Properties of epitaxial lateral overgrowth of GaAsP and GaAs grown by hydride vapor phase epitaxy

SECOND CYCLE, 30 CREDITS STOCKHOLM SWEDEN 2020,

Properties of epitaxial lateral overgrowth of GaAsP and GaAs grown by hydride vapor phase epitaxy

LAKSHMAN SRINIVASAN

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCE DEGREE PROJECT IN SCHOOL OF ENGINEERING SCIENCES ,

KTH Royal Institute of Technology School of Engineering Science Department of Applied Science Stockholm, Sweden

Examiner

Sebastian Lourdudoss

Supervisor Yanting Sun

2020-08-14

Lakshman Srinivasan

Properties of epitaxial lateral overgrowth of GaAsP and GaAs grown by hydride

vapor phase epitaxy

and GaAs grown by hydride vapor phase epitaxy Properties of epitaxial lateral overgrowth of GaAsP

Academic year 2019-2020

European Master of Science in Photonics

Master's dissertation submitted in order to obtain the academic degree of

Supervisors: Prof. dr. ir. Roel Baets, Prof. Sebastian Lourdudoss (KTH)

Student number: 01800686

Lakshman Srinivasan

Abstract

Direct heteroepitaxy of III-Vs on silicon (Si) has always been a challenge and there are various strategies to integrate these materials. This thesis deals with one such strategy known as Epitaxial lateral overgrowth (ELOG) which is exten- sively supported by experiments. For an application such as a multijunction solar cell, with silicon as a bottom cell, the highest efficiency can be achieved with a top cell having a bandgap of 1.7 eV and hence GaAsP as a material suits the profile.

The ELOG GaAsP and GaAs samples were grown using the epitaxial growth technique known as hydride vapor phase epitaxy (HVPE). With its near equilibrium operation capacity, high quality layers were grown. To specifically focus on the crystal defects and dislocations of the atoms, GaAsP was grown on GaAs substrate. Samples with varying growth parameters are investigated using several characterization techniques such as scanning electron microscopy (SEM), Photoluminescence (PL) spectroscopy and Raman spectroscopy. Composition variations (group V elemental incorporation in GaAsP) and crystalline quality are the two major factors that are analyzed. Additionally, ELOG GaAs samples grown on GaAs substrate using HVPE are studied as a reference to observe any strain effects due to the ELOG profile and compare with the GaAsP samples.

The ideal goal of this thesis is to optimize the crystalline quality of the ELOG GaAsP samples and to verify that GaAsP grown using ELOG technique has a better crystallinity than the planar growth (direct epitaxy of GaAsP on GaAs sub- strate) using two major optical characterization tools - PL and Raman spectroscopy. This work is a step towards the development of high efficiency multi-junction solar cells with GaAsP and Si as the respective top and bottom cells.

Keywords: GaAsP, ELOG, HVPE, Photoluminescence, Raman.

Abstrakt

Direkt heteroepitaxi av III-V p˚a kisel (Si) har alltid varit en utmaning och det finns olika strategier f¨or att integrera dessa material. Den h¨ar avhandlingen behandlar en s˚adan strategi som kallas Epitaxial lateral overgrowth (ELOG) som st¨ods externt av experiment. F¨or en applikation som en multi junction solcell, med kisel som bottencell, kan den h¨ogsta effektiviteten uppn˚as med en toppcell med ett bandgap p˚a 1,7 eV och d¨armed GaAsP som ett material som passar profilen.

ELOG GaAsP- och GaAs-proverna odlades med anv¨andning av den epitaxiella tillv¨axttekniken k¨and som hydriddamp- fasepitaxi (HVPE). Med dess n¨ara kapacitet f¨or j¨amviktsdrift odlades lager av h¨og kvalitet. F¨or att specifikt fokusera p˚a kristalldefekter och dislokationer av atomerna odlades GaAsP p˚a GaAs substrat. Prover med varierande tillv¨axtparametrar unders¨oks med anv¨andning av flera karakteriseringstekniker s˚asom skanningselektronmikroskopi (SEM), Photolumi- nescence (PL) -spektroskopi och Raman-spektroskopi. Kompositionvariationer (grupp V elemental inkorporering i GaAsP) och kristallin kvalitet ¨ar de tv˚a huvudfaktorerna som analyseras. Dessutom studeras ELOG GaA-prover od- lade p˚a GaAs-substrat med anv¨andning av HVPE som en referens f¨or att observera eventuella belastningseffekter p˚a grund av ELOG-profilen och j¨amf¨ora med GaAsP-proverna.

Det ideala m˚alet med denna avhandling ¨ar att optimera den kristallina kvaliteten p˚a ELOG GaAsP-proverna och att verifiera att GaAsP som odlas med ELOG-teknik har en b¨attre kristallinitet ¨an den plana tillv¨axten (direkt epitaxi av GaAsP p˚a GaAs underlag) med tv˚a huvudsakliga optiska karakt¨ariseringar verktyg - PL- och Raman-spektroskopi. Detta arbete ¨ar ett steg mot utvecklingen av h¨ogeffektiva multi junction solceller med GaAsP och Si som respektive topp- och bottenceller.

Nyckelord: GaAsP, ELOG, HVPE, Photoluminescence, Raman.

Acknowledgement

I wish to express my gratitude to Prof. Sebastian Lourdudoss who is also my examiner here at KTH for guiding me through the tenure of my thesis. The liberty and regard provided by him to master students such as myself is an inspiration. His serene presence was essential for me to ensure that I was on the right track.

Secondly, I would like to sincerely extend my appreciation to Dr. Yanting Sun, my supervisor here at KTH. His knowl- edge on the current research topic involving III-V materials is vast. His keen sense of curiosity and his intuitive approach to every question/problem always inspires me to think more like him. His support and freedom was crucial to me.

I would like to thank Axel Str¨omberg, PhD student working in the HMA group. His efforts and encouragement were pivotal in my thesis work. His patience to my long and strenuous bombardment of questions is unparalleled. He is my go to solution for every problem that I faced during this thesis. I envy his calm and jovial nature. He is one of the best colleague/senior that one can hope for and I wish him all the luck in his doctorate.

Special thanks to Prof. Anand Srinivasan. His encouragement and hope during the tough times was much needed.

Thanks to all the members of the Electrum lab group who helped me with the license for the tools that I used in the clean room and for training me on the characterization and fabrication equipment. I would also like to thank Prakhar Bhargava, a fellow master student from the nanotechnology department at KTH working in the HMA group. It is always good to have a fellow Indian working with you on a similar topic.

I would like to thank my home university (Ghent) and my supervisor Prof. Roel Baets for providing me the opportunity to work on my thesis in the laboratory of semiconductor materials, ”HMA” group here in Stockholm, Sweden.

Lastly, I would like to thank my parents and friends back in India who have provided me with the moral support that I

very much needed during the course of my thesis.

