Advanced electron beam techniques for solar cell characterization

a SE image; b conventional EBIC image; and SCEBIC image c without and d with a metal enclosure at 5.7 kHz electron beam modulation frequency of a mc- Si solar cell.. Plan view a secondar

FOR SOLAR CELL CHARACTERIZATION

MENG LEI

(B Eng (Hons.), NUS)

A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF ELECTRICAL AND

COMPUTER ENGINEERING NATIONAL UNIVERSITY OF SINGPAORE

2014

DECLARATION

I hereby declare that this thesis is my original work and it

has been written by me in its entirety

I have duly acknowledged all the sources of information

which have been used in the thesis

This thesis has also not been submitted for any degree in

any university previously

MENG LEI

04 May 2014

Acknowledgements

My first and also my most sincere gratitude goes to my Ph.D supervisors, Professors Charanjit Singh Bhatia and Jacob Phang from Department of Electrical and Computer Engineering, National University of Singapore (ECE, NUS), for their continuous guidance and support throughout my doctoral studies Professor Bhatia is someone you will instantly love and never forget once you meet him His mentorship has always been paramount in providing a well-rounded experience consistent with my long-term career goals He has given me the freedom to pursue various areas that I am interested in and has been very supportive in all my Ph.D projects Professor Phang had always been motivating and inspiring me to take up new challenges and had made one of the biggest difference in my life His attitude of living every moment to its fullest and his strong determination has helped me come a long way and will always guide me in future

My special thanks also go to my Ph.D mentor, Alan Street, for always being so kind, helpful and motivating I have always enjoyed the personal discussion with him and the time I spent with him during dry runs of my presentations His technical inputs and friendly nature has always made me feel at ease with him

I would like to express my deep gratitude to Professor Armin Aberle, Dr Bram Hoex and

Dr Johnson Wong from Solar Energy Research Institute of Singapore (SERIS); and Professors Aaron Danner and Yang Hyunsoo from Spin Energy Lab (SEL) The discussion and suggestions from them are always valuable to me My special appreciation goes to Johnson for his kind help in reviewing my thesis chapters on short notices

I am very much thankful to Dr Steven Steen, Dr Satyavolu S Papa Rao, Dr Ron Nunes and Dr Harold Hovel from IBM Thomas J Watson Research Centre for their valuable support and collaboration with Professor Bhatia (ECE, NUS) during the period of NUS-IBM Joint Study Agreement # W0853529 It provided me with the unique opportunity to gain a wider breadth of research experience while I was still a graduate student

I would like to thank the ECE and SERIS for offering me the NUS Research Scholarship

as well as equipment support during my Ph.D candidature

My acknowledgement will never be complete without the special mention of my lab seniors at the Centre of Integrated Circuits Failure Analysis and Reliability (CICFAR):

Dr Xie Rongguo, Dr Hao Yufeng, Dr Huang Jinquan, Dr Wong Chee Leong, Dr Jason Teo, Dr Zhang Huijuan, Dr Pi Can, Dr Wang Ziqian, Dr Wang Rui and Dr Ren Yi for all their personal and professional help during the initial days of my stay in the lab I would also like to extend my sincere thanks to Mrs Ho, Mr Koo and Linn Linn for keeping a friendly and healthy lab atmosphere and bearing with me all these days

I am grateful to my fellow lab mates and friends: Liu Dan, Yihong, Jiayi, Wei Sun, Bai Xue, Yuya, Yunshan, Dr York Lin, Dr Ma Fusheng, Baochen, Mridul, Fajun, Cangming, Yang Yue for always being there and bearing with me for the good and bad times during the wonderful days of my Ph.D life I find myself lucky to have friends like them

Finally, I would like to acknowledge my parents, grandparents and all elders to me in my family for their constant support and strong faith in me I cannot imagine a life without their love and care

Table of Contents

DECLARATION i

Acknowledgements ii

Table of Contents iv

Abstract vii

List of Figures viii

List of Tables xiv

List of Symbols xv

Chapter 1 Introduction and Motivation 1

1.1 Photovoltaic Technology and Challenges 1

1.2 Current Characterization Techniques for Solar Cells 3

1.3 Strengths of Electron-Beam Based Techniques 4

1.4 Organization of thesis 5

Chapter 2 Theory and Literature Review 7

2.1 Introduction 7

2.2 Electron Beam and Sample Interaction 7

2.3 Secondary Electron Imaging in SEM 9

2.4 Scanning Electron Acoustic Microscopy (SEAM) 10

2.4.1 Physical Principles 10

2.4.2 Applications of SEAM 14

2.5 Conventional Electron Beam Induced Current (EBIC) 19

2.5.1 Physical Principles 19

2.5.2 Applications of EBIC Imaging 20

2.5.3 Quantitative EBIC Measurements 24

2.6 Single Contact Electron Beam Induced Current (SCEBIC) 41

2.6.1 Physical Principles 41

2.6.2 Applications of SCEBIC 44

2.6.3 Limitations and Challenges of SCEBIC 45

2.7 Strength and Challenges for Solar Cell Characterization 46

Chapter 3 Experimental Setup 48

3.1 Introduction 48

3.2 Experimental Setup (SEAM, EBIC and SCEBIC) 49

3.3 Summary 52

Chapter 4 SEAM Imaging on SDE Multicrystalline Silicon Wafers 53

4.1 Introduction 53

4.2 SEAM Signal Detection 55

4.3 Sample Procedures of Saw Damage Etch (SDE) 56

4.4 Defect Characterization of Saw-Damage-Etched Wafers 56

4.5 Optimization of SDE Duration 60

4.6 Summary 64

Chapter 5 Defect Characterization of Solar Cells 65

5.1 Morphological and Electrical Defects in Multicrystalline Silicon Solar Cells 65

5.1.1 Principle of Signal Detection 65

5.1.2 Defect Characterization in Isolation Trenches 68

5.1.3 Distinguishing Morphological and Electrical Defects 70

5.2 Defect Characterization of Amorphous Silicon (a-Si:H) Thin Film Solar Cells 76 5.2.1 Device Fabrication and Performance 77

