with the radius o = 0.494 m (1/e ) calculated for the focused Gaussian beam. At this radial position (o), the internal laser intensity modulates axially from 0.84 to 3.74 TW/cm2 as shown in Figure 2-1 (b), suggesting a threshold intensity exposure of

3.74 TW/cm2 for internal structuring of the film.

2.2.1 Theoretical Calculation of Electron Density inside the Film

This intensity profile (Fig. 2-1(b)) was applied to predict the electron density profile generated inside the film at the observed threshold for internal structuring. Because of

the short duration laser pulse, nonlinear light interactions inside the dielectric film will be dominated by multiphoton absorption and electron avalanche that ionize atoms to create

19 an electron density ne according to equation (1) :

(1)

where Na is atomic density of the dielectric thin film and r is the electron relaxation time. The impact ionization rate (wimp) and multiphoton ionization (MPI) rate (wmpi) at the incident laser intensity (I), are given, by equation (2) and (3), respectively19,

(2)

( )

⁄ ( ) (3)

and the effective electron collision time (eff) and the electron quiver energy (osc) are calculated by equation (4)20 and (5)19:

√ [ ] (4)

√

(5) [ ] [ ] [ ⁄ ]

The electron relaxation (r term in Eq. (1)) is insignificant for the short duration (laser =

22 -3 200 fs) laser pulse considered here. For Si3N4, values of Na = 8×10 cm for the atomic

21 density, Eg = 5.3eV for the bandgap , Ji = Eg for the ionization potential, and me* = me for the effective mass of electron were used for computing the electron density. The laser frequency is given by  = 2πc/and the order of nonlinear MPI, N = ГJi/ħ = 3.

The time dependent equations (1), (2) and (4) were simultaneously solved to follow the temporal rise of the electron density expected for the spatial intensity profile in

Figure 2-1(b). At the end of the laser pulse, the electron density is seen in Figure 2-1(b) to peak strongly at the fringe maxima to a value of 5.87 × 1021 cm-3. This value surpasses

21 -3 the critical plasma density (ncr ~ 4.10 × 10 cm ) where the plasma becomes opaque to the laser and is typically expected to initiate material damage13, 19, 20. Hence, the radial extent of the ejected disks at the laser-defined cleavage planes (0.494 m in Fig. 2-1(c)) match closely with the typical laser-plasma conditions found to damage materials.

2.2.1.1 Contribution of Multiphoton Ionization and Impact Ionization

The full width at half maxima (FWHM) of the electron density profile at different times () during the laser = 200 fs long pulse is shown in Figure 2-1(d). The width of the laser plasma zones decreases from 46 nm at the beginning of the pulse to 24 nm at critical

= 84 fs and broadens back to 46 nm at the end of the pulse at  = 200 fs. The width of the laser plasma zones depends on the relative difference in electron generation rate at fringe maxima position and low intensity points within Fabry-Perot interference pattern. At times t < 24 fs, despite of the high rate of impact ionization compared with multi-photon ionization (Fig. 2-1(e)), the electron generation is dominated by multi-photon ionization.

This is because very few electrons are initially available by multi-photon ionization at the beginning of the pulse for significant electron generation by avalanche ionization. In this time regime, equation 1 reduces to

(6)

and the electron density in the interference pattern at a certain time varies with (I)N, N =

3, due to multi-photon ionization, reducing the thickness of the plasma zones to 46 nm.

At > 23 fs, avalanche ionization dominates at fringe maxima positions and the electron density starts growing relatively faster at maxima positions in contrast with lower intensity points that in combination further reduces the width of the laser plasma zones to

24 nm at critical.

At  > critical, the effective collision frequency of electrons in plasma exceeds the laser

 frequency (eff > × Hz) (Fig. 2-1(f)). Hence, in time regime  > critical to  =

laser, equation 2 reduces to

(7) ( ) and the rate of the impact ionization is now decreasing at the maxima positions with the increasing veff as the pulse progresses. However, the density of electrons at low intensity points in the inference pattern continues to rise rapidly to due to impact ionization, therefore broadening the laser-plasma zones to 46 nm at the end of the pulse.

Figure 2-1 A cartoon (a) depicting the division of the focal interaction volume of incident 522 nm laser light into thin interaction zones (orange discs) inside a 500 nm thick Si3N4 film as a result of the interference of Fresnel reflections of the incident light from the boundary interfaces. (b) The modulated intensity profile calculated for a Gaussian-shaped beam of 0.495 m (1/e2) radius (line) together with the electron density profile (shaded) expected inside the film at the threshold exposure of 9×1012 W/cm2 average incident intensity. The electron density exceeds the critical plasma density (ncr) at the fringe maxima positions that manifests in the internal structuring of the film observed in cross-sectional and oblique top SEM views in (c). The threshold intensity results in the ejection of segments S1, S2 and blistering of fused segments S3 and S4 to

form a nano-void at the fourth fringe maximum, and blistering of S5 overlying a second nanovoid. The positions of the cleavage planes align with the fringe maxima positions as depicted by red lines. The thickness of laser-plasma zones (FWHM of the electron density profile) at different times during the () during the laser = 200 fs long pulse (d) shows the minimum width at criticalfs that corresponds to the time at which the rate of impact ionization (wimp, green curve) starts diminishing at maxima positions relative to low intensity positions within the interference maxima while the multiphoton transition (wmpi) remains constant (blue curve) (e). The reduction in the rate of impact ionization occurs at critical when the effective collision frequency of electrons exceed the laser frequency (f).

Hence, the simulations showed the impact and multiphoton ionization to thin the laser-plasma zone to ~46 nm thick disks in Figure 2-1(b) that is significantly narrower than the Fabry-Perot fringe width of 91 nm. Hence, an array of thin heating disks are predicted to have formed (Fig. 2-1(a)) on the /2nfilm fringe spacing on time scales shorter than thermal transport to serve as a new means for machining inside thin transparent films on size scales much smaller than the ~ 2.3 m depth of focus.

2.3 Quantized Internal Structuring and Ejection of a Transparent Film

The definitive evidence of the confinement of the laser-generated plasma into thin disks to create sharp and periodic cleavage planes inside the film is the observed alignment of the annular structures, the ejected membranes and the nano-voids in Figure

2-1(c) with the calculated fringe maxima positions Figure 2-1(b). The ejection of the first membrane structure (segment 1) is evidenced by the annular ring seen in the top view at ~

34 nm depth that matches closely with the expected 29 nm deep position of the first

fringe. The partially attached membrane (segment 2) was formed by plasma-cleavage at the 1st and 2nd fringe maxima, defining a ~ 135 nm thick membrane that matches the

rd th expected /2nfilm = 131.8 nm fringe spacing. The 3 and 4 expected membrane structures (segment 3 and 4) are seen to be fused into a double layer of ~ 267 nm thickness to form a non-punctured blister with thickness that matches the expected double-fringe spacing (2/2nfilm = 263.8 nm). Underlying this blister, a microexplosion from a thin disk plasma zone is inferred to have expanded into an ~ 800 nm diameter nano-void of ~ 138 nm height. A deeper nano-void is seen to have opened at the silicon- film interface to ~ 46 nm height. These nano-voids define the fifth and final membrane

(segment 5) whose observed ~ 64 nm thickness matches closely with the expected quarter-fringe thickness (65.7 nm). One may therefore understand the fusion of the third and fourth segments to be an anomaly associated with opposing forces of microexplosions in the top two cleavage planes against the powerful shock and pressure driven from laser microexplosions at the fourth fringe position and the film-silicon interface.

2.3.1 Film Morphology as a function of Laser Exposure

Once critical plasma density is reached at the first fringe position, strong light reflection and attenuation will reduce the forward propagating beam intensity, diminishing the interaction strength at deeper fringe positions. This presents the opportunity for controlling the number of laser-heating zones to vary the number of ejected segments and nano-voids formed inside the film with varying laser exposure. These principles were examined for a thicker 945 nm Si3N4 film in Figure 2-2 for the Gaussian-shaped beam to the maximum available fluence of 21.65 J/cm2. The top and cross-sectional SEMs (Fig.

2-2(a) (i-viii)) show an expected widening of the laser modification zone for such beam shape from 1.3 m to 2 m diameter with the increasing fluence. Near the modification threshold of 3.50 J/cm2, a first segment of ~ 64 nm thickness was completed ejected while segment 2 formed into a punctured blister with a ~ 135 nm diameter open hole. At a higher fluence of 6.37 J/cm2, Figure 2-2(a) (ii) shows the complete ejection of segment

2. The 3rd and 4th segments are seen (sideview in Fig. 2-2(a) (iii)) here to be fused into a

~ 270 nm thick blister overlying a ~ 260 nm deep nano-void, as similarly observed in the previous case of the 500 nm film. These fused segments form into a punctured blister at

10.2 J/cm2 fluence (side view in Fig. 2-2(a) (iv)) and are partially ejected at the higher laser exposure of 15.9 J/cm2 (side view in Fig. 2-2(a) (vi)), leaving an annular ledge clearly visible within the via. This sequence of blistering, puncturing and ejection of segments to quantized depths advances deeper inside the film with further increase in laser fluence (i.e. Fig. 2-2(a) (vi), (vii), (viii)).

The developing morphology with increasing fluence is summarized graphically in

Figure 2-2(b) by the threshold fluences observed to form a solid blister (purple), a punctured blister (yellow), and an ejected blister (orange-red) and were aligned vertically for each segment according to the observed cleavage plane (dashed line). The cleavage positions were again found to align closely (≤ ± 6 nm) with the calculated Fabry-Perot intensity maxima as aligned graphically on the right. Segment 1 was found to eject at

4.46 J/cm2 threshold fluence (orange-red) together with the perforated blistering of segment 2 (yellow) without revealing a blistering phase for the first segment. The laser- plasma generated at the first fringe may have burnt through thin first segment (64 nm) to prevent such blistering. A nano-void was not observed to open between segments 3 and 4,

resulting in blistering, perforation and ejection of the two segments together in the respective fluence ranges of 5.41 - 9.24 J/cm2, 9.24 - 14.97 J/cm2 , and ≥14.97 J/cm2 , respectively, as shown in Figure 2-2(a) ((ii)-(vi)) and (b). This anomalous fusion was observed also in films varying from 500 nm to 1545 nm thickness. Figure 2-2(a) (i) and

2(b) also shows the early onset of void formation at the Si3N4-Si interface at a low fluence threshold of 3.50 J/cm2.

blistering, puncturing and ejection of each segment with increase in laser fluence is summarized graphically in (b) by the threshold fluences observed to form a solid blister (purple), a punctured blister (yellow), and an ejected blister (orange-red). Each segment was found to align between the cleavage planes (horizontal dashed lines) predicted by the positions of the Fabry-Perot intensity maxima shown graphically on the right.

2.4 Verification of Quantized Film Ejection with an Intensified CCD camera

The observed remains of the ejected Si3N4 segments (Fig. 2-1(c) and 2 (ii), (vi)) suggest the array of laser-induced plasma zones do not burn through and vaporize the forming membranes. Thus, one anticipates the quantized ejections of segments in a temporal sequence as the plasma planes heat and microexplode, beginning from the near-surface to the lower cleavage positions as depicted in supplementary Video A1. The evidence for this sequential ejection is seen in time-gated ICCD images recorded from a 500 nm Si3N4 film shown in Figure 2-3, revealing time-delayed repeating ejections of ablation plume whose number matched well with the number of segments found by SEM to be ejected for a given laser fluence. For example, the SEM image of the film (Fig. 2-3(b)) reveals the ejection of 5 segments when irradiated with 13.67 J/cm2 fluence. The clusters of plume were observed appearing in time zones of 3 - 9 ns (green), 173 - 193 ns (orange),

233 - 393 ns (red) and 393 - 1493 ns (blue), marked in Figure 2-3(a). The plume positions were followed up to 180 m distance from the film surface, with their observed positions recorded as a function of time in Figure 2-3(c). We infer the 5th segment to be the last ejected plume (blue). The partial ejections of the fused 3rd (orange) and 4th (red) segments are nearly indistinguishable, appearing together with the 5th segment at ~ 200 ns, but being ejected forward more quickly than the 5th segment. Therefore the bright

emissions observed in the 3 - 10 ns zone are ascribed to plume expansion and membrane ejection of the 1st and 2nd segments promptly after the laser exposure. At this fluence, the first two segments appeared bright and promptly, moving at ~2.9 km/s speed, while the ejection of deeper layers were delayed (173 - 1500 ns) and appeared with much lower speeds reducing to 0.1 km/s for the 5th segment. Hence, the directly ablated surface material and first ejected segments appear promptly with the highest speeds, while the inertia of pushing against the upper layers leads to much delayed ejection and ~ 30-fold lower ejection speeds for the deeper segments of the film.

Figure 2-3 Time-resolved ICCD images (a) of ablation plumes recorded transversely 2 from a 500 nm Si3N4 film exposed to 13.67 J/cm fluence and the top and cross-sectional SEMs (b) of the film after exposure. The plume emissions (a) appear in several clusters associated with the ejection of segments S1 and S2 (green) in the 3 - 9 ns window, partial ejection of S3 (orange) and S4 (red) in 173 - 193 ns and 233 - 393 ns window, respectively, and ejection of S5 (blue) in 393 - 1493 ns window. The observed position of clusters (c) from the surface as a function of time.

The combination of quantized surface ejection and nano-void formation directly inside a thin transparent film opens a new means for fabricating novel combinations of optical, nanofludic, and MEMs components with facile delivery of varying laser exposure. Figure 2-4(a) presents this concept for fabricating micro- and nano-fluidic 30

devices that include various reservoir designs connected with different types of open and buried channels, including a mixing channel with embedded barriers and serpentine channels. Fresnel lenses and blazed gratings are further depicted together with a large area membrane sensor. The laser-structured devices may potentially be fabricated in films coated over microelectronic and CCD devices on silicon wafers that collectively offer a flexible and attractive integration platform.

2.5 Large Area Processing of Films

Expanding the interferometric fabrication principle to larger processing area may be approached by scaling up the pulse energy and blistering the film into a large-diameter

MEMS device (M) as identified in Figure 2-4(a). However, such uniform processing requires focusing to a flat-top beam profile. With the present laser system, a uniform beam was only available up to a maximum radius of ~ 0.75 m, which yielded more uniform ejected structure as presented in supplementary Figure A2. Structuring beyond this focal 1.5 m spot size was therefore approached by stitching together arrays of individual exposure spots formed with the uniform square beam. Various grid patterns were examined at variable laser exposures to optimize this stitching and generate a uniform morphology over a larger area as shown in supplementary Figure A3. In this way, individual laser ejection zones to the first, second, or third segments could be stitched with high reproducibility over large scanned areas to depths aligned closely with the expected fringe positions at 29 nm, 161 nm, and 293 nm depth, respectively, as shown in Figure 2-4(i), (ii), and (iii), respectively. Interestingly, the first segment was readily removed only when patterned in closely packed arrays and exposed below the

ejection threshold for an isolated laser spot. Laser shock may be dislodging weakly bonded first segments in the neighboring exposure sites. Removal of this first segment layer offered the smoothest morphology, manifesting in the expected color shift by thin- film interference from green of the deposited 500 nm film to red for the remaining 471 nm film as seen by the inset image of Figure 2-4(i) and characterized in supplemental

Figure A4. However, the deeper ejected segments showed dramatically increasing ablation debris on the sub-wavelength scale that is most prominent in Figure 2-4(iii) for the third segment ejection. This debris leads to strong optical scattering, overshadowing the thin-film interference effect to give the grey coloring seen in the case of the second and third segment ejections (Fig. 2-4(ii) and (iii) inset, respectively).