Contents

Abstract ii

Abstrakt iii

Acknowledgement iv

List of Figures vi

List of Tables viii

Acronyms ix

1 Introduction 2

1.1 Research on solar cells . . . . 2

1.2 Epitaxial lateral overgrowth (ELOG) . . . . 4

1.3 Current status and scope of the thesis . . . . 4

2 Experimental setup and designing a solar cell 6 2.1 Hydride Vapor Phase Epitaxy (HVPE) . . . . 6

2.2 Characterization methods . . . . 8

2.2.1 Scanning electron microscopy (SEM) . . . . 8

2.2.2 Photo-luminescence (PL) spectroscopy . . . . 9

2.2.3 Raman spectroscopy . . . . 10

2.3 Design of III-V solar cells with PC-1D . . . . 11

2.3.1 Device simulation and analysis . . . . 13

3 PL characterization of ELOG GaAsP and GaAs 17 3.1 Structure of the ELOG samples . . . . 17

3.2 PL measurements . . . . 18

3.2.1 Comparison between the ELOG and Planar GaAsP samples . . . . 19

3.3 ELOG GaAsP . . . . 23

3.4 ELOG GaAs . . . . 36

4 Raman analysis of ELOG GaAs and GaAsP 40 4.1 Selection rules for Raman scattering by phonons . . . . 40

4.2 ELOG GaAs . . . . 41

4.3 ELOG GaAsP . . . . 49

5 Conclusion 55

References 57

List of Figures

1.1 NREL database for best research cell efficiency [4]. . . . 3

1.2 Schematics of Epitaxial lateral overgrowth (ELOG) [11]. . . . 4

2.1 a) The LP-HVPE reactor from the electrum lab at KTH used in this thesis. b) Schematics depicting the parts of the reactor. . . . 7

2.2 Schematic of a Scanning Electron Microscope [18] . . . . 8

2.3 Principle of photoluminescence spectroscopy (PL) [23]. . . . 9

2.4 Energy level diagrams of Rayleigh, Stokes and anti-stokes Raman scattering [24]. . . . 10

2.5 Quantum efficiency spectrum [29]. . . . 11

2.6 Individual thickness and doping variation for p-n InP emitter and base . . . . 13

2.7 The influence of emitter thickness on efficiency under different emitter doping of p-n InP solar cell . . 14

2.8 The influence of base doping on efficiency under different base thicknesses of p-n InP solar cell . . . . 15

2.9 Influence of base doping and thickness on emitter thickness variation for a p-n InP solar cell . . . . 15

2.10 The designed structure and the I-V curve obtained from PC1D simulation . . . . 16

3.1 Structure of the ELOG profile - Top-view and cross-sectional view . . . . 18

3.2 PL single scan spectra of ELOG GaAsP samples taken at the center of the ELOG in cross-sectional view 19 3.3 PL single scan spectra of planar GaAsP samples taken from top-view . . . . 20

3.4 PL spectra of Planar sample 3748 taken from top-view and cross-sectional view . . . . 20

3.5 Cross-sectional view of one of the ELOG patterns of sample 3748: a) SEM image of the cleaved cross- section. b) Captured image of the ELOG cross-section during PL measurements . . . . 23

3.6 PL spectra across the ELOG cross-section for sample 3748 at sites corresponding to the SEM image in Figure 3.5-a. . . . 24

3.7 Vertical line scan of ELOG GaAsP 3748: a) SEM image corresponding to the vertical line scan. b) Peak wavelength and intensity variation across the line scan. c) PL spectra at 3 different sites in the line scan. 25 3.8 Horizontal line scan of ELOG GaAsP 3748: a) SEM image corresponding to the horizontal line scan. b) Peak wavelength and intensity variation across the line scan. c) PL spectra at 3 different sites in the line scan. . . . 26

3.9 PL mapping result of ELOG GaAsP 3748 - Maximum-wavelength mapping across a single ELOG growth area. . . . 27

3.10 Cross-sectional view of one of the ELOG patterns of sample 3746: a) SEM image of the cleaved cross- section. b) Captured image of the ELOG cross-section during PL measurements . . . . 28

3.11 PL spectra at various points across the ELOG cross-section for sample 3746. The sites in this figure correspond to the SEM image in Figure 3.10-a. . . . 29

3.12 PL mapping result of ELOG GaAsP 3746 - Maximum-wavelength mapping across a single ELOG growth area. . . . 29

3.13 Images of ELOG GaAsP 3762 - a) Cross-sectional SEM image of the ELOG structure of the cleaved sample

3762. b) Image of the ELOG cross-section during the measurements. c) 2-D structure of the ELOG GaAsP

3762 . . . . 30

3.14 PL results of single scans along the cross-section of ELOG GaAsP 3762 corresponding to the sites in the SEM image from Figure 3.13-a. . . . 31 3.15 PL mapping result of ELOG GaAsP 3762 - Maximum-wavelength mapping . . . . 32 3.16 PL mapping result of ELOG GaAsP 3762 - Maximum-intensity mapping corresponding to the maximum

wavelength in figure 3.15 . . . . 32 3.17 Cross-sectional view of one of the ELOG patterns of sample 3810: a) SEM image of the cleaved cross-

section. b) Captured image of the ELOG cross-section during PL measurements . . . . 33 3.18 PL results of single scans along the cross-section of ELOG GaAsP 3810 corresponding to the sites in the

SEM image in Figure 3.17-a. . . . 33 3.19 PL mapping result of ELOG GaAsP 3810 - a) Maximum-wavelength mapping b) Maximum-intensity

mapping corresponding to the maximum wavelength mapping. . . . 34 3.20 PL single scans at site A and B of ELOG GaAsP 3810 from the Maximum-wavelength mapping . . . . . 35 3.21 ELOG GaAs cleaved cross-sectional images taken during the PL measurements - a) Sample 3607 16

b)

3611 16

. . . . 36 3.22 ELOG GaAs PL spectra across the epitaxial GaAs profile - a) Sample 3607 16

b) 3611 16

. . . . 37 3.23 a) ELOG GaAs PL vertical line scan across the epitaxial GaAs profile for sample 3607. b) PL spectra

linewidth across the same vertical line scan from the ELOG GaAs region to the GaAs substrate. . . . . 37 3.24 PL peak wavelength mapping across the ELOG GaAs region - a) Sample 3607 16

b) Sample 3611 16

. . 38 3.25 Cleaved cross-sectional images of ELOG GaAs sample 3611 - a) 16

off-cut b) 31

off-cut c) 63

off-cut. 39 4.1 Cleaved cross-sectional image of one of the ELOG growths of GaAs 3607 . . . . 41 4.2 Micro-Raman spectra from the GaAs substrate - Reference value equivalent to a bulk GaAs crystal . . . 42 4.3 Raman mapping with respect to the wavenumbers (cm

) across the ELOG cross-section for sample 3607. 42 4.4 Difference in shifts in the two Raman peaks from the epitaxial GaAs layer and the GaAs substrate for

sample 3607 . . . . 43 4.5 Visualisation of the Raman line-width of the TO GaAs phonon peak for sample 3607 . . . . 44 4.6 TO GaAs line-width variation in sample 3607: a) Raman mapping across the ELOG area. b) Comparison

of the line-width variation between the GaAs substrate and the ELOG GaAs . . . . 45 4.7 Cleaved cross-sectional image of one of the ELOG growths of GaAs 3611 . . . . 46 4.8 a) Raman mapping w.r.t wavenumber (cm

) across the cleaved ELOG cross-section for sample 3611. b)