5.2.2 Defect Characterization Using LBIC Imaging and FIB Cross-Sectioning 79 5.3 Studies of Photon Emission at Defects in Multicrystalline Silicon Solar Cells 90 5.4 Summary 93

Chapter 6 SCEBIC Imaging on Solar Cells 95

6.1 Introduction 95

6.2 SPICE Model of SCEBIC 96

6.2.1 SCEBIC Transient Phenomenon 97

6.2.2 Factors of SCEBIC Transient Signals 99

6.3 Experimental Verification of SCEBIC Model 101

6.4 SCEBIC Imaging on Multicrystalline Silicon Solar Cells 104

6.5 SCEBIC Imaging on Partially-Processed Solar Cells 106

6.6 Summary 107

Chapter 7 Extraction of Surface Recombination Velocity 108

7.1 Introduction 108

7.2 One-dimensional Numerical Approach for SRV 110

7.3 Three-dimensional Simulative Approach for SRV 115

7.4 Sample Preparation and Experiment Setup 118

7.5 Results and Discussion 120

7.6 Summary 127

Chapter 8 Conclusions 129

8.1 Summary 129

8.2 Future Work 131

References 134

Appendix A: List of Publications 148

Abstract

This dissertation presents a detailed comparative study of advanced electron-beam based

techniques for solar cell characterization Firstly, the advantage of the subsurface imaging

of scanning electron acoustic microscopy (SEAM) was utilized to characterize the

structural properties of saw-damage-induced defects and the non-destructive nature of

SEAM could enable accurate optimization of saw-damage etch process duration SEAM

was also employed together with electron beam induced current (EBIC) to investigate

defects in photovoltaic devices It was found that combination of these two techniques

could provide complementary information that clearly distinguishes the morphological

and electrical nature of the defects The first demonstration of single contact EBIC

(SCEBIC) on solar cells is then reported and the experimental results were supported

with an analytical model and clearly explained using SPICE simulations The

requirement on only one contact enables SCEBIC to be performed on partially processed

solar cells, thus allowing a high degree of flexibility of SCEBIC and its potential

applications in photovoltaic industry Lastly, highly localized quantitative EBIC were

demonstrated to measure surface recombination velocity (SRV) for solar cells with

different surface passivation conditions A three-dimensional Monte Carlo simulation for

electron-beam sample interaction was first employed to create a three-dimensional carrier

generation profile for accurate modelling of EBIC using Sentaurus TCAD These

simulation results were then verified using experimental data that were almost perfectly

matching, clearly demonstrating the capability and benefit of the high resolution and

accuracy of quantitative EBIC for the extraction of SRV for solar cells

List of Figures

Figure 2-1 Electron scattering in silicon using CASINO Monte Carlo simulation at an electron beam energy of 10 keV 9Figure 2-2 Schematic of SEAM thermo-elastic mode 11Figure 2-3 Schematic comparison of (a) SEAM (< 1 MHz), whose acoustic wavelength

is longer than the sample thickness, and (b) conventional SAM (~ few GHz), whose acoustic wavelength is much smaller than the sample thickness 13Figure 2-4 (a) Secondary electron (SE) and (b) SEAM images (at 165 kHz) of the domain structure in Polycrystalline Mn50Ni28Ga22 alloy 15Figure 2-5 SE images of a multi-level IC (a) before and (b) after removing the top metal layer; and corresponding SEAM amplitude images prior to the top-down de-processing at electron beam energy of 30 keV and electron beam modulation frequency of (b) 25 kHz, (c) 60 kHz, (d) 173.8 kHz and (e) 200 kHz 16Figure 2-6 (a) SE image of an IC; and SEAM phase images at modulation frequency of 173.2 kHz and different phases respect with the reference signals when b(1) θ = 40o

, b(2)

θ = 80o, b(3) θ = 100o, b(4) θ = 120o, b(5) θ = 160o

17Figure 2-7 (a) SEAM image taken at 71.9 kHz of a multi-level IC, (b) SE image of the cross-section of the sample after focus ion beam (FIB) milling at the highlighted location indicated at the SEAM image 18Figure 2-8 EBIC images of (a) a continuous junction; and (b) a discontinuous junction regions created by different laser diode currents 21Figure 2-9 Temperature dependence of EBIC contrasts of dislocations for different concentrations of contaminating impurities 22Figure 2-10 Comparison of EBIC (30 keV) and band-to-band luminescence or SiPHER (532 nm) on block-cast mc-Si 23

Figure 2-11 Schematic of excess-charge collection geometries of a p-n junction (a) normal; and (b) planar geometries 25Figure 2-12 Schematic of an ideal point source at z = ξ within a semi-infinite semiconductor that has a planar surface at z = 0 29Figure 2-13 Schematic of an ideal point source at z = ξ within a semi-infinite semiconductor that has a planar surface at z = 0 A p-n junction is inserted parallel to the semiconductor surface at z = zj 31Figure 2-14 Δp(z) as a function of S for steady-state uniform electron-beam excitation when (a) ξ/Lp = 1, i.e ξ ≈ Lp, (b) ξ/Lp= 3, i.e ξ >> Lp, and (c) ξ/Lp = 0.3 35

Figure 2-15 Schematic of EBIC measurement on a p-n junction solar cell 37

Figure 2-16 Experimental results on the effective diffusion length, Leff versus the penetration depth, ξ of the electron beam in GaAs p-n junction structure 38Figure 2-17 Schematic of electron-hole pair excitation in a p-n junction structure 40Figure 2-18 A plot of EBIC current versus ξ for a planar p-n junction silicon device 40Figure 2-19 Schematic of (a) conventional EBIC and (b) SCEBIC configurations 42Figure 2-20 A typical SCEBIC transient of IC when electron beam is turned on at a modulation frequency of 1 kHz 44Figure 2-21 (a) EBIC image of a CMOS transistor array with connections to power pins

