UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Electronic states and transport in GaNAs and GaAsBi Beaton, Daniel A. 2011

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2011_spring_beaton_daniel.pdf [ 1.95MB ]
Metadata
JSON: 24-1.0071670.json
JSON-LD: 24-1.0071670-ld.json
RDF/XML (Pretty): 24-1.0071670-rdf.xml
RDF/JSON: 24-1.0071670-rdf.json
Turtle: 24-1.0071670-turtle.txt
N-Triples: 24-1.0071670-rdf-ntriples.txt
Original Record: 24-1.0071670-source.json
Full Text
24-1.0071670-fulltext.txt
Citation
24-1.0071670.ris

Full Text

Electronic States and Transport in GaNxAs1-x and GaAs1-xBix by Daniel A. Beaton B.Sc., St. Francis Xavier University, 2001 M.Sc., Univ. of British Columbia, 2003 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in The Faculty of Graduate Studies (Physics)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) March, 2011 Daniel A. Beaton, 2011  Abstract This thesis explores the effect of incorporation of nitrogen or bismuth in GaAs on electronic states and transport, namely shallow in-gap states associated with nitrogen clusters in GaNx As1-x and the hole mobility in GaAs1-x Bix . Lowered in-gap emission intensity in GaNx As1-x epi-layers grown under atomically rich operation of the plasma source indicates that atomic nitrogen results in more optically efficient GaNx As1-x films. Two significant additions made to the growth procedure of GaAs1-x Bix allowed for growth of thick, doped epi-layers: reduced growth rate and growth interrupt prior to the GaAs1-x Bix epi-layer. Hole mobility is found to decrease with increasing Bi concentration. This decrease is smaller than the analogous decrease of electron mobility in GaNxAs1-x . The temperature dependence of the hole mobility is modelled using terms for scattering from phonons, ionized impurities and a temperature independent bismuth related term. There is an increasing effect from the bismuth related term with increasing bismuth content. Calculation of the scattering crosssection based on the fitted coefficient for the 1% GaAs1-x Bix sample is in good agreement with a theoretically predicted value. For samples above 4.4% the hole mobility was found to be temperature independent and scattering cross-sections are more than an order of magnitude larger. These are indications that bismuth clusters play a significant role in reducing the hole mobility at high concentrations. GaAs, GaNx As1-x and GaAs1-x Bix were incorporated in to device structures: Schottky diodes and p-i-n LEDs. Several defects were detected using DLTS and an increasing trend of defects for decreasing growth temperature found. Two hole traps were found to be related to growth of GaNx As1-x , where only one was related to nitrogen incorporation. The device structure of the LEDs prevented an in-depth analysis and poor surface morphology of GaAs1-x Bix epi-layers resulted in leaky devices. While further work is required, it was found that the inclusion of a growth interrupt introduced many shallow traps. Optical measurements on the LEDs showed that the electroluminescence from the GaAs1-x Bix layer was temperature insensitive, while the emission from the GaAs layers followed the expected Varshni relation.  ii  Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ii  Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  iii  List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  vi  List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  vii  Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ix  Chapter 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  1  Chapter 2. Background . . . . . . . . . . . 2.1 Semiconductor physics . . . . . . . . 2.1.1 Bands . . . . . . . . . . . . . 2.1.1i) Band gap as a function 2.1.2 Defects in semiconductors . . 2.1.3 Extrinsic semiconductors . . . 2.2 III-V Semiconductor alloys . . . . . . 2.3 Mobility . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . . . . . . . . . . . . . . of temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  5 5 5 8 9 10 12 17  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . method . . . . .  . . . . . . . . . . . . .  . . . . . . . . . . . . .  . . . . . . . . . . . . .  21 21 23 25 26 28 28 29 29 30 31 31 34  Chapter 4. Growth and Properties of GaNx As1-x . . . . . . . . . . . . . . . . . 4.1 Dilute nitride GaNx As1-x growth . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Film composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  36 37 38  Chapter 3. Experimental Techniques . . . . . . . . . . . . . . . . 3.1 Molecular beam epitaxial crystal growth . . . . . . . . . . . . 3.1.1 GaAs growth by molecular beam epitaxy . . . . . . . . 3.2 In-situ sample characterisation . . . . . . . . . . . . . . . . . 3.2.1 Reflection high energy electron diffraction . . . . . . . 3.2.2 Light scattering . . . . . . . . . . . . . . . . . . . . . . 3.3 Ex-situ sample characterisation . . . . . . . . . . . . . . . . . 3.3.1 X-ray diffraction . . . . . . . . . . . . . . . . . . . . . 3.3.2 Atomic force microscopy (AFM) . . . . . . . . . . . . . 3.3.3 Photo- and electroluminescence . . . . . . . . . . . . . 3.3.4 Secondary ion mass spectroscopy . . . . . . . . . . . . 3.3.5 Resistivity and carrier mobility using the van der Pauw 3.3.6 Deep level transient spectroscopy . . . . . . . . . . . .  iii  . . . . . . . .  . . . . . . . .  . . . . . . . .  . . . . . . . .  4.2.1 X-ray diffraction . . . . . . . . . . . . . . . . . . . 4.2.2 Depth profiling by secondary ion mass spectroscopy Photoluminescence measurements . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  . . . .  38 39 41 43  Chapter 5. Growth and Properties of GaAs1-x Bix . . . . . . 5.1 Dilute bismide GaAs1-x Bix growth . . . . . . . . . . . . . . 5.1.1 Initial growth procedure . . . . . . . . . . . . . . . 5.1.2 Modified GaAs1-x Bix growth procedure . . . . . . . 5.2 Film composition . . . . . . . . . . . . . . . . . . . . . . . 5.2.1 X-ray diffraction . . . . . . . . . . . . . . . . . . . 5.2.2 Photoluminescence . . . . . . . . . . . . . . . . . . 5.2.3 Depth profiling by secondary ion mass spectroscopy Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . .  . . . . . . . . .  . . . . . . . . .  . . . . . . . . .  . . . . . . . . .  . . . . . . . . .  . . . . . . . . .  . . . . . . . . .  . . . . . . . . .  . . . . . . . . .  45 46 47 48 51 51 53 54 58  Chapter 6. Electrical Transport in GaAs and GaAs1-x Bix . . . . . . . . . 6.1 Fabrication of resistivity and van der Pauw samples . . . . . . . . . . . . 6.1.1 Carrier concentration calibration . . . . . . . . . . . . . . . . . . 6.1.2 GaAs and GaAs1-x Bix van der Pauw samples . . . . . . . . . . . . 6.2 Hole mobility measurements . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1 Effect of growth conditions and film composition . . . . . . . . . . i) Effect of bismuth surfactant on carbon incorporation . . . . . ii) GaAs hole mobility for varied growth conditions . . . . . . . iii) GaAs1-x Bix hole mobility as a function of bismuth content . . 6.2.2 Temperature dependence of hole mobility in GaAs and GaAs1-x Bix iv) Carrier concentration as a function of temperature . . . . . . i) Contributions to hole scattering in GaAs . . . . . . . . . . . . ii) GaAs hole mobility as a function of temperature . . . . . . . iii) GaAs1-x Bix hole mobility as a function of temperature . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . .  . . . . . . . . . . . . . . .  60 60 60 61 62 62 62 64 65 67 67 67 70 72 74  Chapter 7: Device Fabrication and Characterisation . . . . . . 7.1 Light emitting diodes . . . . . . . . . . . . . . . . . . . . . . . 7.1.1 Device fabrication . . . . . . . . . . . . . . . . . . . . . 7.1.2 Optical emission from GaAs1-x Bix light emitting diodes 7.2 GaNx As1-x and GaAs1-x Bix Schottky diodes . . . . . . . . . . . 7.2.1 Device fabrication . . . . . . . . . . . . . . . . . . . . . 7.2.2 Deep level transient spectroscopy measurements . . . . i) GaAs . . . . . . . . . . . . . . . . . . . . . . . . . ii) GaNxAs1-x . . . . . . . . . . . . . . . . . . . . . . iii) GaAs1-x Bix . . . . . . . . . . . . . . . . . . . . . . iv) Light emitting diodes . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . . . . . . . . . . . .  . . . . . . . . . . . .  76 76 77 79 82 83 86 86 88 89 90 92  Chapter 8. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  94  4.3  iv  . . . . . . . . . . . .  . . . . . . . . . . . .  . . . . . . . . . . . .  . . . . . . . . . . . .  . . . . . . . . . . . .  . . . . . . . . . . . .  Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  98  APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 A. Clarification of GaAs wafer conventions . . . . . . . . . . . . . . . . . . . . 105 B. Sheet resistance from contacts on the perisphery of a thin film . . . . . . . . . . . . . . . . . . . . . . . . . 106 C. Hall measurement wiring schematic . . . . . . . . . . . . . . . . . . . . . . . 108 D. Overview of deep level transient spectroscopy measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 D.1 Tables of DLTS measurement results . . . . . . . . . . . . . . . . . . . . . . 114 E. Details of nitrogen plasma source operation and use for GaNx As1-x growth . . . . . . . . . . . . . . . . . . . . 120  v  List of Tables 2.1 2.2 2.3 2.4 3.1 5.1 6.1 6.2 6.3 6.4 D.1 D.2 D.3 D.4 D.5  Varshni fit parameters for GaAs . . . . . . . . . . . . . . . . . . . . . . . . . GaAs n- and p-type dopants and ionization energies . . . . . . . . . . . . . . III:V semiconductor lattice parameters and band gaps . . . . . . . . . . . . . III:V atomic covalent radii . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ionization efficiencies of some III-V elements . . . . . . . . . . . . . . . . . . Expected and measured doping concentration in GaAs1-x Bix samples compared to carbon (12 C) and silicon 75 As+28 Si) SIMS counts . . . . . . . . . . Carrier concentration for p-GaAs samples grown with and without bismuth used as a surfactant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of p-type GaAs1-x Bix sample details and hole mobility . . . . . . . Best fit parameters for the temperature dependence of the hole mobility in GaAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Best fit parameters for the temperature dependence of the hole mobility in GaAs1-x Bix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of p-type GaAs Schottky device DLTS results . . . . . . . . . . . . Summary of p-type GaNx As1-x Schottky device DLTS results . . . . . . . . . Summary of bulk GaAs1-x Bix DLTS results . . . . . . . . . . . . . . . . . . . Summary of GaAs LED DLTS results . . . . . . . . . . . . . . . . . . . . . . Summary of GaAs1-x Bix LED DLTS results . . . . . . . . . . . . . . . . . . .  vi  9 11 14 14 22 58 63 66 72 72 115 116 117 118 119  List of Figures 2.1 2.2 2.3 2.4 3.1 3.2 3.3 3.4 3.5 4.1 4.2 4.3 4.4  4.5  5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 6.1 6.2 6.3 6.4 6.5 6.6 6.7  GaAs unit cell, Brillouin zone and band structure . . . . . . . . . . . . . . . III-V band gaps as a function of lattice parameter . . . . . . . . . . . . . . . Nitrogen and bismuth resonant states within the band structure of GaAs . . Diagram of a conductor with applied E and B fields . . . . . . . . . . . . . . Example RHEED surface reconstruction images . . . . . . . . . . . . . . . . Examples of possible atomic surface reconstruction configurations . . . . . . Observation of RHEED oscillations for varied Ga flux . . . . . . . . . . . . . Diagram of prepared square van der Pauw sample . . . . . . . . . . . . . . . Iterative procedure to find the sheet resistance . . . . . . . . . . . . . . . . . Examples GaNAs XRD spectra . . . . . . . . . . . . . . . . . . . . . . . . . SIMS analysis of GaNx As1-x samples for O, C and N . . . . . . . . . . . . . . Nitrogen SIMS signal for several GaNx As1-x samples . . . . . . . . . . . . . . Comparison of GaNxAs1-x PL from samples grown before and after the plasma discharge tube was cleaned and for both on- and off-resonance operation of the plasma source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . GaNx As1-x PL peak energy for various N content before and after cleaning the plasma discharge tube for both on- and off-resonance operation of the plasma source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Diffusely scattered light from the growth surface of a GaAs and GaAs1-x Bix epi-layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . AFM scans of two quenched surfaces produced by the two growth procedures GaAs and GaAs(Bi) surface reconstructions phase map . . . . . . . . . . . . Example GaAs1-x Bix XRD spectra . . . . . . . . . . . . . . . . . . . . . . . . Reciprocal space map of 115 peak of a 3.5% GaAs1-x Bix sample . . . . . . . Example GaAs1-x Bix PL spectra . . . . . . . . . . . . . . . . . . . . . . . . . Representative GaAs1-x Bix SIMS analysis profiles . . . . . . . . . . . . . . . Epi-layer thickness for the 1×3 to 2×1 surface reconstruction transition and uniform bismuth content as measured by SIMS . . . . . . . . . . . . . . . . . Comparison of bismuth content from XRD and SIMS measurements . . . . . Resistivity and van der Pauw sample shadow masks . . . . . . . . . . . . . . Schematic of a Hall sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hole mobility of GaAs for various growth conditions . . . . . . . . . . . . . . Hole mobility for GaAs1-x Bix samples . . . . . . . . . . . . . . . . . . . . . . Carrier concentration for GaAs and GaAs1-x Bix films as a function of temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Temperature dependence of the hole mobility of GaAs . . . . . . . . . . . . . Temperature dependence of the hole mobility for GaAs1-x Bix samples . . . . vii  6 13 17 19 26 27 27 32 33 38 40 41  42  43 47 50 51 52 53 54 55 56 57 61 62 64 65 67 71 73  7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 7.10 A.1 B.1 C.1 D.1 D.2 D.3 D.4 E.1 E.2 E.3  Schematic of LED sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV curves for GaAs1-x Bix LED’s . . . . . . . . . . . . . . . . . . . . . . . . . GaAs1-x Bix LED PL spectra over the temperature range 8 K to 300 K . . . . GaAs1-x Bix LED EL spectra at various current densities . . . . . . . . . . . . GaAs1-x Bix LED EL spectra with 50 A/cm2 for temperatures ranging from 100 K to 300 K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Schematic of DLTS sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . GaAs1-x Bix DLTS Schottky device IV characteristics . . . . . . . . . . . . . . GaAs1-x Bix DLTS Schottky device CV characteristics . . . . . . . . . . . . . Summary of DLTS measurements of defect activation energy and concentration for GaAs, GaNx As1-x and GaAs1-x Bix Schottky diodes . . . . . . . . . . Summary of DLTS measurements of defect activation energy and concentration for GaAs and GaAs1-x Bix p-i-n LED heterostructures . . . . . . . . . . . US and EU wafer conventions . . . . . . . . . . . . . . . . . . . . . . . . . . Diagram used in deriving the van der Pauw equation . . . . . . . . . . . . . Wiring schematic used in the Hall measurement apparatus . . . . . . . . . . Band diagram schematics for a) p-i-n and b) metal-semiconductor (Schottky) junction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Schematic diagram of DLTS procedure . . . . . . . . . . . . . . . . . . . . . Illustration of a DLTS measurement and spectrum . . . . . . . . . . . . . . . Example of DLTS spectra and Arrhenius plot . . . . . . . . . . . . . . . . . Diagram of rf plasma source and baffle . . . . . . . . . . . . . . . . . . . . . Frequency dependence of the forward and reflected power of the rf plasma nitrogen source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Frequency spectra for operation of the plasma source in on-resonance and off-resonance mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  viii  77 79 80 81 81 84 85 86 87 91 105 107 108 110 111 112 113 121 122 123  Nomenclature BEP CBM DFT DLTS EL EU HH LH LS MOCVD MOVPE PL PBN PF PL PMT PR rf RHEED SI SIMS UHP UHV US VBM XRD YLF  Beam Equivalent Pressure Conduction Band Minimum Density Functional Theory Deep Level Transient Spectroscopy ElectroLuminescence European Heavy Hole Light Hole Light Scattering Metal Oxide Chemical Vapour Deposition MetalOrganic Vapour Phase Epitaxy Power to Load Pyrolytic Boron Nitride Power Forward PhotoLuminescence PhotoMultiplier Tube Power Reflected Radio Frequency Reflection High Energy Electron Diffraction Semi-Insulating Secondary Ion Mass Spectroscopy Ultra-High Purity Ultra-High Vacuum United States Valence Band Maximum X-Ray Diffraction Yttrium Lithium Fluoride  ix  chapter 1:  INTRODUCTION The ability to produce new materials and/or to apply known materials to new uses has marked advances made by mankind from our earliest beginnings. So much so that eras of human history are demarked by the dominant material(s) of the time: for example stone, bronze or iron ages. We now live in the information age, which could be more aptly referred to as the semiconductor age. Semiconductor devices, primarily those made from silicon, have been at the forefront of modern technology; computers built from integrated circuits which are almost entirely based on silicon, have transformed the way people act and interact. Improvements to present day technology to overcome the problems of today and tomorrow will require new and more specialised materials. Combinations of group III and group V or group II and group VI elements have been explored as potential alternatives to silicon in many areas. This has lead to even greater variety in available materials and applications. Gallium arsenide (GaAs) is one such material and has several proven advantages over silicon, namely a direct band gap (transitions from the conduction band to the valence band do not require a phonon) and high carrier mobilities. Beyond the simple one-to-one alloying of gallium and arsenic, ternary and quaternary III-V alloys have also been explored. Combinations of GaAs with aluminum, phosphorus, indium and antimony have been well studied[1, 2]. The properties of these alloys obey simple, often linear, relationships of their constituent end members and any deviations are not overly dramatic. Incorporation of nitrogen and bismuth stand apart from other III-V GaAs ternary alloys for their anomalously large perturbations to the electric and optical properties of the host GaAs for relatively small incorporated amounts. Both exhibit large band gap bowing and relatively large changes in the band gap energy for large incorporated amounts. In many ways the two materials systems, GaNx As1-x and GaAs1-x Bix , are analogous: as nitrogen is the smallest of the group V elements, whereas bismuth is the largest, and therefore both elements cause significant strain to the lattice of GaAs; both elements have states resonant with the band structure of GaAs, nitrogen effects the conduction band of the host GaAs, bismuth effects the valence band; and both elements appear to produce states near band edges within the band gap associated with clustering of the element. 1  For these reasons the material systems have remained associated, and the dilute nitride GaNx As1-x alloys are used as a road map for expectations in the GaAs1-x Bix alloys. Clusters consist of two or more nitrogen (or bismuth) atoms in close proximity to one another within the lattice[3–5]. This occurs when two or more nitrogen atoms are bonded to a single Ga centered tetrahedron or by simply having nitrogen atoms neighbouring or near other nitrogen atoms on different Ga centers. Neighbouring nitrogen atoms bonded to different Ga centers which have a linear arrangement are referred to as chains. These clusters result in shallow states within the band gap of GaNxAs1-x [6]. Incorporation of bismuth may also lead to formation of clusters and shallow in-gap states. The work presented in this thesis on the dilute nitride GaNx As1-x alloy represents work on a material system that is well understood in terms of growth and optoelectronic properties. On the other hand, the bismide GaAs1-x Bix alloy is relatively unexplored and prior results from GaNx As1-x are used a guideline. Nitrogen incorporation is known to result in several structural and optoelectronic problems, severe reduction of the electron mobility and luminescence efficiency as examples[7–9]. These negative effects of nitrogen incorporation have limited GaNx As1-x for use in semiconductor applications[10]. Quarternary alloys using a combination of nitrogen with indium allowed for reduced incorporated amounts of nitrogen for a given change in band gap energy, however the amounts necessary for practical use in telecommunications as light emitters still resulted in material degradation. Bismuth was proposed as an possible alternative to indium in the quaternary alloy. As it is both larger and has a much stronger effect on the band gap compared to indium, the GaNx As1-x-y Biy quaternary alloy requires less nitrogen for a given reduction of the band gap of GaAs[11]. A better understanding of the effects of bismuth incorporation in GaAs is necessary before the quaternary alloy is further explored. Prior to the undertaking of this thesis work, GaNx As1-x was understood in terms of the band-anti-crossing model[12] where an isolated nitrogen produces a state within the conduction band near the band edge. Interaction of this state and the conduction band of GaAs results in the shift in band gap with increasing nitrogen content. The scattering of electrons associated with this nitrogen resonant state results in a dramatic reduction of the electron mobility[7, 13, 14]. The hole mobility in GaNx As1-x is not strongly affected[14]. From this it is expected that the electronic transport in GaAs1-x Bix should be similarly affected by the bismuth resonant state in the valence band. The measurement of the hole mobility in gaasbi is the focus of the work presented in this thesis. The electron mobility in GaAs1-x Bix (x = 1.4%) samples was measured and found to be similar to the expected electron mobility for GaAs[15]. In terms of crystal growth, GaNx As1-x is well understood and films could be reliably produced. In fact, I was trained in the use of the growth chamber growing GaNx As1-x epi-layers by Dr. N. Zangenberg. It remained unclear as to the preferred species of active nitrogen which resulted in the best quality films, something that is explored in this thesis. On the other hand, the growth process window of GaAs1-x Bix was relatively unexplored 2  at the time of the undertaking of this thesis. GaAs1-x Bix was first made by metalorganic vapour phase epitaxy using low growth temperatures by Oe et al. in 1998[16]. This work followed from work on other bismuth containing semiconducting alloys[17–19], where the combination of semiconductor alloys (i.e. GaAs) and a semimetal (i.e. GaBi) is proposed to give temperature insensitive band gaps[20]. Oe et al.reported that the band gap energy shifts / K , much smaller than the change observed in GaAs, -0.54 meV/ K [16, 21]. It should -0.1 meV be noted that GaBi has not yet been synthesised, though it is believed to be a semimetal with a band gap energy of -1.45eV[3]. Materials with band gap energies that changed very little with temperature would be highly useful in telecommunications as light emitters. A complicating factor faced during the growth of GaAs1-x Bix alloys is that bismuth tends to surface segregate due to its large size. This leads to the formation of droplets on the surface when significant amounts of surface bismuth are present. It is also the reason bismuth is such an effective surfactant. Initial difficulty in incorporation of bismuth lead to the discovery of its effectiveness as a surfactant[22, 23]. The use of bismuth as a surfactant during the growth of GaNx As1-x is further explored here through measurements of the density of deep level defects, which were performed in collaboration with Dr. P. Mooney at Simon Fraser University. GaAs1-x Bix was first grown by molecular beam epitaxy at UBC by Tixier et al.