Shifts in the two Raman peaks - GaAs substrate and the ELOG GaAs . . . . 47 4.9 TO GaAs line-width variation in sample 3611: a) Raman mapping across the ELOG area. b) Comparison

of the line-width variation between the GaAs substrate and the ELOG GaAs . . . . 47 4.10 a) Two different sites in sample 3607 indicating the edges of the ELOG GaAs growth profile. b) Raman

Shifts at the two sites and the appearance of the forbidden LO GaAs phonon peak. . . . 48 4.11 TO and LO phonon peaks of GaAsP taken from Planar (top-view measurements) and ELOG (cross-

section measurements) GaAsP samples respectively. . . . 50 4.12 ELOG GaAsP - TO GaAs phonon peak shifts from the nominal TO GaAs value for different samples: a)

3746 b) 3748 c) 3762 and d) 3810 . . . . 51

4.13 TO GaAs line-width variation for all ELOG samples . . . . 53

List of Tables

2.1 InP material parameters for PC1D modelling . . . . 11

3.1 Growth condition recipes for GaAsP ELOG samples grown for crystalline quality optimization . . . . . 17

3.2 Features of the ELOG GaAsP samples . . . . 17

3.3 Planar vs ELOG peak wavelength variation . . . . 19

3.4 The peak wavelengths, calculated bandgap energy and compositions of the GaAsP ELOG samples . . . 22

4.1 TO GaAs phonon peak shift for ELOG GaAsP samples and their composition from PL measurements . 52

4.2 TO GaAs line-width and intensity variations . . . . 53

Acronyms

Ar Argon

AsH

Arsine

ELOG Epitaxial Lateral Overgrowth

eV Electron Volt

FWHM Full Width at Half Maximum

GaAs Gallium Arsenide

GaAsP Gallium Arsenide Phosphide

GaCl Gallium Chloride

GaP Gallium Phosphide

HVPE Hydride Vapor Phase Epitaxy

H

Hydrogen

InP Indium Phosphide

LO Longitudinal Optical

LPE Liquid phase epitaxy

MBE Molecular Beam Epitaxy

MOCVD Metal Organic Chemical Vapor deposition

µm Micrometer

nm Nanometer

N

Nitrogen

PH

Phosphine

PL Photoluminescence

PV Photovoltaics

SEM Scanning Electron Microscopy

SiO

Silicon dioxide

Si

N

Silicon Nitride

TO Transverse Optical

VPE Vapor Phase Epitaxy

Chapter 1

Introduction

As the world is turning more and more onto renewable energy sources, solar is probably the only efficient clean energy source that is available in abundance. Silicon (Si) is the second most abundant element on earth and photovoltaics (PV) owes a great deal to this material. Practically, silicon is being used as the prime material for solar cells. But due to its bandgap energy of 1.1 eV, from the detailed balance theory limit, it is not optimum and tends to have a limitation on the efficiency [1]. III-V materials such as gallium arsenide (GaAs) and Indium phosphide (InP) have a direct bandgap and their energy gap is close to the ideal value of their 1.33 eV for maximum power conversion efficiency from the sun.

Hence their use as a solar cell would yield a greater efficiency when compared with silicon as a material. But the III-V materials are less abundant and more expensive than Si, hence their hindrance to reach the market level in PV. The dominance of Si will continue in the PV market. But in space, there is a need for high radiation resistance and Si fails to provide that. The combination of high electrical conversion performance and high radiation resistance makes InP a very attractive material for space solar cells [2].

1.1 Research on solar cells

Conversion efficiency and high costs are the major bottlenecks in the potential of solar energy becoming a main source of energy. New methods of harnessing the full spectrum of the wavelength of the sun, new materials and multi-junction solar cells are paving way for solar power to be the primary power resource for the world. Solar cells in general are categorized into three generations majorly based on their prominence. The ’First generation’ corresponds to large area, high quality and single junction devices and owes mostly to Si as a material for its high energy yield and low cost.

Although it has a broad spectral absorption range, the high energy photons in the blue and violet end of the spectrum is wasted as heat [3]. The second generation focuses on thin film materials. Over time, these solar cells are expected to bridge the gap between them and the first generation with respect to the power conversion efficiency as mentioned in the National Renewable Energy Laboratory’s (NREL) research cell efficiency data [4]. The first generation Si and the current generation materials are all focused on the energy received by earth. III-Vs solar cell structures on the other hand are optimized for specific applications such as satellites, PV concentrator systems and laser power beaming.

Limits in efficiency

The maximum theoretical limit for a single junction solar cell under one sun (without sunlight) is about 33.16% as stated

by the detailed balance theory limit (Shockley-Quiesser limit). Under maximum amount of sunlight, a single junction

solar cell has a maximum efficiency of 31% [5].The analysis of the detailed balanced theory were based on assumptions

that included a single electron-hole pair excited per incoming photon, illumination with non-concentrated sunlight and

Figure 1.1: NREL database for best research cell efficiency [4].

thermal relaxation of the electron-hole pair energy in excess of the band gap. There are several considerations to account for the limit in efficiency - black body radiation, recombination, spectrum losses and impedance matching. These factors all together contributes to the losses in the amount of electrical energy that is extracted from the photons [6]. Power conversion efficiency (PCE) can be increased by using multi-junction (tandem) solar cells by capturing more of the solar spectrum. The use of multiple semiconductor materials allows enhances the absorbance to a broader wavelength range thus improving the cell’s sunlight to electrical energy conversion efficiency. The true limit of efficiency is the thermodynamic limit of 87% for maximum solar concentration.

Working of a multi-junction solar cell

As spectrum losses creates the majority of the losses in solar cells, a system can be created with multiple junctions by using materials with different band gaps. The ordering of these materials is done by placing the one with the greatest bandgap first. Photons that cannot excite electrons from the valence band to the conduction band in the first material travel through to the next material which can hopefully absorb it and generate current. This technique forms the basis of a multi-junction solar cell and vastly reduces the amount of energy loss due to electron relaxation and heat since larger bandgap materials can be used to absorb photons with higher energies improving the efficiency of solar cells.

Practically, creating a tandem solar cell has its own challenges - mainly dealing with extremely thin layers and arduous process of extracting current between the top and bottom cell. The best probable solution to this is to stack the top and bottom cell and create a tunnel junction in between them [7]. Apart from this, the choice of material for the cells is a major factor. It should be chosen in such a way that the electrical characteristics of each layer is carefully matched. For the electrons not to be absorbed between the layers, the photo-current generated in the top and bottom cell needs to be matched. The material to be chosen is determined by the requirements of lattice-matching, current matching and high performance opto-electronic properties [8]. These conditions are best met by the III-V semiconductors.

GaAsP as a material for multi-junction solar cells

GaAsP on Silicon multi-junction solar cells points out to a promising path in achieving high efficiency. Si is a widely

used material for solar cells and the use of it as a bottom cell enables low cost. With silicon’s badngap at 1.1 eV, the

top cell has to be chosen in such a way that it has complimentary effects to that of the bottom cell. From the detailed

balance theory limit, the ideal bandgap for the top cell should be around 1.7 eV to obtain maximum efficiency. GaAsP

with a respective phosphorus and arsenic composition of 0.25 and 0.75 has a bandgap of 1.7 eV and hence these two materials can be combined to obtain a high efficiency solar cell.