Vdd and Vss; and (b) SCEBIC image of the same device with substrate as single contact 45Figure 3-1 Overview of the electron-beam based characterization techniques: (a) Conventional EBIC, (b) single contact EBIC (SCEBIC), and (c) SEAM 50Figure 3-2 Block diagram of the experimental setup of EBIC and SEAM 51Figure 3-3 Modification of the setup for single-contact EBIC (SCEBIC) 52

Figure 4-1 SEAM signal detection from a mc-Si wafer 55Figure 4-2 (a) SE and (b) SEAM images of a mc-Si wafer after SDE for 30 minutes 57Figure 4-3 (a) SE image of a mc-Si wafer after SDE for 5 minutes; (b) electron backscatter diffraction (EBSD) grain map on the mc-Si wafer; and SEAM images of the wafer at modulation frequencies of (c) 83 kHz, (d) 99 kHz, (e) 217 kHz and (f) 377 kHz 59Figure 4-4 (a) SE image and (b) SEAM image of a mc-Si wafer as-cut after sawing; (c)

SE image and (d) SEAM image of mc-Si wafer after saw-damage etch (SDE) for 20 minutes; (e) SE image and (f) SEAM image of wafer after SDE for 90 minutes 61Figure 4-5 (a) SE image and (b) SEAM image of another location of the mc-Si wafer as-cut; (c) SE image and (d) SEAM image of mc-Si wafer after SDE for 20 minutes; (e) SE image and (f) SEAM image of the wafer after SDE for 90 minutes 63Figure 5-1 SEAM and (b) EBIC signal detection of the mc-Si solar cell 67Figure 5-2 (a) SE image at 30 keV The parallel lines are the aluminium contacts (b) EBIC image at 20 keV; (c) SEAM amplitude image at 30 keV and modulation frequency

of 261 kHz 68Figure 5-3 SE image and (b) EBIC image at 20 keV; (c) SEAM amplitude image and (d) SEAM phase image at 30 keV and modulation frequency of 363 kHz 69Figure 5-4 (a) SE image and (b) SEAM amplitude image of p- Si wafer before junction formation at 30 keV and modulation frequency of 292 kHz 70Figure 5-5 (a) SE image at 30 keV; (b) EBIC image at 20 keV; (c) SEAM amplitude image and (d) SEAM phase image at 30 keV and modulation frequency of 550 kHz 71Figure 5-6 (a) EBIC image; and (b) SEAM amplitude image at 30 keV and modulation frequency of 391 kHz; (c) Line profile at X-X’ and Y-Y’ of EBIC and SEAM images 73Figure 5-7 (a) EBIC image; and (b) SEAM amplitude image at 30 keV and modulation frequency of 292 kHz; (c) Line Profile at Z-Z’ of both EBIC and SEAM images 74

Figure 5-8 SE image of the cross-section at the location of the line-like defect 75

Figure 5-9 Schematic of a-Si:H thin film solar cell 78

Figure 5-10 Schematic of LBIC imaging setup 80

Figure 5-11 LBIC images of (a) Sample 1; (b) Sample 2; and (c) Sample 3 81

Figure 5-12 (a) LBIC image at the edge of Sample 1; and (b) LBIC line profile across AA’ SE images of the cross section at (c) Location a1; and (d) Location a2 and a3 83

Figure 5-13 (a) SE image; and LBIC images of Sample 2 (b) before and (c) after the FIB cross-sectioning across BB’ and CC’ LBIC line profiles (d) at BB’; and (e) at CC’ 85

Figure 5-14 (a) FIB cross-sectioning across BB’ and CC’ shown in Figure 5-13 SE images of these cross sections with different LBIC contrasts indicated in Figure 5-13(c) and (d) at (a) Location b1; (b) Location c1; and (c) Location c2 86

Figure 5-15 LBIC images of defective area in Sample 3 (a) before and (b) after the FIB cross-sectioning across DD’ and EE’ (c) LBIC line profiles at DD’ and EE’ 87

Figure 5-16 FIB cross-sectioning at DD’ and EE’ shown in Fig 6 SE images of the cross section at (a) Location d1; (b) Location d2; (c) Location d3; and (d) Location e1 88 Figure 5-17 (a) Optical and (b) Electroluminescence images of the mc-Si solar cell 91

Figure 5-18 (a) SE and (b) EBIC images of the mc-Si solar cell at the location that is highlighted in red-dashed box in Figure 5-17 92

Figure 5-19 (a) EL image; and the corresponding LBIC images of the device at a wavelength of (b) 1064 nm; and (c) 633 nm 92

Figure 6-1 (a) SCEBIC configuration; (b) its equivalent circuit diagram of a typical p-n junction solar cell 96 Figure 6-2 SPICE simulation of the SCEBIC transient response ISCEBIC(t) at a modulation frequency of 200 Hz, where Ig = 100 µA, Rsh= 5 kΩ, Rs= 1Ω, Cj = 200 nF, Cs = 100 pF

The values assigned for each parameter are typical for solar cells with a sample size of about 1 cm2 98Figure 6-3 SCEBIC transient characteristics of a typical single-junction solar cell using SPICE simulations with the same model parameters as Figure 6-2, by varying only one parameter each time: (a) shunt resistance Rsh; (b) parasitic capacitance Cs; (c) junction capacitance Cj ; and (d) generation current Ig 99Figure 6-4 Comparison of experimental (blue stars) and simulated (red-solid line) SCEBIC transient responses to a pulsed electron beam (black dash) at the electron beam energy of 30 keV and the modulation frequency of 60 Hz 101Figure 6-5 Comparison of experimental single contact laser beam induced current (SCLBIC) (red dot) and the same SCEBIC (blue dot) transient responses 103Figure 6-6 Experimental SCEBIC (colour dots) transient responses with different values

of generation current Ig at the electron beam energy of 30 keV and the modulation frequency of 60 Hz 104Figure 6-7 (a) SE image; (b) conventional EBIC image; and SCEBIC image (c) without and (d) with a metal enclosure (at 5.7 kHz electron beam modulation frequency) of a mc-