[24], low temperatures and low arsenic overpressures were used. The starting point for this was the work done by Dr. S. Tixier and Dr. E.C. Young[24, 25]. The conditions under which bismuth incorporates in GaAs are significantly different than standard GaAs growth conditions: low growth temperatures and low III:V flux ratios are required. The low growth temperatures are necessary to reduce the tendency of bismuth to desorb from the surface, while the III:V flux ratio reduces the competition with arsenic. The small process window makes the growth of GaAs1-x Bix challenging, and much work was needed before growth of suitably thick doped epi-layers could be grown for electronic transport characterisation. With the use of in-situ UV light scattering and RHEED it was possible to monitor both the surface roughness and the onset of metallic surface droplet formation. This work on the GaAs1-x Bix growth process window is discussed in detail. The focus of the material characterisation in this thesis is the electronic states present in the material due to incorporation of either nitrogen or bismuth. As stated, at the beginning of the thesis the growth of dilute nitride GaNx As1-x epi-layers was well established, though it remained unclear if the use of atomic or excited molecular nitrogen was preferred for high quality film growth. What is meant by high quality in this case is improved optical efficiency as measured by comparing the luminescent intensity of the GaNxAs1-x films. This was explored by manipulation of the operating conditions of the rf-plasma source to preferentially produce increased relative amounts of atomic or excited molecular nitrogen for incorporation[26]. Also in the case of the nitride alloys, the effect of growth conditions on the density of shallow gap states associated with clusters of nitrogen atoms was explored. In the case of the bismide alloys, the focus was the effect of incorporation on the electronic transport properties. Despite the strong luminescence shown in the GaAs1-x Bix alloys[25, 27], it was still reasonable to assume that effects observed in the dilute nitride GaNx As1-x alloys in relation to the conduction band and electrons may manifest themselves in the bismide 3  GaAs alloys in relation to the valence band and holes. There is one reason the mobility of holes in GaAs1-x Bix was of primary interest. Hole mobility is also important to understand in GaAs1-x Bix in terms of device applications. As an example, high hole mobility in the base of an npn HBT allows for for faster operation and the band offset reduces the backflow of holes from the base to the emitter, as well lowering the threshold voltage (reduces power consumption). The electron mobility of GaAs1-x Bix has been measured[15, 28] and bismuth incorporation found to not have a strong effect. As the hole mobility in GaNx As1-x was also not strongly effected by the incorporation of nitrogen[7, 14], This is consistent with the analogy of these two material systems. Through exploration of the temperature dependence of the mobility, various scattering mechanisms can be identified and their relative effects related to the incorporation of bismuth. Deep levels present in both GaNx As1-x and GaAs1-x Bix epi-layers were explored using deep level transient spectroscopy. Thin GaAs1-x Bix films were incorporated into p-i-n heterostructures to evaluate the materials potential in a light emitting device. These pi-n heterostructures were also looked at with deep level transient spectroscopy. While the DLTS analysis was limited by poor surface morphology of the GaAs1-x Bix epi-layers and the two-sided nature of the p-i-n diodes, several deep traps were found in these devices not found in similar GaAs devices. This thesis begins with some of the basics of semiconductor physics, focusing on aspects most relevant to transport and device applications. All experimental techniques used will be discussed, again with more in-depth discussion on techniques pertinent to mobility measurements in chapter 3. Also discussed in chapter 3 is molecular beam epitaxy, both in general and in specific relation to the homoepitaxial growth of GaAs, as well as some of the particulars of molecular beam epitaxy done within our research group. The growth of both dilute nitride GaNx As1-x and bismide GaAs1-x Bix will be discussed in detail in chapters 4 and 5, respectively. Specific to GaAs1-x Bix growth in-situ light scattering and reflection high energy electron diffraction will be covered. Material quality will be discussed initially in relation to x-ray diffraction and photoluminescence measurements. Hole transport in GaAs1-x Bix epi-layers was explored using the van der Pauw method, both at room temperature and as low as 25K, experimental details and results are given in chapter 6. The design and fabrication of devices incorporating these materials will also be explained in chapter 7. Device performance was assessed based first on current-voltage characteristics. Schottky diode devices for deep level transient spectroscopy were further evaluated using capacitance-voltage measurements, and electroluminescence measurements used in the case of the light emitting diodes. Results from deep level transient spectroscopy measurements are present in chapter 7 as well. The final chapter will summarize and discuss the conclusions regarding both the growth and measurements performed.  4  chapter 2:  BACKGROUND The following chapter will provide an overview of semiconductor physics, with a focus on GaAs, GaNx As1-x and GaAs1-x Bix . The first section will cover some of the fundamentals of semiconductor physics: band structure, Fermi energy, doping and defects. The following sections discuss semiconductor alloys and transport properties.  2.1  Semiconductor physics  Semiconductors can simply be defined as materials with resistivity values in the range 1x10−3 to 1x109 Ω cm; GaAs has an intrinsic resistivity of 3.3x108 Ω cm. There are also several other physical characteristics that are unique to semiconductors: rectifying effects in devices, sensitivity to impurities, both negative and positive charge carriers, decreasing resistance with increasing temperature. These characteristics are all due to the small number of electrons that are excited to the conduction band because of the relatively small band gap at finite temperatures. The band gap is defined as the energy difference between the top of the valence band and the bottom of the conduction band. In a metal the conduction band is partially filled. In the case of insulators bands are sufficiently separated as to prevent electrons from being thermally excited to the conduction band. Typical values for the band gaps of semiconductors are 0.1 to 3.0 eV; GaAs has a band gap at room temperature of 1.42 eV.  2.1.1 Bands As a result of having a periodic arrangement of the atoms in a crystal, referred to as the lattice, the allowed energy states of the crystal’s electrons exist in bands. Bands are described by a dispersion relation, E(k), which is found from solving Schr¨odinger’s equation in a periodic potential using a one electron approximation. Solutions take the form of periodic Bloch states[29], which are also periodic in reciprocal space. This allows us to represent the 5  (a) GaAs Unit cell  (b) First Brillouin zone of GaAs  EF  L  Γ  X  Γ  (c) Band structure of GaAs  Figure 2.1: The unit cell, first Brillouin zone[30] and band structure of GaAs. The band structure shown here was calculated using DFT and the band gap corrected to 1.42 eV.  band structure of our crystalline system in the reduced zone scheme (i.e. one Brillouin zone). The first Brillouin zone for a GaAs (zinc-blende) crystal is shown in Fig. 2.1, along with the unit cell and band structure. The lower bands are referred to as the valence bands, the upper bands the conduction bands and the energy difference between the valence band maximum (VBM) and the conduction band minimum (CBM) is defined as the band gap, Eg . We will be mostly concerned with the regions very near the VBM and CBM. At these points the bands can be approximated  6  to be isotropic (spherical) and parabolic[31]: 2 2  k 2m∗c 2 2 k = Ev − 2m∗lh 2 2 k = Ev − 2m∗hh  Ec,k = Ec + Ev,k  (2.1a) heavy holes  (2.1b)  light holes  (2.1c)  where Ec is the energy of the minimum of the conduction band, Ev is the energy of the top of the valence band, is Planck’s constant divided by 2π, k is the electron wavevector, m∗c is the electron effective mass, and m∗lh and m∗hh are the effective mass of light holes and heavy holes respectively. Note that the valence band in GaAs is made up of four overlapping bands: light holes; heavy holes (degenerate) and the split-off band. I have left out the third (split-off) valence band in the above description of the band edge as it lies below the VBM by a factor ∆o (see Fig. 2.3) and does not play a role in the carrier transport measurements discussed here. The effective mass of the charge carriers is found from the curvature of the band(s): 1 1 ∂ 2 Ek = (2.2) m∗ ∂2k The effective mass of the conduction band electrons is m∗c = 0.066mo, the effective mass of the valence band light holes is m∗lh = 0.082mo and the effective mass for the valence band heavy holes is m∗hh = 0.52mo , where mo is the mass of a free electron, 9.1091x10−31 kg. The effective mass of holes in the split-off band is 0.15mo . An effective mass tensor is defined in the case non-spherical bands[31]. The number of electrons excited into the conduction band at a given temperature is given by[31]: n=e  ∞  −(Ec −EF )/kB T  Nc (E)e  −E/kB T  dE  (2.3)  Ec  where E is  2 k2 , 2m∗c  EF is the Fermi energy of the semiconductor, the Fermi distribution has be  approxiamted by e −(Ec,k −EF )/kB T and Nc is the density-of-states (DOS) given by: 1 Nc (Ec,k ) = 2 2π  2m∗c  3 2  2  Ec,k − Ec  (2.4)  When the above integral is performed we write: n = N¯c e −(Ec −EF )/kB T  (2.5)  where N¯c is referred to as the effective DOS: N¯c = 2  m∗c kB T 2π 2 7  3 2  (2.6)  which for GaAs has a value of 4.2x1017 cm−3 at 300 K. A similar argument is made for the valence band and gives the density of hole as: p = N¯v e (Ev −EF )/kB T  (2.7)  where N¯v is the effective density of hole states, similar to equation 2.6 with the density-ofstates effective mass, m∗v defined as: 3/ 2  3/  m∗v = (mlh ) 2 + (mhh )  2/ 3  = 0.53 mo  (2.8)  For GaAs at 300 K, N¯v is 9.0x1018 cm−3 . This is a sum of the contributions from the heavy and light hole bands, p = plh + phh (Nv = Nv,lh + Nv,hh ). To find the intrinsic carrier concentration, ni , and the position of the Fermi energy in an intrinsic semiconductor we use the relationship, n = p = ni to write: n2i = np = N¯c N¯v e −Eg /kB T  (2.9)  where we have used the definition of the band gap energy, Eg = Ec − Ev . The intrinsic n carrier concentration for GaAs is 2.1x106 cm−3 at room temperature. Using the ratio of / p we find the Fermi energy to be: ∗ 1 EF = Ev + Eg + /3 4 kB T ln mv/ m∗c 2  (2.10)  At 0 K the Fermi energy lies directly at the mid point of the band gap (mid-gap) and changes at higher temperatures depending on the ratio of the effective masses; in the case of GaAs we find that the Fermi energy is displaced from mid-gap towards the conduction band by 40 meV at room temperature.  2.1.1i) Band gap as a function of temperature The band gap of a semiconductor is dependent on the lattice parameter and electronphonon interaction; therefore, the temperature dependence of the band gap is related to the thermal expansion of the lattice and the temperature effects related to the electron-phonon interaction[1]. The latter effect is the majority contributor to the shift in the band gap. This gives rise to a linear temperature dependence at high temperatures which becomes non-linear at temperatures below the Debye temperature, θD : ∆Eg ∝ T 2 , ∆Eg ∝ T,  f or T ≪ θD f or T ≫ θD  (2.11a) (2.11b)  The above behaviour can be fit by the Varshni equation for Eg (T )[32]: Eg (T ) = Eg (0) − 8  γ T2 T +β  (2.12)  GaAs  /K γ, meV 0.5408  β, K  θD , K  204  344  Table 2.1: Varshni fit parameters for GaAs[21] and the known Debye temperature.  where Eg (0) is the T=0 K band gap energy, and γ and β are material-specific constants. The constant β is reported to be related to the Debye temperature, θD , though there is no strong correlation between the two; in some cases β may take on negative values[33]. Table 2.1 gives the Varshni fit parameters for GaAs. The Varshni equation has proven to be highly useful experimentally and is widely reported in the literature, however it should be noted that its theoretical basis is weak. An expression with more physical significance (and improved fit to experimental data) was proposed by O’Donnell and Chen[34]: Eg (T ) = Eg (0) − S ω  coth  ℏω 2kB T  −1  (2.13)  where S is a dimensionless coupling constant and ℏω is the average phonon energy. This expression is derived from the entropy and enthalpy of the formation of electron hole pairs in a material.  2.1.2 Defects in semiconductors Semiconductor defects can have the form of vacancies (missing atoms), anti-sites (atom at incorrect location), substitutions (different atom) or interstitial (atom not on a lattice site). The following discussions of extrinsic semiconductors and semiconductor alloys concern special cases of substitutional impurities: intentional addition of impurities to controlably change the material properties. Most often impurities are not added intentionally, can be detrimental to the material properties and are best avoided. In order to improve semiconducting materials and devices it is important to understand the source, nature and effects of all defects in the semiconducting materials. Defects that result in electronic states in the band gap of the material can be separated into two classes: those producing states near the middle of the band gap (deep traps or recombination centers) and those producing states near the band edges (shallow traps). Dopant elements are usually cases of shallow traps, where the impurity does not significantly perturb the lattice or band structure. The next section discusses doping in semiconductors in more detail. More often defects produce deep traps which are effective (non-radiative) recombination/generation centers, shorten carrier lifetimes or pin the Fermi energy at mid band gap. As an example of the last effect, semi-insulating GaAs is produced by having many near mid-gap defects associated with As anti-sites (AsGa ). Transition metal impurities also typically produce trap states deep in the band gap. Deep traps may also be defined as states produced by defects where the short range central part of the potential determines 9  the energy level of the defect[35]. This is sometimes the case for isovalent impurities, i.e., nitrogen and bismuth. This will be discussed in more detail further on in the semiconductor alloy section, section 2.2. In the case of defects that produce states within the band gap, but far from the band edges, we define: electron traps as traps nearer the CBM; hole traps as traps nearer the VBM; and recombination centers as traps close to the middle of the band gap. The relative capture rate into an empty state of a deep trap for electrons, cn is given by: cn = σn vn n  (2.14)  where σn is the capture cross-section, vn is the average thermal velocity of the electrons and n is the density of electrons. The above equation can be re-written for holes by simply replacing the subscript n by p. Traps are further denoted as either majority carrier traps or minority carrier traps in a doped material according to their likelihood to capture either an electron or a hole. For example, in n-type doped materials electrons traps are majority carrier traps and hole traps are minority carrier traps. The recombination and lifetime of charge carriers is strongly affected by the rate of capture of carriers into deep traps which act as non-radiative recombination centers. The emission rate, en , of an electron in a trap is an intrinsic property of the trap whereas the capture rate depends also on the carrier concentration. In thermal equilibrium these processes will be equal (detailed balance). Using this, we can find the emission rate of an electron from a trap, with energy ET , to be: en = σn vn Nc e −(Ec −ET )/kB T  (2.15)  where we can see that the emission is set by the capture cross-section,σn and the energy difference from the appropriate band edge. A measurement of the emission rate as a function of temperature gives the trap energy and cross-section.  2.1.3 Extrinsic semiconductors Doping impurities can be thought of as hydrogenic, where the dielectric constant and effective mass greatly reduces the ground state electron binding energy to a few meV’s and as a result the net concentration of free electrons or holes is increased. When donor or acceptor impurities are incorporated into a semiconductor, it is said to be doped either n- type or p-type; a donor impurity contributes an electron to the conduction band and an acceptor impurity takes an electron from the valence band (donates a hole). The concentrations of donors and acceptors are denoted by ND and NA respectively. The extra charge carriers must be taken into account to maintain charge neutrality, i.e. in the case of a n-type doped material, we now have: n = ND+ + p (2.16) 10  Table 2.2:  n-type, eV  p-type, eV  Si  0.0058  Si  0.035  Ge  0.006  Ge  0.04  S  0.006  C  0.026  Se  0.0059  Zn  0.031  Sn  0.006  Be  0.028  Te  0.03  Mg  0.028  Common GaAs n- and p-type dopant elements and their respective  ionization energies to the either the conduction band (n-type) or valence band (ptype).  where ND+ is the ionized donors. Similarly for p-type doped materials and ionized acceptor impurities, NA− . The number of ionized donors is given by: ND+ = ND NA− = NA  1 1 + ge  -(ED − EF )/ kB T  1 1 + ge  (EA − EF )/ kB T  (2.17a) (2.17b)  where ED and EA are the impurity level energies and g is the degeneracy of the impurity level, equal to 2 in the case of donors and 4 in the case of acceptors[31]. The degeneracy of the impurities levels is due to the fact that a donor impurity may have either a spin up or a spin down electron. In the case of acceptor impurities, the impurity may have either a light hole spin up, a light hole spin down, a heavy hole spin up or a heavy hole spin down occupying it giving a degeneracy of 4. At elevated temperatures all of the donor or acceptor impurity atoms can be considered ionized and the number of ionized donors or acceptors is approximately equal to the total number of donors or acceptors, ND ≃ ND+ . For convenience I will simply use ND or NA to refer to the number free carriers due to incorporated donors or acceptors. Table 2.2 gives a list of common donor and acceptor impurities used in GaAs and their ionization energies, i.e. |Ec − ED | and |Ev − EA |. In the case of extrinsic semiconductors the Fermi energy is shifted from the position of the intrinsic Fermi energy (eqn. 2.10) due to the additional charge carriers. Its position must be adjusted for doped semiconductors in order to preserve charge neutrality. Using eqn. 2.16 to find the concentrations of free electrons and holes in a doped semiconductor, we find for  11  the example of n-type dopant: nn =  1 2  (ND − NA ) +  (ND − NA )2 + 4n2i  (2.18a)  where in the above equation the subscript n has been used to indicate that this is the case of n-type doping. For ND − NA ≫ ni and ND ≫ NA (uncompensated doping), we have that nn ≈ ND . These two assumption will be valid for the doping concentrations and temperatures used in this thesis. This gives for p-type doped materials: pn =  n2i n2 ≈ i nn ND  (2.19)  The same arguments can be made in the case of p-type doping. Again, the above assumptions will always be true in the samples discussed in this thesis. Following the same calculation used to find eqn. 2.10, we find the Fermi energy of a n-type doped semiconductor (where I have assumed that the material is not compensated): EF n = EF + kB T ln  nn ni  (2.20)  pp ni  (2.21)  and similarly for a p-type doped semiconductor: EF p = EF − kB T ln  where EF is the Fermi energy in the intrinsic case, given above in eqn. 2.10. With changing temperature the Fermi energy of a doped semiconductor can be calculated using the above with the appropriate Nc , Nv and ni for the given temperature. At high temperatures where n, p ND , NA the Fermi energy approaches the intrinsic level. The Fermi energy as a function of temperature is typically found graphically and examples can be found in the references [1][30][31].  