Regardless of this roughness, different ejection levels could be interfaced to form pillars as shown in Figure 2-4(iv). This multi-level structuring was applied to demonstrate the concept of a microfluidic mixer with embedded barriers, a Fresnel optical lens and a blazed optical grating as shown in the optical images Figure 2-4(v),

(vi), and (vii), respectively. The integration of these fabrication principles is demonstrated in Figure 2-4(b), where deep reservoirs (R) were connected with various crossed (CC), serpentine (SC) and mixing open channels (Mixer), written in a single run of three laser exposures to pattern over three different ejection levels.

As previously seen (Fig. 2-1 and 2), nanovoids are expected to have opened inside the film below the ejected segment layers (Fig. 2-4(i-iii)). The potential for linking the buried nanovoids below a closed-packed array of exposure spots is clearly demonstrated at the third fringe position in Figure 2-4(viii), thus opening the means for writing buried nano-fluidic channels that may link the various reservoirs as proposed in Figure 2-4(a).

Figure 2-4 A schematic (a) of a multifunctional device design consisting of combinations of optical, nanofluidic, and MEMs components over a large area together with SEM (gray) and optical (color) images of sample components constructed inside a film by interferometric laser processing. Uniform ejection of S1 (i), S2 (ii), or S3 (iii) segments over a large area by raster scanning a top hat laser beam that represent formation of single level reservoirs (R) and open serpentine channels (SC). The film color in the S1, S2 and S3 ejection zones was shifted from green (insets in (i) – (iii)). Different level ejections can also be combined to create multi-level reservoirs (R), mixing channels with pillars (v), and optical components such as a blazed grating (vi) and a Fresnel lens (vii). Nano- cavities at the 3rd Fabry Perot fringe position were stitched together (vii) to represent the writing of buried nanofluidic channels. A large area membrane structure (M) is anticipated with a large beam diameter. Optical image of a multi-component device (b) showing a Fresnel lens, grating, single and multilevel reservoirs (R), and open serpentine

(SC), crossed (CC) and mixing (Mixer) channels, fabricated with interferometric laser fabrication.

2.6 Summary

In this chapter, we present a novel method for highly resolved axial processing inside a thin transparent film on a high index substrate with a femtosecond laser by confining laser-material interaction to an array of narrow zones inside the film as was shown in

Figure 2-1 (a). This confinement is anticipated in transparent films of thickness ≥ /4nfilm, where the optical interference of Fresnel reflections from air-film and film-substrate interface creates a Fabry-Perot intensity modulation of the laser light on /2nfilm fringe spacing (Fig. 2-1(b)). Nonlinear laser interactions predicted the electron density profile to narrow to ~46 nm thick plasma disks that are more than 50-fold narrower than the laser depth of focus. At the threshold exposure for internal structuring, the electron density reaches critical at the predicted fringe maxima positions to facilitate the quantized ejection of the film or the formation of thin nano-voids inside the film at these laser cleaving planes as shown in Figure 2-1 and Figure 2-2. This geometry for internal laser cleaving has not been previously reported inside a transparent material and greatly extends the control over the laser modification in contrast with internal structuring over the whole laser focal volume4 or structuring confined at a film-substrate interface7, 16-18, 22.

Further, the predicted plasma disks were shown by intensified CCD imaging (Figure 2-3) to validate the quantized ejection of multiple segments in a temporal sequence. Both internal structuring and quantized ejection of films was observed in 500 - 1500 nm thick films with either uniform (Fig. A2) or Gaussian beam (Fig. 2-1, 2 and 3) shape.

2.7 Discussion

This partial and digital removal of a thin transparent film opens new directions in selectively texturing and surface micromachining to /2nfilm precision inside the film and in finely pitched patterns of 1 m lateral resolution. This opens new means for marking, coloring and multi-level structuring of thin transparent films (Fig. 2-4 and A3). Further, the combination of this laser-direct writing of multilevel patterns with buried nano-voids or channels can be exploited as demonstrated in Figure 2-4 for fabricating MEMS, optofluidic and other optical components in thin films on wafers or flexible silicon foils.

This approach is very attractive from transforming lab-on-a-chip (LOC) devices to flexible lab-in-a-film (LIF) structures that are compatible with today’s state-of-the-art manufacturing facilities for CMOS microelectronics or glass display and create novel chip-scale biosensors, minimally invasive implantable devices, portable point-of-care medical products, compact diagnostic platforms, or interactive sensor display. One can envision such laser-structured films over microelectronic chips, light sources or optical sensors, and ultrathin wafers to create epidermal biosensors in ultra-thin foldable and stretchable integrated circuits or to shape vascular-type networks into bio-implants that mimics natural structures.

2.8 Experimental Methods

Si3N4 films of thicknesses ranging from 20 nm to 1545 nm were grown by Plasma

Enhanced Chemical Vapor Deposition (PECVD) on single-side polished p-doped (001) crystalline silicon wafers of 400µm thickness in a PlasmaLab 100 PECVD system

(Oxford Instruments) at 300°C and 650 mT chamber pressure using a gas mixture of 5%

silane in nitrogen (400 sccm), ammonia (20 sccm) and pure nitrogen (600 sccm). The deposition was carried out at the rate of 14 nm/minute by using alternate combinations of high frequency (13.56 MHz) for 13 seconds and low frequency (100 kHz) for 7 seconds, successively. The radio frequency (RF) power was set to 50 W and 40 W for high and low frequencies, respectively.

A fiber laser (IMRA, FCPA µJewel D-400-VR) operated at 100-kHz repetition rate was frequency doubled to generate 200 fs duration pulses at 522 nm wavelength. By monitoring the back reflection on a CCD camera, a plano-convex lens of 8 mm focal length (New Focus, 5724-H-A) was positioned to focus the Gaussian-shaped laser beam

2 to a spot size of o = 0.495 μm radius (1/e ) onto the sample surface. Alternatively, a uniform exposure profile was attempted by masking the ~ 4.5 mm diameter beam with either a 0.6 mm × 0.6 mm square aperture or a circular aperture (1 mm diameter) positioned ~115 cm before an aspheric lens of focal length f = 2.8 mm to image to a comparatively uniform 1.5 m × 1.5 m square beam or 2 m diameter top-hat beam profile, respectively. A computer controlled linear polarizer attenuator varied the laser pulse energy between 5 and 70 nJ and single pulses were applied to each site by scanning the sample with an XY motorized stage (Aerotech, ABL1000). Laser raster scanning was employed to separate (speed > 15 µm/s) or to stitch together laser modification structures while an acousto-optic modulator (AOM) (Neos, 23080-3-1.06-LTD) further offered flexibility in patterning the surface with computer control.

Time-resolved 2-dimensional side-view images of the laser ablation plume were captured through a microscope objective (50×) onto an intensified CCD camera (ICCD)

(Andor, iStar DH734-18U-03). The ICCD trigger gating was synchronized to the laser

pulse with a digital delay generator (DDG) (Stanford Research Systems, DG535) while the laser repetition rate was down counted to 1 Hz with an AOM. Plume emissions were recorded with gate width varied from 3 to 50 ns and time delays from 0 to 2 s, and were examined for a wide range of laser exposure conditions (50 to 380 nJ) in a 500 nm thick film.

The morphology of the laser processed samples was inspected using scanning electron microscopy (SEM) and select samples were cut axially by a focused ion beam

(FIB) for cross-sectional SEM imaging.

Chapter 2 References

1. Schermelleh, L. et al. Laser microdissection and laser pressure catapulting for the generation of chromosome-specific paint probes. BioTechniques 27, 362-367 (1999).

2. Miura, K., Qiu, J., Inouye, H., Mitsuyu, T. & Hirao, K. Photowritten optical waveguides in various glasses with ultrashort pulse laser. Appl. Phys. Lett. 71, 3329-3331 (1997).

3. Gattass, R. R. & Mazur, E. Femtosecond laser micromachining in transparent materials. Nature Photon. 2, 219-225 (2008).

4. Glezer, E. N. & Mazur, E. Ultrafast-laser driven micro-explosions in transparent materials. Appl. Phys. Lett. 71, 882-884 (1997).

5. McDonald, J. P., Thouless, M. D. & Yalisove, S. M. Mechanics analysis of femtosecond laser-induced blisters produced in thermally grown oxide on Si(100). J. Mater. Res. 25, 1087-1095 (2010).

6. Rublack, T., Hartnauer, S., Kappe, P., Swiatkowski, C. & Seifert, G. Selective ablation of thin SiO2 layers on silicon substrates by femto- and picosecond laser pulses. Appl. Phys. A 103, 43-50 (2011).

7. McDonald, J. P., Mistry, V. R., Ray, K. E. & Yalisove, S. M. Femtosecond pulsed laser direct write production of nano- and microfluidic channels. Appl. Phys. Lett. 88, 183113-1-183113-3 (2006).

8. Kumar, K., Lee, K. K. C., Herman, P. R., Nogami, J. & Kherani, N. P. Femtosecond laser direct hard mask writing for selective facile micron-scale inverted-pyramid patterning of silicon. Appl. Phys. Lett. 101, 222106-1-222106-5 (2012).

9. Bohandy, J., Kim, B. F. & Adrian, F. J. Metal deposition from a supported metal film using an excimer laser. J. Appl. Phys. 60, 1538-1539 (1986).

10. Westphal, G. et al. Noncontact laser catapulting: a basic procedure for functional genomics and proteomics. Meth. Enzymol. 356, 80-99 (2002).

11. Mero, M., Sabbah, A. J., Zeller, J. & Rudolph, W. Femtosecond dynamics of dielectric films in the pre-ablation regime. Appl. Phys. A 81, 317-324 (2005).

12. Mero, M. et al. On the damage behavior of dielectric films when illuminated with multiple femtosecond laser pulses. Opt. Eng. 44, 51107-1-51107-7 (2005).

13. Jasapara, J., Nampoothiri V.V. A. & Rudolph, W. Femtosecond laser pulse induced breakdown in dielectric thin films. Phys. Rev. B 63, 045117-1-045117-5 (2001).

14. Rublack, T., Hartnauer, S., Kappe, P., Swiatkowski, C. & Seifert, G. Selective ablation of thin SiO2 layers on silicon substrates by femto- and picosecond laser pulses. Appl. Phys. A 103, 43-50 (2011).

15. Rublack, T., Schade, M., Muchow, M., Leipner, H. S. & Seifert, G. Proof of damage- free selective removal of thin dielectric coatings on silicon wafers by irradiation with femtosecond laser pulses. J. Appl. Phys. 112, 023521-1-023521-7 (212).

16. Tull, B. R., Carey, J. E., Mazur, E., McDonald, J. P. & Yalisove, S. M. Silicon Surface Morphologies after Femtosecond Laser Irradiation. MRS Bull. 31, 626-633 (2006).

17. McDonald, J. P., McClelland, A. A., Picard, Y. N. & Yalisove, S. M. Role of a native oxide on femtosecond laser interaction with silicon (100) near the damage threshold. Appl. Phys. Lett. 86, 264103-1-264103-3 (2005).

18. McDonald, J. P. et al. Femtosecond-laser-induced delamination and blister formation in thermal oxide films on silicon (100). Appl. Phys. Lett. 88, 153121-1-153121-3 (2006).

19. Gamaly, E. G., Rode, A. V., Luther-Davies, B. & Tikhonchuk, V. T. Ablation of solids by femtosecond lasers: Ablation mechanism and ablation thresholds for metals and dielectrics. Phys. Plasmas 9, 949 (2002).

20. Jing, X. et al. Modeling validity of femtosecond laser breakdown in wide bandgap dielectrics. Appl. Surf. Sci. 258, 4741-4749 (2012).

21. Wang, Y. et al. Visible photoluminescence of Si clusters embedded in silicon nitride films by plasma-enhanced chemical vapor deposition. Phys. E 27, 284-289 (2005).

22. Rublack, T. & Seifert, G. Femtosecond laser delamination of thin transparent layers from semiconducting substrates. Opt. Mater. Express 1, 543-550 (2011).

Chapter 3 Femtosecond Laser Direct Hard Mask Writing for Selective Facile Micron-Scale Inverted-Pyramid Patterning of Silicon

3.1 Introduction

The micro-patterning of silicon surface has been a major field of research for several decades, motivated broadly by the trends for miniaturization in microelectronic, photonic, photovoltaic, and microfluidic applications1-3. High resolution patterning of silicon has been demonstrated with a wide range of techniques varying from well-established optical and electron beam lithography to recent alternative approaches such as nanosphere lithography4, 5, block copolymer lithography6, nanoimprint lithography7, and soft lithography8 that all require a cleanroom processing environment to achieve defect free patterning. In contrast, laser-materials processing is rapidly emerging as a versatile method for non-contact micro-processing of a variety of materials, including silicon, offering high reproducibility in a non-cleanroom environment.

Direct laser machining of silicon wafers has been widely studied3, 9-13, particularly with nanosecond duration pulses, offering flexible means for surface structuring, but with limitations due to damage, amorphization or ablation debris. Hence, many groups have applied isotropic chemical etching to remove such laser damage and thus facilitate micro-scale patterning of silicon such as a honeycomb array of holes to sizes of 30 m9. Alternatively, anisotropic KOH etching is attractive since it provides clean (111) surfaces on c-Si (001) in the form of inverted- pyramid structures, but this requires a conformal mask to be patterned without inducing undue

damage to the underlying silicon. While patterned photoresist is used to direct chemical etching of c-Si into patterns of inverted-pyramids1, 14, 15 photoresist coatings are not easily extensible to silicon wafers with thickness less than 100 m, unless post-pattern wafer thinning is used1.

Hence, hard mask patterning techniques that avoid spin coating are desirable.

3.2 Laser Patterning of a Hard mask

Many approaches in laser patterning of hard masks have been tried, but laser interactions typically create large heat affected zones (HAZ) of damaged amorphous or cracked silicon that

16, 17 degrade high resolution patterning. For example, Niinobe et al. studied laser ablation of SiNx films at 355 nm to take advantage of high UV opacity, and wrote high contrast hard mask patterns that were then transferred into multicrystalline Si by isotropic etching. Although the

40 ns laser interaction extended thermal damage up to 2 m into the underlying Si substrate, the acid etching was able to remove the damage, at the expense of reduced resolution. In contrast, similar substrate damage by 200 ns pulses at 1064 nm inhibited KOH etching, preventing the full formation of inverted-pyramids18.

A favourable approach to reducing laser damage is through direct ablation with ultrashort (ps and fs) pulsed lasers19. While such short pulse interaction has been very effective for high resolution patterning of both opaque and transparent thin films on silicon20-22, a more desirable approach to reduce HAZ follows for transparent dielectric films with ultra-short pulses where laser absorption and interaction is directed at the dielectric-substrate interface. Such narrow-zone interaction leads to a favourable blistering or delamination of the layer at low exposure and film ejection at higher exposure, as demonstrated first in silicon oxide films by

Yalisove et al.23, 24 and extended to other transparent dielectric films by Rublack et al.25. A

further benefit of fs-laser delamination and ejection is the low damage that largely maintains the crystalline structure of the underlying silicon when an appropriate fluence is used26.

The objective in this work is to extend fs-laser induced localized blistering to eject small submicron apertures within the hard-mask films with minimal collateral damage in the underlying c-

Si substrate, so that high resolution pattern transfer is possible by anisotropic etching. Using an appropriate combination of laser fluence and dielectric film thickness, we create a pattern of high resolution apertures in a SiNx hard mask layer. The c-Si wafer is then subjected to an optimized anisotropic etch procedure to render an almost defect-free pattern of inverted-pyramids on the micron scale. This flexible top-down patterning control can create connected inverted pyramids or arrays in a range of feature sizes to serve in applications such as low-reflectivity PV texturing and marking or labelling, as well as a new approach for creating microfluidic components, optical devices or laboratories on silicon wafers.

3.3 Texturing Method

A schematic of the inverted-pyramid patterning method is shown in Figure 3-1, beginning with laser ejection of thin dielectric film in a desired pattern (Figure 3-1(a)), followed by selective anisotropic etching (Figure 3-1(b)) of c-Si to form inverted pyramids and finishing with HF removal of the dielectric hard-mask (Figure 3-1(c)). By varying the laser written patterns, laser exposure and etching conditions, various structures such as inverted-pyramid arrays, V-channels, and V-reservoirs of varying size can be created, as shown in Figure 3-1(c).