Si solar cell All the images are taken at 30 keV electron beam energy 105Figure 6-8 SCEBIC image of the same mc-Si solar cell after removing the bottom contact The image was taken with a metal enclosure at the electron beam energy of 30 keV, and modulation frequency of 5.7 kHz 106Figure 7-1 Theoretical charge collection efficiency as a function of electron beam penetration depth Ldiff and zjare assumed to be 1 μm and 0.5 μm, respectively 114Figure 7-2 Electron-beam energy dissipation volume in silicon at 10 keV simulated by CASINO software programme 116Figure 7-3 Schematic illustration of the cross-section of the (a) fully passivated and (b) partially passivated bifacial n-type monocrystalline silicon (mono-Si) solar cell device structure and the IEBIC connections 117

Figure 7-4 The simulated carrier generation profiles (in the x-z plane when y = 0) resulting from a 5 keV electron beam for an n-type silicon wafer solar cell (a) with, and (b) without 80 nm AlOx/SiNx surface passivation film The electron beam current and radius are 300 pA and 10 nm, respectively 118Figure 7-5 Mono-Si solar cells after hot phosphorus acid bath for durations of (a) 0 minute, (b) 2 minutes, (c) 5 minutes, and (d) 10 minutes 119Figure 7-6 (a) SE and (b) EBIC images of the partially etched mono-Si solar cell The white region at the bottom left in the SE image and the corresponding dark region in the EBIC image is the metal finger 120Figure 7-7 Plan view (a) secondary electron (SE) and (b) plan-view EBIC images of the textured monocrystalline silicon solar cell at electron beam energy of 5 keV Cross-sectional view of (c) SE image, and (d) SE and EBIC overlap image of the same sample

at electron beam energy of 3 keV 121Figure 7-8 Comparison of experimental and simulated EBIC gain (IEBIC/Ibeam) for the solar cell without passivation Surface recombination velocity is equal to the maximum-possible value of 107 cm/s 123Figure 7-9 Simulated EBIC gain (IEBIC/Ibeam) as a function of electron beam energy for a n-type silicon wafer solar cell (a) with and (b) without a 80 nm AlOx/SiNx surface passivation film, assuming various values of SRV 125Figure 7-10 Comparison of experimental and simulated EBIC gain (IEBIC/Ibeam) for the solar cell with AlOx/SiNx passivation Surface recombination velocity is 2.8 × 105 cm/s 126

List of Tables

Table 5-1 Summary of performance of the three solar cell samples 79

E b Energy of electron beam (keV)

f Modulation frequency of electron beam

G Total charge induced due to SCEBIC phenomenon

G eff Effective generation strength

G o Generation strength of electron beam

I b Incident beam current

I EBIC Electron beam induced current (EBIC) signals

I Diff Diffusion current

I g Electron beam generation current

I m (t) Current transient as a function of time

I max Maximum value of current transient

I s Diode saturation current

I sc Short-circuit current

I SCEBIC Single contact electron beam induced current (SCEBIC) signals

J h Current density of holes

J sc Short-circuit current density

k 1 , k 2 Constant

K Thermal conductivity of the material

L b Minority carrier bulk diffusion length

L diff Minority carrier diffusion length

n Diode ideality factor

Δp Excess minority carrier density (holes)

R G Gruen range of electron

R s Series resistance of solar cell

R sh Shunt resistance of solar cell

v s Surface recombination velocity

V dd Positive DC supply voltage terminal

V oc Open-circuit voltage

V ss Ground terminal (substrate of IC)

x Distance between electron beam and p-n junction

Chapter 1 Introduction and Motivation

1.1 Photovoltaic Technology and Challenges

Sunlight [1] is a vastly abundant and relatively evenly distributed source of renewable energy The key techniques to harvest solar energy include photovoltaics (PV) that use solar cells to convert sunlight to electricity [2], solar-thermal relying on solar thermal collectors to convert sunlight to heat [3], and solar-chemical, which uses photochemical cells to collect sunlight to sustain chemical reactions such as electrolysis [4] Compared

to the combustion of fossil fuels, these technologies are far less impactful on the environment, especially in terms of the air pollutants and greenhouse gases emitted during their life cycles Despite holding great promise, solar energy, especially PV, is still considered uneconomical due to its relatively low electricity-generating capacity, and high production and installation costs It currently contributes less than 1% to the global energy supply [5] The solar energy industry has thus been focusing on cost reduction to add more competitive advantages to the various PV technologies In particular, owing to the growing environmental concerns and governmental supports for “green” energy, interest and effort on PV research has intensified and significant developments have been achieved recently This has made the PV technologies a more economically viable alternative energy option [6-10]

The main focus of the photovoltaic industry over the past decades has been on bulk silicon crystalline solar cells, known as the “first generation” wafer-based technology Recently, major investments have been made in new manufacturing facilities for

monocrystalline and multicrystalline wafer based solar cells, as well as for the closely related silicon ribbon and sheet approaches A “second generation” of thin-film solar cell technology has also emerged during the past 15 years [11-13] Thin-film solar cells offer strong advantage as a major reduction in material cost by eliminating the need for expensive wafer substrates Thin film solar cells also offer other benefits, particularly in terms of the increase in the unit of manufacturing from a wafer of about 100 cm2 to a glass sheet of about 1 m2 Thin film solar cells, however, suffer from substantially lower energy conversion efficiencies as compared to those of bulk silicon crystalline cells