2.2  III-V Semiconductor alloys  A major limiting factor, common to all crystal growth techniques, is the availability of suitable substrates. From a commercial standpoint substrates need to be both large and inexpensive. This problem can be overcome by creating materials structurally similar to available substrates, but varying in optical or electronic properties. The easiest and most common way to do this is to alloy with a third or fourth element. The incorporation of small amounts of nitrogen or bismide within GaAs are two examples of ternary alloys. The fact that the necessary amounts are small for large changes in optoelectronic properties 12  Figure 2.2:  Band gaps as a function of lattice parameter for various III-V  semiconductor compounds. The trends for both GaNx As1-x and GaAs1-x Bix are shown by dashed lines.  make this possible; the crystal structure of the GaAs is only slightly affected, whereas the band gap changes significantly. Another III-V alloy used extensively as an example below is Inx Ga1-xAs, which is lattice matched to InP (see Fig. 2.2) for x = 0.47. Shown in Fig. 2.2 are the band gaps for several III-V semiconductor materials in relation to their lattice parameter. The band gaps of the ternary alloys is shown by the lines adjoining the constituent end members, i.e. Inx Ga1-xAs is indicated by the line connecting GaAs and InAs, where the amount of In increases in the alloy as you follow the line from pure GaAs to pure InAs. The band gap of a semiconductor alloy depends on the constituent elements through the lattice parameter and the electron-phonon interaction. The combination of three or more elements makes it possible to tune the resultant material’s band gap by changing the composition. The following section will discuss some of the effects of alloying in III-V semiconductor materials. The lattice parameter of an alloy can be approximated by a linear function of the composition, commonly known as Vegard’s law[36]: aA1−x Bx (x) = xaB + (1 − x) aA  (2.22)  where the superscripts A and B designate the constituent end-members. Table 2.3 gives values of the lattice parameter and band gap for common III-V semiconductors, and their respective band gap energies.  13  III-V Alloy ˚ Lattice parameter, A Band gap at 300 K, eV Table 2.3:  GaAs  InAs  GaP  AlAs  InP  GaN  GaBi  5.65  6.06  5.45  5.66  5.87  4.50  6.33  1.43  0.359  2.76  3.01  1.35  3.23  -1.45  Lattice parameter and band gap of common III-V semiconductor  materials[1]. The values for GaBi are theoretical predictions as the material has yet to be synthesized[3].  Vegard’s law is known to work very well for group III-V alloys[1]. To illustrate this point, consider In1−x Gax Asy P1−y which can be made lattice matched to InP for: x=  0.1896y 0.4176 − 0.0125y  (2.23)  which is valid for 0 ≤ x ≤ 0.45 and 0 ≤ y ≤ 1[37]. The above can be approximated by the simpler form, x = 0.47y. In this case the arsenic which is larger than the phosphorus it replaces, compensates for the smaller gallium which replaces indium (see table 2.4). For y = 1 we have Inx Ga1-xAs, which is lattice matched to InP for x = 0.47. element covalent radius, ˚ A Table 2.4:  N  Al  As  Ga  In  Bi  0.75  1.18  1.20  1.26  1.44  1.46  Covalent atomic radii for commonly used group III and V  semiconductor materials.  In GaNx As1-x-y Biy , nitrogen is much smaller and bismuth much larger than the arsenic they substitute for and again there is a possibility for lattice matching. For GaNx As1-x-y Biy on GaAs substrates the relation for lattice matched concentration is[38]: x = 0.588y  (2.24)  This assumes a linear relation in the lattice constants of GaAs1−x−y Nx Biy to GaN and GaBi. The above examples demonstrate that it is possible to lattice match quaternary alloys to a substrate over a range of compositions. This is generally not the case for ternary alloys; a counter example is the In0.53 Ga0.47As grown on InP (as in the example above for y = 1). For a non-lattice matched alloy, the epi-layer’s crystal structure must be deformed in order to match the lattice parameter of the substrate; in this case films are said to be pseudomorphically strained. Using known values of the lattice parameters for the substrate (in table 2.3) and unstrained epi-layer (from eqn. 2.22) a calculation of the strain, ǫ , or lattice mismatch, f , can be found: f =ǫ =  |aunstrained − asubstrate | asubstrate 14  (2.25)  For films in which the in-plane lattice parameter is larger than that of the substrate, the strain is compressive and the lattice parameter in the out-of-plane (growth) direction is expanded. In the opposite case, a smaller epi-layer lattice parameter, the strain is tensile and the lattice parameter in the out-of-plane (growth) direction is reduced. This change in the out-of-plane lattice parameter allows for a measurement of the compostion of an epi-layer based on the position of epi-layer x-ray diffraction peaks relative to substrate diffraction peaks. In both cases the expansion or reduction of the out-of-plane lattice parameter tends to conserve the volume of the epi-layer’s unit cell. The volume change in the epi-layer is given by: ∆V ∆a = (1 − 2ν) (2.26) V a where a is the in-plane lattice parameter and ν is the Poisson ration, which is 0.30 for GaAs. As a film grows thicker more strain energy is stored. This increases the probability of the formation of misfit dislocations, increased surface roughening and/or cracks. The critical thickness, hc , for the formation of misfit dislocations can be found from the MatthewsBlakeslee relation[39], which has been here simplified using known values for GaAs[40]:  hc =  0.4 1 − /4 ν πǫ (1 − ν)  ln h/c 4 + 1  (2.27)  This is known experimentally to be an underestimate of the critical thickness as the energy required to form dislocations is typically less than the thermal energy available. As a result it is possible to grow films thicker than this limit. In other words the formation of dislocations is kinetically limited and the grown films are metastable. The accessible film thicknesses will be negatively affected by increased growth temperatures, decreased growth rates or larger lattice mismatch. The incorporation of either nitrogen or bismuth will lead to strained films. The expected critical thickness based on the above for 2% incorporation are 154 nm and 308 nm, for nitrogen and bismuth respectively. As mentioned these are underestimates and for the purposes of this thesis GaNx As1-x films with similar concentrations have been grown to thicknesses greater than 500 nm and GaAs1-x Bix films with even higher concentrations have been grown to thickness of 400 nm without forming misfit dislocations. The band gap of an alloyed material can be approximated by a simple parabolic function[41], which is related to Vegard’s law: EgAB (x) = xEgB + (1 − x)EgA − bx(1 − x)  (2.28)  where the superscripts A and B designate the constituents of the alloy, x is the composition, and the constant b is known as the bowing parameter. Table 2.3 gives band gap values for common III-V semiconductors. Consider the example of Inx Ga1−x As. As more and more 15  indium is incorporated x varies from 0 (pure GaAs) to 1 (pure InAs). The band gap increases as shown in Fig. 2.2. The bowing of the band gap is seen in the bending of the band gap away from a linear relationship between the end-members. In the case for indium incorporated / %In [42]. into Inx Ga1-xAs, the parameter b has a value of 0.477 eV In both the nitride and bismide GaAs alloys, the host lattice is sufficiently perturbed as to require special treatment. Recall that nitrogen is the smallest of the group V atoms and that bismuth is the largest. The band gap bowing parameter of nitrogen and bismuth incorporation in GaAs is anomalously large: 7-16 eV for nitrogen[43] and 5.6 ± 1 eV for bismuth[11] in GaAs. Furthermore, it has been found to be necessary to introduce a concentration-dependent bowing parameter[44][27]. For these reasons, GaNx As1-x and GaAs1-x Bix are considered extreme or anomalous alloys of GaAs. The reason for the anomalous behaviour is due to states resonant with the band structure of GaAs; nitrogen has a resonant state near the bottom of the conduction band of GaAs and bismuth has a resonant state near the valence band maximum of GaAs. This is shown schematically in Fig. 2.3. These resonant states are considered deep traps as their energy is determined by their central potential[35]. Because of the relatively strong role of the central potential, the impurity-like nature of the incorporated element persists to high concentrations[43]. This is not the case in other isovalent impurities (i.e In, Al, P, etc) where the central potential is relatively weak and bands are formed for small incorporated amounts. The interaction of these resonant states with the band structure of GaAs causes the strong change in band gap in these materials. In the case of nitrogen the state lies 1.65 eV above the CBM[45]. The position of the bismuth resonant state is estimated to be 0.4 eV below the VBM[46]. The change in band gap energy with the incorporation of either nitrogen or bismuth is understood in terms of band anticrossing[47, 48]. For GaNxAs1-x we have: E − EM (k) −VN M (2.29) −VN M E − EN where EM (k) are the extended states of GaAs, EN is the localised energy state of the nitrogen or bismuth atom and VN M is the coupling potential between the states. The interaction potential is used as a material- and composition-dependent fitting parameter. It should be noted that there is some debate over the validity of the band anti-crossing model for GaNx As1-x (and therefore for GaAs1-x Bix ), see Mascarenhas et al.[49]. While the repulsion of the two bands resulting in the shift in CBM in the case of GaNx As1-x or VBM in GaAs1-x Bix is widely used, one would not normally expect a randomly placed alloying element, in this case nitrogen or bismuth, to result in a band. Inclusion of a nitrogen (or bismuth) will result in symmetry breaking and subsequent intraband coupling, which in turn would give rise to the level repulsion in the CBM (or VBM). However to be clear, the inclusion of either nitrogen or bismuth in GaAs does result in a state within either the conduction band or valence band near the band edge and a discussion of scattering from these resonant states follows further on in the thesis in relation to the analysis of the composition dependent mobility measurements.  16  Figure 2.3:  Nitrogen and bismuth resonant states within the band structure of  GaAs. Interaction of these states causes the observed strong change in band gap energy.  2.3  Mobility  Starting from the assumptions of the Drude model for conductivity[50], the mobility of a charge carrier can be defined as the constant of proportionality between the velocity of the carrier and the applied electric field. It is shown to depend on the carrier density, charge and resistivity of a material. Drude’s basic assumptions are:  1) A gas of conduction electrons moves under the influence of an electric field against a background of stationary positive ions. The motion of the electrons is given by Newton’s second law. 2) Electrons undergo instantaneous collisions and these collisions are the only means of achieving thermal equilibrium. 3) The average time between collisions is τ , referred to as the relaxation time.  From Ohm’s law we know that the potential drop along a conductor is proportional to the current carried and the resistance of the conductor, V = IR. This may also be written in terms of the electric field, E and current density, j; E = ρj, where the constant of proportionality is the resistivity of the material, ρ. 17  Consider a conductor of length, L, and cross-sectional area, A. The potential drop and current density may be written as: V = |E|L I |j| = A  (2.30a) (2.30b)  where the current, I (= nqA|v|) is due to a particle with charge, q, with a density, n, passing through the area, A, with and average velocity, v. Note that the sign of the carrier is not explicitly given; q = −|e| for electrons and both I and j will be in the opposite direction to /A. the velocity. This yields the well known relation for the resistance, R = (ρ L) From the above we see that the current density may be written as j = nqvav , where the qE velocity used in this case is the average drift velocity, vav = m ∗ τ . From this we can define the resistivity of a material as: −1 nq 2 τ (2.31) ρ= m∗ From the definition of the relaxation time in terms of the resistivity, we may define the mobility, µ, of a charge carrier as the constant of proportionality between the average velocity and the applied electric field: vav = µE;  µ=  1 q = ∗τ nqρ m  (2.32)  As the relaxation time is difficult to determine independently for a given material, typically what is done is use the measured resistivity to find the relaxation time or mobility. For conduction in a magnetic field, B, one can find the Hall coefficient. First let us define the average rate of change in momentum of an electron, where the average momentum for an electron at time, t is given by p = m∗ v (note that the subscript av has been dropped, all vector quantities will now be an average over all charges). The equation of motion is written as: p dp = F+ (2.33) dt τ where F is the applied force. Now if we apply a magnetic field along the zˆ direction, Bz and an electric field along the xˆ direction, Ex , the force in the above equation is given by the Lorentz force[51]: p F = −q E + ∗ × B (2.34) m As charges are deflected into the yˆ direction, a transverse field, Ey will build up to balance the Lorentz force. See Fig. 2.4. The Hall coefficient is then defined as: RH =  Ey 1 = jx B nq 18  (2.35)  Figure 2.4:  A Schematic diagram of a conductor with applied electric and  magnetic fields. Charges deflected along yˆ by the Lorentz force build up on the sides of the conductor inducing a field in this direction, indicated by the darker (negative) and lighter (positive) regions of the conductor. This field will increase until the Lorentz force is balanced.  where n is the density of the charge carriers; conventionally n is used in the case where the charge carriers are electrons and p is used when the charge carriers are holes. Combining this with the result above (eqn. 2.32), we see that it is possible to find the mobility from the Hall coefficient and the resistivity. The procedure used to perform such a measurement is discussed in the next chapter (see section 3.3.5). From the generalised form of µ in eqn. 2.32, we can see that the temperature dependence of the mobility lies in the effect of temperature on the relaxation time. Below, two factors that contribute to the temperature dependence of the relaxation time are discussed: scattering from ionized impurities and phonons. Both of these are unavoidable and are necessary in any discussion of carrier mobility as a function of temperature. As GaAs1-x Bix is an anomalous alloy of GaAs, further terms will be considered to limit the mobility related to the incorporation of bismuth. These will be discussed along with the results in presented chapter 6. As a reminder we are restricting our consideration of only isotropic (spherical) bands and non-degenerate carriers. When an electron from a donor impurity is promoted into the conduction band the remaining positive impurity ion will act as a scattering center for negative charge carriers. The same is true of acceptor impurities, the valence band and positive charge carriers. The interaction is treated as a screened Coulomb potential. This is the Brooks-Herring approach[31]. The approach is similar to Rutherford scattering and includes a term for the screening of the impurity by free carriers. The temperature dependence of the mobility  19  related to ionized impurity scatterers with a concentration of nI is[31]: µ I = CI  3/ T 2;  3/  4(4πǫo ǫ)2 (2kB ) 2 CI = 3/ √ π 2 m∗ nI Z 2 q 3  (2.36)  where ǫ is 12.9 for GaAs and Z is the atomic number of the impurity (dopant) atom. This result can be qualitatively understood when you consider that scattering cross-section is reduced for faster moving charge carriers and that higher temperatures result in faster moving carriers. Experimental measurement of ionized impurity scattering is complicated by the fact that at high temperatures scattering from phonons dominates and at low temperatures most impurities are neutral. Motion of atomic nuclei results in a change in the periodicity of the carrier potential which leads to scattering. This is simplified if we consider scattering from only longitudinal acoustic phonon modes[31]. Also we will restrict ourselves to intraband scattering. These assumption works very well in the case of holes. While this scattering process is inelastic, resulting in the emission or absorption of a phonon, the energy differences involved are generally much lower than the carrier energies and again we may simplify by treating this process as elastic. The total inverse relaxation time is given by the sum of the scattering for absorbed and emitted phonons and the mobility is found to be: µph = Cph  -3/ T 2;  1/  Cph  e 8π 2 ρc2l = ∗ m 3C12  2  2m∗ kB  3/ 2  (2.37)  here ρ is the density of the material (not to be confused with with ρ as the resistivity), cl is the longitudinal speed of sound in the material and C1 is the coefficient from the strain tensor. This result is clearly understood in more physical terms if one considers that the relaxation time in this case goes as the mean free path divided by the carrier velocity, where the mean free path is inversely related to temperature (more phonons at higher temperatures) and the average velocity of the carriers goes as the square root of the temperature, which gives the -3/ T 2 relation observed in the relaxation time and mobility above. The measured mobility (or total scattering rate) for a given sample will be the summation of all contributing terms: 1 1 1 1 1 1 = + + + + +··· µ µph µI µα µβ µγ  (2.38)  where each component of the mobility is assumed to act simultaneously and independently, i.e., Matthiessen’s rule. From the temperature dependence, the contributions to the mobility from the various scattering processes can be quantified.  20  chapter 3:  EXPERIMENTAL TECHNIQUES The following chapter discusses material growth by molecular beam epitaxy, the techniques used to evaluate the grown samples and the methods by which their optoelectronic properties were explored. The author does not attempt to comprehensively cover each topic, but rather the basic principles of each are explained and any aspects relevant to discussions further on will be noted. The chapter is divided into three sections: molecular beam epitaxy, where the growth of materials is discussed; in-situ characterisation, with covers the analysis techniques used during the growth process; and ex-situ characterisation, where the methods of exploring the material properties once they have been grown are discussed.  3.1  Molecular beam epitaxial crystal growth  In the most general use of the word, epitaxy refers to growth of a crystal on a crystalline substrate that determines its orientation∗ . In many practical applications, the in-plane lattice parameters will be the same as those of the deposited film. The case where the deposited film is the same material as the substrate material is known as homoepitaxy and the deposited film’s crystalline characteristics will be the same as those of the substrate. Deposited films typically have slightly different optoelectronic properties than the substrate due to small amounts of impurities incorporated in the growth process. For example, epitaxial GaAs is commonly slightly p-doped, a result of incorporating background carbon from the growth environment[52]. The deliberate addition of a significant level of another atomic species allows for the controlled modification of the properties of the deposited film as discussed in earlier (sections 2.1.3 and 2.2). Molecular beam epitaxy is carried out under ultra-high vacuum conditions (UHV), where the background pressures are typically as low as 1x10−10 Torr. The UHV conditions are required for very high purity in the epi-layer. Molecular beams are produced by heating ultra-pure elemental sources, which are arranged about the substrate in such a way as to produce good sample uniformity. The low background pressures mean that the beams of ∗  from the Oxford English dictionary  21  atoms or molecules have a large mean free path and therefore will not interact with each other. Sources are most often arranged in a circular pattern, where each is directed at the substrate position. By varying the relative fluxes of the elemental sources the composition of the epitaxial grown films can be controlled. The introduction of source flux to the substrate is typically controlled by a shutter; opening/closing the shutter turns the source flux on/off. A source’s flux can be calculated knowing the source’s temperature, T , and orifice area, A[52]: pA −2 −1 J = 1.118 × 1022 s (3.1) 1/ molecules cm 2 l (MT ) 2 where p is the vapour pressure of the source material at temperature T , l is the distance from the source to the substrate or ion gauge (20 cm), M is the atomic mass of the source material and the prefactor takes into account the geometry of the source[52]. More often the relative fluxes of sources are determined from beam equivalent pressures (BEP’s) measured using a retractable ion gauge. It is first necessary to correct the measured pressure for the measurement efficiency, η, of the ion gauge for the source material[53]: ηi =  0.6Zi + 0.4 14  (3.2)  where Zi is the atomic number of the element i. The above is based on an linear fit to measured data where the value for N2 is taken to be unity. The above equation is known to work well for sources that produce atomic species, but less so for molecular source species[53]. The values of η for the source materials used in the MBE growth chamber used in this thesis are given in table 3.1. The ratio of fluxes can then be found from[54]:  Table 3.1:  source  Ga  N  Bi  In  Al  As2  η  1.73  0.7  3.96  2.5  0.96  4.0  Calculated values of the ionization efficiencies of various elemental  source material used in the growth of GaAs alloys from Preobrazhenskii[54]. The value for the arsenic dimer is found experimentally.  