The major hindrance here is the high dislocation density (> 10

cm

) between the two materials [9]. For that, a technique is implemented to facilitate direct heteroepitaxy of III-Vs on Si called as Epitaxial lateral overgrowth (ELOG).

1.2 Epitaxial lateral overgrowth (ELOG)

It is a technique where the desired layer is grown on a substrate which is partially masked. Prior to the growth, the substrate is covered by a thin Si

N

or SiO

film/layer. This layer is then patterned using photo-lithography to create openings from where the epitaxial layer growth can begin. The concept of selectivity comes into play here as the growth of the epitaxial layer starts selectively in mask-free seeding areas and the gradually proceeds in the lateral direction over the dielectric film, hence the name - Epitaxial lateral overgrowth (ELOG) [10]. The schematics of the growth is shown in Figure 1.2 where the epitaxial layer to be grown is a III-V semiconductor and the substrate is silicon [11].

Figure 1.2: Schematics of Epitaxial lateral overgrowth (ELOG) [11].

Since the growth starts only from the mask openings, the dislocations starting from the seed are hindered by the SiO

mask.The defects which reach the window openings are the ones that are propagated into the grown layer and hence the surface which is grown laterally should be defect free [10]. Since direct epitaxy of III-Vs on Silicon leads to a high defect density due to lattice mismatch, ELOG is an effective way to integrate III-Vs on silicon.

1.3 Current status and scope of the thesis

The work on this thesis majorly focuses on the optical characterization of ELOG GaAsP samples, with an emphasis on understanding how the uniformity and composition of a ternary compound varies within the growth itself. In addition, planar samples of the same run are used to identify the variations exclusively from the ELOG samples. The two major characterization techniques used here are Photo-luminescence (PL) and Raman spectroscopy.

This thesis is the continuation of the pre work done for fabricating a tandem solar cell with GaAsP as a top cell and Si

as a bottom cell. With Si as a bottom cell, the ideal bandgap for the top cell is ∼ 1.7 eV, which is easily attainable with GaAsP [12].

Overall, the focus is more towards the optimization of the crystalline quality of the ELOG GaAsP layers rather than the fabrication part and the substrate used for all the samples is GaAs. Although, to understand the design of a solar cell, a PC-1D (Simulation software for the modelling of solar cells) model of an InP solar cell is simulated and reported in this thesis.

This thesis is organized as follows:

Chapter 1 covers the introductory part about the research on solar cells, working of a multi-junction solar cell, GaAsP as a suitable material for such applications and the ELOG growth technique.

Chapter 2 gives an overview of the HVPE reactor that was used to grow the epitaxial layers, basic information about the tools used in characterizing the epitaxial layers and the design of a solar cell using a modelling program called PC-1D.

Chapters 3 and 4 are the heart of this thesis and they explain about the analysis done using PL and Raman spectroscopy to characterize the ELOG GaAsP and GaAs samples.

Chapter 5 provides a general conclusion of the whole thesis highlighting the major aspects of it.

Chapter 2

Experimental setup and designing a solar cell

This chapter mainly focuses on the reactor used to grow the GaAsP samples and the tools that were used to characterize them. Apart from that, to understand the design of a solar cell for application purposes, an InP solar cell model is built using the PC-1D software.

2.1 Hydride Vapor Phase Epitaxy (HVPE)

There are several methods to prepare materials for optoelectronic devices namely metal-organic vapor phase epitaxy (MOVPE), liquid phase epitaxy (LPE), molecular beam epitaxy (MBE) and vapor phase epitaxy (VPE). The most widely spread technique among these has been MOVPE. However, to carry out selective growth on semi-insulating layers, and to grow thick layers in a short span of time, vapor phase epitaxy is considered [13]. In a template such as the ELOG, selectivity plays a major role and hence VPE is preferred over MOVPE. Higher growth rate for HVPE results in a reduction of the capital equipment costs and hence the prepared samples are cheap and easy to manufacture.

Another unique advantage of HVPE over MOVPE which is beyond the scope of this thesis is its ability to grow con- ductive transition layers which can act as buffer layers during hetero-structures growth for power electronic devices.

Whereas in MOVPE, high resistance, low temperature nucleation layers are used for device heterostructure growth [14].

The reactor

In this thesis, the growth of both the planar and ELOG GaAsP epitaxial layers were realized by Hydride vapor phase epitaxy (HVPE) technology.

The system has two chambers separated by a gate valve. One of the chambers consists of the load-lock system and the

other contains the reactor itself. The furnace which surrounds the reactor is separated into five regulated temperature

zones. The III-Chlorides are generated in zones 1 and 2 which is also called as the source zones. Zone 3 is where the

gases get mixed, zone 4 is the deposition zone and finally zone 5 acts as an interface to the loading chamber which is

kept at the room temperature. [15].

On comparing with all the other growth techniques, HVPE is the only III-V semiconductor crystal growth process which works almost close to an equilibrium. Dechlorination frequency of GaCl and InCl (chloride vapor precursors) is high and there is no kinetic delay. This means that close to the surface (substrates’) there is an increase of the vapor phase supersaturation and an immediate reactivity can be observed, in other words, surface reactions occurring on the substrate displays faster kinetics [16].

For GaAsP, the group III precursors are GaCl generated from HCl and Ga (molten). Group V precursors include PH

and AsH

. As they enter the reactor chamber, they partially decompoise into As

and P

at high temperatures [15].

GaAs

P

is formed by the chemical reaction:

GaCl(g) + (1 − x)AsH

(g) + xP H

(g) * ) GaAs

P

+ HCl(g) + H

(g) (2.1)

Figure 2.1: a) The LP-HVPE reactor from the electrum lab at KTH used in this thesis. b) Schematics depicting the parts of the reactor.

Since the precursors are in their gaseous phase, the desired composition of a ternary compound such as GaAsP can

be easily obtained by changing these gas phase components. HVPE being an equilibrium process has a lot of salient

features involving the growth of thick layers in a short span of time, attaining flexible selective growth on non planar

substrates and the low cost of its precursors. Overall with such major advantages over other prevailing growth methods,

HVPE is an important hetero-epitaxial growth technique.

2.2 Characterization methods

2.2.1 Scanning electron microscopy (SEM)

With the necessary amount of light, the human eye has a resolution of ∼ 0.2 mm. The resolving power of the microscope is governed by factors such as the quality of the lens and the wavelength of the illuminated light. Visible light (White light) has a wavelength in the range of 400-700 nanometers (nm) and the average is around 550 nm. Hence the resolution of the light microscope is around 200-250 nm. This is hardly enough to see the subtle variations and shifts in wavelengths in the nanometers scale and hence the use of an electron microscope [17].

The main components of a scanning electron microscope (SEM) are the source of electrons, electromagnetic lenses present in the column down which the electrons travel from the electron gun to the sample. A computerized system is used to analyse the recorded images from the microscope. A more detailed overview of the schematics is presented in Figure 2.2.

Figure 2.2: Schematic of a Scanning Electron Microscope [18]

The produced electrons from the electron gun are accelerated downwards while passing through the electromagnetic lenses and the apertures towards the sample chamber. The function of the apertures and the lenses are to focus the electron beam as much as possible before they reach the sample. There are specific scan coils situated above the objective lens that aid in controlling the position of the electron beam and hence the name scanning. When the electron interacts with the sample, numerous signals are produced which are then detected using the various detectors present [18].