The Carnot limit on the conversion of sunlight to electricity is 95% as opposed to the theoretical upper limit of 33% [14] for a standard solar cell This suggests that the performance of solar cells could be further improved 2 - 3 times if different concepts are used to produce a “third generation” of high-performance cells [15, 16] For example, novel structural design was employed to produce what is best known as a tandem cell, where efficiency can be increased by adding more cells of different band gaps to a stack, although it remains to be seen if this approach is economically viable for large scale deployment Apart from structural approaches, efficiency of a PV device can also be enhanced through the use of new materials Regardless of the approach, be it structural or material or a combination of both, the success in enhancing PV efficiencies relies on a clear understanding of the PV materials, optimization of the material quality and the fabrication processes All these require the right characterization techniques that are effective and efficient in providing the corresponding useful information that would aid

in achieving PV cells with better conversion efficiencies

1.2 Current Characterization Techniques for Solar Cells

According to the “Shockley-Queisser (SQ) limit” [14], also termed as the “radiative recombination limit” for PV energy conversion, the efficiency of an ideal bandgap solar cell whose charge carriers undergo solely radiative recombination is 33% For real devices, additional recombination channels come into play which further reduces the efficiency Ultimately, the performance of solar cells is limited by material properties

such as defect density [17], and p-n junction related parameters [18] like minority carrier

lifetime and surface recombination velocity The abilities to characterize and optimize these parameters are thus critical in achieving a cost effective PV process There has been much attention devoted in developing characterization procedures of solar cells recently through a variety of techniques that include luminescence imaging [19, 20], which has the advantage of high throughput and large-area imaging [21]; quasi-steady-state photoconductance (QSSPC) [22] that is a widely used for the determination of minority-carrier lifetime in semiconductor and trapping effects at a precisely calibrated carrier density [23]; and surface photovoltage (SPV) [24], which calculates the minority-carrier diffusion length and surface recombination velocity by measuring the illumination induced change of the photovoltage at the surface of solar cells [25] While these techniques are straightforward and capable of large throughput characterization, they are

of relatively poor resolution due to the large spot size of the photon beam, and thus may not be suitable for accurate analysis of microscopic material properties that are often the fundamental origin behind a specific material characteristic

1.3 Strengths of Electron-Beam Based Techniques

Apart from the various methods above-mentioned, there has been an increasing trend of extending electron-beam based techniques for characterization of PV material properties, such as carrier recombination activities within defects, and device properties like excess-carrier generation and illuminated current-voltage characteristics [26-30] Electron-beam based techniques rely on the bombardment of electrons [31] and their interaction with the materials Owing to the relatively small probe size of the electron beam (about few nanometres), these techniques can be applied at localized areas of the device with a much higher resolution than the conventional PV characterization methods that use photon as

an excitation source For instance, electron beam induced current (EBIC), a technique

traditionally used for integrated circuit (IC) failure analysis, makes use of p-n junctions or

Schottky barriers to collect the charge carriers generated as a result of the electron-beam interaction [32] within the selected scan areas of the devices The flows of those charge carriers are subsequently sensed as a current as a function of position, allowing accurate mapping of recombination centres and electrically-active-defect locations of the devices

Conventional EBIC techniques require electrical contacts to both sides of the p-n junction

for the measurement of the induced current so as to form a closed-loop in the external circuit This requirement poses a severe limitation to the application of EBIC imaging on partially processed devices In order to overcome this limitation, single contact EBIC (SCEBIC) has been developed [33] for IC failure analysis As the name suggests, SCEBIC requires only one connection to the device and thus has the potential to characterize a partially processed solar cell as long as one side of electrode is formed Moreover, another electron-beam based technique, scanning electron acoustic

microscopy (SEAM), which is based on the detection of electron acoustic signals [34] generated within the materials by a periodic intensity-modulated electron beam, is a well-established technique used for subsurface defect imaging and depth discrimination of multi-level ICs

While EBIC, SCEBIC and SEAM have been proven to be versatile techniques in IC and other material systems, application of these techniques on solar cells remain relatively less established This project aims to adapt EBIC, SCEBIC and SEAM from the IC world

to the solar cells in the PV industry and to develop them for in-depth characterization with both qualitative and quantitative analysis, where conventional EBIC techniques can

be employed for both material and device analyses and SCEBIC and SEAM methods can

be applied prior to the completion of the solar cells Successful application of these techniques on solar cells could offer a great potential in establishing the links between the material properties and the performance of the corresponding photovoltaic devices made from them

1.4 Organization of thesis

This thesis consists of eight chapters describing a detailed comparative study of different electron-beam based techniques such as EBIC, SCEBIC and SEAM for solar cell characterization

Following the present chapter (Chapter 1) on the background and motivation of the project, Chapter 2 gives a detailed literature survey together with an in-depth discussion

on the theories and working principles of the key electron-beam based characterization

techniques mentioned in this thesis whereas Chapter 3 describes briefly the overall system design and setup of the key techniques, i.e EBIC, SCEBIC and SEAM as well as their typical imaging results of solar cells

There are four main chapters that discuss the experimental findings SEAM imaging on silicon wafers processed at various solar cell fabrication stages is first presented in Chapter 4 This is followed by an in-depth discussion on novel applications of EBIC, SEAM and other complementary techniques on defect characterization of solar cells in Chapter 5 Chapter 6 presents the first demonstration of SCEBIC imaging on solar cells with a detailed theoretical explanation on its physical phenomenon and current characteristics The results are also supported with an analytical model that is verified by SPICE simulations, which show a close match with the experimental results The last technical chapter (Chapter 7) illustrates the application of the EBIC technique for quantitative analysis A three-dimensional simulation is employed to fit the measured EBIC data as a function of electron-beam energy, for the extraction of surface recombination velocity of solar cells The experimental data matches very well with the simulation results

Last but not least, Chapter 8 summarizes the findings reported in this project The thesis then concludes by suggesting a number of possible directions for future works