ηi Pi Ji = Jk ηk Pk  Ti Mk Tk Mi  (3.3)  where P is the measured BEP, T is the source temperature and m is the source atomic mass. Absolute values of flux can be found from knowledge of the growth rate and the above equation used to find values for other source materials. It is not uncommon for BEP’s and their ratios to be quoted instead of actual fluxes as they are used on a daily basis and more readily available to material growers. Throughout this it will be made clear if BEP or flux ratio is being referred to. 22  Dopant sources are used at conditions where their fluxes are below the limits of the sensitivity of the ion gauge. Dopant sources therefore need to be calibrated by other means. By using an estimate of the mobility for doping concentrations near the expected incorporated levels, the sources are calibrated by measuring the resistivity of epi-layers. These calibration values are then later confirmed/corrected for by more accurate and direct measurements of the doping concentration from Hall measurements.  3.1.1 GaAs growth by molecular beam epitaxy The standard growth conditions used for the growth of GaAs bulk epi-layers, including the buffer layer, are Tsubstrate ⋍ 560◦ C and As:Ga flux ratio in the range 3-8. GaAs samples can be grown with substrate temperatures in the range 200◦ C to 650◦ C. Higher quality samples are grown near the higher end of this range. Using growth temperatures at the lower end of this range leads to films that are non-stoichiometric, with excess As incorporated into Ga sites or interstitially degrading the material properties[55]. Temperature during growth was monitored using optical band gap thermometry with an accuracy of approximately ±2.5◦ C[56, 57]. The growth rate is controlled by the arrival rate of group III (Ga) atoms, while an over pressure of group V (As) atoms is maintained; it is necessary to keep the As:Ga flux ratio above stoichiometric levels. A 2×4 surface reconstruction is observed with reflection high energy electron diffraction (RHEED) under these conditions for the substrate temperature in the approximate range 450-650◦C[58]. See discussion below on RHEED set-up and observations, section 3.2.1. In this reconstruction the surface unit cell lattice parameter is twice that of the bulk along 011 and four times as long along [011]. These in-plane crystallographic directions are identified according to US convention flats, which are adopted in the thesis (see appendix for the correct directions of EU convention wafers). When the As:Ga flux ratio approaches unity a 1×1 reconstruction is observed. This is attributed to a random distribution of both the 2×4 and 4×2 surface constructions[59]. The diffusion length of the group-III atoms along the surface is determined by the substrate temperature, growth rate, and step density (roughness). High substrate temperatures and slow growth rates allow for greater diffusion and therefore smoother surfaces. Under the standard conditions described above Ga adatom diffusion is 2-dimensional. Away from the standard conditions, low temperature and/or low As overpressure, Ga adatom diffusion occurs more readily in the [1¯10] direction and not so readily in the [110] direction. These are therefore referred to as the fast and slow directions. The surface of a GaAs crystal is Asterminated under the conditions typically used for growth. There exists a higher density of bonds along the [1¯10] direction, allowing for greater motion of the gallium adatoms along this direction. It is also known that the presence of bismuth on the surface also tends to smooth the surface[23]. The enhanced surface diffusion in the presence of bismuth is confirmed by the observation of RHEED oscillations at low temperatures and low As:Ga flux ratio conditions 23  only when bismuth is present on the surface[60]. The observation of RHEED oscillations is a strong indication of 2-dimensional growth. A typical mode of growth for GaAs homoepitaxy under standard conditions is step flow growth. Step flow growth is considered 2-dimensional as the surface features are relatively flat. Adatoms will diffuse around the surface and attach themselves to a step edge. For large terrace sizes, i.e. low step edge density, small nucleated islands may form in the regions between the steps, these will also act as attachment sites for adatoms. An increase in step edge density is the same as to an increase in surface roughness. When growth is interrupted the sample is maintained at the growth temperature with an overpressure of arsenic, the adatoms from these islands will migrate to the step edges and in this way the surface is smoothed[61, 62]. At low temperatures the growth mode deviates from step flow growth. This is related to the inability of the Ga adatoms to readily diffuse around the surface. Again the presence of bismuth as a surfactant helps maintain smooth surfaces. All films were grown for this thesis in a VG-V80H molecular beam epitaxy deposition system using standard thermal Knudsen effusion cells for gallium, aluminum, indium and bismuth. A valved two zone cracker source was used to produce arsenic dimers. An in-house designed and built rf plasma source was used for active nitrogen[59]. Dopant elements were introduced via an external pressure controlled system; CBr4 was used as a carbon p-type dopant source and SiBr4 as a silicon n-type dopant source. A silicon Knudsen cell was also eventually introduced to the chamber. BEP’s of the sources at the surface of the substrate during a typical growth ranged from 1x10−10 to 1x10−6 Torr. Sample substrates were rotated at 0.2 Hz during growth process to ensure uniform deposition. Films were grown on polished, (001)-oriented GaAs 2” substrates cleaved into quarter (or smaller) pieces. Substrates of thickness 350 µm or 450 µm were used with either US or EU conventional orientation indicating flats (see appendix for clarification of substrate orientation and conventions). Substrates used were either semi-insulating (SI), doped ptype with zinc (1x1018 -1x1019 cm−3 ) or doped n-type with silicon (1x1018 cm−3 ). Before each substrate was loaded into the growth chamber, they were degassed for at least one hour at temperatures above 200◦ C in the MBE preparation chamber. Once loaded into the growth chamber and prior to growth each substrate is heated to above 600◦ C for more than 10 minutes, under an overpressure of As2 , to remove the native surface oxide layer. This desorption process is rather violent and tends to leave the surface in a roughened state with submicron pits observed on the growth surface. The substrate temperature is then ramped down to appropriate levels and the growth of the GaAs buffer is started. The buffer layer is necessary to smooth the pitted surface. GaAs buffer layers of at least 100 − 200 nm were grown before epi-layers were deposited on all substrates. During the growth an overpressure of group V elements is maintained (V:III flux ratios were kept between ≃ 1 and 8) and the arrival of the group III elements at the crystal 24  surface is used to control the growth rate. Group III fluxes were typically on the order of / hr . The fluxes 4x1014 cm−2 s−1 , which corresponds to growth rates of approximately 1.0 µ m for the arsenic and gallium source correspond to source cell temperatures of 400◦ C and 950◦ C respectively. Growth rates were determined from the Pendellosung fringes observed in the high resolution x-ray diffraction of GaAs epi-layers where a thin interruption layer (a single InGaAs or GaNAs quantum well) is included. The combination of growth rate and source fluxes was used to predict epi-layer compositions and thicknesses. The specific conditions under which the dilute nitride and bismide alloys are grown will be discussed in greater detail in the next chapter. Samples were doped by simply introducing carbon or silicon flux into the growth chamber. This is achieved using a Varian mass flow controller for either CBr4 or SiBr4 , where the molecule cracks on the surface of the heated substrate and the bromine is then pumped away by the vacuum system. A silicon knudsen cell was also added as a source, several small pieces of un-doped silicon wafer were used as the dopant source material. Dopant atoms were incorporated into the growing films under the growth conditions of the epilayers. No significant effect was observed for the various growth conditions used on the doping concentrations measured.  3.2  In-situ sample characterisation  The characterisation that takes place in-vacuo during the growth of the epi-layers is essential to the production of high quality films. Described below are two of the techniques, RHEED and UV light scattering, used to collect information about the as-growing epilayer. The use of both of these techniques was particularly useful in exploring the growth process window for the GaAs1-x Bix epi-layers, as growth of thick, doped films proved difficult to produce without significant build up of metallic surface droplets (discussed in more detail, section 5.1). The growth of GaNx As1-x films was well understood at the time of the undertaking of this thesis, and the use of the surface analysis played a less significant role. The optical band gap thermometry set-up used to measure the substrate temperature during growth will not be discussed in detail here as the effect of temperature on the band gap of GaAs having been discussed in the previous chapter. Suffice to say that the GaAs substrate growth temperature is found from fitting the measured optical transmission spectrum. More info on the technique can be found in the following references[56, 57, 63–65].  25  (a) x1  Figure 3.1:  (b) x2  (c) x3  Example RHEED images:  a) 1x, b) 2x and c) x3 surface ¯ directions reconstructions. Observations are made along either the [110] or the [110] (US convention wafers).  The reconstruction is observed in the extra streaks  appearing in the RHEED diffraction pattern.  3.2.1 Reflection high energy electron diffraction RHEED is a widely used surface analysis technique, especially for the measurement of surface structure during the growth of epitaxial films. Electron diffraction requires only a simple set-up: an electron gun and a phosphor screen[66]. The analysis of the diffraction patterns can prove to be rather difficult, but for the purposes of this thesis we are only concerned with some of the relatively simple information available from RHEED diffraction patterns: surface roughness, growth rate and surface reconstruction. In the VG-V80H system the electron gun is situated on one side of the growth chamber and the phosphor screen on the other. Electrons are incident on the substrate at the grazing angle of 3◦ . At the energies used (15 keV) and the grazing incidence the electrons are sensitive to only the outermost atomic layers. Both the shape and the intensity of the diffracted beams are affected by variations of the surface structure on the atomic scale. The native oxide present on all GaAs prior to growth is amorphous and therefore gives no RHEED pattern. When the oxide is thermally desorbed, the surface remaining is rough but crystalline, and gives a RHEED diffraction pattern. This pattern is spotty due to the 3-dimensional structured surface. Growth of the GaAs buffer layers smoothes the surface and the RHEED pattern is seen to elongate into streaks. A streaky pattern is associated with a smooth, 2-dimensional surface. The appearance of chevron shaped diffraction streaks is an indication of a facetted (roughening) surface. Reorganization of the structure of the growth surface is necessary to minimize surface energy. This reorganization, referred to as the reconstruction of the surface, makes itself evident in extra diffraction spots/streaks appearing in the RHEED pattern, see Fig. 3.1. A doubling of the periodicity at the surface relative to that of the bulk, results in one additional line in the RHEED pattern; tripling of the surface periodicity results in two additional lines, 26  underlying substrate topmost atomic layer 1x1 2x1 1x3  [1 1 0]  [1 1 0]  Figure 3.2:  Examples of possible atomic surface reconstruction configurations  that would result in the RHEED patterns shown in Fig. 3.1. By convention the reconstructions are labelled as [1¯10]×[110]. 175 TGa=950 C TGa=925 C TGa=910 C  RHEED Intensity (arb. units)  Ga open 170  165  160  155  0  5  10  15  time (s)  Figure 3.3:  Observation of RHEED oscillations for varied Ga flux. The growth / hr rate can be determined form the period of the observations: TGa = 950◦ C, ≃ 1 µm / hr ; TGa = 910◦ C, ≃ 0.5 µm/ hr . ; TGa = 925◦ C, ≃ 0.6 µm  and so on (see examples in Fig. 3.2). Surface reconstructions depend on the substrate temperature and arsenic overpressure[58]. This makes it possible to ascertain information about the conditions at the surface of the growing film. As atomic layers start, fill and are completed the RHEED diffraction pattern will dim and brighten as each successive monolayer is deposited, see Fig. 3.3. In this way it is possible to directly measure the growth rate in terms of monolayers per second. Smooth starting surfaces are required to observe this phenomenon. It is necessary to stop the growth and 27  anneal before restarting the growth to observe RHEED oscillations, which damp out over time.  3.2.2 Light scattering Optical monitoring of the substrate, using diffusely scattered light, is a convenient means of observing the growth surface in-situ. Features on the order of the wavelength of the light being used can be detected using this method. There is also a dependence on the scattering angles[61]. The set-up presently in use uses an Ar+ laser at a wavelength of 514 nm and is sensitive to features with a spatial frequency of 15 µm−1 with the ports being used. The diffusely scattered light is collected by a photomultiplier tube (PMT). The relative intensity of the collected light is a measure of the relative roughness at this scale throughout the growth process. The intensity of the signal from the polished substrate surface and from the roughened oxide free surface are used as benchmarks for a given measurement. Much work was done by Dr. M.B. Whitwick in the set-up and maintenance of the light scattering apparatus. At the spatial frequencies available in the light scattering, the light scattering apparatus is sensitive to the surface roughness associated with the thermal desorption of the native oxide and the formation of metallic surface droplets. The use of light scattering to monitor the surface for droplet formation has proven invaluable in the exploration of the growth process window for GaAs1-x Bix [67].  3.3  Ex-situ sample characterisation  The following section describes the analysis techniques used post growth to characterise the epi-layers, where the goal may be to measure the change in epi-layer quality with changing growth conditions or to measure the effect of changing composition on the material’s optoelectronic characteristics. Both the x-ray diffraction and photoluminescence should be considered as standard techniques used to evaluate material quality, based on the features of the respective spectra (i.e. intensity, peak position and width). The other techniques were chosen to evaluate the materials in specific ways, with the ultimate goal of understanding any performance limiting factors related to growth conditions or composition.  28  3.3.1 X-ray diffraction High resolution X-ray diffraction (XRD) is the primary tool used to evaluate the structure of the grown films. θ−2θ scans of the GaAs (004) peak give both qualitative and quantitative information about the epi-layer. The position of a given peak in an XRD spectra depends on the spacing betwen the planes: nλ sin(θ) = (3.4) 2dhkl where θ is the angle of the diffracted beam from the plane of the atoms, n is an integer, λ is the wavelength of the x-rays and dhkl is the spacing of the planes. The GaAs (004) peak is found at 33.0511◦. As discussed above in section 2.2 the growth of an epi-layer on a mis-matched substrate leads to strain, which in turn causes the lattice parameter in the out-of-plane or growth direction to change. The assumption is made here that the lattice mis-match is sufficiently small (< 1%) and the epi-layers as-grown are not relaxed. Epi-layer compositions can then be found from the separation of the GaAs (004) peak and the epilayer (004) peak. Note that the growth direction used throughout the thesis is (001). This assumes that the epi-layer is un-relaxed, that is the epi-layer has the same in-plane lattice spacing as the substrate. Off-axis diffraction peaks, typically (115) for GaAs, are used to verify that the films are not relaxed. The epi-layer thickness may also be estimated from the relative intensity of the peaks and the width of the epi-layer peak gives information about the uniformity and thickness of the epi-layer. The observation of Pendellosung fringes is an indication of uniform composition, un-relaxed films, and smooth interfaces. They are the result of thin film interference of x-rays between interfaces in the heterostructure, in bulk thin films the GaAs/epi-layer interface and the epi-layer surface. The spacing of the interference oscillations is a direct measure of the epi-layer thickness. Measurements were carried out with a Philips X’Pert XRD system using either a hybrid 4bounce or a 4-bounce monochrometer, with resolutions of 26 and 15 arcseconds respectively. Typically peaks were measured over a range of 1 − 2◦ using step sizes of 0.003◦ and count times of 2 − 4 seconds/step. Use of the higher resolution 4-bounce monochrometer was necessary only to observe the Pendellosung fringes of very thick ( 1000 nm) epi-layers. Data were modeled using Bede RADS software which is based on the dynamical theory of x-ray scattering. Composition and thickness values are taken from the results of fitted peaks whenever Pendellosung fringes are observed. Thicknesses are determined from growth rates and growth times otherwise.  3.3.2 Atomic force microscopy (AFM) The surfaces of the as-grown films were studied using a Digital Instruments Multimode Scanning Probe Microscope operated in tapping mode using 10 nm radius silicon tip or a 1 nm radius tungsten tip on a silicon cantilever; the latter is used where high resolution of the surface features is required. AFM gives a real space image of the film’s surface. 29  3.3.3 Photo- and electroluminescence Luminescence spectroscopy is a non-destructive method of investigating the electronic structure of a material. In photoluminescence, light incident on a sample optically excites the electrons. It is also possible to excite the electrons electrically, this is known as electroluminescence. In this case a fixed current or current pulse is passed through the material. In both cases it is possible to alter the density of the excited carriers by changing either the intensity of the incident light or increasing the current density. As electrons fall back into the lower energy states (i.e. recombining with holes) some of their energy is released in the form of radiation. The spectrum of the emitted photons contains information about the electronic structure of the material. The peak of the measured spectrum gives the band gap of the material as electron-hole pairs are most likely to recombine from the band edges. Higher energy photons are a result of thermal excitation above the band minimum, giving rise to a high energy tail in the photoluminescence spectra with a width proportional to kB T. Shallow states in the band gap can also radiate and this gives rise to a low energy tail in emission spectra. Phonons are emitted when the decaying electron interacts with the surrounding atoms and excites a vibrational mode in the lattice. Phonon emission is typically associated with defects or impurities in the material which tend to give states deep inside the band gap. The Auger effect is a result of an electron-electron interaction. The recombining electron-hole pair give their energy up to another electron or hole in either the valence or conduction band. Auger recombination will occur more frequently for larger excited carrier concentrations. The efficiency of the luminescence depends on the lifetimes of the excited carriers with respect to both radiative and non-radiative recombination processes. In practice, the efficiency of the radiative recombination processes can be quite low if there are competing non-radiative processes. Strong luminescence is a good measure of the quality (lack of defects leading to non-radiative recombination) of a grown epi-layer in terms of low recombination centers. A comparison of emission spectra of different samples is a good indicator of the relative quality of films. For photoluminescence measurements, 20 ns pulses of a 523 nm diode pumped frequency doubled YLF laser focused to approximately a 1 mm diameter is used at a rep rate of 1400 Hz. Each pulse has an energy of 2 µJ. A Keithly 220 programmable current source is used for electroluminescence. The emitted light is collected by a 512 element InGaAs linear diode array cooled to approximately -100◦C. Emission spectra are corrected for both the background light, dark current of the detector and the throughput of the measurement apparatus. Temperature dependent measurements are carried out in an ARS low vibration closed-cycle He optical cryostat.  