Depending on the accelerating voltage, the electrons penetrate the sample to a depth in the micron scale which results in back-scattered electrons, secondary electrons and X-rays based on the depth of penetration and these signals are recorded and the corresponding images are produced on the computer screen.

In this thesis, the SEM tool is mainly focused on getting a high resolution image of the cleaved cross-section of the

ELOG GaAsP samples. As a result, the samples are mounted upright and the cleaved area which is of a few microns are

focused and the images are recorded.

2.2.2 Photo-luminescence (PL) spectroscopy

Photoluminescence spectroscopy is a technique that is vastly used for characterisation of the electronic and optical properties of semiconductors. Photo-luminescence is descried as the process in which the light energy/photon stimu- lates the emission of another photon from a material/ any matter. It is basically a nondestructive, non-contact method of probing materials. Photo-excitation occurs when a light which is directed onto a sample is absorbed. This enables the material to go to a higher electronic state and as the relaxation of the material to a lower energy state is accompanied by a release of energy (photons) [19].

The PL setup used to study the optical properties of ELOG GaAs and GaAsP was equipped with an Ar laser which has a peak wavelength at 514 nm for optical excitation with a spot size of ∼ 3 µm. The objective had a numerical aperture of 0.45 and a peltier cooled charged coupled device (CCD) detector was used.

Typically, a light source with an energy greater than the bandgap energy of the semiconductor excites it. Bloch equa- tions [20] [21] describes the polarization of the excited incoming light after it interacts with the semiconductor. These absorbed photons results in the formation of electrons and holes that have a finite momenta ’k’ in the respective conduc- tion and valence bands. After this process, energy and momentum relaxation occurs towards the band gap minimum.

In the end, electron-hole recombination occurs resulting in the emission of photons. PL spectroscopy is an important technique for measuring the crystalline quality of semiconductors such as GaAs and InP or in this case GaAsP and for quantification of the level of disorder in a system [22].

Figure 2.3: Principle of photoluminescence spectroscopy (PL) [23].

The interpretation of the PL spectra are done by examining the emission at a desired wavelength where the absorption spectrum is visible due to the excitation of electrons. The scan range in wavelengths is chosen around this desired value to obtain a spectrum. Although, it is important to note that PL spectra and absorption spectra are different in the sense that PL focuses on transitions from the excited to the ground state and the absorption spectra deals with transitions from ground to the excited state. The duration of absorption and emission is short.

In this thesis, PL spectroscopy is used to qualitatively analyse the ELOG GaAsP samples i.e. their crystalline quality

and to determine their composition. Several PL spectra, line scans and mappings are measured across the cleaved cross-section of a single ELOG GaAsP growth to effectively focus on the epitaxial layer alone.

2.2.3 Raman spectroscopy

Similar to the PL, Raman spectroscopy involves illumination of a sample with a monochromatic light and examining the scattered light. Hence, the same instrument is used as in PL with some minor variations involving the SI units, the grating and a change in the calibration. It is a non destructive technique that provides detailed information about the chemical structure, crystallinity and the molecular interactions in sample and requires no sample preparation. Raman spectroscopy depends on the inelastic scattering of photons, known as Raman scattering. In Raman scattering, there is an exchange of energy and a deviation in the direction of light. These effects constitutes to vibrational energy being gained by a molecule as incident photons from a visible laser are shifted to higher(Anti-stokes raman scattering) and lower energy levels (stokes raman scattering). The shift in energy gives information about the vibrational modes in the system. The schematics of the stokes and anti-stokes Raman scattering are shown in Figure 2.4.

Figure 2.4: Energy level diagrams of Rayleigh, Stokes and anti-stokes Raman scattering [24].

In our analysis, we consider only the Stokes raman scattering. Conforming to the Boltzmann distribution, electrons with normal temperature ranges tend to be in their lowest energy state and hence anti-stokes Raman scattering are not as common as stokes Raman scattering [24].

Optical phonons are basically out-of-phase movements of the atoms in lattice consisting of two or more atoms. Raman active optical phonons interact with light by the means of Raman scattering. They are classified as Longitudinal optical (LO) and transverse optical (TO) phonon modes and their splitting is described by the Lyddane-Sachs-Teller relation [25]. The measuring units of the optical phonon frequencies are given in spectroscopic wavenumber notation (cm

).

The occurrence of the LO and TO phonon modes in Raman measurements depends on the crystal plane orientation

while measuring the sample. In this case, all the ELOG samples are measured in the cross-sectional plane with the

GaAsP growth towards the [110] plane.

2.3 Design of III-V solar cells with PC-1D

As the factors of a cell operation extend, the analytical approach by hand becomes difficult to solve. The basic operation of a modelling program involves setting up the model with user defined parameters, a generation of nodes to solve and then iterating these parameters to get a solution that is consistent with the entire array of nodes. PC1D was developed to permit one-dimensional analysis of cells in an interactive array. The simulation work in this thesis focuses mainly on homo-junction InP solar cells. Relative parameters of the material (InP) is necessary for this simulation which are built in along with the software. Since InP solar cells are an ideal candidate for both space and terrestrial applications, a partial aim of the thesis is to perform modelling studies in one-dimension and to determine their maximum attainable AM0 efficiencies. The program follows a finite-element approach.

The best cells made to this date have an efficiency of 19% (AM0) which is equivalent to 21% (AM1.5 global) [26]. The modified material parameters for PC1D simulation and design modelling is given in table 2.1.

Table 2.1: InP material parameters for PC1D modelling

Parameters Symbols Value

Energy gap E

1.35 eV

Intrinsic carrier concentration n

1.2 × 10

cm s

Band-to-band recombination τ

100 ns

Electron diffusion length L

3 µm

Hole diffusion length L

0.01 µm

Surface recombination velocity S

10

cm s

Although the default value for the intrinsic carrier concentration (n

) is 8 × 10

cm

, extensive survey and previous modelling literature experience [27] has indicated that a value of 1.2 × 10

cm

is more appropriate based on the reported values of effective mass [28].

Figure 2.5: Quantum efficiency spectrum [29].

There are substantial gains to be made by increasing the minority carrier diffusion length of the emitter to a modest

value. Also, it is important to supplicate a high surface recombination velocity (S

) and an extremely low minority carrier lifetime in the emitter. The lifetime is however the dominant factor [27].

For InP p-n junction, the quantum efficiency in the red part of the spectrum in Figure 2.5 is almost near to 100% and a long minority carrier diffusion length in the base is required to account for this. So values from 1 µm to 30 µm can be chosen for the electron diffusion length without any significant change in the efficiency from the simulations.

Typically, the minority electron lifetime in bases are of the order of 100 ns as mentioned in the photo-luminescence decay [30]. This value together along with the electron and hole diffusion length gives a diffusion coefficient for InP which is consistent with the expected mobility.