Chapter 2 Theory and Literature Review

2.1 Introduction

The scanning electron microscope (SEM) is one of the most widely used scientific instruments for characterizing morphology and topography of materials in various technology fields ranging from engineering and physics to biology and medicine etc For morphology and topography imaging, SEM relies on the mechanism of secondary electron emission from the sample surface as a result of the interaction between the highly energetic electrons and the sample Other than providing secondary electrons for morphology and topography analysis, the interaction between the highly energetic electrons and the sample also gives rise to other useful information/signals, whose manifestation can be exploited for other applications In this chapter, advanced electron-beam based characterization techniques including scanning electron acoustic microscopy (SEAM), electron beam induced current (EBIC) and single contact EBIC (SCEBIC) are discussed in detail When applied on solar cell characterization, these techniques are capable for detailed localized analysis with a relatively high resolution given the small probe size of the electron beam In addition, these imaging techniques are often performed simultaneously with the conventional SEM imaging procedure, thus allowing meaningful comparison or correlation with the secondary electron (SE) images

2.2 Electron Beam and Sample Interaction

The essence of electron beam-sample interaction is the generation of electron-hole pairs within the materials and their subsequent behaviours In brief, highly focused electrons

accelerated at several or tens of kilovolt energy impinge into the sample and, ionize the sample atoms to produce electrons in the conduction band and holes in the valence band These electron-hole pairs, together with the impinging electrons would penetrate deeper into the sample material Along the way, they scatter randomly and recombine until all excess carriers are eliminated, resulting in typically a “pear-shape” interaction volume of the electron beam and the sample [35]

This interaction volume has been well studied during the past decades and has been found

to be highly sensitive to the ionization energy and thus the penetration depth of the electron beam Penetration of electron beam into solid materials can be computed by the

Gruen range RG where the centre of the electron cloud in the sample is usually estimated

to be 0.4 RG [36] The Gruen range can be further applied in a semi-empirical analytical expression for energy dissipated per unit depth known as the Everhart and Hoff depth dose function [37] Such analytical expression describing electron beam penetration into solids and the depth dependent energy dissipation have been proven to be valid for certain beam energy ranges [38], and the accuracy can be greatly improved with CASINO Monte Carlo simulation program [39] Figure 2-1 illustrates the typical CASINO simulated electron trajectories in silicon at an electron beam energy of 10 keV

As demonstrated by Czyzewski and Joy [40], Napchan and Holt et al [41], the Monte Carlo program simplifies the process of computing energy losses in various layers of the materials that are present in a physical device It also yields a reasonable agreement with the experiments [42-44] The size of the interaction volume has been found to be strongly dependent on the electron beam energy, and the shape of the volume is often

approximated as a Gaussian curve, or a sphere located below the surface, or even a point source at the beam entry point [45]

Figure 2-1 Electron scattering in silicon using CASINO Monte Carlo simulation at an electron beam energy of 10 keV The red lines represent elastic backscattered electrons that preserve their energy and escape from the silicon surface; whereas the blue lines are the primary electrons, which experience inelastic scattering, penetrate deeper and dissipate all their energy into the material

2.3 Secondary Electron Imaging in SEM

The electron-beam and sample interaction produces various signals that can be detected and that contain information about the sample's surface topography and composition In general, the electron beam is scanned in a raster scan pattern and the beam's position is combined with the detected signal to produce an image The most common mode of detection is by secondary electrons emitted by atoms at or near the surface of the sample excited by the electron beam In the most common or standard detection mode, secondary

Electron Beam

Si

Backscattered electrons

Primary electrons

electron imaging, the SEM can produce very high-resolution images of a sample surface, revealing details less than 1 nm in size Typically, a wide range of magnifications is possible, from about 10 times (about equivalent to that of a powerful hand-lens) to more than 500,000 times, about 250 times the magnification limit of the best light microscopes.

2.4 Scanning Electron Acoustic Microscopy (SEAM)

Scanning electron acoustic microscopy (SEAM), invented by Brandis and Rosencwaig [46], and Cargill [47] in 1980, has been referred to as thermal wave microscopy by the former and electron acoustic microscopy by the latter The former name is however considered to be slightly misleading because it focuses on the thermal aspect and ignores the important role of acoustic phenomena During the past decades, SEAM technique has been further developed by Rosencwaig [48-51], Cargill [52, 53], Davies [54, 55], and Holstein [56-58], and finalized by Balk [34], which has become the foundation of the modern SEAM technique

2.4.1 Physical Principles

In SEAM technique, an electron beam, typically accelerated at 5-30 kV, is focused onto the sample surface and its intensity is modulated via blanking plates at a certain frequency in the range of 10 kHz to 5 MHz The waveform used to control the beam blanking is typically a square wave which in turn results in a periodically varied injection

of energetic electrons into the sample Figure 2-2 shows a schematic of the sectional view of the SEAM signal generation phenomenon due to the modulated electron beam bombardment within the sample These primary electrons dissipate their kinetic energy gradually as they penetrate further into the sample material until they lose all their

cross-kinetic energy The energy lost causes periodic local heating and thus cooling (when the beam is not present, i.e blocked by the blanking plates) of the sample, which in turn result in the periodic thermal expansion and contraction of a relatively small interaction volume that is in the vicinity of the beam entry point into the material The thermal waves, due to fairly large attenuation [51], travel only a relatively short distance within the sample and generate an acoustic force through thermo-elastic effect [56] Unlike the thermal waves, these elastic sound waves penetrate all the way through the sample and form the key imaging mechanism of the SEAM technique

Figure 2-2 Schematic of SEAM thermo-elastic mode [34]

The penetration depth of the thermal waves is limited to one or two thermal diffusion

lengths dT, and such penetration depths determine the region where conversion of thermal

into elastic waves occurs [34] In general, d T could be estimated as follows:

𝑑𝑇 = �2𝜋𝑓∙𝜌∙𝐶2𝐾 𝑡ℎ (1)

where K is the thermal conductivity of the material, f is the modulation frequency, 𝜌 is

material density and C th is specific heat

As shown in Figure 2-2, the volume contributing to the SEAM signal generation is

determined by d T that is typically several to a few tens of micrometres As a result, thermal waves, which are the origin of the generation of the SEAM signals, cannot be

detected directly before they dissipate and disappear within d T Instead, the resultant acoustic waves with their much longer wavelengths serve simply as the carriers of the information that is generated from the interaction of the thermal waves with the sample [57] As the modulation frequencies of SEAM signals are often less than 1 MHz, the corresponding acoustic wavelength is in the order of mm range Attenuation of sound is thus unlikely to affect the SEAM signals when they propagate through the entire sample thickness, unless the propagation of sound is strongly influenced by large imperfections such as a delamination between different layers within the sample This is the unique feature of SEAM technique comparing to the conventional scanning acoustic microscopy (SAM) [59] with modulation frequencies typically more than 1 GHz such that the acoustic wavelengths become too short to propagate through the entire sample thickness and, as a result, usually attenuate within the sample The schematic comparison of these two imaging techniques is shown in Figure 2-3 [60] Unlike ultrasonic acoustic microscopy, whose resolution is determined by the acoustic wavelength [59, 61] and

often expressed as c/f (where c is the speed of light), SEAM at the same modulation

frequency has essentially much better resolution since its resolution depends on the size

of the thermo-elastic conversion volume For instance, at 1 MHz, SEAM can detect structural features in the sample with a resolution of ~ 1 μm for nearly all solids, whereas the resolution of sound waves that are actually detected at the same frequency is ~ 0.5 cm, which is not sensitive enough to show any microscopic structure information of the sample

Figure 2-3 Schematic comparison of (a) SEAM (< 1 MHz), whose acoustic wavelength is longer than the sample thickness, and (b) conventional SAM (~ few GHz), whose acoustic wavelength is much smaller than the sample thickness [60]

Acoustic wave generation through the thermo-elastic effect by periodic heating of a local spot on sample surface strongly depends on the material’s thermal and elastic properties Therefore, any change in material physical properties, such as material density, morphological imperfection or mechanical deformation, of the sample material within the

thermal interaction volume can influence SEAM signals According to Equation (1), d T is

inversely dependent on f 1/2 Thus, by varying the modulation frequency, one can perform depth-profile scanning for providing both surface and subsurface visualization through the interaction of thermal wave with the sample

Both Davies [55] and Balk [62] have demonstrated how further imaging contrast due to different mechanisms can be generated by setting the detection phase-lock frequency to both multiples and odd harmonics of the modulation frequency However, in this dissertation only detections of the generated acoustic signals at given driven frequencies that utilize the harmonic primary electron amplitude modulation [63] (also known as linear signal generation), are concerned because the excitation conditions are low enough not to change the thermal properties of the materials

2.4.2 Applications of SEAM

Since the invention of the initial works of SEAM as mentioned above, the technique has been extended for various material research applications, including imaging dopant regions in semiconductors and grains in metals [64] as well as for detecting subsurface flaws [65, 66] and ferroelectric and ferromagnetic domains [67-72] Figure 2-4 shows one example of the SEAM imaging on ferromagnetic shape memory alloy Mn50Ni28Ga22 [69, 73] without any pre-treatment of the sample It was found that SEAM is able to detect the

grain boundaries and stripe martensite variants, which match well with the magnetic domain walls shown in the secondary electron (SE) image

Figure 2-4 (a) Secondary electron (SE) and (b) SEAM images (at 165 kHz) of the domain structure in Polycrystalline Mn 50 Ni 28 Ga 22 alloy [73]

It is beyond the scope of this chapter to give a complete view of all published applications of SEAM since it is not the aim of this chapter to present an exhaustive review on SEAM applications Instead, the following sub-sections discuss two typical examples to illustrate on how SEAM can be used for semiconductor technology as a tool for non-destructive process control

2.4.2.1 Depth Discrimination

One of the most well-known applications of SEAM is to provide high resolution images

of subsurface features in multi-level DRAM integrated circuits (ICs) [74] For example, owing to its non-destructive nature and depth profiling capability, it can be used to demonstrate subsurface imaging at different depths of a fully passivated multi-level IC sample with six layers of metallization and a thickness of 1 mm Figure 2-5(a) and (b) show the secondary electron (SE) images of an IC sample surface before and after the top-down de-processing to remove the top metal layer, respectively As depicted in the

SEAM amplitude images (Figure 2-5(c)-(f)), which were taken at the same location with different modulation frequencies, both top surface details and subsurface metal lines that are perpendicular to the top lines are clearly captured due to the depth-discriminating capability of SEAM at different frequencies When compared to the SE images, these SEAM images at different depth can provide useful correlation and extra meaningful information

Figure 2-5 SE images of a multi-level IC (a) before and (b) after removing the top metal layer; and corresponding SEAM amplitude images prior to the top-down de-processing at electron beam energy of 30 keV and electron beam modulation frequency of (b) 25 kHz, (c) 60 kHz, (d) 173.8 kHz and (e) 200 kHz [74]

In addition to variation of the frequency, SEAM’s depth profiling can also be observed

by varying the phase angle [75], which describes the time delay between the SEAM signals and the reference signals of the lock-in amplifier As an illustration, Figure 2-6 shows a series of the SEAM images taken at the same modulation frequency but at different phases with respect to the reference signals on an IC chip It is found that the imaging depth varies when the phase angle changes and this observation is then verified

by calculation, which also gives a comparable match with the theoretical thermal diffusion length

Figure 2-6 (a) SE image of an IC; and SEAM phase images at modulation frequency of 173.2 kHz and different phases respect with the reference signals when b(1) θ = 40 o , b(2) θ = 80 o

, b(3)

θ = 100 o , b(4) θ = 120 o , b(5) θ = 160 o

[75]