Figure 3-1 A schematic illustrating (a) selective removal of thin dielectric film coated on crystalline silicon (001) by 522 nm femtosecond laser ejection to form a pattern of open apertures and (b) the resulting inverted-pyramid structure following anisotropic etching of exposed silicon. Variation in the laser written aperture pattern spacing, laser exposure and KOH etching time yield (c) V-channels, V-reservoirs, and arrays of inverted pyramids of diverse sizes as seen following HF removal of the dielectric mask.

SiOx and SiNx dilectric layers with thickness < /4nfilm (to avoid the Fabry-Perot fringe effect discussed in Chapter 2) were tested as hard masks for high resolution mask writing and successful selective etching of silicon. Owing to the high etching rate of 18nm/min of SiOx in

KOH compared with 1.2nm/min for SiNx, a minimum of 100nm thick SiOx and 20nm thick SiNx were required for hard-masking Si during KOH etching. Hence, 20-100 nm thick PECVD SiNx films and 100 nm thick PECVD and commercial thermally grown Silicon oxide (SiOx) films were tested to establish fluence thresholds and exposure windows that produced the smallest ablation craters in the hard mask with high reproducibility and minimal collateral damage to expose Si for alkaline etching.

SiNx films were grown on 100 mm diameter (100) c-Si wafers by Plasma Enhanced

Chemical Vapor Deposition (PECVD) in a PlasmaLab 100 PECVD system (Oxford Instruments) at 300 C and 650 mT chamber pressure using a gas mixture of 5% silane in nitrogen (400 sccm), ammonia (20 sccm) and pure nitrogen (600 sccm). The film deposition was carried out at the rate of 14 nm/minute using alternating combinations of high frequency (13.56MHz) power for 13 seconds and low frequency (100 KHz) power for 7 seconds. The RF power was set to 50 W and

40 W for high and low frequencies, respectively. For SiOx, PECVD procedure was modified to

1000mT chamber pressure and a gas mixture of 5% silane in nitrogen (170 sccm) and nitrous oxide (710sccm). The deposition was carried out at the rate of 55nm/minute by using 30W RF power. The film thickness was determined by ellipsometry and subsequently confirmed by cross- sectional SEM imaging.

The femtosecond fiber laser (FCPA µJewel D-400-VR, IMRA) beam was frequency doubled to 522 nm via second harmonic generation in Lithium Triborate (LBO) crystal, and

applied at 100-kHz pulse repetition rate to avoid cumulative multi-pulse heating effects by such rapidly arriving laser pulses. The beam was linearly polarized and of 200 fs pulse duration and

M2 = 1.3 beam quality. By monitoring the back reflection on a CCD camera, a plano-convex lens of 8 mm focal length (5724-H-A New Focus) was positioned to precisely focus the near-

Gaussian-shaped laser beam to a spot size of 1.25 μm diameter (1/e2) onto the sample surface that was mounted on a XY motorized stages (ABL1000, Aerotech). A linear polarizer and waveplate power attenuator varied the single pulse energy in a preferential range of 0.2 to 2.0 nJ while the number of pulses applied per site was varied from 1 to 20 with an acoustic optical modulator. These exposure ranges drove various levels of surface modification and ejection of the dielectric thin film that depended on the film thicknesses. The laser exposure was tuned for formation of the smallest mask apertures in the film while at the same time causing minimal damage in the underlying c-Si substrate.

The morphology of the laser patterned SiNx films was studied with optical, atomic force microscopy (AFM), and scanning electron microscopy (SEM). The laser exposed wafers were then submerged in 30% KOH at 60 oC for 30s to remove ablation debris, cleaned and washed in deionized water, followed by nitrogen drying. The samples were KOH etched again, in repeated steps of 2 to 10 minutes to form inverted pyramids, and finally finished with a 30 seconds dip in

1% HF to remove the hard-mask. The optimization of the anisotropic etching of laser exposed silicon is detailed in Appendix B.

3.4 Optimization of Single Pulse Laser Exposure and Hard Mask Thickness for High Resolution Writing

The fluence (F) threshold for blistering and ejection of SiNx film with single laser pulses was found to increase strongly from 0.29 Jcm-2 and 0.41 Jcm-2, respectively, for a 20 nm thick film to

0.53 Jcm-2 and 0.65 Jcm-2, respectively, for a 100 nm thick film. Over this 20 to 100 nm thickness range, the minimum open aperture diameter in the SiNx film increased monotonically from 0.6 µm to ~2 µm. Generally, higher fluence was necessary to generate larger internal ablation pressures to delaminate and lift thicker films, albeit at the cost of creating larger diameter craters and damage to the c-Si substrate. As an example, the top SEM view (Figure

3-2(a) (i), (iii)) and the corresponding ion-milled cross-section view (Figure 3-2(a) (ii), (iv)) show the mask features at this minimum etching aperture as produced in 70 nm (Figure 3-2(a)

(i), (ii)) and 20 nm (Fig. 2(iii), (iv)) thick SiNx. The mask patterning by laser ejection is shown to yield shallow craters in the underlying Si substrate and small diameter open apertures in SiNx film varying from 40 nm and 0.6 m, respectively, for 20 nm thick film to 340 nm and 1.45 m, respectively, for 70 nm thick film. The blister dynamics observed over this range of SiNx thicknesses with increasing laser fluence were similar to that reported 25 for a 100 nm silicon oxide film; the blister grows in diameter until a threshold for the perforation of the blister is reached, at which point a nano-hole is opened, followed by mechanical ejection of the blistered

SiNx film at higher fluences. The morphology of these features was confirmed for the case of the

20 nm film as shown by the 0.29 to 0.45 J/cm2 fluence sequence of SEM and AFM top views together with the AFM line profiles shown in Figure 3-2(b) (i), (ii) and (iii), respectively. Here, blistering, blistering with a nano-hole, collapsing of a blister with a nano-hole, and ejection of

the blister are demonstrated at F = 0.29, 0.31, 0.41, and 0.45 Jcm-2, respectively, while similar morphology in thicker films required higher fluence exposure.

Figure 3-2 (a) SEM top views ((i) and (iii)) and ion-milled cross-section views ((ii) and (iv)) of the laser ejected mask aperture and ablation crater produced in 70 nm ((i) and (ii)) and 20 nm -2 -2 ((iii) and (iv)) thick silicon nitride (SiNx) film at 0.52 Jcm and 0.45 Jcm fluence, respectively.

(b (i)) SEM images of laser modifications in a 20nm SiNx film on crystalline silicon (c-Si) with increasing laser fluence showing (left to right) blistering, blistering with a nanohole, collapsed blister and ejected blister together with the corresponding (b (ii)) atomic force micrographs and (b (iii)) line profiles.

The observed single pulse fluence thresholds and the mask aperture diameters were nearly identical in 100 nm thick thermal and PECVD SiOx films in comparison with SiNx films

27 of the same thickness and were in accord with results reported for 147nm SiOx films. Atomic force micrographs in Figure 3-3 show the surface morphology of 100 nm thick PECVD SiNx (i) and PECVD SiOx (ii) and thermal SiOx (iii) at the ejection threshold. All three films are ejected at the same single pulse fluence, F= 0.65 J/cm2, and result in similar mask aperture diameters of

~ 2 m.

Figure 3-3 AFM micrographs of apertures produced in 100 nm thick (i) PECVD SiNx (ii)

PECVD SiOx, and (iii) thermal SiOx at the film ejection threshold. The single pulse threshold fluence for ejecting a 100 nm thick dielectric film (F= 0.65 J/cm2) is independent of tested dielectric material and deposition method and results in a mask aperture diameter of ~ 2 m in all three films. In (i) PECVD SiNx film remains partially attached to the aperture wall after ejection. Such partially opened apertures do not hinder anisotropic wet etching of the exposed silicon.

The results for single-pulse mask patterning favor the selection of thinner dielectric films for higher resolution mask formation together with benefits of smaller crater size and reduced damage in the silicon substrate. Balancing with the need for sufficient film thickness for selective masking of the c-Si during KOH etching, a SiNx film thickness of 20 nm was deemed optimal.

Further optimization of patterning the 20 nm thick SiNx mask examined the formation of inverted pyramids against the single-pulse fluence exposure as shown in Figure 3-4(a). The blistered SiNx layer exhibiting intact, perforated and collapsed morphologies was found to protect the underlying Si from KOH etching, whereas the unprotected Si crater formed due to blister ejection (F ≥ 0.42 Jcm-2) resulted in typical inverted-pyramid structure after KOH treatment. The graph (Figure 3-4(a)) shows the open aperture diameter in SiNx and the resulting inverted-pyramid width observed after KOH etching for 2.5 minutes to increase from 0.6 µm and

1.13 µm, respectively, at 0.43 Jcm-2 to 1.25 µm and 1.31 µm, respectively, at 0.53 Jcm-2 fluence exposure. The amount of undercutting during the KOH etch increases dramatically as the fluence decreases towards the ejection threshold, below which only laser blistering without silicon etching was observed. The decrease in undercutting is attributed to the decrease in laser-induced

Si damage at low fluence. The undercutting was essential to etch beyond the laser damage zone in the c-Si surface to expose damage free (111) planes as seen in the SEM image of Figure 3-4(a)

(inset) for the case of 0.45 Jcm-2 exposure. Otherwise, Si damage is apparent in the SEM images

(Figure 3-4(a) inset) for the higher exposure cases of 0.49 and 0.53 Jcm-2. These cases required longer etching times to compensate for the additional damage but with the trade-off of forming larger sized inverted-pyramid structures that facilitated a useful reproducible tuning of clean inverted-pyramid sizes in the 1.07 to 2.0 m range, which can easily be extended to larger sizes with longer etch times. On the other hand, for the case of 0.45 Jcm-2 exposure, Figure 3-4(b) (i) shows the formation of (111) facets impeded by a small laser damage zone at 1 minute of KOH etching, but which open into clean inverted pyramid structure after 2 minutes of KOH etching

(Figure 3-4(b) (ii)) to yield after HF mask removal the smallest (1.13 m) clean inverted- pyramid as seen in Figure 3-2(b) (iii). Such structures could be reproducibly formed at 0.45 Jcm-

2 whereas smaller 1.07 m wide inverted pyramids were formed at 0.43 Jcm-2 with only 33% occurrence (low reproducibility zone in Figure 3-4(a)). Hence, 1.13 m is the smallest size available in single pulse laser hard-mask ejection to reproducibly form arrays of inverted pyramids with the tightest packing density.

3.5 Hard Mask Writing with Multiple Pulse Exposure

The optimization of SiNx film ejection with varying number of laser pulse numbers shows that clean inverted pyramids could also be formed after KOH and HF etching, showing a useful progression towards smaller pyramid sizes from 1.13 m at one pulse (Figure 3-4(d) (iii)) to 623 nm at 10 pulses as seen in Figure 3-4(d) (iv). The single pulse fluence for this threshold of reproducible film ejection was lowered by 74% (from 0.45 to 0.33 J/cm2) with the increase in the number of applied pulses from N = 1 to 10, attesting to potential incubation effects that damage the substrate and thus strengthen the interaction of subsequent laser pulses. Ashkenasi et al.28 observed a large 4-fold reduction in the fluence damage threshold of silica glass from one to 20 pulses with 100 fs pulse duration. However, the formation of defects and/or mechanical stress within the silicon is more likely according to Bonse et al.29, who reported a N0.16 scaling reduction of the fluence damage threshold for c-silicon with 130 fs laser pulses. This scaling predicts a 69% decrease from N = 1 to 10 pulse that closely corresponds with our 74% observation. Moreover, the additional pulses offered an attractive 45% reduction in feature size to yield the highest packing density of inverted pyramids, but with the cost of longer laser writing times.

(~33%) reproducibility ejection of SiNx are identified. (b) SEM views of (i) partially etched c-Si in KOH for 1 minute. The formation of fresh (111) planes is impeded by laser induced damage which is finally removed after 2 minutes of KOH etching to form (ii) clean 1.13 m inverted pyramid shown in (iii) finished with HF removal of dielectric mask. The inverted pyramid size can be further reduced to 623 nm by lower the fluence (0.33 J cm-2) and compensating with the increase in number of pulses (10 pulses) as shown in b (iv).

3.6 Facile Patterning for Large Area Inverted Pyramidal Texture, V-Grooves and Reservoirs

A high density grid of inverted-pyramids was fabricated using the optimized 0.45 Jcm-2 fluence for single-pulse ejection of 20 nm SiNx film coated on silicon. Scanning the 100 kHz repetition

-1 rate laser at speeds varying between 120 and 150 mms (yielding x-direction period of x = 1.2 to 1.5 µm) and line-to-line offsets (y direction) of y = 1.2 µm to 2 µm produced tightly packed two-dimensional arrays of isolated laser ejected apertures, with minimum collateral damage for

x = y ≥ 1.5 µm spacing. The densest grid of laser written apertures are shown in Figure

3-5(a), which in turn, was successfully etched into a high-fidelity array of inverted-pyramids seen in Figure 3-5(b). Within this grid size, the inverted-pyramid size could be reproducibly varied from 1.13 µm to 1.4 µm by increasing the KOH etching time from 2.5 to 5 minutes, leaving 370 nm to 100 nm wide flat mesas, respectively, between the inverted-pyramids. Larger area SEM observation of the inverted-pyramidal array did not reveal any pyramid defects over our whole sample set (~14,000 holes viewed) in spite of a non-cleanroom processing environment, showing the reproducibility of this technique, with less than one defect per 104 holes. To demonstrate flexible top-down pattern control, the University of Toronto crest was written on c-Si (Figure 3-5(c)) where the crest is clearly delineated by areas of high and low reflectivity corresponding to un-patterned and inverted-pyramid structure, as shown in the magnified inset.

Additionally, by varying the spacing and arrangement of the laser written apertures, structures such as a V-channel (Figure 3-5(d) right) and a V-reservoir (Figure 3-5(d) left) can be created. They resulted from fusion of 1×2 and 4 ×4 arrays of neighboring pyramids. Such

structures are extensible to much larger lengths or sizes, enabling microfluidic channels and reservoirs to be flexibly patterned into the c-Si substrate.

Figure 3-5(a) SEM top views of a crystalline silicon wafer with a grid of laser ejected apertures at 1.5 μm periodic spacing and following (b) KOH etching, yielding an array of 1.3 μm inverted- pyramids. (c) Optical micrograph of University of Toronto crest pattern crest patterned with areas of high and low reflectivity corresponding to un-patterned and inverted-pyramid structure as seen in the SEM micrograph (outlined in red). (d) SEM images of 4x4 and 2x1 inverted pyramid arrays that have coalesced into a V-reservoir (left) and V-channel (right), respectively, as controlled by the laser pattern and KOH etching time.

3.7 Discussion

The success of the above described technique for facile patterning of variable sized inverted pyramids and V-channels or reservoirs centers on the use of femtosecond laser processing and

optimized laser fluence exposure for the ejection of thin transparent dielectric layers that leads to low thermal damage in underlying silicon; an essential requirement for effective alkaline etching of crystalline silicon. Through a series of systematic experiments we have optimized fluence exposure on SiNx film that was scaled to only 20 nm thickness, to achieve ~1 m mask apertures with minimal damage to the c-Si substrate and thereby offers high resolution patterning of c-Si.