An advantage of InP over GaAs is its lower surface recombination velocity. As a more realistic approach, to fit the quantum efficiency in the blue part of the spectrum, a maximum value of S

should be chosen and the hole diffusion length can be adjusted to a very short value of 0.1 µm. Although S

in the range of 10

cm s

and 10

cm s

high values) gives us a more realistic analysis, it requires an emitter doping in the range of 10

cm

which is highly unrealistic in our HVPE reactor. The reason for choosing a high surface recombination velocity and a low minority carrier lifetime is to account for the relatively low IQE (internal quantum efficiency) in the blue part of the spectrum [27]. Further, during simulations, such values approaches to a convergence failure . Hence to obtain a tangible result and avoid this error, doping concentration in the range of 10

cm

and a low S

in the range of 10

cm s

is chosen.

A trade off is observed here - the convergence failure error can be resolved by choosing a thinner emitter (< 0.1 µm)

with the other parameters intact, but that hinders the purpose of using HVPE for our growth. Another approach would

be to select an even lower value for the hole diffusion length and then to fit the S

as an adjustable parameter. But

physically this would point out to an emitter with totally dead bulk properties but a reasonable surface properties and

this seems intuitively questionable.

2.3.1 Device simulation and analysis

Once the parameters for the model are set, the simulation work can be started. As mentioned before, the thickness and the doping concentration of the epitaxially grown layers (base and emitter) are of main concern. For the initial part of the simulation, to observe a trend for the above mentioned parameters, each of them are varied separately by keeping the other parameter constant (kept at optimum values for maximum efficiency from literature [31]).

Figure 2.6: Individual thickness and doping variation for p-n InP emitter and base

The main parameters that are modified during the simulation work are the thickness and the doping concentration of the epitaxially grown layers. This thesis work strives towards the maximum efficiency of a solar cell. But it is necessary to limit ourselves to practical considerations for these parameters in accordance with our HVPE reactor as the end goal is to integrate the III-V with Si using this technology. Unlike Metal Organic Vapor Phase epitaxy (MOVPE), our HVPE reactor cannot have epitaxially grown layers at the nanometer scale range as the quality will be compromised. The trade off here is that for better performance, the emitter thickness should be <0.1µm.

From Figure 2.6, it is clear that a thin emitter results in a higher efficiency of almost 24%. Ideally this would be the case with an emitter thickness in the range 0.05 - 0.1 µm. But the use of HVPE technology for epitaxial growth hinders this.

In our HVPE reactor at KTH, for a uniform and high quality InP growth, a minimum thickness of 0.2 µm is considered

(from the previous growths and calibration runs).

Once the approximate values were created from these individual variation simulations, optimization was done by vary- ing both parameters simultaneously and observing the effect.

Figure 2.7: The influence of emitter thickness on efficiency under different emitter doping of p-n InP solar cell First, the emitter thickness is varied for different doping concentrations as seen in Figure 2.7 - a. High doping concen- tration of 1 ×10

cm

results in a high efficiency for a very thin emitter ( 0.1 µm). As the thickness increases, higher doping results in lesser efficiency. This can be seen clearly in Figure 2.7, as the doping concentration increases from high 10

cm

to 10

cm

, the efficiency decreases. Increase in the doping concentration beyond a certain level leads to bandgap narrowing which in turn results in an increase in the V

c (open circuit voltage) and a decrease in the reverse saturation current, which is nothing but a decrease in the overall efficiency [32].

An anomaly here is observed for the doping concentration of 1 ×10

cm

where the efficiency is much lesser even for a thin emitter, but also follows a different trend from the other doping concentrations - this is due to the low doping concentration which is almost equal to that of the base. Ideally, the open circuit voltage and therefore the efficiency increases for increased doping concentration.

As seen from Figure 2.8, base thickness variation from 2- 10 µm) does not have a significant impact on the efficiency of

the solar cell, so ideally a thickness within this range would be suitable for the growth in our HVPE reactor.

Figure 2.8: The influence of base doping on efficiency under different base thicknesses of p-n InP solar cell For the first test run on our HVPE reactor, to be on the safer side, an emitter thickness of 0.4 µm was chosen at a growth rate of 12 µ/min. Our HVPE reactor cannot grow InP with a high doping (10

cm

). The doping concentration feasible with our reactor is in the early 10

cm

range and an p-type doping value in the range of 1.5 ×10

cm

- 3 ×10

cm

(emitter) was aimed at for the first run. From the simulations, n-type doping (base) from 10

cm

to 3 ×10

cm

yielded almost the same efficiency. Once again, a reactor feasible value of 1 ×10

cm

was aimed for the first run.

Additionally, some more simulations are run to verify the variation in emitter thickness on efficiency if base parameters are varied and vice-versa.

Figure 2.9: Influence of base doping and thickness on emitter thickness variation for a p-n InP solar cell

As seen from figure 2.9, no significant changes are observed in the emitter thickness variation curve due to thickness

and doping variation of the base.

The final structure and the corresponding I-V curve are shown in Figure 2.10. The simulated p-n InP homo-junction solar cell has an efficiency of 23% with a fill factor of 85.5%. This is close to the InP state of the art solar cell design from spire [33].

Figure 2.10: The designed structure and the I-V curve obtained from PC1D simulation

Chapter 3

PL characterization of ELOG GaAsP and GaAs

Photo-luminescence (PL) and Raman measurements were carried out for different GaAsP Epitaxial lateral overgrowth (ELOG) samples to study and analyse the material’s optical properties. To focus more on the ELOG growth area of GaAsP, the analysis was done mainly on the cross-section of the samples. All the samples were grown on a semi- insulating GaAs substrate. Before moving on to the analysis, it is important to know the growth conditions of these GaAsP samples to link and justify any observed trend in the measurements. The growth conditions are listed in table 3.1.

Table 3.1: Growth condition recipes for GaAsP ELOG samples grown for crystalline quality optimization Run number Temperature (

C) GaCl flow (sccm) AsH

flow (sccm) PH

flow (sccm)

3746 660 5 20 80

3748 660 15 10 90

3762 710 5 20 80

3810 660 5 15 85

GaAsP growth on GaAs substrate with an ELOG template was done using the Hydride vapor phase epitaxy reactor. The run duration for all the four samples was the same - 15 minutes. Some additional features during the growth run helps in distinguishing the different ELOG samples clearly, it is given in table 3.2.

Table 3.2: Features of the ELOG GaAsP samples

Run number Features

3746 PH

in the stabilization flow to form P-rich nucleation sites on substrate before growth.

3748 Same as 3746, but compensate for composition (increase PH

) and reduce V/III ratio.

3762 Same as 3746, at a higher temperature and a 2 minute GaAs buffer layer at the start of the growth 3810 Same as 3746, but targeting GaAs

P

. Sulphur doped GaAsP.

3.1 Structure of the ELOG samples

In ELOG, the opening of the seed layer to be grown is surrounded by a dielectric mask. The initial growth (Selective area

growth) of the epitaxial layer takes place in the openings until the brink of the mask is reached. The prolongation of

growth results largely in lateral growth, which also involves some forcible vertical growth. The structure of the ELOG

profile along with the pattern template is shown in Figure 3.1.

Figure 3.1: Structure of the ELOG profile - Top-view and cross-sectional view

The stripes from the top-view of this figure depicts the pattern by which the openings for the ELOG growth are made.