30  3.3.4 Secondary ion mass spectroscopy SIMS can have a depth resolution down to 2 − 4 nm and detection limits of a few parts per billion depending on the cleanliness/purity of the sample and the element being measured. This makes it a highly sensitive analytical tool. It is however also a destructive technique. A primary ion source is used to sputter the surface of a material, the secondary ions produced by this process are then analysed with a mass spectrometer. The choice of primary source ion depends upon the sample being analysed and the element being profiled. Ga ions are a typical source for SIMS profiling. The charged secondary ions sputtered from a material represent only a small fraction of the particles ejected from the sample. Careful calibration of the sputtering rate and detection efficiency are needed to fully understand a SIMS profile. In the absence of such calibrations, knowledge of the growth parameters can be used to understand a SIMS measurement.  3.3.5 Resistivity and carrier mobility using the van der Pauw method To initially roughly calibrate the dopant sources, bulk doped epi-layers at least 500 nm thick are grown on SI substrates and long finger-like contacts are deposited. Current is applied to the outermost two contacts and the voltage measured across two adjacent inner contacts. The sample resistance is measured based on a fit of the current-voltage data, and the resistivity found from the known geometry of the sample. From the measured value of the resistivity, the doping concentration is calculated using an estimate of the carrier mobility[1]. These values are then used as guidelines for future growths. The doping source calibration is then confirmed and corrected when more direct measurements of the carrier concentration are performed. In practice the measurement of the resistivity is simple, quick, does not require ohmic contacts and gives reliable, accurate estimates of the true carrier concentration. The mobility of a charge carrier is found from measurements of the resistivity and the carrier concentration using the relationship found earlier (2.3): µ = (qnρ)−1  (3.5)  where q is the fundamental charge of an electron or hole, n is the free electron (or hole) density and ρ is the bulk resistivity. If the sheet resistivity, Rs , and sheet carrier density, ns are used the above equation becomes: µ = (qns Rs )−1  (3.6)  Using the van der Pauw method, both the sheet resistance and carrier density can be measured on a single sample of arbitrary shape with four contacts[68]. Samples used for 31  Figure 3.4:  Square van der Pauw sample schematic with the corner contacts  labelled in a clockwise fashion. The small quarter-circular contacts are shown to scale.  the measurements discussed in this thesis are all 7 × 7 mm squares with small quartercircular ohmic contacts at the corners, see Fig 3.4. Fabrication of the samples is discussed later on (see section 3.3.5). All measurements were carried out in the dark. To perform a measurement of the sheet resistance, a current is first applied to contacts 1 and 2, I1,2 and the voltage across contacts 4 and 3 is measured. R12,43 =  V4,3 I1,2  (3.7)  The current is then applied in the opposite direction to find R21,34 . The measured values of the resistances are checked to be approximately equal (within 5%). The current and voltage contacts are now interchanged and the resistances measure again; R43,12 and R34,21 . Again the values are checked to be approximately equal. These four values of the resistance in this sample direction are now averaged to give RA . This procedure is repeated for the other sample direction to give RB . The sheet resistance of the sample is given by the van der Pauw equation[68]. A derivation of this equation is given in the Appendix:  e  −πRA RS  +e  −πRB RS  =1  (3.8)  which must be solved iteratively. The procedure used is shown in Fig. 3.5 was used† . As values of yi approach 1, the sheet resistance also converges; typically the process required less than five iterations for the value of yi to be within 0.1% of unity. †  this  procedure  is  outlined  on  a  NIST  http://www.nist.gov/eeel/semiconductor/hall.cfm  32  webpage  for  Hall  measurements:  zo =  2 ln(2) π(RA +RB )  ✲  yi = e −πzi−1 RA+ e −πzi−1 RB  ✛  ❄ ✲  zi = zi−1 −  1−yi π  RA e −(πzi−1 RA )+ RB e −(πzi−1 RB )  −1  ✛  ❄  ❅   ❅   ❅ NO   ❅   ❅ 1-y i < 0.001?   ❅   ❅   ❅   ❅ ❅   YES  ❄  RS =  Figure 3.5:  1 zi  Iterative procedure used to calculate the sheet resistance from van  der Pauw resistance measurements on Hall samples. The above was repeated until the value of yi was sufficiently close to one.  In order to find the sheet carrier density the sample is placed in a magnetic field of strength 2.65 kG perpendicular to the sample surface with the current applied to contacts on opposite corners and the voltage measured across the other contacts. For example current in contact 1 and removed at contact 3, I1,3 , with the voltage measured across contacts 2 and 4, V2,4+ . The positive sign in the subscript is used to indicate the field direction. V2,4− is also measured and the difference of the two voltages gives VA . Again this is repeated for all other contact configurations giving VB , VC and VD . The sample majority carrier type is determined from the sign of the sum of the four voltages; if the sum is positive then the carrier is holes and electrons if the sum is negative.  33  The sheet carrier density is found from: ns or ps = 8x10−8 IB  1 |q (VA + VB + VC + VD ) |  (3.9)  where the absolute value of q (VA + VB + VC + VD ) is used in the above only to find the value of the sheet carrier density regardless of the sign of the carrier. The mobility of the sample can now by found using equation 3.6. Bulk values of the carrier density are found by dividing by the sample effective thickness, where the sample effective thickness takes into account the depletion width at the surface. To simplify the above procedure, a 4-pole/4-throw (4P/4T) switch was used to toggle through all the necessary combinations of current and voltages connections. A wiring diagram is included in the appendix. A Labview program was used to control the sourced current and to record the voltages. All resistances were based on a least squares fit to the current-voltage data. The same program was used for Hall measurements, where fits to the resistance were used to confirm the voltage value measured at a given current. A spread sheet was set up to record the data and calculate the sheet resistance, carrier concentration and mobility. Temperature dependent measurements of the mobility were carried out in an ARS low vibration closed-cycle He optical cryostat. Sample mounts were thermally anchored to the cold head using a thermal paste. Brass wiring was used and was thermally anchored to the heat shielding to lower the heat load on the sample mount. Base temperatures of approximately 8 K were achieved, though due to high resistivity found in the samples at these temperatures data collection began near 25K. Two calibrated diode thermometers were used to monitor temperature throughout the measurements; one was placed directly on the cold head, while the other was placed near the sample on the sample mount. The samples were shielded from the ambient light in the room using aluminum foil anchored to the heat shielding. This also acts to lower the radiative heat load on the sample stage.  3.3.6 Deep level transient spectroscopy This is a capacitance transient technique that is used to detect charges trapped in deep levels in the material within a junction. A more detailed overview of the measurement technique is found in the appendix, and only the basics of the method are presented here. The depletion region of a junction can be thought of as a parallel plate capacitor, with a well defined capacitance for a given doping concentration(s) and device area. The capacitancevoltage characteristics of the device yields values for the built-in potential,Vbi , of the device, and in the case of the the one-sided junctions the doping concentration, NA or ND . This is the reason that abrupt one-sided junctions are preferred in this measurement technique. Junctions are intially held in reverse bias and the capacitance is said to be at its equilibrium 34  value. By applying a forward bias pulse to the junction, the width of the depletion region is reduced and carriers can now become trapped by defects in this now accessible region of the junction. As these trapped charges are emitted back to the band edges when the forward biasing pulse is removed the measured capacitance decays exponentially back to the equilibrium value. The device’s doping concentration and leakage current, place limits on the deep level transient spectroscopy (DLTS) measurements’ ability to observe traps with low concentration or high activation energy, respectively. Defects are classified according to the type of carrier they trap, i.e. there are hole traps and electron traps. Holes will be emitted from hole traps back to the valence band edge and electrons emitted from electron traps back to the conduction band edge. A measurement of DLTS spectra, the change in the capacitance measured at two distinct times as a function of temperature, yields information about the activation energy of the defect level, its concentration and capture cross-section. The measurement and analysis of DLTS spectra can be quite difficult, if for example there are several traps with similar activation energies. Analysis can be further complicated by the fact that the capture cross-section may also be temperature dependent, adding a second parameter to the measured energy difference between band edge and the trap activation energy. The DLTS measurements were performed at Simon Fraser University by Dr. P. Mooney and her students Ingrid Koslow, Adam Royle, Eric Chen and Zenan Jiang on a SULA DLTS system. Two types of junctions were explored: Schottky metal-semiconductor junctions, p- and n-type, to measure bulk epi-layers; and the p-i-n junctions of light emitting diodes. Diagrams for these junctions are shown in the Appendix (Figs. D.1a and b). Samples were cooled to liquid nitrogen temperatures (77 K) under low vacuum (10 mTorr). Various reverse bias voltages were used based on the leakage current of the device being measured. The maximum usable temperature was also determined by the leakage current. Device fabrication is discussed later in sections 7.2 and 7.1.  35  chapter 4:  GROWTH and PROPERTIES of GaNxAs1−x During my Master’s degree, it was found that there are shallow gap states in the as-grown GaNx As1-x alloys. These states were attributed to cluster states near the conduction band edge and were found to be distributed in energy[6]. The use of bismuth as a surfactant reduced the density of these traps, based on the lowered PL emission from these shallow in-gap states. The reduction in shallow gap states was expected when bismuth was used as a surfactant, as surfactant use is known to increase luminescence intensity in GaNx As1-x films[23]. The goal of my Ph.D. work on the dilute nitride GaNx As1-x was to further investigate the effect of growth conditions on the shallow gap states observed in the material and to establish whether to use of active atomic or molecular nitrogen is preferred for higher quality epi-layers, i.e. decreased amounts of shallow gap states associated with nitrogen clusters. The use of two possible operation modes of the active nitrogen rf plasma source, onand off-resonance, was explored, as well as the effect of the condition of the plasma discharge tube. On-resonance operation produces more active atomic nitrogen, whereas off-resonance operation shows weaker atomic nitrogen and increased N∗2 (excited nitrogen dimer) production. On-resonance operation was the typical mode of operation as it was known to historically give increased nitrogen incorporation. A more complete description of the plasma source and its operation can be found in the Appendix. There was a noted decrease in incorporated amounts of nitrogen over a period of several growth campaigns for similar operation conditions of the plasma source. It was found that continued use of the plasma source had deposited a film of unknown composition (likely arsenic or arsenic-related) on the inside surface of the plasma discharge tube. This contaminant coating was found to greatly reduce the efficiency of the plasma source to produce active nitrogen species. The plasma source was only occasionally cleaned during periodic vents of the MBE chamber. During one such vent of the chamber the plasma source was thoroughly cleaned and the PBN discharge tube was soaked in an aqua-regia  36  bath (HNO3:HCl;1:3) to remove all surface contamination. Film composition and emission from samples grown before and after this cleaning were compared. This chapter will begin by describing the growth of the GaNx As1-x epi-layers and then present film characterisation results from SIMS and PL measurements. Discussion of the growth will start where the previous chapter left off - at the growth of a 100-200 nm thick buffer layer of GaAs using the standard conditions of Tsubstrate = 550-580◦C and an As:Ga flux ratio of 5 or greater. Example XRD spectra from GaNxAs1-x samples will also be presented along with the growth discussion and used to determine film nitrogen content. SIMS and PL are used to further evaluate the quality of the grown films, and in the case of the SIMS measurements to explore any possible contaminant incorporation.  4.1  Dilute nitride GaNx As1-x growth  The active nitrogen species are introduced to the growth chamber using an rf plasma source designed and built at UBC. The source and its operation are discussed in more detail in the Appendix. Nitrogen will only be incorporated in the epi-layer after it has passed through the ignited plasma source to become activated. Two modes of operation of the plasma source are used during the growth of GaNx As1-x films: on-resonance and off-resonance. These definitions are based on measurements of the frequency and power required to maintain the plasma, where the forward and reflected rf power are monitored with a directional watt-meter and the difference is the net power to the load (see Fig. E.2). On-resonance operation is defined as the frequency where the net forward rf power to the plasma is maximised and there is a local minimum in the reflected power. Away from this helix resonance we define another high plasma power operating mode, which we refer to as off-resonance. In this case the reflected power is not at a local minimum, however the net power to the plasma is at a local maximum. The activated nitrogen beam is comprised of excited atomic nitrogen and excited nitrogen dimers, N2∗ . Energetic ions are removed from the beam by the inclusion of a baffle at the end of the plasma discharge tube which increases the number of collisions undergone before exiting the tube. The ion current is reduced by a factor of 103 by this method[26]. The ratio of atomic nitrogen to excited dimers is different under the two operational modes; ‘on resonance’ produces more atomic nitrogen and ‘off resonance’ operation results in more excited nitrogen dimers (see Fig. E.3). The first step in the growth of GaNx As1-x films after the growth of the GaAs buffer layer is to reduce the temperature from the GaAs standard conditions to 450◦ C. This growth temperature is used to lower loss of nitrogen to the formation of volatile AsN on the surface of the substrate[59]. A further 200 nm of GaAs is grown during this period and the plasma will be prepped for firing. The growth of the buffer layer is not affected by this change in temperature or by the presence of molecular nitrogen. While the substrate is cooling, the gas lines leading to the plasma source are flushed a number of times. High purity N2 gas is introduced through a leak valve into the PBN discharge tube inside the resonator. The 37  r1732, [N]=0.19% r1736, [N]=0.33% r1735, [N]=0.45%  -100  -50  Figure 4.1:  0  50  100  150 200 Theta  250  300  350  400  Examples of GaNx As1-x XRD spectra showing the GaAs substrate  (004) peak and the epi-layer split-off peak. Pendellosung fringes, when observed, provide information about epi-layer thickness.  UHP grade N2 gas is also passed through an in-line purifier (SGT Super-Clean filters, triple filter part no. F0301). A PBN plug with a small hole was placed at the back end of the discharge tube. This constriction prevents the plasma from extending back into the stainless steel gas feed tube. The foreline nitrogen pressure in the plasma source, measured by a pirani gauge, is correlated to the measured nitrogen pressure in the growth chamber. The growth chamber pressure and pirani gauge values are set to achieve the desired nitrogen incorporation. The setting is based on BEP measurements and incorporation in previous growths. When the substrate temperature has stabilized at 450◦ C, the plasma source is fired to begin the growth of the GaNx As1-x film. To stop the growth, the plasma source is simply turned off. The ability to quickly turn on and off the plasma source is a major advantage over commercially available plasma sources.  4.2  Film composition  4.2.1 X-ray diffraction Shown in Fig. 4.1 are three XRD spectra of GaNx As1-x samples. Composition is based on fits to the spectra using dynamical scattering theory; the incorporation may also be 38  quickly estimated based on the fact that 1% nitrogen incorporation shifts the split-off peak approximately 525 arcseconds. A split-off peak on the right side (larger θ) of the (004) GaAs XRD peak indicates that the epi-layer has a smaller lattice constant than the substrate. These films are doped p-type and as a result of the relative size of the two doping elements used, Zn in the substrate and C in the epi-layers, the substrate peak is split. The relative intensities of the substrate peaks are due to the optimization algorithm used. Pendellosung fringes, most clear away from the split-off GaNx As1-x peak at far right of the figure, allow for a measurement of the film thickness. For the samples shown in Fig. 4.1, the thicknesses are: r1732, 580 nm (green); r1735, 750 nm (blue); and r1736, 580 nm (red). These fringes also indicate that the buffer/epi-layer interface and the epi-layer surface are quite smooth. GaNx As1-x films were grown with and without a bismuth surfactant. The presence of the surfactant increases the amount of incorporated nitrogen, produces smoother interfaces and improves luminescence[6, 23, 69]. No effect was observed with the use of bismuth as a surfactant on the incorporation of dopants during the growth of GaNx As1-x films.  4.2.2 Depth profiling by secondary ion mass spectroscopy In order to eliminate the contamination on the plasma discharge tube as the explanation for the differences in the observed PL spectra before and after the cleaning of the discharge tube, several samples were analysed using SIMS. The SIMS analysis was performed by Irwin Sproule at the NRC in Ottawa. Samples were depth profiled for gallium (69 Ga), arsenic (75 As+75 As), nitrogen (69 Ga+14 N), oxygen (16 O), carbon (12 C), carbon monoxide (28 CO), boron (11 B), and silicon (75 As+28 Si); the atomic and molecular masses in parentheses indicate the species that were measured for each element. A mass of 28 amu was correlated with 12 C and is therefore labelled as 28 CO and not 28 Si. We are mainly concerned with levels of contaminants in the GaNx As1-x films, however the other features in the spectra (buffer/substrate interface and relative strength of the nitrogen signal) also give some insight into the material. Shown in Fig. 4.2 are SIMS scans of GaNx As1-x epi-layers showing the effect of the plasma source on the incorporated levels of carbon and oxygen. Silicon was observed only at the substrate/buffer interface and boron was not observed in any of the samples grown using either on-resonance or off-resonance operation of the plasma source or for a clean or contaminated plasma discharge tube, and therefore these data have not been included in the plots for clarity. The substrate/epi-layer interface is clearly seen by the large increase of oxygen and carbon from the surface of the substrate, for example at 1500 nm in Fig. 4.2b. The increased signal of both C and O near the surface of the film in Fig. 4.2 is due to residual signal from surface contamination, which occurs post growth when samples are removed from the UHV environment. The SIMS spectra showed no measurable increase in carbon and oxygen after the plasma source is ignited. This would appear as an abrupt increase in the SIMS signal correlated with the start of the epi-layer. All samples, pre- and 39  Concentration (cm-3)  1021  1022  r1581, GaNxAs1-x x=0.25% Carbon Oxygen GaNAs  Nitrogen  GaAs buffer  1021  substrate/epi-layer interface  1020 1019  Concentration (cm-3)  1022  plasma ignited  1018 1017  r1664, GaNxAs1-x x=0.28% Carbon Oxygen GaNAs  substrate/epi-layer interface  1020 1019  plasma ignited  1018 1017  500  1000  1500  300  600  Depth (nm)  (a) 0.25% GaNx As1-x  Concentration (cm-3)  1021  1022  r1526, GaNxAs1-x x=0.61% Carbon Oxygen GaNAs  Nitrogen GaAs buffer  1019  1021  substrate/epi-layer interface  1020  900 Depth (nm)  1200  1500  1800  (b) 0.28% GaNx As1-x  plasma ignited  1018  Concentration (cm-3)  1022  Nitrogen  GaAs buffer  1017  r1623, GaNxAs1-x x=0.1.12% Carbon Oxygen GaNAs  Nitrogen  GaAs buffer substrate/epi-layer interface  1020 1019 1018  plasma ignited  1017 300  600 Depth (nm)  900  500  (c) 0.61% GaNx As1-x  1000 Depth (nm)  1500  2000  (d) 1.12% GaNx As1-x  Figure 4.2: SIMS analysis of a GaNx As1-x films for oxygen, carbon and nitrogen. Figures on the left are prior to plasma discharge tube cleaning and figures on the right are post plasma discharge tube cleaning. The epi-layer and buffer layers are indicated. Sample was depth profiled to slightly past the substrate/buffer interface. High levels of both carbon and oxygen near the surface are due to post growth surface contamination. SIMS analysis was performed by Irwin Sproule at the NRC in Ottawa.  