At optimized fluence exposure, KOH solution effectively etches the c-Si substrate through the laser ejected mask apertures to form (111) facets, thus confirming the minimal presence of amorphous silicon within the laser formed crater. Evidence of such minimal laser damage in c-Si substrate after the femtosecond laser removal of thin dielectric coatings was also recently reported by Rublack et al.26

The present areal writing rate of ~0.1 mm2s-1 for single pulse aperture writing can be increased dramatically (×105) by scaling to higher repetition rate, using beam multiplexing, and adopting more powerful ultrafast lasers that are available commercially today. In this way, laser patterning of 200 cm-2 hard mask can be anticipated in 1 second exposure times, for purposes of low-reflectivity texturing in photovoltaics through to patterning of microfluidic and optical components with ~ 1 m resolution. With multi-pulse laser ejection, SiNx apertures as small as

~ 400 nm were formed, offering much higher resolution than the >10m features previously demonstrated in laser ejected dielectric films27, 30. The smaller inverted pyramids of 623 nm demonstrated here open opportunities of writing optical gratings, multiplexers and diffusers for silicon photonics in the near-IR spectrum.

3.8 Conclusions and Outlook

In summary, patterning of dielectric films by femtosecond laser ejection was extended to enable low-damage and high resolution hard mask formation on c-Si, and thus open a new direction for alkaline etching of silicon into inverted pyramid structure in controllable patterns or arrays at ~1

m resolution. This approach is extensible to thin silicon wafers that cannot withstand spin- coating of traditional resist masks. The facile laser patterning of small to large V-channels and

V-reservoirs further facilitates micron-scale integration of microfluidic, photonic, and MEMs structures onto silicon microelectronic circuits while the formation of dense ~ 1 m periodic arrays of isolated inverted pyramids offer new means for silicon marking and labeling as well as texturing of low-reflectivity photovoltaic cells.

Chapter 3 References

1. Mavrokefalos, A., Han, S. E., Yerci, S., Branham, M. S. & Chen, G. Efficient Light Trapping in Inverted Nanopyramid Thin Crystalline Silicon Membranes for Solar Cell Applications. Nano Lett. 12, 2792-2796 (2012).

2. Jokerst, N. M. et al. Progress in Chip-Scale Photonic Sensing. IEEE Trans. Biomed. Circuits Syst. 3, 202-211 (2009).

3. Kam, D. H., Shah, L. & Mazumder, J. Femtosecond laser machining of multi-depth microchannel networks onto silicon. J Micromech Microengineering 21, 045027-1-045027-8 (2011).

4. Sun, C. H., Min, W. L., Linn, N. C., Jiang, P. & Jiang, B. Templated fabrication of large area subwavelength antireflection gratings on silicon. Appl. Phys. Lett. 91, 231105-1-231105-3 (2007).

5. Sun, C. H., Jiang, P. & Jiang, B. Broadband moth-eye antireflection coatings on silicon. Appl. Phys. Lett. 92, 061112-1-061112-3 (2008).

6. Qiao, Y., Wang, D. & Buriak, J. M. Block Copolymer Templated Etching on Silicon. Nano Lett. 7, 464-469 (2007).

7. Hauser, H. et al. Honeycomb Texturing of Silicon Via Nanoimprint Lithography for Solar Cell Applications. IEEE J. Photovoltaics 2, 114-122 (2012).

8. Shestopalov, A. A. et al. Soft-Lithographic Approach to Functionalization and Nanopatterning Oxide-Free Silicon. Langmuir 27, 6478-6485 (2011).

9. Abbott, M. & Cotter, J. Optical and electrical properties of laser texturing for high-efficiency solar cells. Prog. Photovoltaics 14, 225-235 (2006).

10. Bärsch, N., Körber, K., Ostendorf, A. & Tönshoff, K. H. Ablation and cutting of planar silicon devices using femtosecond laser pulses. Appl. Phys. A 77, 237-242 (2003).

11. Dobrzański, L. A. et al. Laser surface treatment of multicrystalline silicon for enhancing optical properties. J. Mater. Process. Technol. 201, 291-296 (2008).

12. L.A. Dobrzański & Drygała, A. Laser texturization in technology of multicrystalline silicon solar cells. Journal of Achievements in Materials and Manufacturing Engineering 29, 7-14 (2008).

13. Kam, D. H. & Mazumder, J. Three-dimensional biomimetic microchannel network by laser direct writing. J Laser Appl. 20, 185-192 (2008).

14. Zhao, J., Wang, A., Altermatt, P. P., Wenham, R. S. & Green, A. M. 24% efficient perl silicon solar cell: Recent improvements in high efficiency silicon cell research. Sol. Energ. Mat. Sol. C. 41-42, 87-99 (1996).

15. Hantschel, T. et al. Tip-on-tip: a novel AFM tip configuration for the electrical characterization of semiconductor devices. Microelectron. Eng. 46, 113-116 (1999).

16. Niinobe, D. et al. Large-size multi-crystalline silicon solar cells with honeycomb textured surface and point-contacted rear toward industrial production. Sol. Energ. Mat. Sol. C. 95, 49- 52 (2011).

17. Morikawa, H., Niinobe, D., Nishimura, K., Matsuno, S. & Arimoto, S. Processes for over 18.5% high-efficiency multi-crystalline silicon solar cell. Curr. Appl Phys. 10, 210-S214 (2010).

18. Alavi, M., Fabula, T., Schumacher, A. & Wagner, H. -. Monolithic microbridges in silicon using laser machining and anisotropic etching. Sensor Actuat. A-Phys 37, 661-665 (1993).

19. Küper, S. & Stuke, M. Femtosecond uv excimer laser ablation. Appl. Phys. Lett. 44, 199-204 (1987).

20. Kim, J. & Na, S. Metal thin film ablation with femtosecond pulsed laser. Opt. Laser Technol. 39, 1443-1448 (2007).

21. Sonntag, S., Roth, J., Gaehler, F. & Trebin, H. -. Femtosecond laser ablation of aluminium. Appl. Surf. Sci. 255, 9742-9744 (2009).

22. Venkatakrishnan, K., Tan, B. & Ngoi, B. K. A. Femtosecond pulsed laser ablation of thin gold film. Opt. Laser Technol. 34, 199-202 (2002).

23. McDonald, J. P., McClelland, A. A., Picard, Y. N. & Yalisove, S. M. Role of a native oxide on femtosecond laser interaction with silicon (100) near the damage threshold. Appl. Phys. Lett. 86, 264103-1-264103-3 (2005).

24. McDonald, J. P. et al. Femtosecond-laser-induced delamination and blister formation in thermal oxide films on silicon (100). Appl. Phys. Lett. 88, 153121-1-153121-3 (2006).

25. Rublack, T., Hartnauer, S., Kappe, P., Swiatkowski, C. & Seifert, G. Selective ablation of thin SiO2 layers on silicon substrates by femto- and picosecond laser pulses. Appl. Phys. A 103, 43-50 (2011).

26. Rublack, T., Schade, M., Muchow, M., Leipner, H. S. & Seifert, G. Proof of damage-free selective removal of thin dielectric coatings on silicon wafers by irradiation with femtosecond laser pulses. J. Appl. Phys. 112, 023521-1-023521-7 (212).

27. Tull, B. R., Carey, J. E., Mazur, E., McDonald, J. P. & Yalisove, S. M. Silicon Surface Morphologies after Femtosecond Laser Irradiation. MRS Bull. 31, 626-633 (2006).

28. Ashkenasi, D., Lorenz, M., Stoian, R. & Rosenfeld, A. Surface damage threshold and structuring of dielectrics using femtosecond laser pulses: the role of incubation. Appl. Surf. Sci. 150, 101-106 (1999).

29. Bonse, J., Baudach, S., Krüger, J., Kautek, W. & Lenzner, M. Femtosecond laser ablation of silicon–modification thresholds and morphology. Applied Physics A Materials Science & Processing 74, 19-25 (2002).

30. McDonald, J. P., Mistry, V. R., Ray, K. E. & Yalisove, S. M. Femtosecond pulsed laser direct write production of nano- and microfluidic channels. Appl. Phys. Lett. 88, 183113-1- 183113-3 (2006).

Chapter 4 Wave-optical Study of Inverted pyramidal Texture on Thick Silicon Wafers and Ultra-thin foils

4.1 Introduction

In order to meet grid-parity and compete with energy derived from fossil fuels the crystalline silicon (c-Si) Photovoltaic (PV) industry faces the challenge of lowering the cost of installed PV panels while retaining their high solar-to-electrical energy conversion efficiencies. One avenue of reducing the cost of c-Si based PV panels is to reduce the thickness of their constituent cells; the material cost of c-Si wafers represents ~ 35 percent of the cost of the installed panel1-3. In this context, a c-Si wafer which is typically 180 m thick1, 2, 4, could be replaced with less expensive thinner c-Si foils of thickness < 20 m5. However, increased optical losses in PV cells made from thin c-Si foils necessitates a light trapping scheme to be integrated into their design in order to achieve high efficiencies, in addition to minimizing Fresnel reflections at the front air-silicon interface.

4.1.1 Surface Textures for Enhanced Light Trapping

The most effective light trapping approach is to incorporate an ideal rough surface for

Lambertian scattering at the front or both surfaces bounding the absorbing layer. Geometrical optics predicts that the optical path length near the absorption band edge (i.e. in the weakly absorbing region) can be increased by a maximum factor of 4n2 over that of a single pass by integrating a Lambertian surface at front and a lossless reflector at the back surface, where n is the refractive index of the absorbing medium. This 4n2 enhancement limit, also known as the

Lambertian or Yablonovitch limit6-8, suggests the possibility of a ~50 fold increase in the path length in silicon (n ~ 3.5) and if achieved, the prospect of a few micron thick c-Si solar cell with similar efficiencies to current 180 m thick c-Si PV cells.

A periodically structured surface or grating is one type of optical structure that can be used to enhance light absorption within a cell. It was first proposed by Sheng et al. in 1983 for amorphous-silicon thin film cells9 and later studied by Heine and Morf for wafer-based high efficiency c-Si cells10. Prior studies have shown that a high performance grating structure consists of tapered features highly packed in a two dimensional array with grating period slightly smaller than the wavelength range of interest11-13. The tapered features reduce reflection losses at the front surface by producing a refractive index profile that gradually decreases from n ~ 3.5 in the bulk of the Si wafer to n = 1 in the medium (air) directly above the wafer. Furthermore, the grating period approximating the wavelength increases the optical thickness of the cell by coupling the light into diffractive modes that propagate at large off-normal angles within the silicon wafer. In this case ray optics is not applicable and coherent wave optics analysis is required. In the wave optics regime the Lambertian limit is no longer strictly applicable and absorption can surpass the Lambertian limit in a limited spectral range at normal incidence at the cost of absorption at other incident angles, keeping the angle-averaged absorption below the

Lambertian limit. 9, 11, 13, 14.

4.1.2 Inverted Pyramidal Grating Texture

Grating structures with hemispherical12, triangular15, conical16, pyramidal11, 17, inverted pyramidal18, nano-holes19 and rod-like19 features have been previously studied, aimed at maximizing optical absorption within a cell. By comparison, an inverted pyramid grating offers

good light trapping and also minimally increases the surface area upon texturing which is required to achieve low surface recombination losses. The inverted pyramidal features decrease surface reflection by providing a graded refractive index from air to silicon. Furthermore, since the fabrication steps are easy to implement, inverted-pyramidal grating textures have been applied on both thick and thin crystalline silicon using various fabrication methods18, 20-23 to demonstrate high efficiency PV cells 18, 20, 24. Prior theoretical and experimental studies on inverted pyramid grating texture have been focused on specific grating periodicities and feature size on either very thick (~ 400 m)22, 23, 25 or very thin silicon membranes (< 50 m)18, 24. In

2009, 20.7% efficient solar cells were demonstrated by Kray et al. on 40 m thick high quality c-Si with an inverted pyramid grating of period24, ~2 m and recently, Mavrokefalos et al. employed a 700 nm period inverted pyramidal grating to enhance absorption in10 m thick c-Si foils anticipated to produce ~25% efficient cells 18. However, heretofore the understanding of the precise interplay between the grating periodicity, wafer thickness, and absorption enhancement is still unclear.

4.2 Wave optical Study of Inverted Pyramidal Texture on Thick c-Si wafers and Thin Foils

We use a wave-optical approach to study inverted pyramidal textures with grating periods ranging from subwavelength to ~ 2 × typical wavelengths (300 – 1100 nm) on the front surface of 2 μm to 400 μm thick crystalline silicon with the objective of maximizing photoabsorption.

We first study absorption in a planar silicon slab of thickness ranging from 2 – 400 m without and with a perfect back reflector (a perfect mirror) under front-side illumination, as shown in

Figure 4-1(a) and (b), respectively, to identify the weakly absorbing spectral region. This sets the

reference point for further calculations. We then add an inverted pyramidal texture of grating period () at the front surface of the wafer and optimize the periodicity to reduce reflection and increase light trapping at the same time. We evaluate the performance of the grating texture based on the maximum calculated photocurrent produced. The aspect ratio (width of inverted pyramid: depth) is kept constant at 1.404, which is typical of inverted pyramids produced by anisotropic chemical etching of Si(001) wafers. The cell structure considered in this case is depicted in Figure 4-1(c). The photoabsorption in textured silicon subjected to normal incidence

AM1.5 solar spectrum26 is calculated in the wavelength () range from 280 - 1107 nm using the scattering matrix method, 27, 28 (detailed in Appendix C) and the optical constants of silicon29.

The photocurrent is calculated with the assumption of 100% internal quantum efficiency, i.e. each absorbed photon with energy greater than the silicon band gap produces one electron-hole pair. To ensure accuracy of the solution, the individual layer thicknesses and the number of modes in the calculations were set to /20 and 121, respectively. Later we investigate photocurrent loss due to flat regions (mesas) in the texture (Fig. 4-1(d)), which are unavoidable for practical reasons, and optimize anti-reflective coating on the front surface to cover the optical losses due to the mesas (Fig. 4-1 (d)). Our work identifies a “one size fits all” front texture that leads to maximum photo-absorption in thin and thick wafers. With the identification of such texture, a common fabrication process can be designed to manufacture high-efficiency devices on c-si of various thicknesses in the industry.

Figure 4-1 Schematic depiction of the solar cell architectures investigated in this work. A planar silicon slab in air (a) without and (b) with a perfect back surface reflector (BSR, orange), with front inverted pyramid texture (c) without mesa; filling fraction = 1 (d) with mesa; fill fraction <1 (e) with mesa and an antireflective coating (ARC, blue) and BSR with front side illumination.

4.2.1 Optical Loss in Bare 2 – 400 m Thick c-Si wafers

First we study spectral losses in a standard 400 m thick silicon wafer to understand the optical loss mechanisms. Then we extend our knowledge to identify weakly absorbing spectral regions in thin silicon wafers. This is particularly helpful in understanding the physics underlying the absorption enhancement by a grating texture. The spectral reflectance (R), transmittance (T) and total optical loss (R+T) for a 400 m thick planar slab of crystalline silicon in 300 nm -

1100 nm wavelength region is shown in Figure 4-2(a). Light with wavelength between 300 -

924 nm is almost completely absorbed (T < 0.1%) and integrated optical losses of 39% over this spectral region is solely due to the reflection of light at the front air-silicon interface. The transmittance increases to ~ 16 % in the 924 - 1107 nm wavelength region, showing insufficient silicon thickness to absorb these wavelengths due to their large penetration depths. Such weakly absorbed light gets partially reflected inside the slab from the rear silicon-air interface and leaves the slab, causing optical loss from the rear surface of the slab. In this spectral region, the loss plots (R, T, R+T) in Figure 4-2(a) display a series of peaks that arise due to Fabry-Perot interference patterns generated from the internal reflection of electromagnetic waves from the 64

two interfaces bounding the silicon wafer. Thus, the presence of these Fabry-Perot peaks in the spectral loss plot indicates the incomplete absorption of wavelengths in the available thickness of active layer and loss from the rear surface in addition to the optical loss from front surface due to reflection. The front grating of a silicon solar cell needs to be designed to facilitate diffraction in this spectral region. In contrast, the reflection spectra does not exhibit Fabry-Perot interference patterns for wavelengths between 300 – 924nm, which is indicative of the fact that the optical path length through the 400 m thick Si wafer is sufficient for complete absorption of photons entering the wafer. However, it is noted that the graded index profile of the front texture enhances absorption over this spectral region by reducing reflection losses.