The cross-sectional view depicts the growth profile of the ELOG structure that is triangular in shape indicating a forcible vertical growth and the lateral growth extending gradually.

3.2 PL measurements

To get an idea of the wavelength emitted by the GaAsP different samples, the PL spectra for a single point at the center on the ELOG cross-section is measured. A material’s crystalline quality and its composition can be determined by analysing its PL spectra. In this case, the material under observation is a ternary compound - GaAsP grown on GaAs substrate having an ELOG template.

The three major parameters to focus on from the PL measurements are the peak wavelength shift (nm), the peak in-

tensity variation and the line-width (FWHM) of the PL spectra. For the GaAsP samples, the wavelength shift majorly

corresponds to a change in the composition between the group V elements (Arsenic and phosphorus). The peak inten-

sity of the Photo-luminescence emission directly correlates with the defect density . High luminescence intensity is

an indication of good crystalline quality layer. The intensity of the emitting state is a measure of the strength of the

absorption of the excited state, so peaks in a PL spectra provides a map of the electronic structure of the material. Bulk

transitions in semiconductors similar to the electron-hole recombination and any defects such as dislocations, strain,

impurity atoms etc affect the emission either by changing the emission properties or by eliminating all the luminescence

properties [34]. The line-width (FWHM) of the PL spectra is another factor that corresponds to the crystalline quality

of the material - narrower the line-width, better the crystalline quality. This parameter is more dependable than the

peak intensity variation while comparing samples which were measured at different timelines with different calibration

setups.

3.2.1 Comparison between the ELOG and Planar GaAsP samples

PL spectra single scan measurements are done for various ELOG GaAsP samples and are compared with that of the planar GaAsP PL measurements. The growth condition for the Planar samples are similar to that of the ELOG samples except the template and the pattern since both were placed simultaneously inside the HVPE chamber. Unlike the ELOG samples, for the planar samples of the same run number, the PL measurements were taken from the top-view as it is hard to distinguish the epitaxial GaAsP layer and the GaAs substrate in cross-sectional view for the planar samples.

Figure 3.2: PL single scan spectra of ELOG GaAsP samples taken at the center of the ELOG in cross-sectional view In figure 3.2, the PL single scan spectra of the ELOG GaAsP samples is seen. The peak wavelength shifts between various samples is an indication of varying group V elemental incorporation in the samples. The cause and effects of these variations will be more elaborated in the coming sections on individual sample analysis.

On comparing the Planar and ELOG samples, it is obvious that there is a red shift in wavelength for all the four samples.

The difference in wavelength is given in table 3.3.

All the ELOG samples have a red shift in the same range (60 nm) except sample 3748 which has a slightly higher red shift (90 nm). It is obvious that sample 3748 as to why sample 3748 has a peak wavelength at a range (in nm) much lesser than that of the other samples. Although, it is important to note here that this is a single point scan analysis and there is a clear variation in peak wavelengths across different areas of the ELOG samples - especially for sample 3748 which has a higher phosphorus composition.

Table 3.3: Planar vs ELOG peak wavelength variation

Run number Planar peak wavelength (nm) ELOG peak wavelength (nm) Difference (nm)

3746 737 800 63

3748 643 734 90

3762 715 773 58

3810 722 784 62

Another interesting phenomenon observed here in table 3.3 is that the constant red shift is observed between the ELOG

Figure 3.3: PL single scan spectra of planar GaAsP samples taken from top-view

and Planar GaAsP samples albeit the growth conditions are similar. The reason for this shift might be due to either of the two factors: 1) Measurements from a different crystal plane (top-view and cross-section of respective planar and ELOG samples). 2) Strain from the ELOG growth profile. Both top view and cross-section PL analysis of the same sample is hardly ever done and the connection between the crystal plane and the bandgap is not clear, hence this reason is investigated first to negate the first hypothesis and move on to the strain effects.

Investigating reason 1

Cross-sectional PL single scan measurement is taken for sample 3748 and is compared with the top-view measurement of the same sample.

Figure 3.4: PL spectra of Planar sample 3748 taken from top-view and cross-sectional view

No change in the PL peak wavelength for the 3748 Planar sample between top-view and cross-section measurement.

Hence the wavelength shift between the ELOG and planar GaAsP samples might either be due to the lower phosphorus

incorporation or the strain effects.

The dielectric mask (SiO

) introduces strain in the substrate as well as the epitaxial layer that is grown. This might be due to the difference in thermal expansion coefficients between SiO

and GaAsP or from the volume expansion during the oxide deposition [10].

The thermal strain in ELOG GaAsP layers over the SiO

mask can be estimated from



=

(α

− α

)dT (3.1)

where α is the thermal expansion coefficient and the integration is over the growth temperature and the measurement temperature [35]. The thermal expansion coefficient (TCE) in

C

of GaAs

P

is given as α

= (5.73 − 0.16x) × 10

[36]. The TCE of SiO

is 5 × 10

C

. The growth temperature for the samples is 660

C

and they were measured at room temperature (27

C ). Taking the phosphorus concentration to be x = 0.25 to relate to all the samples and substituting these values in equation 3.1, the thermal strain is calculated as 

= 3285 ppm. The thermal strain in ELOG layers is also affected by the aspect ratio - width to height ratio of the epitaxial layer. A wider and a thinner ELOG GaAsP layer will reduce the strain in GaAsP by either the deformation of the ELOG layer at the edge or by the slip of the ELOG layer on the SiO

mask [35].

The relation between strains and the PL peak shifts is given by the deformation potential model [37]. The energy gap change under tensile stress is given by equation 3.2.

∆E

= [2a(C

− C

)/C

+ b(C

+ 2C

)/C

]

(3.2) In equation 3.2, the values were calculated for GaAsP with a phosphorus composition of 0.25. The calculated values a = -7.16 eV and b = -1.73 eV are the hydrostatic and shear deformation potential respectively. C

= 12.42 × 10

dyn/cm

and C

= 5.58 × 10

dyn/cm

are the elastic stiffness constants and 

is the thermal strain calculated manually.

These values were taken from literature where the GaAsP were calculated by interpolating the physical formulas for GaAs and GaP [38]. Substituting these values in equation 3.2, the energy gap change is calculated to be 36.6 meV. This change in energy gap corresponds to a change in peak wavelength from the PL spectra.

The red shift from the Planar to the ELOG GaAsP samples corresponds to band-gap narrowing and is attributed to the bi axial tensile strain in the ELOG GaAsP layers [39] [40]. The dependence of the band edges under strain has already been investigated from theoretical and experimental stand points in binary and ternary compound semiconductors [41]

[42] [43]. In the planar sample, GaAsP is grown directly on GaAs substrate indicating a lesser thermal strain, whereas in the ELOG, the thermal strain is caused by the difference of thermal expansion coefficients of GaAsP and SiO

mask [39].

Band-gap variation and the corresponding composition

From the peak wavelengths of the various ELOG samples, their corresponding band-gap energy can be calculated from

equation 3.3 which can further be used to calculate the composition.