post-cleaning of the plasma discharge tube, showed no measurable increase in any of the impurities (C, O, B, Si) for an ignited plasma source. It was found that at the start of the GaNx As1-x growth, as the plasma source conditions are being tuned, there is higher than average nitrogen incorporation, see Fig.4.3. The relative size of the overshoot is larger for periods when the plasma source was most strongly contaminated with As. The heights of the SIMS signals, away from the initial overshoot, are in good agreement with the nitrogen composition found using XRD.  40  r1664, [N]=0.28% r1623, [N]=1.12% r1581, [N]=0.25% r1526, [N]=0.61%  Nitrogen (cm-3)  1020  1019  1018  1017 200  400  600  800 1000 Depth (nm)  1200  1400  1600  Figure 4.3: Nitrogen SIMS signal for several GaNx As1-x samples. Samples r1526 ans r1581 are prior to plasma discharge tube cleaning and samples r1623 and r1664 are post plasma discharge tube cleaning. The initial peak is observed in all samples and is larger when the discharge plasma tube is contaminated.  4.3  Photoluminescence measurements  Several samples were grown in the on-resonance mode and several other were grown in the off-resonance mode before the discharge tube was cleaned. These samples had nitrogen concentrations of 0.19% to 0.74%. A further 18 samples were grown on-resonance and 10 off-resonance after the discharge tube was cleaned. These samples had N concentrations of 0.28% to 5.5%. PL spectra from these samples were studied. The PL spectra shown in Fig. 4.4 are representative of these groups of samples and were chosen because the samples have similar concentrations of nitrogen from before and after the plasma discharge tube was cleaned. Both room temperature and 150K PL spectra emission are shown. For the low temperature the low energy emission is stronger and the effect of plasma source operation mode, as well as the state of the discharge tube can be more easily seen. It can be seen that off-resonance operation of the plasma source showed an increased PL intensity from shallow gap states which are related to nitrogen clusters. This is to be expected, increased production and incorporation of N2 * should result in a higher probability of the formation of nitrogen clusters in the material. The thin region of increased nitrogen found in the SIMS analysis, particularly when the plasma discharge is contaminated, does not account for the low energy emission. The amount of increased nitrogen would move the peak deeper into the band gap than the observed in-gap emission peaks, furthermore the intensity would be weak due to the thinness of the layer. A marked difference in sample quality, based on increased PL intensity and increased nitrogen incorporation was found after the discharge tube was cleaned. The lower PL intensity of the figures on the right hand side is attributed to the increased nitrogen incorporation. Growth using a cleaned discharge tube results in strongest emission at the 41  before discharge tube cleaned  after discharge tube cleaned [N]=1.01% [N]=1.15%  Intensity (arb. units)  [N]=0.56% [N]=0.58%  1.1  1.2  1.3 1.4 Energy (eV)  1.5  1.2  1.3 1.4 Energy (eV)  1.5  (a) Room temperature PL emission  before discharge tube cleaned  after discharge tube cleaned [N]=1.01% [N]=1.15%  Intensity (arb. units)  [N]=0.56% [N]=0.58%  1.1  1.2  1.3 1.4 Energy (eV)  1.5  1.2  1.3 1.4 Energy (eV)  1.5  (b) PL emission at 150K  Figure 4.4:  PL spectra of GaNx As1-x samples at room temperature and 150 K.  Samples were chosen for their similar nitrogen content. The relative intensity of the spectra are correct for each temperature measurement. Spectra at 150K are approximately a factor of 5 more intense. The expected band gaps for the 150K emission are shown by the arrows (from Tisch et al.[44]). Solid lines indicate onresonance operation and dashed lines indicate off-resonance operation. A similar figure was presented in J. Vac. Sci. Technol. A 25, 850 (2007)[26]  42  1.45  , ,  PL peak (eV)  1.4  On-resonance Off-resonance secondary peak  1.35 1.3 1.25 1.2  Figure 4.5:  0  0.2  0.4  0.6  0.8 [N] (%)  1  1.2  1.4  GaNx As1-x PL peak energy at 150 K for various nitrogen  concentrations before and after cleaning the plasma discharge tube for both onand off-resonance operation of the plasma source. The expected band gap is shown by the solid line, from Tisch et al.[44]. Solid symbols indicate samples grown with the contaminated discharge tube, hollow symbols are for sample grown after the discharge was cleaned. Secondary peaks are also indicated where they could be resolved in the PL spectra. A similar figure was presented in J. Vac. Sci. Technol. A 25, 850 (2007)[26]  band gap for on-resonance operation. Fig 4.5 shows the measured PL peak energies at 150 K for a series of samples grown using on- and off-resonance operation of the plasma source either before or after the discharge tube was cleaned. Samples grown before the cleaning of the tube show stronger emission away from the expected band gap energy from in-gap cluster states. In some samples secondary peaks were observed and these peaks are in agreement with the expected band gap energy. After the discharge tube is cleaned both on- and off-resonance operation show good agreement with the expected band gap energy.  Conclusions GaNx As1-x epi-layers grown using an rf plasma source both before and after the contamination of the discharge was removed and for operation at on- and off-resonance conditions were analysed using SIMS and PL. It was found that the plasma discharge tube becomes contaminated over the period of several campaigns, however no contaminants were detected to be incorporated in the grown epi-layers by SIMS associated with the firing 43  of the plasma. Growth with a contaminated discharge tube results in increased emission observed from in-gap states, especially for off-resonance operation where there is a higher ratio of excited dimers to atomic nitrogen (see Fig. E.3). As the contamination of the plasma discharge tube is unavoidable, growth at on-resonance is preferential over off-resonance to reduce low energy emission from shallow in-gap states related nitrogen clusters. Growth of epi-layers after the plasma discharge tube was cleaned resulted in a further reduction of the emission from shallow in-gap states associated with clusters of nitrogen atoms in the material. Growth of GaNxAs1-x epi-layers with an uncontaminated plasma discharge tube in the on-resonance mode resulted in the highest measured PL intensities at the expected band gap. These results and the lower probability of cluster formation in the case of on-resonance operation of the plasma source because of the increased ratio of excited atomic and dimers suggests that atomic nitrogen is the preferred species for improved material quality. As was noted in the introduction, results on the GaNx As1-x material system can be used a guideline for expectations for GaAs1-x Bix . From the above we can expect that the incorporation of bismuth is likely to result in clusters of bismuth atoms. These clusters would result in wide PL spectra and should limit the hole mobility, as they do in the case of GaNx As1-x . These results will be discussed in the next two chapters.  44  chapter 5:  GROWTH and PROPERTIES of GaAs1−xBix At the on-set of my Ph.D. the growth conditions best suited to growing thick GaAs1-x Bix epi-layers without the build up of metallic surface droplets were not known. Initially much work was required to find growth conditions for thick (300 nm or thicker), doped GaAs1-x Bix epi-layers. The use of in-situ light scattering and RHEED were instrumental in exploring the growth process window[60, 67]. While ideal growth conditions have not yet been realized, the growth process originally used did not allow for suitably thick GaAs1-x Bix epi-layers for transport measurements. A few keys changes were made to the growth procedure that did allow for the production of thick doped epi-layers with minimal degradation of the surface from metallic surface droplet formation. The focus of my work on GaAs1-x Bix epi-layers was to explore hole transport related to bismuth incorporation. As noted the incorporation of nitrogen, which perturbs the conduction band of GaAs, greatly reduces the electron mobility[7, 14] and it was therefore reasonable to expect a similar degradation of the hole mobility in GaAs1-x Bix . Also, as the growth conditions are quite different from the conditions typically used to grow GaAs, the effect of the growth conditions on the hole mobility in GaAs were also of interest. This chapter will cover in detail the investigation of the growth procedure focusing on difficulties with the growth of GaAs1-x Bix and how they were overcome or minimised. Results from both light scattering and RHEED surface analysis will also be presented in relation to their use in the exploration of the growth process window of GaAs1-x Bix . SIMS measurements were performed on the as-grown films to investigate the effect of various growth conditions and dopant incorporation/activation. Transport measurements on GaAs and GaAs1-x Bix are discussed in the following chapter.  45  5.1  Dilute bismide GaAs1-xBix growth  Due to bismuth’s large size and tendency to surface segregate, the conditions required for the growth of GaAs1-x Bix are drastically different from those used for either GaAs or GaNx As1-x . Low growth temperatures (300-390◦C) are needed in order to decrease the desorption rate of bismuth adatoms from the growth surface. Low arsenic overpressures (As:Ga ≃ 1 − 2) are necessary to reduce the competition between the group V atoms and reduce loss of bismuth adatoms by evaporation. Recall that growth of GaAs under these conditions leads to excess arsenic incorporation and rough surfaces. Recall also that the presence of surface bismuth tended to mitigate surface roughening. There exists a narrow process window for GaAs1-x Bix at a given substrate temperature and growth rate. The flux of bismuth must be balanced between the incorporation and surface desorption and the arsenic overpressure kept low enough to allow for significant incorporation but not so low that the V:III ratio is less than unity[24]. From a practical stand point, this makes the growth of GaAs1-x Bix very difficult, especially in relation to thick films. Small changes in source fluxes from week to week require close monitoring and the high variability in the valve of the arsenic source is particularly difficult to account for. A vernier scale actuator is used to set the position of the As valve and any errors introduced by the mechanical set-up is minimised by always closing to the desired setting. However, because there exists some play in mechanical actuation of the valve the same settings of the arsenic valve result in variations of the source flux. This play positioning of the As valve has been measured to result in flux variations up to 30%. For these reasons, growth of GaAs1-x Bix films can typically only be carried out under conditions that only approximate ideal conditions. This problem can be overcome by first measuring the gallium flux and then while measuring the arsenic flux adjusting it to the desired flux ratio during the growth interrupt prior to starting the epi-layer. Growth of GaAs1-x Bix under non-ideal conditions leads to the build up of metallic surface droplets, either gallium or bismuth. If the arsenic overpressure is too low the droplets are likely gallium and the arsenic flux should be increased. If however the arsenic overpressure is too high the droplets are likely bismuth and the arsenic flux should be decreased or the substrate temperature raised. The latter case also results in low bismuth incorporation. The inability to discern the elemental composition of the droplets during the growth, does not allow for a correction to the arsenic overpressure in-situ. Furthermore, once droplets are observed on the surface of the growing film, either visibly or using light scattering, the growth conditions can not be changed without changing the composition. Therefore the growth parameters must be set as close to the ideal conditions as possible at the on-set of the epi-layer growth in order to allow for the growth of suitably thick films. Moderate amounts of surface bismuth can be removed by annealing the wafers above 500◦ C for a few minutes. This cannot be done in the case of excess surface gallium.  46  (a) Light scattering for variable As2  (b)  overpressures  GaAs1-x Bix growth  Figure 5.1:  Light  scattering  during  Diffusely scattered light from the growth surface of a GaAs and  GaAs1-x Bix epi-layer. With decreasing arsenic overpressure the surface roughness increases abruptly. The increase in surface roughness at longer time scales during GaAs1-x Bix growth is attributed to metallic surface droplet formation. These figures were originally presented in Phys. Stat. Sol. C 4 1707 (2007) [67]  5.1.1 Initial growth procedure At the beginning of this work, the growth of GaAs1-x Bix films was accomplished by first setting all the growth parameters, source cell temperatures and substrate temperature to the correct levels during the growth of the latter part of the GaAs buffer layer. Growth of the GaAs1-x Bix epi-layer was started by then simply reducing the arsenic overpressure, by partially closing the source valve, to nearly stoichiometric levels. Work by E.C. Young / hr . Using in-situ light on the growth of GaAs1-x Bix was done at a growth rate of 1.0 µm scattering the surface roughening of the GaAs1-x Bix epi-layer was monitored. When the surface roughness was determined to be approaching levels detrimental to further growth, the growth was stopped (see Fig 5.1). Using this method, the grower has no control over the final thickness of the film. Later work by Xianfeng Lu and myself found that a reduced growth rate resulted in improved control over the film growth. The growth procedure in this case was to ramp down the gallium and arsenic source temperatures to 850◦C and 350◦C respectively during the latter part of the GaAs buffer layer. The gallium flux is reduced to lower the growth rate / hr ; this growth rate is used for all samples discussed in this thesis. The ability to 0.1 µm of the gallium adatoms to diffuse along the surface is reduced due to the lowered substrate growth temperature. The lowered growth rate allows for a longer time for the adatoms to 47  diffuse along the surface. The arsenic cell temperature is lowered to improve the dynamic range of the accessible As2 fluxes available by tuning the arsenic source valve. Bismuth flux was introduced during the period when the substrate was cooling to the growth temperature, 280 − 320◦ C. A growth temperature of 300◦ C was typically used. This procedure worked satisfactorily for undoped films of thickness no more than 80 − 100 nm. For the growth of thicker films with longer growth times, the build-up of metallic surface droplets proved too difficult to avoid. The build-up of metallic surface droplets was also exacerbated by the introduction of dopant flux. In many cases when bismuth was incorporated the resultant epi-layers were highly resistive. It was thought that the dopants were either not being incorporated due to a thick bismuth wetting layer on the surface or were being incorporated but not electrically active. It should be noted that during this period, the chamber was not equipped with a functioning RHEED screen. A systematic exploration of the effect of the substrate temperature and varying the As2 flux was undertaken to find growth conditions that could reliably produce doped GaAs1-x Bix epi-layers for transport measurements. Growth temperatures near 350◦ C was found to give the best results with a As:Ga flux ratio near unity as above. Other changes to the growth procedure are described below.  5.1.2 Modified GaAs1-xBix growth procedure A new approach was required in order to produce thick doped films. RHEED was used as a means of monitoring the substrate surface during growth and several small, but significant changes were needed in the growth procedure. Of particular importance is the inclusion of a growth interrupt between the GaAs buffer and GaAs1-x Bix epi-layer to produce a smooth starting surface for the GaAs1-x Bix epi-layer. The first step in growing a GaAs1-x Bix film using the modified procedure is again to reduce both the gallium and arsenic fluxes; this is done as above and for the same reasons. Both sources are ramped down at the same rate over a ten minute period. When the sources have reached the desired levels for GaAs1-x Bix growth, the substrate is lowered to the desired growth temperature in a two step procedure. First the substrate is cooled to roughly 440◦ C. Again this has no strong adverse effects on the GaAs buffer layer as in the case of GaNx As1-x . Second, the substrate heater power is reduced to the approximate setting for the growth of the bismide film. The typical growth temperature used with the modified growth procedure was 350◦ C. While the substrate is still cooling to the GaAs1-x Bix growth temperature, the growth is interrupted for approximately 10 minutes to avoid growth of GaAs at low temperatures and low arsenic overpressures. To interrupt a growth, the gallium flux is switched off (shutter closed). During this growth interrupt the substrate temperature is set to the correct level for the growth, the arsenic overpressure is lowered and bismuth is introduced to the surface. This interruption, where the arsenic overpressure is maintained, allows the surface to anneal and therefore smooth through a reduction in step edge density (see section 3.1.1). The 48  surface reconstruction is also monitored during the interruption in growth. The growth is restarted by opening the gallium shutter. Surface roughness is believed to be a major contributer to the problem of a build up of metallic surface droplets. While the growth of a reasonably thick GaAs buffer layer over the roughened thermally desorbed oxide surface dramatically improves the surface smoothness, it was found to be necessary to include the growth interruption to allow the surface to further smooth microscopically before beginning the growth of the GaAs1-x Bix epi-layer. It is speculated that bismuth and gallium atoms present at the surface will tend to nucleate at small growth islands. Build up of these adatoms may result in the formation of droplets. Consolidation of monolayer islands into the atomic terraces[61] during the growth interruption reduces the possibility of metallic droplets concentrating in a particular region, therefore delaying the formation of surface droplets. Fig. 5.2 shows two GaAs surfaces produced by either: the original growth procedure (a); or the modified growth procedure with a two step substrate cooling and a growth interrupt (b). The AFM scans represent the would be starting surface for a GaAs1-x Bix film. Qualitatively, it can be seen that the modified procedure produces a smoother surface. This is further verified by the measured rms roughnesses, which are 0.225 nm and 0.159 nm respectively for the sample with and without a growth interrupt. Shown in Fig. 5.2c, are histograms of the heights of the surface features. Again it can be seen that the starting surface from the modified growth procedure is smoother. Additionally shown in Fig. 5.2c are histograms of the surface heights from two samples grown with the starting surfaces produced in the same way as in the shown AFM scans, but with approximately 25 nm of GaAs1-x Bix deposited on top. After deposition the surface was quenched. It can be seen that the relative smoothness of the modified growth procedure is maintained. The histogram of the surface heights shows that there is a lower number of step edges for the modified growth procedure. In other words the inclusion of the growth interrupt did smooth the surface. This is believed to be the reason for the ability to grow thicker films before metallic surface droplets form, with the growth interrupt. To ensure that the arsenic flux is set to an appropriate low level, RHEED (section 3.2.1) is used as an indicator. It is well known that the surface of GaAs shows a 1×1 surface reconstruction for As:Ga flux ratio near unity[59]. A new 2×1 surface reconstruction is observed when bismuth is present on the surface of the growing film. A transition in the surface re-construction from 1×3 to 2×1 is observed for V:III ratios in the range of 2 − 1 at a growth temperature of 350◦C (see Fig. 5.3) when bismuth is present on the surface of the growing GaAs or GaAs1-x Bix epi-layer. Note that the convention for the ordering of the directions indicated in the re-construction are 110 × [110] and that these are the fast and slow directions of gallium surface diffusion. RHEED oscillations were observed on the GaAs surfaces with incident bismuth flux. RHEED oscillations are not typically seen on GaAs surfaces at these conditions and are an indication of a smooth surface and 2-dimensional growth.  49  (a) initial growth procedure  (b) modified growth procedure  rms roughness 0.225 nm  rms roughness 0.159 nm  counts  starting surface +25nm GaAsBi  0  1  2 height, nm  3  4  (c) surface height histogram  Figure 5.2:  AFM scans of two quenched surfaces produced by the two growth  procedures described above; a) shows a surface from the original growth procedure and b) shows a surface due to the modified growth procedure. Both images are shown for the same z-scale range (1.4nm). The sample on the right, which was annealed under As overpressure for 10 minutes, appears smoother. Shown in c) is the height histogram for the surfaces from a (red) and b (blue). The increased number of higher features is an indication of more step edges when the original growth procedure is used. Also shown in c) is a histogram of the heights of the surface features for the scan shown and for samples produced in the same way with 25 nm of GaAs1-x Bix grown on top showing that the surface smoothness is retained in the case of the modified growth procedure.  