The spectral losses for 2 - 400 m thick planar silicon slabs for the cases in which it has no reflector and a hypothetical perfect reflector at the rear surface, are plotted in Figure 4-2(b) and 4-2(c), respectively. In the latter case we assume no absorption at the back reflector and the spectral losses are due to reflection and escape of unabsorbed light from the front surface. With the addition of a the perfect reflector at the back, the amplitude of Fabry-Perot peaks increases because of the increase in the reflection from the mirror at the rear surface and the integrated

AM1.5 loss only decreases to 62 % compared with 69 % in the case with no back reflector.

However, for both cases, the weakly absorbing region extends into the visible region of the spectrum with decreasing silicon thickness. This is indicated by the presence of sharp peaks associated with Fabry-Perot resonance modes in the spectral loss plots. For example, this region extends from 420 nm – 1100 nm in the case of a 2 m thick silicon membrane.

The situation remains the same when we consider absorption of AM1.5 solar spectrum in the given thickness (plotted in in Figure 4-2(d)). The optical loss extends into the visible region

and increases as the silicon thickness is reduced from 400 m to 2 m. The weighted absorption calculated by normalization of the absorption spectrum with the AM1.5 spectrum in the 300 –

1107 nm wavelength range decreases from 62.4 % to 37.8 % with the decrease in the silicon thickness from 400 m to 2 m, respectively.

The above observations suggest that the front grating structure should be optimized to facilitate diffraction in the 900-1100 nm range for 400 m thick Si, and over a broader 450- 1100 nm region for 2 m thick silicon. The front texture should also reduce reflection over a broad wavelength scale at the same time. Sai et al.30 suggested submicron scale – 0.8 m periodicity in the array for the features with aspect ratio close to unity for broad-band anti-reflection.

However, such structures are not good diffractive elements for longer wavelengths with large penetration depths in silicon. For this reason, Wang et al.16 suggested a dual (front and back) surface texturing approach and optimized the front and rear grating for broad-band antireflection and trapping long wavelengths, respectively. Dual surface texturing not only adds to fabrication costs but it is difficult to control the relative alignment of front and back textures. The relative alignment of front and back gratings is reported to have an effect on the efficiency of the cell17.

Therefore we only consider front inverted-pyramidal grating and vary its periodicity with the aim to maximize photocurrent in the given thickness of silicon.

Figure 4-2 (a) The spectral reflectance (R), transmittance (T) and R+T loss in 400 m thick wafer. Spectral loss due to reflectance and transmittance in 2 - 400 m thick silicon (b) without and (c) with a perfect back surface reflector. (d) Absorption of AM1.5 spectrum in 2- 400 m thick silicon with a perfect reflector at rear surface.

4.2.2 Optimal Front Texture Parameters for Maximum Absorption in 2 – 400 m Thick c-Si wafers

We now consider adding a tightly packed two-dimensional array of inverted pyramids to the front surface of the slab to enhance photoabsorption. Here we consider the cell architecture shown in Figure 4-1(c), where the inverted pyramids touch each other and there is no flat region

(mesa) in between the features, i.e. the filling fraction is 1. The calculated photocurrent in 2 -

400 m thick silicon with a perfect back reflector and grated front surface with period () ranging from 100 – 2000 nm is plotted in Figure 4-3(a). In 400 m thick silicon, the minimum

photocurrent (31.6 mA/cm2) was calculated for grating period,  = 100 nm. The photocurrent increased to a local maximum value of 40.8 mA/cm2 as the period is increased to nm, and saturated at 41.2 mA/cm2 for grating periods above  = 1500. A similar trend in photocurrent was observed for all thicknesses as the grating period was increased from

tonm; the photocurrent increased with increasing grating period and was maximized for ~ 1000 nm. The photocurrent dropped for grating periods larger than ~ 1000 nm except for the case of 400 m thick silicon where gradual degradation of photocurrent was observed at very large periods beyond the periodicity scale shown in the Figure 3(a). The maximization of photocurrent at ~ 1000 nm is expected because such grating periods introduce strong diffraction in the long wavelength region to increase optical path length and thus absorption in silicon. When the grating period becomes larger than the wavelengths in the weakly absorbing spectral region, the light couples out of the cell through higher order external channels and decreases the photocurrent in the cell. This drop in photocurrent at larger grating periods became more evident in thinner silicon slabs, suggesting that the absorption is largely driven by diffraction in thinner wafers. The maximum photocurrent at the periodicity of  ~

1000 nm dropped to 30.79 mA/cm2 with the decrease in thickness of silicon to 2 m.

A second peak in the photocurrent originated at  = 850 nm for wafer thickness of 10 m producing 36.45 mA/cm2 photocurrent similar to 36.46 mA/cm2 at  = 980 nm. The position of this peak changed to  = 650 nm and  = 680 nm in 5 and 2 m thick silicon membranes, respectively. However the first peak remain positioned between  = 980-1000 nm for all thicknesses. The origin of the second peak in the photocurrent can be derived from the loss curves from different silicon thicknesses (Figure 4-2(b) – (d)). As previously discussed, with the

decrease in the silicon thickness the optical losses extend to smaller wavelengths in visible region of the spectrum. In such thin silicon, the smaller grating periods in 650- 850 nm range increase light absorption by reducing surface reflection in the broad spectrum range and facilitating strong diffraction in visible region of the spectrum. However, we find that the absorption enhancement by smaller grating periods in thin silicon is equivalent to the comparatively large  ~ 1000 nm grating period that reduce optical losses mainly by facilitating strong first order and second order diffraction in the IR and visible region of the spectrum, respectively, and produce the same photocurrent in silicon.

The photocurrent at grating periods corresponding to the two peaks in photocurrent graph of each silicon thickness with inverted pyramidal front texture (Fig. 4-3(a)) and for the ideal case when the grating structure is replaced by Lambertian rough surface is plotted in Figure 4-3(b).

The photoabsorption calculation involving Lambertian surface considers maximum 4n2 increase in the optical length of wavelengths and represents the maximum photocurrent that can be achieved in a given thickness of silicon with a textured front surface and a perfect mirror at the back in geometric optical regime. The grating period  ~ 1000 nm produced maximum photocurrent in 2- 400 m silicon thickness which is close to the Yablonovich limit values at respective silicon thickness (shown by solid colored lines above bars). The photocurrent with grating period corresponding to the second peak ( = 680 nm) is only slightly greater than the photocurrent at  ~ 1000 nm, and this difference is most pronounced for the case of 2 m thick silicon. The calculated photocurrent in 10 m thick silicon with 680 nm grating period is consisent with the calculated short-circuit current reported by Mavrokefalos et al.18 for their 10

m thick silicon samples with 700 nm periodic inverted pyramidal front texture. From the above calculation results, we conclude  = 1000 nm as the optimum grating periodicity for inverted 69

pyramidal front texture to obtain maximum photocurrent from silicon with a back reflector. A high efficiency in a silicon solar cell can be achieved with such inverted pyramidal grating period, irrespective of the silicon thickness.

Figure 4-3 (a) Simulated photocurrent in silicon of thickness 2- 400 m with perfect back reflector and a front inverted pyramidal grating of periodicity ranging from 100 – 2000 nm. (b) Maximum photocurrent at grating period  ~ 1000 nm (wide solid bars) in 2- 400 m thick silicon and  = 850, 680 and 650 nm in 10, 5 and 2 m thick silicon, respectively, (narrow faded bars) corresponding to peak positions in (a). Solid lines above the bars represent Yablonovich limit for each silicon thickness.

4.2.3 Optical Loss Due to Mesas

The flat ridges (mesas) between the inverted pyramids in the texture as shown in Figure 4-1(d) are unavoidable irrespective of the fabrication process, due to the limited accuracy in placing the inverted pyramids precisely with respect to each other. These mesas can be reduced to a practical minimum limit of 100 nm in width in samples prepared with high precision patterning using etch masks fabricated with e-beam lithography. Such flat regions in the texture result in optical losses due to reflection and hence decrease the short circuit current in an actual cell. Figure 4-4 shows absorption in a 400 m thick silicon wafer as a function of the grating period and filling fraction,

i.e. the ratio of feature size and periodicity. Clearly, 90% of the solar spectrum was absorbed for

 = 980 nm and  = 1020 - 2000 nm (saturation region) and filling fraction = 1, when the mesa width is 0. In the case of no mesas, the drop in absorption with the increase in grating periodicity, mentioned before in the text, is evident in this graph. The photoabsoption in general decreased with the decrease in feature size at a given grating periodicity. For grating periods corresponding to maxima in the photoabsorption, i.e.  = 980 nm and  ~ 1500 nm, the absorption decreased by 30% with the decrease in filling fraction from 1 to 0.5, corresponding to the cases where the feature size is equal to and half of the grating period, respectively.

Figure 4-4 Simulated bsorption of AM1.5 solar spectrum in 400 m thick silicon above the band gap of silicon as a function of grating period () and filling fraction (inverted pyramid size/ grating periodicity) of the front inverted pyramidal grating texture.

Figure 4-5 shows the photocurrent in silicon thickness ranging from 2 - 400 m, respectively, for the optimum grating period  ~ 1000 nm with mesa width ranging from 0 - 250

nm. For all thickness, the photoabsorption dropped linearly with increasing mesa width with an average rate of 0.02 mA/cm2 /nm mesa width.

Figure 4-5 Degradation in photocurrent with the increase in mesa width in grating with optimum periodicity  ~ 1000 nm for various crystalline silicon thicknesses.

4.2.4 Optimal Antireflective Coating (ARC) Parameters for the Texture with Minimal 100 nm Wide Mesas

The loss due to mesas can be fully recovered by adding an anti-reflective coating (ARC) on the top of the texture. The cell architecture considered here is shown in Figure 4-1(e). Our simulations optimize the refractive index of the coating to 2.1 for maximum photoabsorption.

This corresponds to the materials such as silicon nitride or silicon rich silicon oxide. Figure 4-6 shows calculated photocurrent in 2- 400 m thick silicon with optimum front texture parameters

with 100 nm mesas coated with different thicknesses of ARC of refractive index 2.1. It is clear that 80 nm thick ARC will maximize the photocurrent in all thicknesses.

Figure 4-7 shows the photocurrent current in 2 - 400 m thick silicon for the case when silicon is textured with optimum front texture with no mesas, with 100 nm wide mesas and 80 nm optimized ARC. It is clear that the optimized ARC recovers the optical losses from mesas and enhances the absorption to result in photocurrent slightly larger than the case with no mesas in all thicknesses. However, the values were still below the photocurrent that can be achieved with the ideal Lambertian texture in all cases.

Figure 4-7 Simulated photocurrent current in 2 - 400 m thick silicon with optimum front texture with no mesas ( ) and with 100 nm mesa without ( ) and with 80 nm ARC ( ), and Lambertian front texture (×)

4.3 Summary

In summary, a wave-optical study of inverted- pyramidal texture on 2- 400 m thick silicon with a perfect back reflector shows that a front grating period of 1000 nm leads to maximum photoabsorption normally incident solar spectrum irrespective of silicon. The photocurrent attained with this texture lies close to the Yablonovitch limit for all silicon thicknesses. The calculated photocurrent from the 1000 nm periodic textures is close to the photocurrent that can be obtained by etching tens of microns silicon for random inverted pyramidal structures, thus is an ideal material reducing texturing solution31. Our calculations show that practically unavoidable flat

ridges in the texture leads to optical losses and decrease in the photocurrent at the rate of ~ 0.02 mA/cm2/nm mesa width. The reflection loss from mesas can be reduced with the addition of 80 nm thick antireflective coating of refractive index ~ 2.1.

Chapter 4 References

1. Powell, D. M. et al. Crystalline silicon photovoltaics: a cost analysis framework for determining technology pathways to reach baseload electricity costs. Energy Environ. Sci. 5, 5874-5883 (2012).

2. Group, C. T. M. International Technology Roadmap for Photovoltaics- Results 2012. ITRPEV. net 2 (2012).

3. Catchpole, K. R. & Polman, A. Plasmonic solar cells. Opt. Express 16, 21793-21800 (2008).

4. Ravi, K.V. (Crystal Solar, Inc.). Thin is in, but not too thin! Future Photovoltaics (2011).

5. Janssen, E. &Kleiman, R. Novel process flow and cell architecture for 10 µm thick membrane single-crystalline silicon solar cell (38th IEEE Photovoltaic Specialists Conference (PVSC), 2012 , IEEE, 2012).

6. Agrawal, M. Photonic design for efficient solid state energy conversion. (2009).

7. Deckman, H. W., Roxlo, C. B. & Yablonovitch, E. Maximum statistical increase of optical absorption in textured semiconductor films. Opt. Lett. 8, 491-493 (1983).

8. Yablonovitch, E. & Cody, G. D. Intensity enhancement in textured optical sheets for solar cells. IEEE Trans. Electron Devices 29, 300-305 (1982).

9. Sheng, P., Bloch, A. N. & Stepleman, R. S. Wavelength-selective absorption enhancement in thin-film solar cells. Appl. Phys. Lett. 43, 579-581 (1983).

10. Heine, C. & Morf, R. H. Submicrometer gratings for solar energy applications. Appl. Opt. 34, 2476-2482 (1995).

11. Han, S. E. & Chen, G. Toward the Lambertian limit of light trapping in thin nanostructured silicon solar cells. Nano Lett. 10, 4692-4696 (2010).

12. Song, Y. M., Yu, J. S. & Lee, Y. T. Antireflective submicrometer gratings on thin-film silicon solar cells for light-absorption enhancement. Opt. Lett. 35, 276-278 (2010).

13. Yu, Z., Raman, A. & Shanhui, F. Fundamental limit of light trapping in grating structures. Opt. Express 18, A366-A380 (2010).

14. Callahan, D. M., Munday, J. N. & Atwater, H. A. Solar Cell light trapping beyond the ray optic limit. Nano Lett. 12, 214-218 (2012).

15. Dewan, R. et al. Light trapping in thin-film silicon solar cells with submicron surface texture. Opt. Express 17, 23058-23065 (2009).

16. Wang, K. X., Yu, Z., Liu, V., Cui, Y. & Fan, S. Absorption enhancement in ultrathin crystalline silicon solar cells with antireflection and light-trapping nanocone gratings. Nano Lett. 12, 1616-1619 (2012).

17. Chutinan, A., Li, C. W. W., Kherani, N. P. & Zukotynski, S. Wave-optical studies of light trapping in submicrometre-textured ultra-thin crystalline silicon solar cells. J. Phys. D 44, 262001-1-262001-4 (2011).

18. Mavrokefalos, A., Han, S. E., Yerci, S., Branham, M. S. & Chen, G. Efficient Light Trapping in Inverted Nanopyramid Thin Crystalline Silicon Membranes for Solar Cell Applications. Nano Lett. 12, 2792-2796 (2012).

19. Han, S. E. & Chen, G. Optical absorption enhancement in silicon nanohole arrays for solar photovoltaics. Nano Lett. 10, 1012-1015 (2010).

20. Zhao, J., Wang, A. H. & Green, M. A. 24.5% efficiency PERT silicon solar cells on SEH MCZ substrates and cell performance on other SEH CZ and FZ substrates. Sol. Energ. Mat. Sol. C. 66, 27-36 (2001).

21. Kumar, K., Lee, K. K. C., Herman, P. R., Nogami, J. & Kherani, N. P. Femtosecond laser direct hard mask writing for selective facile micron-scale inverted-pyramid patterning of silicon. Appl. Phys. Lett. 101, 222106-1-222106-5 (2012).