E

= hc

λ E

= 2.78 − 1.55x + 0.19x

(3.3)

Table 3.4: The peak wavelengths, calculated bandgap energy and compositions of the GaAsP ELOG samples Run number Wavelength (nm) Bandgap energy (eV) Composition

3746 800 1.55 GaAs

P

3748 734 1.69 GaAs

P

3762 773 1.60 GaAs

P

3810 784 1.58 GaAs

P

From table 3.4, it is clear that sample 3748 has higher Phosphorus incorporation at the center. For an ideal GaAsP/Si hetero-junction tandem solar cell, the top cell’s band-gap is close to 1.7 eV [44]. Sample 3748 has its band-gap energy value close to that of the ideal value and hence can the same growth parameters can be considered for the upcoming runs made in the reactor to fabricate a GaAsP solar cell.

The compositions obtained from a High resolution x-ray diffraction instrument (HR-XRD) might be more reliable that

the ones calculated from PL measurements, but the use of a small cross-sectional facet makes it hard to load and measure

it using the HR-XRD tool. It is also important to acknowledge the fact that the intensity measurement at a single point

for an entire ELOG growth will result in trivial information. Hence, a PL mapping across the area of the ELOG could

justify the higher intensity variation for these samples and provide more information on the quality of the material and

the elemental incorporation at different areas of the ELOG.

3.3 ELOG GaAsP

ELOG GaAsP 3748

The PL analysis of sample 3748 here consists of measurements from several single scans, vertical and horizontal line scans and mapping across the ELOG growth area to validate its uniformity and to attest to its high crystalline quality.

Given in figure 3.5 -a is the cross-sectional SEM image of sample 3748. The profile of a single ELOG GaAsP is similar to that of the structure defined in the previous sections consisting of a triangularly shaped facet. The region below this shape is the GaAs substrate and the interface is clearly defined. Figure 3.5 -b shows the image taken from the TV screen during the PL measurements. The focused laser spot and the boundary of the measured area is also depicted in this image. A 50x lens was used during this measurement and hence the ELOG profile is zoomed out and the entirety of the GaAs substrate is clearly visible.

Figure 3.5: Cross-sectional view of one of the ELOG patterns of sample 3748: a) SEM image of the cleaved cross-section.

b) Captured image of the ELOG cross-section during PL measurements

Single scans across the ELOG

Several PL scans at random points across the focused ELOG region (given in figure 3.5 -b) and the peaks are plotted in figure 3.6. The scans across the ELOG points out to the peak wavelength being spread out across a range in nano-meters.

This shows that the GaAs

P

does not have a fixed composition and it varies randomly across the ELOG. Although,

from equation 3.3, it is obvious that the fluctuation in the composition of Arsenic and phosphorus is small.

Figure 3.6: PL spectra across the ELOG cross-section for sample 3748 at sites corresponding to the SEM image in Figure 3.5-a.

The luminescence intensity across the ELOG region is almost uniform and the value is high depicting a high quality epitaxial GaAsP growth.

Line scans

PL line scans are useful for comparing and analysing the lateral and vertical ELOG growth quality of a material. For a ternary compound such as GaAs

P

, the wavelength shift at different points from the line scans can be directly linked to group V elemental incorporation in the compound.

Vertical line scan

The PL measurements for the vertical line scan from figure 3.7 yields some useful information about the quality of the forcefully grown vertical layer during the ELOG growth.

At the top of the ELOG (site-1), the wavelength has shifted from 734 (center of the ELOG - site 2 in figure 3.7-a) nm to

707 nm. As we move away from the GaAs substrate, more P incorporation is observed in GaAsP which is verified by

the wavelength shift away from that of the wavelength of GaAs (866 nm) and towards the wavelength of GaP (550 nm)

at room temperature. This higher incorporation of As near the nucleation site is due to the impact of the inter-facial

strain relaxation corresponding to the lattice pulling/ composition latching effect between the GaAs substrate layer

and the GaAsP ELOG layer. Lattice pulling effect in this case is described as the effect in which the lattice mismatch

strain between the GaAs substrate and GaAsP epitaxial layer impedes the incorporation of phosphorus, thus pulling the

compositions of the growing GaAsP towards that of the substrate, i.e., towards lattice matching conditions [45]. This

same effect is observed from the vertical line scans. In conclusion, as we move away from the substrate vertically, the

composition latching effect is weakened and the Arsenic incorporation is reduced.

Figure 3.7: Vertical line scan of ELOG GaAsP 3748: a) SEM image corresponding to the vertical line scan. b) Peak wavelength and intensity variation across the line scan. c) PL spectra at 3 different sites in the line scan.

In Figure 3.7-b, the emitted wavelength is almost constant from the start of the epitaxial growth to the region near the center (site-2) with a uniform composition of GaAs

P

. From here on, there is a gradual decrease in the wave- length corresponding to a higher phosphorus incorporation which can be verified from figure 3.7-c, where at site-1, the maximum emitted wavelength is around 707 nm with a material composition of GaAs

P

.

Another far-fetched reason might be due to the fact that at the beginning of the ELOG GaAsP growth, Arsenic atoms are readily available as they are mass transported from beneath the masked region (from the GaAs substrate) to the openings where GaAsP nucleation starts. Hence with sufficient As incorporation at the beginning, the wavelength of GaAsP leans towards that of GaAs rather than GaP.

Additionally, in figure 3.7-b, there is a rise in wavelength from 735 nm to 860 nm at Y = 10 µm which is a signal from the GaAs substrate. This steep rise in wavelength with a slope of almost zero is an indication of a well defined interface between the substrate and the epitaxial layer which is grown.

Horizontal line scan

From figure 3.8-b, it is clear that the intensity is higher as we move laterally on both sides of the ELOG growth from

the center - this is desired as one expects high quality GaAsP lateral growth to cover the entire substrate, i.e., while

spreading out laterally. The wavelength at the center of the ELOG has more Arsenic content and it gradually decreases

Figure 3.8: Horizontal line scan of ELOG GaAsP 3748: a) SEM image corresponding to the horizontal line scan. b) Peak wavelength and intensity variation across the line scan. c) PL spectra at 3 different sites in the line scan.

laterally on both sides. This indicates to an impact of strain relaxation between the substrate and the ELOG layer due to the fact that the GaAs substrate has direct contact with the epitaxial layer (GaAsP) only in the center (nucleation region). Thus, towards the center, the growth profile tries to match with the substrate for obtaining lattice matching conditions and the strains due to the mismatch (As and P atoms) gets relaxed.

In figure 3.8-c, there is a clear transition of wavelength from the center (X = 0 µm) towards the lateral edges of the ELOG growth. From equation 3.3, the corresponding composition is calculated. As the growth extends laterally, a change in the material’s composition is observed - from GaAs

P

(site-1) to GaAs

P

(site-2) to GaAs

P

(site-3).

In figure 3.8-b, similar to the vertical line scans, there is a rise in wavelength from 700nm to 860 nm at X = 15 µm.

Although this region is not on the substrate, a very weak signal is still observed corresponding to the substrate. This is an indication of a well defined termination of the ELOG growth of GaAsP laterally.

PL mapping

Photo-luminescence mapping is carried out across the GaAsP ELOG growth area to study the material’s optical prop- erties as a whole. A small area of the substrate is also covered to observe a prominent shift in wavelength and intensity corresponding to the ELOG GaAsP and study the growth more in detail.

Figure 3.9 represents the peak wavelength mapping across an ELOG region of GaAsP. The area also covers a part of the