2.4.2.2 Subsurface Defect Imaging

Another well-known application of SEAM is in the field of failure or defect analysis It has been widely used for detection of defects such as subsurface cracks [60, 65] and dislocations in semiconductor materials especially at mechanically stressed regions [76, 77] due to its thermal wave coupling mechanism For example, SEAM has been successfully applied for IC failure analysis to localize the subsurface anomaly prior to any destructive processes such as focused ion beam (FIB) milling [78] A crack propagating deep into the sample bulk was observed in the trench sidewall at the location

of the extremely high SEAM contrast shown in Figure 2-7(a)

Figure 2-7 (a) SEAM image taken at 71.9 kHz of a multi-level IC, (b) SE image of the section of the sample after focus ion beam (FIB) milling at the highlighted location indicated at the SEAM image [78]

cross-The strong signal contrast from one point to another is often explained as a result of differential signal delays, i.e differences in both amplitude and phase (where phase shifts can take any value in the range 0 ~ 2π [79]) It has been found that SEAM phase angle changes rapidly at the grain boundaries, where the discontinuity causes the image contrast to change from white to black due to graduate oxygen segregation [34]

2.5 Conventional Electron Beam Induced Current (EBIC)

In conventional EBIC techniques, charge carriers generated by the excitation source (in this case an incoming high energy electron beam) are collected by an electric field within the sample and sensed as a current in an external circuit As the principle of EBIC closely

resembles that of short-circuit current of a p-n junction solar cell excited by energetic

electron beam, conventional EBIC has become one of the indispensable characterization techniques in photovoltaic research and industry

2.5.1 Physical Principles

Since its introduction and the pioneering works by Everhart [80, 81], electron beam induced current (EBIC) has been used for semiconductor device studies Following its introduction, EBIC has been rapidly developed for the characterization of both materials and devices in the IC industry within the several decades Typical applications of EBIC

includes the inspection of p-n junctions [82] and effects of electron penetration, as well as

for local determinations of charge carrier diffusion and quantitative measurement of minority carrier lifetime in semiconductors [83] In particular, the theory and applications

of this SEM-based charge collection technique have been excellently summarized in a detailed literature review by Leamy [32]

In SEM, when the incident electron beam impinges upon the device, the energetic beam electrons (also known as the primary electrons) collide with valence electrons in the sample material The collision results in ionization of the atoms and produces electrons and holes in the conduction band and valence band, respectively and the generation of

electron-hole pairs allows subsequent diffusion of electrons and holes towards the p-n

junction Owing to the presence of an internal electric field resulted from the depletion

region of the p-n junction, these carriers would drift accordingly and constitute a current

when the device is short-circuited Such induced current due to the scanning of the

electron beam across the p-n junction is the original EBIC signals used for material and

device characterization It has been established that the energy flux of the incoming beam, the depth dependence of the energy dissipation of the beam, the minority carrier recombination in the semiconductor, the width of the depletion region and the spatial relation of the generation volume with respect to the depletion region dictate the magnitude of the electron beam induced current

2.5.2 Applications of EBIC Imaging

Conventional EBIC imaging is a well-known characterization technique that can be used

to localize defects in semiconductor devices and reveal the recombination activities of the defects Based on the charge collection phenomenon, EBIC images are produced by recording short-circuit current using a two-dimensional scan during the electron bombardment on a device When defective regions are present in the sample, the current intensity varies according to the different strengths of the recombination activities of the defects A direct contrast of the defect and the background signals can thus be obtained

by EBIC imaging EBIC often provides valuable information on both materials and

devices, such as the location and profile of p-n junctions [84], the occurrence of impurity

distribution inhomogeneities [85], and crystallographic defects [86] Figure 2-8 gives an

example of p-n junction profiling obtained using EBIC scans of the cross-section of a

solar cell with laser-assisted dopant diffusion Regions of strong EBIC signals appear

brighter than others, thus providing a clear and intuitive visualization for locating the p-n

junction and detecting discontinuities of the profile possibly due to micro-cracks, which electrically isolate some junction regions and make collection of EBIC signals impossible

Figure 2-8 EBIC images of (a) a continuous junction; and (b) a discontinuous junction regions created by different laser diode currents [84]

In addition to being an ideal technique for p-n junction profiling, EBIC has also been

proven to be extremely versatile in the study of electrically-active defects [87] EBIC can

be used to characterize various defects such as dislocations, stacking faults and precipitates that are formed in the solar cell materials during either crystal growth or device processing stages Some of these defects, classified as electrically-active defects [88], may reduce minority carrier lifetime [89] and in some cases, create shunt paths in device regions that lead to poor cell efficiencies The recombination process at defects and the resultant variation of its EBIC contrast with temperature and beam current has been explained by Kittler [89-96] and Wilshaw [97] using the charge-controlled recombination model , and by Shockley-Read-Hall (SRH) recombination models

Temperature-dependent EBIC contrast signals arise from the recombination activities at dislocations is a convenient tool to clearly demonstrate the role of defect contamination

[93] Both the magnitude of the contrast and its temperature dependence are found to depend on the amount of contamination as shown in Figure 2-9, which represents a fingerprint characterizing the degree of contamination of the crystal defects [90] On the other hand, clean dislocations show only very weak activities (type II) [98], that peaked with a maximum at about 50 K but with activities untraceable at room temperature Weak contamination leads to an increase of the low temperature activity, leaving the room temperature activity below the detection limit (type 2) [96] Upon a further increase in contamination the temperature dependence changes and defect activity becomes detectable at room temperature (type 1) [99] Dislocations decorated with metal silicide precipitates show the highest activity throughout the whole temperature range (type I) [100] A model describing the aforementioned dislocation behaviours allow quantitative access to the concentration of the deep-level impurities at dislocations [28]

Figure 2-9 Temperature dependence of EBIC contrasts of dislocations for different concentrations of contaminating impurities Change of contrast type with increasing contamination is in the sequence: type II (clean) → type2 → mixed → type 1 → type I [86]