22. Sun, C. H., Min, W. L., Linn, N. C., Jiang, P. & Jiang, B. Templated fabrication of large area subwavelength antireflection gratings on silicon. Appl. Phys. Lett. 91, 231105-1-231105-3 (2007).

23. Chaturvedi, N., Hsiao, E., Velegol, D. & Kim, S. H. Maskless Fabrication of Nanowells Using Chemically Reactive Colloids. Nano Lett. 11, 672-676 (2011).

24. Kray, D. & McIntosh, K. R. Analysis of ultrathin high-efficiency silicon solar cells. Phys. Status Solidi (a) 206, 1647-1654 (2009).

25. Parretta, A. et al. Angle-dependent reflectance measurements on photovoltaic materials and solar cells. Opt. Commun. 172, 139-151 (1999).

26. http://rredc.nrel.gov/solar/spectra/am1.5/.

27. Liscidini, M., Gerace, D., Andreani, L. C. & Sipe, J. E. Scattering-matrix analysis of periodically patterned multilayers with asymmetric unit cells and birefringent media. Phys. Rev. B 77, 035324-1-035324-11 (2008).

28. Whittaker, D. M. & Culshaw, I. S. Scattering-matrix treatment of patterned multilayer photonic structures. Phys. Rev. B 60, 2610-2618 (1999).

29. Herzinger, C. M., Johs, B., McGahan, W. A., Woollam, J. A. & Paulson, W. Ellipsometric determination of optical constants for silicon and thermally grown silicon dioxide via a multi- sample, multi-wavelength, multi-angle investigation. J. Appl. Phys. 83, 3323-3336 (1998).

30. Sai, H., Kanamori, Y., Arafune, K., Ohshita, Y. & Yamaguchi, M. Light trapping effect of submicron surface textures in crystalline si solar cells. Prog. Photovoltaics Res. Appl. 15, 415- 423 (2007).

31. Campbell, P. & Green, M. A. Light trapping properties of pyramidally textured surfaces. J. Appl. Phys. 62, 243-249 (1987).

Chapter 5 Experimental and Theoretical Study of Reflectance from Sub-micron to Wavelength Scale Periodic Textures

5.1 Introduction

Due to the high reflectance of bare silicon to incident light, which is around 35%, techniques to reduce reflectance is required to increase light energy that couples into the cell. Two general approaches are used, using single or double antireflective coatings, or developing geometric textures on the incident surfaces of silicon wafers 1-12. Relatively greater interest has been drawn to designing different textures to decrease wafer reflectance. Various textures have been developed such as nipples5, grooves5, holes13, rods14-16, cones13-15, honeycomb17 and pyramids3,

18. These textures provide a graded index of refraction in the near surface region as well as increasing the number of times that light reflects and the length of optical path. An inverted pyramid structure has shown excellent performance11, 18-21. Previous study has also shown the superior quality of inverted pyramid structures on minimizing reflection loss compared to upright pyramid structures 22.

In order to make such a structure, photolithography, nanoimprint, colloidal lithography, laser equipment, or electron beam lithography (EBL) have been applied 3, 16, 21, 23-25. For conventional photolithography, the resolution is limited by light wavelength. Nanoimprint and colloidal lithography are not readily scalable to industrial demands. In addition, neither of them is compatible with device manufacturing due to various limitations such as reproducibility, etc.

Chapter 3 discusses fabrication of an inverted pyramid structure on the surface of Si

(001) wafers using a combination of laser patterning of a hard mask layer and subsequent anisotropic chemical etching. As reported in that chapter, the minimum center to center spacing between the inverted pyramids was limited by a combination of factors, including the minimum mask aperture size produce by the laser pulses, thermal effects and the tolerance in aperture spacing that was set by vibration, and the accuracy of sample stage motion. This chapter reports on the fabrication of high resolution inverted pyramidal textures with electron beam lithography

(EBL) to compare the reflectance from sub-wavelength to sub-micron scale periodic textures.

Reflectance from textures fabricated by laser direct mask writing (Chapter 3) and EBL will be compared to study quality differences.

Figure 5-1 shows an example of using ultrafast laser direct write masking and chemical etching to produce inverted-pyramid structures. The periodicity and the feature size here is limited to about 1.5 m and 1.3 m, respectively, due to the reasons discussed above. Electron beam lithography can produce textures with feature sizes much less than 100 nm. For instance, the effective resolution of the EBL system at the University of Toronto can be as low as 10nm.

Figure 5-1 1.5 m periodic array of 1.3 m inverted pyramids in top and side view, produced by laser patterning and chemical etching.

5.2 Reflectance from Sub-micron to Wavelength Scale Periodic Textures

In chapter 4 the study of inverted pyramidal texture on 2- 400 m thick silicon with a perfect back reflector identified  ~ 1000 nm as the optimum periodicity to maximize photoabsorption with in the cell. Such texture with periodicity close to the weakly absorbing IR wavelengths is expected to enhance absorption by diffractive light trapping as reported in the literature 26, 27.

However, the reduction in surface reflectance by front texture also contributes towards absorption enhancement. In this chapter, we fabricate 500, 1000, and 1500 nm periodic inverted pyramidal textures in order to compare the performance of the finer textures in reducing surface reflection against the results produced by laser texturing. EBL is used to produce the initial pattern in photo resist on silicon nitride (SiNx) hard mask coated on silicon substrate. The pattern is transferred to the silicon nitride film by reactive ion etching (RIE). Finally, anisotropic etching is done on the silicon through the holes in the hard mask in order to produce the final textured surface. The reflectance of the surface is measured and compared with optical simulations. All textures are fabricated on a 400 m thick wafer to minimize the escape of unabsorbed light from the front surface. The rear surface of the wafer is not coated with a reflective metal for the same reason.

Figure 5-2 Simulated reflectance from the  = 100 nm to  = 4000 nm periodic texture of inverted pyramids with no mesas. The simulated reflectance at each periodicity shows a sharp reduction in the reflectance at wavelengths around /p, where p is an integer, for example, shown at 500 nm, 1000 and 1500 nm periodicity (dashed lines).

Figure 5-2 shows the spectral reflectance from the front  = 100 nm to  = 4000 nm periodic texture of inverted pyramids with no mesas, i.e. filling fraction = 1. The simulated reflectance at each periodicity shows a sharp reduction in the reflectance at wavelengths around

/p, where p is an integer and is the periodicity of the texture. However, the depth of minimum, i.e. trough in the reflection spectra, decreases and spreads over broad wavelength range with the increase in the grating periodicity. For example at 500 nm periodicity a single sharp minimum was observed around 500 nm wavelength, whereas for 1000 nm periodicity a broad minimum was observed extending from 1000-1040 nm wavelength range with a shallow

second minimum at ~ 500 nm wavelength, shown clearly in Figure 5-3(a) and (b), respectively.

For 1500 nm periodic texture, three very shallow minima are visible around 380, 500 and 750 nm wavelength corresponding to /4, /3 and /2 positions in the reflectance spectra shown in

Figure 5-3(c). While the sharp minimum in reflection in 500 nm periodic texture aligns with the maximum in AM 1.5 solar spectrum, the calculated solar weighted reflection (SWR) shows that the texture with = 1500 show similar performance to  = 1000 nm periodicity and outperforms the texture with  = 500 nm.

Figure 5-3 Simulated reflectance from the (a) 500 nm, (b) 1000 nm and (c) 1500 nm periodic texture of inverted pyramids with no mesas. 84

On the basis of the simulations results reported above and optimum texture parameters predicted for 400 micron thick silicon in chapter 4, we chose to pattern the silicon surface at 500,

973, and 1500 nm periodicities.

5.3 Inverted Pyramidal Texturing of c-Si wafers with Electron Beam Lithography

The optimized process for fabricating the inverted texture with EBL was as follows:

1. Deposit the masking layer of 70 nm thick PECVD SiNx on 380 nm thick Si wafer.

2. Spin-coat the deposited silicon pieces with HMDS to enhance adhesion to photoresist.

3. Spin-coat the HMDS coated pieces with EBL photoresist ZEP520A, post bake at 180°C for

3 minutes.

4. Apply EBL to define the desired pattern.

5. Develop the sample with HMDS.

6. Apply RIE to transfer the pattern to the substrate.

7. Remove photoresist ZEP520A and HMDS and use AFM to confirm total removal of

photoresist.

8. Apply anisotropic chemical etching (30% KOH at 66 – 70°C) to make the pattern inverted

pyramids.

9. Clean off SiNx layer with buffered oxide etch (BOE).

After confirmation of the proper pattern transfer by RIE, 30% KOH wet etching was done at a temperature in the range 66-70 C. Forty five seconds wet etching was the first duration tested and corresponding SEM images are shown in Figure 4 at different magnifications. In these three images, blunt square etching areas with crosses in the center are clearly seen. The dark crosses

are the result from anisotropic KOH etching on bare silicon. The blunt squares are formed by the initial electron beam writing process and since silicon nitride layer on the surface is not reactive to KOH, the etched area keeps the original written shape. The size of the blunt square can be further estimated based on the SEM image as about 300nm which coincides with the square size in the designed pattern. The light squares show where the etching is undercutting the nitride protective layer (Figure 4 (b) and (c)). Overall, the structure is seen very uniformly distributed over the sample.

Figure 5-4 shows an early trial run while the process parameters were being explored. This particular sample had 9 min of RIE and 45 seconds of KOH etching at a temperature in the range

66-70-°C. Clear, inverted pyramidal structure has been formed in the centre of each pit. The nitride layer has not yet been removed. There is some undercutting of the opening in the nitride layer that is visible as faint white squares around each opening. At this amount of KOH etching, the squares have significant space around them, so that if the nitride were to be removed, there would be significant flat areas or mesas separating every inverted pyramidal pit.

Figure 5-4 SEM images of the sample after 9 minutes RIE and 45 seconds of KOH etching.

Figure 5-5 shows a similar sample after 30s additional KOH etching. Here, there is significantly more undercutting, and the borders of the light squares are much closer together.

Figure 5-5 (a) High and (b) low resolution SEM images of the new sample after 9 minutes of RIE, photoresist removal and 75 seconds of KOH etching.

Figure 5-6 shows the sample after removal of the nitride layer by a dip of 4-6 min in

BOE. Fine inverted pyramidal texture with a lateral periodicity of 600 nm has been produced.

Figure 5-6 SEM images of a sample after 9 minutes of RIE, 75 seconds of KOH etching and

SiNx layer removal with BOE at (a) high and (b) low resolution.

5.3.1 Effect of E-Beam Dose on Etched Inverted Pyramidal Texture

One other process parameter to be varied was the electron beam dose used to produce the initial holes in the photo resist. Although it is possible in principle to produce samples such as those show in Figure 6 regardless of the size of the initial perforation in the photoresist layer, in practice, the time required for the KOH etching was too long if the final pyramid size was much larger that the hole in the photoresist. Increasing the EB dose had the benefit of increasing the initial aperture size.

Using the same EBL pattern used in the above sample, 16 doses from 180 C/cm2 to 480

C/cm2 were tested, in dose steps of 20 C/cm2, and all the samples were processed in a similar manner afterwards, with 9 minutes of RIE and 75 seconds of KOH etching and SiNx layer removal. Figure 5-7 shows the results for the 1st, 8th and 16th doses.

All of the patterns show clearly defined inverted pyramids, but the space between the pyramids decreases strongly with increasing dose, showing the decrease in the amount of undercutting required at higher electron doses. The pattern at the highest dose shows an inverted pyramidal structure with less than 30% of the surface as flat mesas. Interestingly, the width of the mesas is different on the sides versus the tops and bottoms of the pyramids. This is due to the distortion in the electron optics that is more apparent at the higher doses.

Figure 5-7 High (left) and low (right) resolution SEM images of SiNx sample prepared with electron beam doses of (a) 180 C/cm2 (b) 320C/cm2 and (c) 480 C/cm2 followed by KOH etching.

Table 5-1 shows the variation in length and width of the resultant inverted pyramid with electron dose. An average of four pits was measured for each dose. The length to width ratio of resultant inverted pyramids and percentage of etched area (ratio of area of silicon sample covered by inverted pyramids to total area of silicon sample) is plotted in Figure 5-8 and Figure 5-9, respectively.

Table 5-1 The summary of inverted pyramid parameters and etched area for each EBL dose.

Dose # Dose Length x Width y x/y Etching Area Percentage of (μC/cm2) (nm) (nm) per hole (nm2) etched area 1 180 309 291 1.063 89769 18.32% 2 200 423 400 1.058 169300 34.55% 3 220 482 423 1.141 203645 41.56% 4 240 500 428 1.167 214018 43.68% 5 260 508 479 1.061 242958 49.58% 6 280 520 504 1.033 262076 53.48% 7 300 515 472 1.092 242833 49.56% 8 320 543 502 1.081 272461 55.60% 9 340 557 521 1.071 290049 59.19% 10 360 568 517 1.099 293798 59.96% 11 380 580 501 1.159 290290 59.24% 12 400 593 542 1.095 321110 65.53% 13 420 605 554 1.094 335006 68.37% 14 440 610 551 1.108 335667 68.50% 15 460 619 559 1.107 345432 70.50% 16 480 621 557 1.116 345881 70.59%

1.18

1.14

1.10

1.06

widthratio

- to - 1.02

0.98 Length

0.94

0.90 160 200 240 280 320 360 400 440 480 Dose (μC/cm2)

Figure 5-8 Average length-to-width ratio of fabricated inverted pyramids as a function of EBL dose

80%

70%

60%

50%

40%

30%

20%

10%

Percentage area etched of Percentage 0% 160 200 240 280 320 360 400 440 480 Dose (μC/cm2)

Figure 5-9 Average percentage of etched area as a function of EBL dose.

As seen in the above figures, the etched area increases with electron dose. The etched area can be further increased by increasing the duration of the KOH etching time. However, the etching is constrained by the non-unity aspect ratio of the initial holes. Significantly more etching of the sample for the 16th dose would produce overetching in the horizontal direction before the mesas bounding the tops and bottoms of the pits were removed. The length to width ratio is somewhat random, but shows a general increase with electron dose.

5.4 Measured and Simulated Reflectance from Fabricated Samples with and without Antireflective Coating

Figure 5-10 shows total reflectance measurements for three samples and calculated solar weighted reflectance (SWR) at the three different periodicities and feature sizes (s) produced, along with optical simulations at the same texture parameters.

Figure 5-10 Simulated (dashed lines) and measured (solid lines) total reflectance for three samples at the three different periodicities () and feature sizes (s) with calculated Solar weighted reflectance (SWR) from simulated and experimental reflectance spectra in inset.

It is clear that the reflectance from the 500 nm period sample is significantly worse than the larger periodicities. This is partly because of the large percentage of flat region in 500 nm sample (filling fraction = 0.8) compared with 973 nm (filling fraction = 0.92) and 1500 nm samples (0.93). The flat ridges in 500 nm periodic sample could only be reduced to a minimum of 100 nm because of the reasons discussed earlier. For the smaller periodicity, this 100 nm mesa width represents a much larger fraction of the surface. The reflectance data for the 500 nm sample shows a dip that is roughly aligned with the dip predicted by the simulation, but the depth of the dip is far smaller than expected. The conclusion to be drawn is that from the viewpoint of reflectance, there is no particular advantage to making a texture with a period that is as small as

500 nm, i.e. close to the wavelengths of light near the peak of the solar spectrum. The 1000 and

1500 nm period patterns do better. More significantly, the larger periodicity makes EBL unnecessary to make the textured surface in the first place. A 1000 or 1500 nm period texture can be made with optical lithography and deep UV exposure, which would be much more commercially viable.

Figure 5-11 shows the total reflectance measured from the same three samples after application of a 70 nm thick layer of SiNx which acts as an antireflective (AR) coating. Although the performance of all three samples is greatly enhanced by the AR coating, the 500 nm sample still has the worst performance.

Figure 5-11 Simulated (dashed lines) and measured (solid lines) total reflectance for three samples at the three different periodicities and feature sizes coated with 70 nm SiNx film used as an antireflective coating.