50  Figure 5.3: Maps of the a) GaAs and b) GaAs(Bi) surface reconstructions from in-situ RHEED measurements as functions of substrate temperature and As2 :Ga flux ratio. This data is from the M.A.Sc. thesis of M. Masnadi-Shirazi[60]. For low As2 :Ga flux ratios the surface roughens and chevrons are observed in the RHEED pattern. This region of the phase map is labelled as ‘facets’. this figure will be presented in a paper to be published[70]  The 1×3 reconstruction is observed at the start of a growth and then the 2×1 reconstruction emerges over time and remains throughout the growth. The region in Fig. 5.3b where the 2×3 reconstruction is observed is believed to be due to a combination of the 1×3 and 2×1 surface reconstructions. Growth under conditions which show a 2×1 surface reconstruction have been found to result in higher quality films based on XRD and PL data[60]. Continued lowering of the arsenic flux results in roughening of the surface, indicated by the appearance of chevrons in the [110] (slow) direction of the RHEED patterns. Chevrons are associated with rough or facetted (3-dimensional) growth. Observation of chevrons was typically followed by the observation of metallic surface droplets.  5.2  Film composition  5.2.1 X-ray diffraction Shown in Fig. 5.4 are five example XRD spectra of GaAs1-x Bix samples. Composition is based on fits to the spectra using dynamical scattering theory. The incorporation may also 51  2.9% 0.94% 1.95% 5.5%  3.5%  -2400 -2000 -1600 -1200 -800 -400 θ (arcsec) Figure 5.4:  0  400  Examples of GaAs1-x Bix XRD spectra showing the GaAs substrate  (004) peak and the epi-layer split-off peak. Pendellosung fringes, where observed, give information about epi-layer thickness.  be quickly estimated based on the fact that 1% bismuth incorporation shifts the split-off peak approximately 300 arcseconds[24]. A split-off peak on the left side (smaller θ) of the [004] GaAs XRD peak indicates that the epi-layer has a larger lattice constant than the substrate. Pendellosung fringes, due to thin film interference, allow for a measurement of the film thickness. For the samples shown in Fig. 5.4, the thicknesses are: 0.94%, 265 nm; 1.95%, 300 nm; 2.9%, 202 nm. In the absence of Pendellosung fringes, the sample thickness is estimated from the growth rate and time. The samples with 3.5% and 5.5% bismuth are both approximately 350 nm thick. The lack of fringes is attributed to the roughness of the epi-layer surface due to metallic droplet formation. As the spectra for the higher concentration samples is broader than the others, an off-axis peak was scanned to verify that the epi-layers were not relaxed, shown in Fig. 5.5. Alignment of the peaks of the substrate and epi-layer along the Qx (in-plane) indicates that the epi-layer is not relaxed. Furthermore, Pendellosung fringes are observed in the above reciprocal map which tells us that the interfaces of the epi-layer are smooth and the composition is uniform. This was unclear from the rocking curve shown in Fig. 5.4. Unrelaxed GaAs1-x Bix epi-layers of up to 7% have been grown to thicknesses of 800 nm.  52  0.886  GaAs 115  0.884  0.882 Qy (1/A)  0.880  0.878  GaAsBi, 3.5% 0.876 0.248 0.249 Qx (1/A)  Figure 5.5:  Reciprocal space map of [115] peak of the 3.5% GaAs1-x Bix sample  shown in Fig. 5.4. Alignment of the GaAs and GaAs1-x Bix peak along Qx indicates that the films is not relaxed. Observation of Pendellosung fringes shows that the interfaces are smooth and the compoisition is uniform.  5.2.2 Photoluminescence Photoluminescence spectra for several GaAs1-x Bix films are shown in Fig. 5.6. These samples are all nominally 350 nm thick and have similar intensity spectra. The small differences are consistent with variations in film thickness. These measurements are consistent with other measurements of PL spectra on GaAs1-x Bix thin films grown with GaAs capping layers[27]. Increased width of the GaAs1-x Bix PL spectra is attributed to a distribution of localised, shallow states in the band gap of the material related to clusters of bismuth atoms. Shallow in-gap states have also been observed in GaNx As1-x [6]. The effect on the spectra width is greater here for samples with higher concentrations of bismuth. The PL spectra in combination with the XRD data shows that the GaAs1-x Bix epi-layers are of good crystalline and optical quality. The XRD spectra indicate that the films are uniform in composition, and in the cases where Pendellosung fringes are observed that the interfaces are smooth. Strong luminescence, compared to other GaAs1-x Bix films grown on site[27], shows that the films are defect free.  53  5.5% 4.38% 3.5%  1.9%  Normalised Intensity, arb. units  1.68%  dashed spectra x30  0.9  1.0  1.1 1.2 Energy, eV  1.3  1.4  Figure 5.6: Examples of PL spectra from GaAs1-x Bix bulk epi-layers. The spectra have been normalized to unity.  5.2.3 Depth profiling by secondary ion mass spectroscopy In order to better understand the issues encountered while attempting to grow sufficiently thick and (highly) doped epi-layers, several samples were analysed using SIMS (section 3.3.4). Of particular interest was the incorporation of dopant elements or lack thereof when bismuth is incorporated. Several p-type, n-type and undoped GaAs and GaAs1-x Bix samples were analysed to quantify the incorporated amounts of arsenic (150 As2 ), gallium (69 Ga), bismuth (209 Bi+75 As), carbon (12 C), silicon (75 As+28 Si) and bromine (79 Br). The values in parentheses give the measured signals used for each element. Samples were also chosen based on their RHEED reconstruction during the growth in hopes of correlating the incorporation of the dopants with the reconstruction transition. Fig. 5.7 shows SIMS profiles for the GaAs1-x Bix samples analysed. The signal strengths are normalized to the arsenic (150 As2 ) signal. The SIMS signal does not indicate relative concentrations of the various elements accurately because of the element-dependent sensitivity of SIMS. However, similar elements may be compared between samples. The dominant features in the SIMS spectra are the substrate/buffer interface and the onset of the epi-layer. We are most interested in the effect of the surface reconstruction transition on the incorporated amounts. The transition can be correlated with the start of the epi-layer 54  r2156, 4.2%  r2171, 1.71%  0  300  600 900 Depth (nm)  As2 AsSi C AsBi  Normalised SIMS counts  Normalised SIMS counts  As2 AsSi C AsBi  1200  1500  0  (a) n-type 4.2% GaAs1-x Bix  300  900 Depth (nm)  1200  1500  1800  (c) n-type 2.28% GaAs1-x Bix  Figure 5.7:  1500  1800  As2 AsSi C AsBi  Normalised SIMS counts  Normalised SIMS counts  600  1200  r2127, 0.5% Bi  As2 AsSi C AsBi  300  900 Depth (nm)  (b) p-type 1.71% GaAs1-x Bix  r2186, 2.28% Bi  0  600  0  300  600 900 Depth (nm)  1200  1500  (d) p-type 0.5% GaAs1-x Bix  Representative SIMS profiles from for GaAs1-x Bix samples. SIMS  counts have been normalized to the arsenic (150 As2 ) signal. The relative strengths of the various elemental signals depends on the concentration and elemental sensitivity. / s . SIMS analysis was performed by Depth is based on an sputtering rate of 0.7 nm Simona Moisa at the NRC in Ottawa.  from the knowledge of the growth. The substrate/buffer interface can be clearly seen in the sharp increase in the concentration of carbon and silicon. Depths of the features observed in the SIMS spectra are based on / s . Depths measured in the above SIMS spectra are in good sputtering rates of 0.7 nm agreement with the expected epi-layer and buffer layer thicknesses from XRD and growth time. Similar to the case of the SIMS analysis on GaNx As1-x samples surface contamination is observed in the form of slowly declining levels of carbon and silicon with depth. Therefore only abrupt increases in carbon (or silicon) are attributed to the start of deliberate dopant incorporation. Sharp increases in the silicon signal indicate n-type doping, as in the case of Fig. 5.7a, and sharp changes in the carbon signal indicate p-type doping, as in the case of  55  r2156, 4.20%  r2171, 1.75%  r2186, 2.28%  r2153, 4.38% RHEED transition From SIMS data r2182, 1.83% 0  30  60 90 thickness (nm)  120  150  Figure 5.8: Epi-layer thicknesses for the 1×3 to 2×1 reconstruction transition from RHEED observation during growth and thickness difference from start of epi-layer and uniform bismuth content as measured by SIMS; sample r2182 was not doped and therefore no indicator of epi-layer onset existed. The error bars on the RHEED data represent the fact that the RHEED reconstruction may have occurred previous to being first observed.  Fig. 5.7b. The abrupt changes of either carbon or silicon are also used as an indication of the start of the epi-layer growth in doped GaAs1-x Bix samples along with knowledge of the sample growth; typically the dopant source is opened at the same time as the bismuth and gallium. What is first apparent from the above SIMS spectra is a delay of incorporation of bismuth from the start of the epi-layer growth. The transition of the surface reconstruction occurs at approximately 300 nm, and can be clearly seen by its effect on the carbon signal, which is discussed below. The delay is attributed to the amount of epi-layer growth before the transition from the 1×3 reconstruction to the 2×1 reconstruction while the surface bismuth builds up. This is particularly clear for sample r2171 (Fig. 5.7b), where both the bismuth and carbon signals rise sharply separated by approximately 45 nm. Little or no bismuth SIMS signal is observed for growth under the 1×3 reconstruction, both before the reconstruction transition and throughout samples where no 2×1 is observed. Fig. 5.8 shows a comparison of the amount of epi-layer deposited before the RHEED reconstruction is observed and the amount deposited before bismuth content is measured to be uniform by SIMS. A strong correlation is observed between the two thicknesses. Bismuth is only incorporated after the surface reconstructs to 2×1 and the longer the time delay to the reconstruction transition the longer the incorporation is delayed. These delays are believed to be related to the amount of required surface bismuth. As shown in Fig. 5.8, nearly twice as much surface bismuth is required for incorporation as for the transition in the surface reconstruction. As a metal, bismuth on the surface will more effectively reflect infrared radiation and this will reduce the surface’s emissivity. The lowered surface emissivity will result in an increase in substrate 56  Normalised Bi SIMS counts 0  1  2  3  4  5  [Bi] (%)  Figure 5.9: Comparison of bismuth content from XRD and SIMS measurements. The bismuth SIMS signal is averaged over the uniformly alloyed regions.  An  approximately linear relationship is observed. The dashed trend line is added as a guide for the eye.  temperature. Such a step in temperature has been noted for many growths in relation to the surface reconstruction transition. Prior to the surface reconstruction transition the surface may be considered As-rich, whereas the 2×1 surface reconstruction transition indicates a Bi-rich surface. The measured bismuth content from XRD compared to the bismuth SIMS signal is shown in Fig. 5.9. An approximately linear relationship is observed, as expected. The bismuth content as measured by SIMS is averaged over the uniformly alloyed regions. The increased background of the 209 Bi+75 As SIMS signal used for measuring the bismuth content in the films is attributed to sputtering of surface bismuth droplets. For the n-type doped samples the build up to uniform bismuth content is much slower than in the case of the p-type samples, as seen in Fig. 5.7. This is believed to be due to the additional heat load from the Si cell (operated at 1150◦ C) raising the substrate surface temperature which will cause some of the surface bismuth to desorb. There is also an effect on the incorporated amount of carbon as the surface reconstruction transition occurs. The incorporated amount is observed to decrease slightly during the transition from 1×3 to 2×1 at a depth of approximately 300 nm (see Fig 5.7a). This depth is consistent with the measured epi-layer thickness based on Pendellosung fringes observed in XRD. No such effect was observed with silicon n-type doping. The measured carrier concentrations are shown in table 5.1. The carrier concentrations measured are higher than expected based on previous doping calibrations on GaAs samples 57  carrier concentration, cm−3 sample  RHEED recons.  expected  measured  GaAs1-x Bix :C  SIMS Signal  2×1  3x1017  1x1018  0.08  GaAs1-x Bix :C  2×1  3x1017  3x1018  0.06  GaAs1-x Bix :C  1×3  1x1017  1x1015  0.05  GaAs1-x Bix :Si  2×1  4x1017  4x1018  0.038  GaAs1-x Bix :Si  2×1  4x1017  1x1018  0.038  GaAs1-x Bix :Si  1×3  2x1017  1x1018  0.015  Table 5.1: Expected and measured doping concentration in GaAs1-x Bix samples compared to carbon (12 C) and silicon (75 As+28 Si) SIMS counts. Expected carrier concentrations are based on doping calibrations and are indicative of the source flux used. Measured values of carrier concentration are from Hall effect measurements. The SIMS signal has been normalized to the As2 signal after the background counts have been subtracted. The RHEED reconstruction observed during the growth of the doped GaAs1-x Bix epi-layer is also given. Measured values of the carrier concentration are lower where samples have been deposited under conditions where a 1×3 surface reconstruction is observed. In the case of p-type doping, the incorporated  carbon is not electrically active for 1×3 growth conditions.  (grown at Tgrowth = 560◦C). Doping concentrations are not significantly effected by reduced growth temperatures. In the case of GaAs1-x Bix growth under the 1×3 reconstruction, carbon was incorporated but found not to be electronically active. Many highly resistive GaAs1-x Bix films were grown while exploring the growth conditions for thick doped epilayers, of which the above sample is representative. The above result on the resistive GaAs1-x Bix :C film indicates that carbon incorporation under 1×3 surface reconstruction conditions is electronically inactive resulting in highly resistive films. It is unclear if this is caused by the dopants incorporating onto both As and Ga sites, leading to compensated films. Silicon n-type doping is not observed to be similarly affected.  Conclusions Modifications to the growth procedure initially used to produce GaAs1-x Bix epi-layers were necessary to allow for the growth of thick ( 300 nm), doped films . Two key changes were made: 1) the introduction of the growth interrupt prior to the on-set of the GaAs1-x Bix epilayer growth to avoid growth of the GaAs buffer at low temperatures and low As:Ga flux ratio; 58  2) a reduction of the growth rate to compensate for low surface diffusion of the Ga adatoms at low growth temperatures. Work continues on identifying the ‘ideal’ growth conditions for GaAs1-x Bix . Using the modified changes bulk epi-layers were grown to thicknesses greater than 250 nm, suitable to explore the electrical transport properties of GaAs1-x Bix (discussed in the next chapter). RHEED was found to be a good indicator of the growth conditions in-situ through observation of the surface reconstruction. GaAs1-x Bix films grown under conditions where a 2×1 reconstruction was observed resulted in incorporated bismuth and conductive films. Growth of GaAs1-x Bix under conditions where a 1×3 surface reconstruction is observed results in electrically in-active doping with carbon. Silicon doping was found to be unaffected. A delay of bismuth incorporation from the start of the GaAs1-x Bix epi-layer growth was observed to be correlated with the observation of the surface reconstruction transition. This is believed to be related to the time required for sufficient surface bismuth to build up, where it is estimated that approximately twice as much surface bismuth is necessary for incorporation as for a 2×1 surface reconstruction.  59  chapter 6:  ELECTRICAL TRANSPORT in GaAs and GaAs1−xBix Using the modified growth procedure described in the previous chapter I grew GaAs1-x Bix thin films to explore the effect of bismuth incorporation on the electronic transport of the material. As bismuth largely affects the valence band, the mobility of holes was of primary interest. Measurements of the electron mobility in GaAs1-x Bix will also be discussed[15, 28]. Also, as GaAs1-x Bix is grown under conditions quite different from those conventionally used for GaAs growth, the GaAs1-x Bix samples were compared to GaAs samples grown under a variety of conditions. The following chapter covers the fabrication of the Hall and resistivity samples and discusses the effects observed on the hole mobility due to the growth conditions and bismuth content. In order to better understand the effects of bismuth incorporation on the hole mobility, the temperature dependence was measured for several samples. Modelling the temperature dependence allows for insight into the nature of the scatterers. This is discussed in detail below.  6.1  Fabrication of resistivity and van der Pauw samples  6.1.1 Carrier concentration calibration GaAs doping concentration calibration samples were grown on SI GaAs substrates. It is necessary to grown films and measure their resistivity as doping calibrations as the flux from dopant sources is too low to measure reliably using the ion gauge. Calibration samples were grown under a variety of growth conditions, conditions similar to those being used to grow the doped epi-layers, at the start of each new campaign and periodically throughout a campaign. Cr/Au contacts were deposited on the surface of 7 × 7mm squares by electron 60  (a) finger mask  (b) van der Pauw mask  Figure 6.1: Shadow masks used for deposition of contacts on resistivity and van der Pauw samples.  beam evaporation. These contacts are sufficiently conductive and non-rectifying for the 4point measurement of the resistivity. The shadow mask used for these samples is shown in Fig. 6.1a. The sample size was determined by the contact deposition sample holder available.  6.1.2 GaAs and GaAs1-xBix van der Pauw samples GaAs1-x Bix samples for electronic transport measurements were grown on SI GaAs substrates as described above using the modified growth procedure. Again it was necessary to post-growth anneal the samples to remove excess surface bismuth. 7 × 7mm squares were used for all transport measurements. Contacts for the Hall measurement samples were placed at the corners of the square. The sample size was determined by the contact deposition sample holder available. While this sample geometry is not preferred over the cloverleaf, as errors in the sheet resistance measurements go as the ratio of the contact size to the perimeter[68], the fabrication of the samples using the square geometry is far simpler. Estimation of the error in sheet resistance based on sample and contact size is less than 7%. Using a shadow-mask technique small ohmic contacts were deposited by electron beam evaporation. Ti/Pt/Au contacts were used for p-type[71] and Ni/AuGe/Au for n-type ohmic contacts[72]. The p-type ohmic contacts were annealed for 30 s at 450◦ C and n-type ohmic contacts for 120 s at 400◦C to improve their conductivity. Hall samples were mounted on thin copper plates with vacuum grease. Copper was chosen for its high thermal conductivity and these plates were covered with an insulating layer of  61  Ohmic contacts doped GaAs1-xBix GaAs buffer  SI GaAs substrate  Figure 6.2: Hall sample schematic with the device structure indicated by shading: the darkest region shows the undoped substrate; the buffer is show in a lighter shade; and the doped epi-layer is shown in the lightest shade. Contacts are shown in gold. Layer thicknesses and contact size are not to scale.  cigarette paper and GE varnish‡ . Small brass wires were then silver epoxied to the annealed ohmic contacts. Care was taken to ensure that the wiring of each sample was configured such that the contacts could be numbered in a clockwise manner around the sample. This was to make each sample compatible with the set-up of the measurement apparatus. Sheet resistivity of the samples were found using the van der Pauw method outlined in section 3.3.5 and carrier concentrations found from Hall effect measurements, unless otherwise stated. The mobility is then calculated from eqn. 3.6.  6.2  Hole mobility measurements  6.2.1 Effect of growth conditions and film composition i) Effect of bismuth surfactant on carbon incorporation There was speculation that the reduction in nitrogen clusters observed in GaNx As1-x samples when bismuth was used as a surfactant, would also allow for higher doping of GaAs with carbon[73]. Carbon doping in GaAs is limited to approximately 1x1020 cm−3 active ‡  this is a well known technique used in cryogenics  62  p ±5%(cm−3 )  Sample  CBr4 set pt. (Torr)  R (Ω)  ρ (Ωcm)  r1791  2.0  60.8  0.0304  r1792 w/Bi  2.0  62.7  0.0314  1.3x1018  r1832  4.0  16.6  0.00831  5.0x1018  r1836 w/Bi  4.0  18.2  0.00911  4.6x1018  Table 6.1:  1.4x1018  Carrier concentration for p-GaAs samples grown with and without  bismuth used as a surfactant using standard GaAs growth conditions of Tsubstrate ≃ 580◦ C and As:Ga flux ratio of 5 or higher.  The bulk carrier concentration is  calculated using the expected hole mobility of GaAs at these doping levels[1] and using the effective thickness of the doped layer.  