5.5 Comparison of Reflectance from Samples fabricated by Laser Hard Mask Patterning and E-Beam Lithography

Figure 5-12 shows the spectral reflectance from laser and EBL fabricated textures of the same periodicity of  = 1500 nm for the cases with and without a 70 nm AR coating. The filling fraction (ratio of feature size to texture periodicity) is 0.86 and 0.93 for the case of laser and EBL fabricated samples, respectively. The calculated solar weighted reflectance from both textures show that the EBL sample performs better than laser textured samples without an antireflective coating. However, both perform similarly with ~ 70 nm antireflective coating. This suggests that

the difference in the reflectance of uncoated samples is due to larger percentage of flat area for the laser processed sample in comparison with the EBL sample. When the reflection from mesas is reduced with the AR coating they give similar results. This also suggests that the quality of textures produced with the laser hard masking writing technique is similar to the textures produced with the more expensive EBL tool.

Figure 5-12 Spectral total reflectance measured from  = 1500 nm periodic textures fabricated by laser and electron beam lithography with feature size 1300 nm and 1400 nm, respectively, for the case with and without and antireflective coating.

5.6 Summary High resolution inverted pyramidal textures were fabricated with EBL to study the reflectance from submicron to wavelength scale periodic textures. The mesa width could only be reduced to

~100 nm in the fabricated textures by using high precision lithographic method. The reflectance studies on 500, 973 and 1500 nm periodic samples show that the wavelength scale periodic inverted pyramidal textures result in lower reflectance in comparison with sub-micron scale periodicity. This result aligns with the theoretical study on reflectance and light trapping property of periodic textures by Sai et al.4, where texture periodicity of ≥ 0.8 m is suggested for textures of features with aspect ratio close to unity. The reflectance of AM1.5 solar spectrum reduced by only 0.6 % in 1500 nm periodic textures in comparison with 973 nm periodic textures proposing that ~ 1000 nm texture is optimum for reflection reduction as well as diffractive light trapping

(discussed in chapter 4). The comparison of reflectance from 1500 nm periodic samples fabricated by EBL and laser processing showed no significant difference with an antireflective coating. This confirms that the new laser hard -mask writing method produces high quality samples equivalent to expensive high precision lithography tools. Furthermore, the fact that the optimal periodicity is about one micron means that an optimal texurization is also within reach of optical or UV lithography, which would make EBL unnecessary in any case.

Chapter 5 References

1. Zhao, J. & Green, M. Optimized antireflection coatings for high-efficiency silicon solar cells. IEEE Trans. Electron Devices 38, 1925-1934 (1991).

2. Sun, C. H., Jiang, P. & Jiang, B. Broadband moth-eye antireflection coatings on silicon. Appl. Phys. Lett. 92, 061112-1-061112-3 (2008).

3. Sun, C. H., Min, W. L., Linn, N. C., Jiang, P. & Jiang, B. Templated fabrication of large area subwavelength antireflection gratings on silicon. Appl. Phys. Lett. 91, 231105-1-231105-3 (2007).

4. Sai, H., Kanamori, Y., Arafune, K., Ohshita, Y. & Yamaguchi, M. Light trapping effect of submicron surface textures in crystalline si solar cells. Prog. Photovoltaics Res. Appl. 15, 415- 423 (2007).

5. Kanamori, Y., Sasaki, M. & Hane, K. Broadband antireflection gratings fabricated upon silicon substrates. Opt. Lett. 24, 1422-1424 (1999).

6. Xu, H. et al. Biomimetic Antireflective Si Nanopillar Arrays. Small 4, 1972-1975 (2008).

7. Brueck, S. R. J., Gee, J. M., Ruby, D. S. & Zaidi, S. H. Characterization of Si nanostructured surfaces (SPIE's 44th Annual Meeting and Exhibition, NASA, Washington, DC, 1999).

8. Sainiemi, L. et al. Non-reflecting silicon and polymer surfaces by plasma etching and replication. Adv. Mater. 23, 122-126 (2011).

9. Gangopadhyay, U. et al. A novel low cost texturization method for large area commercial mono-crystalline silicon solar cells. Sol. Energ. Mat. Sol. C. 90, 3557-3567 (2006).

10. Hylton, J. D., Burgers, A. R. & Sinke, W. C. Alkaline etching for reflectance reduction in multicrystalline silicon solar cells. J. Electrochem. Soc. 151, G408-G427 (2004).

11. Parretta, A. et al. Angle-dependent reflectance measurements on photovoltaic materials and solar cells. Opt. Commun. 172, 139-151 (1999).

12. Yoldas, B. E. & O'Keeffe, T. W. Antireflective coatings applied from metal-organic derived liquid precursors solar cell coatings. Appl. Opt. 18, 3133-3138 (1979).

13. Kumaravelu, G., Alkaisi, M. M. & Bittar, A. Surface texturing for silicon solar cells using reactive ion etching technique (29th IEEE Photovoltaic Specialists Conference, Institute of Electrical and Electronics Engineers Inc., 2002).

14. Hsu, C. M., Connor, S. T., Tang, M. X. & Cui, Y. Wafer-scale silicon nanopillars and nanocones by Langmuir-Blodgett assembly and etching. Appl. Phys. Lett. 93, 133109-1-133109- 3 (2008).

15. Zhu, J. et al. Optical absorption enhancement in amorphous silicon nanowire and nanocone arrays. Nano Lett. 9, 279-282 (2009).

16. Sun, C. H., Min, W., Linn, N. C., Jiang, P. & Jiang, B. Large-scale assembly of periodic nanostructures with metastable square lattices. J. Vac. Sci. Technol. B 27, 1043-1047 (2009).

17. Zhao, J., Wang, A., Green, M. A. & Ferrazza, F. 19.8% efficient “honeycomb” textured multicrystalline and 24.4% monocrystalline silicon solar cells. Appl. Phys. Lett. 73, 1991-1993 (1998).

18. Campbell, P. & Green, M. A. Light trapping properties of pyramidally textured surfaces. J. Appl. Phys. 62, 243-249 (1987).

19. Zhu, J., Yu, Z., Fan, S. & Cui, Y. Nanostructured photon management for high performance solar cells. Materials Science & Engineering R 70, 330-340 (2010).

20. Smith, A. W. & Rohatgi, A. Ray tracing analysis of the inverted pyramid texturing geometry for high efficiency silicon solar cells. Sol. Energ. Mat. Sol. C. 29, 37-49 (1993).

21. Zhao, J., Wang, A. H. & Green, M. A. 24.5% efficiency PERT silicon solar cells on SEH MCZ substrates and cell performance on other SEH CZ and FZ substrates. Sol. Energ. Mat. Sol. C. 66, 27-36 (2001).

22. Baker-Finch, S. C. & McIntosh, K. R. Reflection of normally incident light from silicon solar cells with pyramidal texture. Prog Photovoltaics Res Appl 19, 406-416 (2011).

23. Mavrokefalos, A., Han, S. E., Yerci, S., Branham, M. S. & Chen, G. Efficient Light Trapping in Inverted Nanopyramid Thin Crystalline Silicon Membranes for Solar Cell Applications. Nano Lett. 12, 2792-2796 (2012).

24. Kray, D. & McIntosh, K. R. Analysis of ultrathin high-efficiency silicon solar cells. Phys. Status Solidi (a) 206, 1647-1654 (2009).

25. Kumar, K., Lee, K. K. C., Herman, P. R., Nogami, J. & Kherani, N. P. Femtosecond laser direct hard mask writing for selective facile micron-scale inverted-pyramid patterning of silicon. Appl. Phys. Lett. 101, 222106-1-222106-5 (2012).

26. Sheng, P., Bloch, A. N. & Stepleman, R. S. Wavelength-selective absorption enhancement in thin-film solar cells. Appl. Phys. Lett. 43, 579-581 (1983).

27. Wang, K. X., Yu, Z., Liu, V., Cui, Y. & Fan, S. Absorption enhancement in ultrathin crystalline silicon solar cells with antireflection and light-trapping nanocone gratings. Nano Lett. 12, 1616 (2012).

100

Chapter 6 Summary

In this thesis, we develop a new method that solves an important barrier in the trend towards texturing thinner silicon wafers that will be the basis for cheaper, high efficiency c-Si PV devices. The two-step technique first uses a laser to pattern a dielectric hard mask on silicon such with insignificant damage and removal of underlying silicon. The exposed Si is then etched in a

KOH solution that removes laser induced damage and exposes (111) planes to form a high quality inverted-pyramidal texture. Inverted pyramidal textures with different feature sizes and pitches were fabricated with this technique to demonstrate complete control over position and feature size. This contactless and non-cleanroom approach also offers precise control of the patterned areas, which can exclude areas for front surface contacts on PV devices. We show that minimum 1500 nm periodic textures consisting of 1300 nm wide inverted pyramids can be obtained with single pulse laser processing. The feature size and periodicity can be further reduced to submicron scale by switching to multiple pulse processing. The developed method can also be extended to pattern small to large V-channels and V-reservoirs that further facilitate micron-scale integration of microfluidic, photonic, and MEMs structures onto silicon microelectronic circuits.

In the course of developing the new texturing technique we extensively study laser processing of thin dielectric films on silicon with femtosecond  = 522 nm wavelength laser pulses. We report a novel observation of quantized internal structuring of films of thickness ≥

/4nfilm with laser, where nfilm is the refractive index of the dielectric film. We show that this is the result of the strong nonlinear interactions driven by ultra-short pulses within the Fabry-Perot interference fringes formed by multi-surface reflections. These strong interactions define narrow

101

nano-scale zones for highly resolved internal structuring of the film to open sub-wavelength internal cavities at single or multiple depths, or to eject nano-disks of film at quantized film depths. The work includes observations of color changes with ejection of film segments relevant to embossing or labeling applications, as well as a means for three dimensional surface structuring and multi-layer catapulting for creating MEMs or microfluidic devices that all hold significant new promise for scientific study and new application.

In contrast, the films of thickness ≤ /4nfilm were completely ejected from the surface of the silicon substrate assisted by laser interaction at the film-silicon interface to expose the underlying silicon and hence were suitable for the hard-mask writing technique. The observation of the confinement of laser interaction at the film-silicon interface aligns with the existing literature1-3. We optimize the hard-mask to 20 nm thick silicon nitride film and single pulse laser exposure conditions to obtain the smallest mask aperture with minimum silicon damage for effective anisotropic etching of exposed silicon into a high resolution inverted pyramidal texture.

The anisotropic etching procedure was also optimized to a two-step procedure in 30 wt % aqueous potassium hydroxide solution at 60 ᵒC for efficient etching of laser processed samples into a high fidelity inverted pyramidal texture.

We also study theoretically the inverted pyramidal grated texture of different periodicities and feature sizes on 2 - 400 m thick silicon with a perfect back reflector. The study identifies a universal texture of periodicity,  = 1000 nm, that leads to maximum photoabsorption in silicon under front illumination and normally incident light irrespective of the material thickness. We also fabricate high resolution inverted pyramidal textures with electron beam lithography. The close correspondence of the measured reflectance from electron beam and laser fabricated

102

samples suggests high quality of the surface texture fabricated with the developed technique. The experimental and simulated reflectance spectra closely correspond to each other; validating our calculations. We show that for the case of inverted pyramids  ~1000 nm periodic textures result in minimum weighted reflectance. This result is consistent with Sai et al’s suggestion of using 

> 0.5 m periodic textures with aspect ratio close to unity for reasonable reduction in surface reflection4.

There are several avenues for future work to extend the work in this thesis, both in the direction of being some of these results closer viability, and also in exploring the range of application of the laser processing results. With respect to the laser patterning for texturizing silicon for PV applications, the main stumbling block is the speed at which these patterns are written. Presently, the hard masks are written with ~ 1 m resolution at the rate of ~0.1 mm2s-1 with single pulse aperture writing. In future, more work needs to be done to scale-up the developed method to mass production level. This involves exploring higher pulse repetition rate, beam multiplexing techniques, and more powerful ultrafast lasers that are available commercially today. From the view point of commercial applicability, it is important to extend this work to high power industrial picosecond lasers. Further, with burst picosecond machining, it is may be possible to write the hard-masks with submicron resolution.

We are currently in the process of fabricating the texture on 10 m thick silicon foils in collaboration with McMaster University to experimentally study the light trapping properties of the fabricated structure.

103

The reported new method to internally structure the transparent films with femtosecond laser has opened new research directions. The immediate projects are included but are not limited to the following:

1. Optimization of the refractive index of the substrate together with multilayered dielectric

reflective coatings to strengthen the finesse of the Fabry Perot fringes and improve the

resolution and the depth control.

2. Investigate interferometric internal structuring of transparent films with picosecond

lasers.

3. Study quantum-like structuring in various single layer transparent films including liquids,

inks and paints.

4. Device fabrication including optical, micro-fluids and MEMS components.

104

References

1. Rublack, T., Hartnauer, S., Kappe, P., Swiatkowski, C. & Seifert, G. Selective ablation of thin SiO2 layers on silicon substrates by femto- and picosecond laser pulses. Appl. Phys. A 103, 43- 50 (2011).

2. McDonald, J. P., Mistry, V. R., Ray, K. E. & Yalisove, S. M. Femtosecond pulsed laser direct write production of nano- and microfluidic channels. Appl. Phys. Lett. 88, 183113-1-183113-3 (2006).

3. McDonald, J. P. et al. Femtosecond-laser-induced delamination and blister formation in thermal oxide films on silicon (100). Appl. Phys. Lett. 88, 153121-1-153121-3 (2006).

105

Appendix A Extension of Quantized Internal Structuring of Transparent Thin Films to Larger Areas

The interferometric internal structuring of a thin transparent film on a high index substrate with a laser is depicted in the supplementary video A1. The video represents the case of 500 nm thick film at an exposure greater than the threshold intensity that ejects the top three segments in sequence and then blisters the fourth (last) segment in accordance with the experimental observations in Figures 1 to 3. According to the experimental arrangement, a long depth of focus

(~ 2.3 µm) greater than the film thickness facilitates multiple reflections of the beam from the air-Si3N4 and Si3N4-silicon interfaces to form into four evenly spaced Fabry-Perot fringes at

/2nfilm = 131.8 nm. The nonlinear laser-material interactions (Eq. 1 to 5) were further found to confine the laser dissipation into flat ~ 46 nm thick disks (Fig. 2-1(b)) that align with Fabry-

Perot maxima positions and define cleavage planes to internally structure the film. In Figure 2-1, a threshold intensity (9×1012 W/cm2) was associated with the electron density reaching critical

21 -3 plasma density (ncr = 4.10 × 10 cm ). In the case of the first fringe position, such dense plasma will attenuate and reflect the incoming laser light to therefore reduce the intensity to below threshold at deeper fringe positions. Hence, for this threshold exposure, one anticipates only the top-most fringe position to cleave open into a single nano-void or to explosively eject only the top segment.

Given the flow of laser energy from above the film in Video A1, an increase in laser exposure to compensate for such plasma shielding will drive the electron density to critical at deeper fringe positions. In this way, several segments will be ejected, but delayed in time as 106

each deeper plasma disk explodes against the shock pressure of the disks explosions above. For each segment, one first anticipates, as seen in Figure 2, a nanovoid to open inside the film at the cleavage plane, and to form a thin blister, followed by perforation of the blister, and finally the ejection of a /2nfilm thick segment from the film. The sequential blistering and ejection from fringe positions deeper in the film, as depicted in Video A1, hence presents a novel opportunity to tailor the laser exposure to excite a controllable number of laser-heating zones and thereby control the film morphology and processing depth in discrete quantum steps.

One practical direction for the digital laser etching of thin films is to extend the blistering and ejection to larger and more uniform processing areas. In one approach, the laser beam was clipped to a top-hat beam profile and demagnified to ~1.5 m diameter. Although limited by the

~1 m resolution of the lens, the resulting film blistering and ejection led to the improved morphology as shown in Figure A2 in contrast with the Gaussian-shaped profile (Fig. 2-1-3).

The blistering onset for segment 1 at 93.5 mJ/cm2 (Fig. A2 (i)) is seen to give way to the partial ejection seen at 140.2 mJ/cm2 in A2 (ii) together with the underlying nanovoids at the second and fourth Fabry Perot fringe maxima positions. Hence, four segments (3 and 4 segments are fused) have been delineated in the film by the interferometric scribing at this laser fluence. Laser induced damage is evident at the Si3N4-silicon interface at the threshold fluence for blistering of

107

the first segment (Fig. A2 (i)), due to the lower damage threshold of silicon. At higher fluence of

303.8 mJ/cm2, a uniform ejection of segment 1 and 2 is noted in Figure A2(iii). At the maximum available exposure (436.2 mJ/cm2) for this beam profile, one sees no further layer ejection in

A2(iv), however, segment 5 has become more fully separated while segments 3 and 4 remain fused as previously noted for the Gaussian-beam exposures in Figures 1 - 3. One notes an improvement in complete and uniform ejection of the segments with the increase in fluence exposure.

Figure A2 shows cross-sectional SEM views of a 500 nm thick Si3N4 film exposed to a top-hat beam profile for fluences of (i) 93.5 mJ/cm2, (ii) 140.2 mJ/cm2, (iii) 303.8 mJ/cm2, and (iv) 436.2 mJ/cm2. The threshold fluence of 93.5 mJ/cm2 shows (i) the onset of blistering for the first ~ 29 nm thick segment of the film, which is seen ejected at the higher fluence exposure in (ii). Segment 1 and 2 are both removed at the higher fluences as shown in (iii) and (iv), yielding a more uniform morphology in contrast with the case of Gaussian beam exposure shown in Figure2-1 - 3.

In a second approach to large and uniform ejection zones, a uniform square beam profile

(1.5 × 1.5 m) was raster scanned in square-grid and hexagonal patterns over a 500 nm thick

Si3N4 film with varying spot-to-spot offsets and laser fluences. The exposure and spacing combination was optimized to ideally bring together uniform ejection layers with minimal collateral damage and ablation debris. Figure A3(i) - (v) shows SEM images comparing 396

108

mJ/cm2 and 339 mJ/cm2 fluence exposures on hexagonal patterns with varying 0.64 - 0.8 m offsets. With 0.8 um offset, an exposure of 396 mJ/cm2 resulted in closely packed but isolated ejections of segment 1, while larger offsets would not eject this 1st segment. Reducing the offset from 0.8 m (i) to 0.72 m triggered the onset for ejecting segment 2 while the most uniform morphology with connected ejection zones of segment 2 was found at 0.68 m offset. At the lower exposure fluence of 339 mJ/cm2, the threshold offset for initiating second layer ejection shifted from 0.72 um to 0.68 m and the optimal offset for uniform ejection shifted from 0.68 to

0.64 m. Hence, one can tune the spot-to-spot offset together with the laser fluence for high process flexibility in controlling the ejection zone morphology.

The quantum ejection of the Si3N4 film segments lead to distinct colour changes observed in the film (Fig. 4(i) - (iii) inset) that arise from thin-film interference effects. Following the

Fresnel reflection and transmission coefficients, r1, r1 , r2 and t1, and t1 , respectively, for internal (

) and external reflections at the air-Si3N4 (1) and Si3N4-silicon (2) interfaces, the reflection

109

spectrum (r) and spectral reflectance (R) for the thin film interference were calculated as a function of film thickness (d) of refractive index nfilm with equations S1 and S2, respectively, for normal incidence ( = 0ᵒ).

(S1)

(S2)

The wavelength dependence in the reflectance is found in the phase difference,

nfilmdcos(, which was calculated over the visible spectrum (lambda = 400-750) and plotted as a function of film thickness in Figure A4.

For a 500 nm thick Si3N4 film, the calculated reflectance at normal incident peaks mainly at 508 nm (green) and 412 nm (violet) wavelengths. The relative 1000 times stronger human eye response at 500 - 520 nm will favor the green wavelength dominating for this thickness, as seen in the microscope image inset in Figure S4. When a first segment is ejected, the remaining 471 nm thick film will shift intensity peaks to 630 nm (brilliant red), 470 nm (blue) and 390 nm

(violet) wavelengths. Human eye is equally responsive to 630 nm and 470 nm wavelengths and shows very low response to 390 nm wavelength light. The brilliant red color is anticipated to be dominant here given the higher spectral intensity of the 630 nm light emitted by tungsten lamp, attesting to the brilliant red color observed from the film ejected to the first segment depth (inset in Figure A4).

Color changes from deeper segments were overshadowed by optical scattering from the surface roughness, but can be expected to improve with further tuning of the laser exposure and/or chemical cleaning.

110

111

Appendix B Anisotropic etching of silicon through laser written dielectric hard mask

Wet-chemical etching has been used for processing mono-crystalline silicon (c-Si) since its introduction in the 1950’s1-4. In this technique, silicon is etched in a highly concentrated (10-50 weight %) aqueous solution of etchant at temperature typically between 40-90 ᵒC. Isotropic etchants such as HF (hydrofluoric acid) and HNA ( hydrofluoric-nitric-acetic acid mixture) etch silicon crystal planes at nearly the same rate and results in the same etching profile irrespective of silicon wafer orientation as shown in Figure B1(a). In contrast, anisotropic etchants such as potassium hydroxide (KOH), tetramethyl ammonium hyroxide (TMAH) and EDP

(ethylenediamine pyrocatechol) are orientation dependant etchants, i.e., the etch rate is different for different crystal planes. Figure B1(b) shows the result of selective anisotropic etching on

(100) and (110) silicon wafer through a hard mask. Typically, the etch rate of (100) > (110) >

(111) however it is possible to etch (110) faster than (100) by varying the etchant concentration and temperature. (111) plane is always the slowest etching plane because of its highest packing density5-7. Etch rate of crystal planes for different etchants, etchant concentrations and temperatures is given in [7].

112

Figure B1 shows cross-section of typical etch profiles in (110) and (100) wafers after (a) isotropic etching and (b) anisotropic etching.

Optimization of Wet Chemical Etching Process

This section summarizes the systematic study that was carried out to optimize the anisotropic etching process for converting exposed silicon craters into smooth square-based inverted pyramid structures and to obtain a defect-free pattern with minimum flat silicon mesas between inverted pyramidal features. Minimum silicon damage was observed in samples with thinnest dielectric hard mask, i.e., 20 nm thick silicon nitride (SiNx) at laser fluence F = 0.41

J/cm2 (threshold fluence for ejection) that resulted in ~ 1 m diameter apertures in the mask

(Figure 3-4). The spacing between the laser ejected apertures (pitch, ) was adjusted to 1.5 m in a pattern to prevent collateral thermal damage (see Chapter 3 for details). Several samples of c-Si with 20 nm thick SiNx film patterned over small patches 22.5 m x 22.5 m in area with above mentioned laser and pattern parameters were used for the etching process optimization study.

TMAH and KOH are the most commonly used anisotropic etchants in the microelectronics industry. Etching of bare silicon in 30 wt% of TMAH and KOH at 60-70 ᵒC is reported to produce the pyramids with smooth (111) planes7 and hence samples were etched in

KOH and TMAH with the same etch recipe parameters and compared for better etch quality.

Figure B2 shows post-etching scanning electron micrographs (SEMs) of a sample etched in 30 wt% KOH at 60ᵒC in a single step for 2.5 minutes. Under these conditions the sample was non-uniformly etched into smooth inverted pyramids, incomplete inverted pyramids and over etched regions (fused neighbouring inverted pyramids).

113

Figure B2 Scanning electron micrograph (SEM) of 22.5 m x 22.5 m laser patterned sample after continuous etching in 30 wt% KOH maintained at 60ᵒC for 2.5 minutes. The sample etched non-uniformly in to regions of clean inverted pyramids, smooth inverted pyramids, and incomplete inverted pyramids and over etched regions, e.g. enclosed by dashed square, polygon and rectangle, respectively. The trench outlining the patterned region is the etched marker made by removing SiNx mask at higher laser fluence.

Next, a two-step etching procedure was developed to improve etch uniformity. In the first step, the sample was dipped in KOH for 30 seconds, washed and cleaned in distilled water and dried with nitrogen. In the second step, the cleaned sample was again etched in KOH for 2 minutes followed by a rinse in distilled water. Figure B3 shows SEM of the sample after two step etching in 30 wt% KOH at 60ᵒC. The exposed silicon craters were etched into complete inverted pyramids with smooth (111) planes, with a very few sites consisting of fused

114

neighbouring inverted pyramids. The fusion of inverted pyramids/over etching can be controlled by etching time. A similar two step etching procedure was followed by various research groups for high quality etching of laser damaged silicon8, 9. The first etching step is thought to remove laser produced debris that would otherwise prevent silicon sites from etching and hence prepares the sample for uniform anisotropic etching in the second step. The evolution of laser exposed silicon into smooth pyramids during the second KOH etching step is shown in Figure 3-4(b).

20 nm thick SiNx hard mask (outlined by dashed circle) used for selective etching.

115

The same etching procedure repeated at a higher temperature of 75 ᵒC lead to a pattern of incomplete inverted pyramids as shown in Figure B4. Fusion of these incomplete inverted pyramids was observed at longer etching times with complete but rough inverted pyramids in only 30% of the patterned area. This suggests the requirement of lower temperature i.e. a slower etch rate to remove silicon damage and to ultimately form the inverted pyramid structure.

Figure B4 Shows the result of etching a sample in two steps in 30 wt% KOH at 75ᵒC. The etching resulted in a pattern of incomplete inverted pyramids. Fresh (111) planes were formed closer to the top of silicon crater, i.e. in the region with minimum laser damage, however, their growth was inhibited by high laser induced damage and debris inside the crater.

Next, a 30 wt% aqueous solution of TMAH was tested at 60ᵒC (optimized temperature for

KOH) using the two-step etching procedure. The post-etching SEM of the sample in Figure B5

116

shows that TMAH was not efficient in removing laser induced silicon damage at the base of silicon ablation craters and thus resulted in a rough pattern of square craters shaped by (111) planes formed by etching of low-damaged crater walls. Fusion of neighbouring structures was observed at higher etching times and the etching quality did not improve at lower temperatures with TMAH.

Figure B5 Shows the result of etching a sample in two steps in 30 wt% TMAH at 60ᵒC. The exposed silicon craters were etched into incomplete inverted pyramids and fused over etched region. Inset shows magnified image of the incomplete inverted pyramids with flat base in a pattern with ~ 170 nm wide mesa.

Based on these experimental observations, the two-step etching procedure with 30 wt% aqueous solution of KOH at 60 ᵒC was deemed optimal to etch exposed silicon craters.

117

However, the etching time required to obtain a damage-free inverted pyramid from a crater depends on the diameter of the mask aperture and laser induced damage, and hence the laser fluence (Figure 3-4 (a)). The small shift in laser focusing conditions from day to day caused the

~ 10% variation in the threshold fluence for ejecting the SiNx film and resulted in mask aperture diameters between 0.95 m and 1.1 m. With the general assumption that the rise in etching rate due to increase in etchant volume in a crater compensates the reduction in reaction rate due to increase in surface area in larger craters, it was predicted that the etching time for our samples was controlled by silicon damage that increased in samples with larger diameter apertures ejected using higher laser exposure fluence. Hence, the time required for etching crater pattern with a specific pitch into a high-density pattern of high fidelity inverted pyramids with minimum over etched area varied from sample to sample.

A process flow (Fig. B6) was developed in which the samples were recurrently inspected under optical microscope for over etching. Figure B7 shows the optical image of large area pattern with 300 nm mesa and 200 nm mesa in (a) and (b), respectively. The formation of a dark black cluster in the optical image suggests the fusion of neighbouring inverted pyramids in the pattern.

118

Figure B6 Schematic of wet etching process flow to achieve high density pattern of high fidelity inverted pyramids from a pattern of laser ablated silicon craters.

119

Figure B7 Large area optical images of etched samples with inverted pyramids textures with (a) 300 nm (b) 200 nm mesa width. Formation of clusters (inside dashed circles) in (b) suggests the fusion of neighbouring inverted pyramids.

120

Appendix B References

1. Petersen, K. E. Silicon as a mechanical material. Proc IEEE 70, 420-457 (1982).

2. Trimmer, W. in Micromechanics and MEMS : classic and seminal papers to 1990 (eds Trimmer, W. & Institute of Electrical and Electronics Engineers.) 720 (IEEE, New York, 1997).

3. Madou, M. J. in Fundamentals of microfabrication and nanotechnology 656 (CRC Press, Boca Raton, FL, 2012).

4. Gad-el-Hak, M. in MEMS : design and fabrication 664 (CRC/Taylor & Francis Group, Boca Raton, FL, 2006).

5. Franssila, S. in Introduction to microfabrication (2nd Edition) 534 (John Wiley & Sons, Chichester, West Sussex [England], 2010).

6. Adams, T. M. in Introductory MEMS: fabrication and applications (ed Layton, R. A.) 460 (Springer, New York, 2010).

7. Hull, R. in Properties of crystalline silicon 1016 (INSPEC, London, 1999).

8. Alavi, M., Fabula, T., Schumacher, A. & Wagner, H. -. Monolithic microbridges in silicon using laser machining and anisotropic etching. Sensor Actuat. A-Phys 37, 661-665 (1993).

9. Alavi, M., Büttgenbach, S., Schumacher, A. & Wagner, H. -. Fabrication of microchannels by laser machining and anisotropic etching of silicon. Sensor Actuat. A-Phys 32, 299-302 (1992).

121

Appendix C Wave Optical Simulation Method

A three-dimensional scattering matrix method is employed to study inverted pyramidal textures with grating periods ranging from subwavelength to ~ 2 × typical wavelengths (300 – 1100 nm) on the front surface of 2 μm to 400 μm thick c-Si with the objective of maximizing photoabsorption of the AM1.5 solar spectrum. Scattering matrix techniques has been proposed recently for solving patterned multilayer structures. The scattering matrix technique has numerical stability compared to transfer matrix technique, which is widely used for analyzing multilayer waveguide structures1, 2. In this method, the structure is divided into a number of vertically uniform layers that can be periodic or uniform in the lateral direction, as shown in

Figure C1(a). Each layer is treated as a separate diffraction grating (Figure C1(b)) that is expressed in terms of periodically varying material properties, such as refractive index and permittivity. The electromagnetic field in each layer is represented by an infinite set of plane waves. The method rigorously solves Maxwell’s equations by imposing matching conditions for the tangential field components at each layer boundary.

122

Ideally, the number of plane waves (modes) in the Fourier transforms and also the number of discretized layers in lateral directions should be infinite for accurate solution.

However, considering memory and time constraints, a finite number of diffraction modes can be used. To ensure accuracy of the solution, the individual layer thicknesses and the number of modes in the calculations were set to /20 and 121, respectively. Figure C2 shows the calculated reflected power at 450 nm (AM 1.5 solar spectrum power peak) from a 1000 nm periodic inverted pyramidal texture on a 400 mm thick wafer. The graph shows negligible difference for

≥ 81 modes.

Figure C2 Reflected AM1.5 power at 450 nm wavelength from 1000 nm periodic inverted pyramidal texture on 400 micron wafer as a function of number of modes.

123

Appendix C References

1. Liscidini, M., Gerace, D., Andreani, L. C. & Sipe, J. E. Scattering-matrix analysis of periodically patterned multilayers with asymmetric unit cells and birefringent media. Phys. Rev. B 77, 035324-1-035324-11 (2008).

2. Whittaker, D. M. & Culshaw, I. S. Scattering-matrix treatment of patterned multilayer photonic structures. Phys. Rev. B 60, 2610-2618 (1999).

124