carriers due to the formation of a C-C split interstitial[74, 75]. Operation of the CBr4 source is limited by a maximum foreline pressure that did not allow for doping concentrations above 5x1018 cm−3 . However, some insights were drawn from comparisons of samples grown with and without a bismuth surfactant at these high fluxes. Two pairs of samples were grown under the same set of conditions (standard GaAs growth conditions of Tsubstrate ≃ 580◦C and As:Ga flux ratio of 5 or higher), both with and without bismuth incident flux, at two relatively high dopant source fluxes (set by the source foreline set point) to investigate this effect. A recent GaAs doping concentration calibration sample where the foreline set point was 100 mTorr resulted in a film with a resistivity of 0.069 Ωcm. This is an estimated doping concentration of 6.1x1017 cm−3 . Each pair of samples was grown on the same day to reduce any changes within the system on a day-to-day basis. These pairs of samples were also grown sufficiently close in time that no significant decrease in dopant source flux was expected. Results on the doping of these samples are given in table 6.1. The values found for the carrier concentration are based on the expected hole mobility in GaAs at these carrier concentrations[1] (these are the only estimated carrier concentrations, all other presented values of carrier concentration are from Hall effect measurements). No significant effect of the use of bismuth as a surfactant can be seen in these pairs of samples. Based on the previous calibration samples, the doping for these samples was expected to be a factor of 2 higher. It is assumed that there is a loss of activation of the carriers, possibly from the aforementioned split interstitials, for this increased dopant incorporation. As no difference in carrier activation is observed with bismuth surfactant use, it is concluded that surfactant assisted growth does not increase the doping limits of carbon in GaAs.  63  µh (cm2V-1s-1)  102 conventional growth Low T Low T, low As:Ga Low T, low As:Ga w/Bi From Adachi, 2005 1017  1018 Hole concentration (cm ) -3  Figure 6.3: Hole mobility of GaAs for various growth conditions:  , conventional  GaAs growth conditions of Tgrowth ≃ 580◦ C and As:Ga≃ 8; , î, low temperature GaAs with Tgrowth ≃ 290, 320◦ C and As:Ga≃ 2, 8; and  350◦ C),  low temperature (≃  low As:Ga GaAs grown with a bismuth flux. The solid line shows the  expected hole mobility for GaAs from eqn. 6.1.  ii) GaAs hole mobility for varied growth conditions The effect of the growth conditions on the mobility of holes was explored in GaAs, Fig. 6.3. GaAs samples grown under conventional conditions of Tgrowth ≃ 580◦ C and As:Ga flux ratio ≃ 8 show mobilities very close to the expected value[1]: µmax − µmin  µ(p) = µmin +  1+  p/  α  (6.1)  pref  which is an empirical model of the trend of reduced mobility with increasing doping concentration. The parameters for GaAs are[1]: µmin = 20, µmax = 491.5, pref = 1.48x1017 cm−3 and α = 0.38. The effect of the doping concentration on the room temperature mobility is often incorrectly attributed to an increase in ionized impurity scatterers. The effect of the ionized impurity scattering does increase with increasing doping concentration. The contribution is directly proportional to the doping concentration. However, as can be seen in section 2.3 in eqn’s 2.36 and 2.37, the contribution to the room temperature hole mobility from ionized impurity scattering is exceedingly small and the room temperature mobility is dominated by phonon scattering. The form of the above equation, does roughly go as the doping concentration to the /1 3 power, i.e. the average physical separation of the dopant impurity atoms. A local lattice perturbation could act as a scatterer for the charge carriers and give a functional form similar to the above. Growth of GaAs under low temperature conditions results in a small reduction in the hole mobility, whereas growth at both low temperature and low As:Ga shows a larger reduction 64  Low T, low As:Ga GaAs w/Bi GaAsBi  µh (cm2V-1s-1)  103  µe (cm2V-1s-1)  102  GaNAs, µe 101  102  0  Figure 6.4:  1  2  3 x (%)  4  5  6  Hole mobility for GaAs1-x Bix samples, as well as GaAs (at x= 0)  samples grown under similar conditions where bismuth is not incorporated. The hole mobility is seen to decrease with increasing bismuth content. Also shown for comparison is the electron mobility for GaNx As1-x , data from [14].  in hole mobility. Growth of GaAs under either low temperature is known to yield epilayers with increased defects, so a reduction in mobility is expected[55]. The hole mobility is observed to return to expected values when bismuth is used as a surfactant during the growth. No bismuth incorporation was measured in these samples by XRD. Previous work on growth using bismuth as a surfactant showed that films showed improved luminescence, decreased interface roughness and in the case of GaNx As1-x lower emission from shallow ingap states[6, 23, 69]. A further material quality improving effect on the mobility is not a surprising result. Based on this and the fact there is a significant amount of surface bismuth present during the growth of GaAs1-x Bix films, it is concluded that any reductions in hole mobility observed in GaAs1-x Bix epi-layers will be solely due to the incorporation of bismuth and not to the non-standard growth conditions.  iii) GaAs1-x Bix hole mobility as a function of bismuth content For increasing bismuth concentration the hole mobility is observed to decrease as shown in Fig. 6.4. This is to be expected as the incorporation of bismuth results in a resonant state within the valence band of GaAs. In the analogous case of GaNx As1-x where the incorporated nitrogen is known to primarily effect the conduction band, the electron mobility is seen to decrease for increased incorporation[7, 14]. The relative decrease in mobility is far less dramatic in the case of the hole mobility in GaAs1-x Bix than observed in the electron mobility in GaNx As1-x ; for 1% nitrogen in GaNxAs1-x the electron mobility is reduced 20×. In contrast the hole mobility is reduced by a factor of 2 or less for 1% bismuth in GaAs1-x Bix , based 65  doping ±2% (cm−3 )  [Bi] (%)  9.2x1017  0  2 µ300K ±5% ( cm/ V s ) 185  r2208  7.0x1017  0  191  r2201  5.1x1017  0.94  86  r2205  4.5x1017  1.48  73  r2178  1.3x1017  1.68  82  r2171  8.2x1017  1.75  118  r2164  4.7x1017  1.95  101  r1913  2.3x1018  3.4  66  r2157  1.1x1018  3.5  60  r2153  2.3x1017  4.4  11  r2148  4.5x1017  5.5  9  sample r2193  Table 6.2:  Summary of p-type GaAs1-x Bix samples and their measured room  temperature hole mobility.  on the results shown here (see Fig. 6.4). This may be due to the differences in the effective masses of the charge carriers, electrons and holes, and the relative position of the resonant states in GaNx As1-x and GaAs1-x Bix compared to their respective band edges. Scatter in the data results from sample to sample variability, and possibly the growth date in relation to the start of the growth campaign. Samples grown near the start of a growth of a campaign are expected to be poorer due to chamber contaminants that will dissipate with time and chamber usage. Small variations in the doping will also result in variability in the data. However these fluctuations in mobility due to doping differences are much smaller than the changes measured here for increasing bismuth content. It is also known from the SIMS measurements that there exists a layer of doped GaAs grown after the growth interrupt but prior to the surface reconstruction transition, see fig 5.8. This thin layer could act as a parallel conductor in the measurement and as a result values shown Fig. 6.4 and table 6.2 would overestimate the hole mobility in GaAs1-x Bix . Also recall that carbon incorporated under 1×3 surface reconstruction growth conditions was not electrically active. A calculation of the magnitude of the effect of the thin conducting GaAs buffer layer, treating the two epilayers as parallel conductors and using the worst possible case (thinnest, 200 nm, GaAs1-x Bix epi-layer and thickest, 25 nm, GaAs conducting layer), limits the effect to less than a 10% error in the measured value.  66  carrier concentration cm-3  1019  1018  1017  r2208, GaAs r2200, GaAs r2171, [Bi]=1.75% r2164, [Bi]=1.91% r2157, [Bi]=3.5%  1016  50  100  150 200 Temperature, K  250  300  Figure 6.5: Carrier concentration for GaAs and GaAs1-x Bix films as a function of temperature from Hall measurements.  6.2.2 Temperature dependence of hole mobility in GaAs and GaAs1-xBix iv) Carrier concentration as a function of temperature All doping concentrations quoted below are from room temperature measurements of the sheet carrier concentration where the effective thicknesses of the layer is taken into account. The effective thickness is based on the assumption that the Fermi energy is pinned at mid band gap at the surface of the epi-layer and the surface depletion region needs to be subtracted from the measured epi-layer thickness. Over the temperature range measured the carrier concentrations of the samples did not vary much as shown in Fig. 6.5. The increase in the measured hole concentration at low temperatures is an indication that the hole Fermi energy is approaching the band edge and lies within the bismuth cluster states. This would lead to hopping conduction between the bismuth states and a related change in measured Hall voltage, which is observed here as an apparent increase in carrier concentration.  i) Contributions to hole scattering in GaAs This section is a continuation of the discussion of the temperature dependence of the mobility of charge carriers in section 2.3. Recall that the measured mobility as a function of temperature will have contributions from several scatterers: 1 1 1 1 1 1 + + + + +··· = µ µph µI µα µβ µγ  (6.2)  Each component of the mobility is assumed to act simultaneously and independently, where µph is the contribution to the mobility from acoustic phonon scattering and µI is the 67  contribution due to scattering from ionized impurities. Their temperature dependencies are: µph = Cph T µ I = CI  -3/ 2  3/ T 2  (6.3) (6.4)  As we are dealing with hole transport and the valence band, the prefactors of the above take into account the more complex nature of the valence band (i.e. influence of light holes and interband scattering)[1]. This is accomplished by assuming that the LH and HH bands are not coupled. In this case, the mobility of each band can be simply added and the concentration of carriers in each band depends on the ratio of the effective masses[31]. Heavy holes will dominate in this case as they are approximately 10× heavier than light holes. Three other possible scattering mechanisms are necessary to complete the discussion of the hole mobility of GaAs1-x Bix : alloy deformation potential, neutral impurities and scattering from the bismuth related resonant state. The latter is based on theoretical work done on GaNx As1-x [13, 76]. Alloy scattering of charge carriers results from the potential fluctuations due to local composition variations (disorder). This type of scattering can play a significant role for ternary and quarternary alloys. For a ternary alloy the mobility contribution from alloy scattering is[2]: √ 2πe 4 Nal µal = (6.5) √ 5 3(m∗ ) /2 kB T x(1 − x)(∆U)2  where Nal is the density of alloy impurities, x and (1 − x) are the mole fractions of the constituent end members and ∆U is the alloy scattering potential. The scattering potential is used as an adjustable fitting parameter. Alloy scattering will only play a significant role when the mole fractions are large. Often a further term, S(α), is included in the denominator to account for the degree of randomness in the alloy. As an example, the electron mobility in Inx Ga1-x As with x ≃ 0.5 is strongly limited by alloy scattering[1] and S(α) is near unity. However, Inx Ga1-x As and some other III-V semiconductor alloys can be grown in such a way as to have an ordering of the group III atoms[77]. This ordering has been observed in Inx Ga1-xAs to suppresses the alloy scattering and increase the carrier mobility[78]. In other words, the term S(α) approaches 0 and the material is no longer limited by alloy scattering. GaAs1-x Bix has also been observed to exhibit this behaviour[79] and this would be easily observed as anisotropic resistivity if ordering were significant in the samples measured here. Measured resistivities along the two directions were typically very similar. At very low temperatures the dopant impurities will no longer be ionized or bismuth may act as a neutral impurity, as well as any other unintentionally incorporated neutral impurities. Scattering from a neutral impurity in a semiconductor is treated in the same way as elastic scattering of electrons from neutral hydrogen, where the Bohr radius and binding energy are modified by the effective mass and the dielectric constant of the semiconductor. The 68  mobility found using this approach is temperature independent[80]: q m/mo 20αB ǫ NN I ∗  µ N I = CN I =  (6.6)  where αB is the Bohr radius, ǫ is the dielectric constant and NN I is the concentration of neutral impurity scatterers. At the temperatures measured in this thesis all dopant impurities can be considered to be ionized. However, there may exist other neutral impurities related to bismuth incorporation. Measurements of the electron mobility in GaAs1-x Bix with concentrations as high as 2.5% showed a temperature independent mobility attributed to strong neutral impurity scattering and modelled using the above equation along with terms for phonon and ionized impurity scattering[28]. The neutral impurity electron scatterers were found to have a density of 2.5x1017 cm−3§ . The neutral impurities were attributed to clusters of bismuth atoms as the density was much lower than the density of bismuth atoms (1x1020 cm−3 ). There is also the possibility that there are other unintentionally incorporated neutral defects. Modelling the hole mobility using the above neglects scattering from the resonant state in the valence band known to exist for bismuth alloying with GaAs, as discussed in section 2.2. The contribution due to scattering from the resonant state could be much larger than the effect of any neutral impurities on the hole mobility in GaAs1-x Bix . Therefore, to more correctly model the hole mobility in GaAs1-x Bix it is necessary to consider the effect of the state(s) within the valence band due to incorporation of bismuth. The general approach to scattering from impurities, using the Born approximation, underestimates the scattering cross-section in GaNx As1-x by a factor of more than 100[13]. The approach does not work in the case of nitrogen in GaNx As1-x because nitrogen is isovalent with the arsenic it replaces and its scattering rate is therefore dominated by its central potential which is ignored in the Born approximation. This too will be the case for bismuth. In cases where resonant states exist near band edges, the scattering of a charge carrier with an energy near the resonance will be stronger as the cross-section is proportional to the deBroglie wavelength of the charge carrier[81]. A complete treatment of the wave function at the band edge in the presence of a nitrogen (or bismuth) atom is required. This results in the following scattering cross-section for an isolated nitrogen impurity in GaNxAs1-x [13]: σ=  π 4  m∗ 2π 2  2  d Ec dx  2  (ao )6  (6.7)  where ddxEc is the change in the conduction band energy with nitrogen incorporation,x, and ao is the lattice parameter of GaAs. The above is valid only in the dilute limit where the mean free path of charge carriers is smaller than the average distance between nitrogen atoms. The §  There is a miscalculation in the article and the density of neutral impurities is given as ≃ 3x1016 cm−3 .  This is due to h being used instead of  in the calculation.  69  mobility that results from this is, using kinetic theory, (m.f.p = (N σ)−1 = µN =  √  3π 2 m∗ kB T e  = CN T  m∗ 2π 2  2  d Ec dx  −1  2  -1/ 2  vth m∗ µBi )[13]: e  (ao )3 x  (6.8) (6.9)  To apply this equation to GaAs1-x Bix samples the density of states effective mass for holes will be used and the shift in conduction band edge with increasing nitrogen content will be replaced with the change in band gap energy with increasing bismuth content, d dxEg = 8.4 eV/%. This assumes the shift in the band gap in entirely due to changes in the valence band edge when bismuth is incorporated. The result for the scattering cross-section for an isolated bismuth impurity is 2.2x10−14 cm2 using eqn. 6.7. We can use eqn. 6.9 to estimate the mobility due to a density of scatterers with this cross-section. For 1% incorporation of 2 bismuth in GaAs, this results in a room temperature mobility of 78 cm . This value is quite V s close to several of the measured values for room temperature GaAs1-x Bix samples. The above equation also assumes only one resonant state within the band, and will only approximate the mobility if more than one resonant state is present in the band. Away from the dilute limit, more complicated complexes of the alloying atoms (i.e. clusters) need to be considered[76]. This is the case for nitrogen where there are a number of states near the conduction band edge due to clusters of nitrogen atoms in the material. For example, there is a state at 1.51 eV above the VBM due to N-N pairs[45]. This affects the temperature 1/ dependence of the related mobility, weakening the T 2 dependence above[76]. This will be treated as temperature independent based on the GaAs1-x Bix mobility measurement results. For large incorporated amounts the mobility was found to be constant over the temperature range 75-300 K. The resultant model used to fit the measured temperature dependence of the mobility includes terms for phonon scattering (T −1.5 ), ionized impurity scattering (T 1.5 ) and a temperature independent term CBi : 1 1 1 1 = + + 1.5 −1.5 µ CI T Cph T CBi  (6.10)  where the subscripts I and ph indicate ionized impurities and phonons, respectively. The GaAs hole mobility will be fit in the same way using only the terms for phonon and ionized impurity scattering.  ii) GaAs hole mobility as a function of temperature Shown in Fig. 6.6 are the temperature dependences for three GaAs samples from Fig. 6.3. As stated above, the temperature dependence of the mobility gives insight into the nature 70  103  104 T-3  T1.5 T-1  103  r2199, GaAs  102  30  Low T, low As:Ga, w/Bi Low T, low As:Ga  µ, cm2/V s  µ, cm2/V s  T1.5  T-1.5  102  100 Temperature, K  300  (a) Tgrowth =580◦C As:Ga≃8  50  100  150 200 Temperature, K  250  300  (b) Tgrowth =350◦C As:Ga≃2.0 with and without bismuth flux  Figure 6.6:  Temperature dependence of the hole mobility of GaAs: a) grown  under conventional conditions and b) grown under low temperature and low As:Ga conditions with (red curves) and without (black curves) a bismuth surfactant. The 3/ 2  temperature dependences are shown on each of the plots; T impurity scattering and  -3/ T 2  is related to ionized  is related to phonon scattering. The best fit parameters  for the low temperature, low As:Ga flux ratio sample in b) are given below in table 6.4.  of the scatterers present in a material, though not all temperature dependences have been associated with specific scattering mechanisms. For the GaAs samples grown under conventional conditions: a T−3 trend is observed at T= 175 − 300 K; an approximate 3/ T−1 trend is observed for 75 <T< 175; and at low temperatures T 2 is observed. The 3/ T 2 is related to ionized impurity scattering. Neither the T−3 nor T−1 are related to any known scattering mechanism, though similar temperature dependencies in GaAs are reported elsewhere[2]. For the GaAs samples grown at low temperature and low As:Ga -3/ 3/ (GaAs1-x Bix -like growth conditions) only T 2 and T 2 are observed, where the latter is related to acoustic phonon scattering. For the sample where bismuth has been used as a surfactant, the hole mobility is larger over the entire temperature range. As surfactant use has been known to improve other material properties, this result is not unexpected[23, 69]. These low temperature and low As:Ga GaAs samples will be used as a reference for all GaAs1-x Bix samples as its growth conditions more closely resemble those of GaAs1-x Bix . In other words, the data in Fig. 6.6b should be regarded as a [Bi]= 0% measurements. The fitting parameters for these GaAs1-x Bix x = 0 samples will be discussed below along with other GaAs1-x Bix (x = 0) samples.  71  doping ±2% (cm−3 )  µ300K ± 5%  Cph ±5%  CI ±10%  r2208: Low T, low As:Ga w/Bi GaAs  7.1x1017  190  1.0x106  0.838  r2200: Low T, low As:Ga GaAs  7.8x1017  133  8.5x105  0.158  sample and details  8.6x1016  r2199: conventional GaAs  300  -  Table 6.3: Best fit parameters for the temperature dependence of the hole mobility in GaAs samples grown under GaAs1-x Bix -like conditions. Units for the coefficients can be inferred from eqn. 6.10.  sample  doping±2% (cm−3 )  r2201: [Bi]=0.94%  5.1x1017  r2171: [Bi]=1.75%  Cph ±5%  CI ±10%  8.2x1017  7.8x105  1.21  r2164: [Bi]=1.91%  5.4x1017  1.0x106  0.281  r2157: [Bi]=3.5%  1.2x1018  6.3x105  0.337  r2153: [Bi]=4.4%  2.5x1018  -  -  r2148: [Bi]=5.5%  5.1x1017  -  -  1.0x106  0.283  CBi 237 ±20  790 ±150 567 ±130 155 ±10 11 ±0.5 9 ±0.5  Table 6.4: Best fit parameters for the temperature dependence of the hole mobility in GaAs1-x Bix . Units for the coefficients can be inferred from eqn. 6.10. Empty boxes in the table indicate parameters that were not included in the fit of the data.  iii) GaAs1-x Bix hole mobility as a function of temperature The temperature dependence of the hole mobility was explored for several GaAs and GaAs1-x Bix samples. Data are shown in Fig. 6.7 for samples containing 1.75%, 1.91%, 3.5% and 5.5% bismuth. The best fit parameters from the model given above in eqn.6.10 are given in Table 6.4 for the GaAs1-x Bix samples shown in Fig. 6.7. Best fit parameters for the low temperature, low As:Ga (GaAs1-x Bix x = 0) samples shown in Fig. 6.6b are given in Table. 6.3. The coefficients for the contributions from phonon and ionized impurity scattering are similar in all samples, and these values are also similar to values as those found for the low temperature and low As:Ga flux ratio GaAs samples above (see Table 6.3). Recall that the contribution to the mobility from ionized impurity scattering is difficult to quantify due to lack of data at sufficiently low temperatures. Small variations in the doping concentrations of the samples will also contribute to the scatter in this coefficient of the ionized impurity scattering (see eqn.2.36). 72  0.13  103  T-1.5 T1.5  T-1.5  µ, cm2/V s  µ, cm2/V s  T1.5  102  1.75% Bi  1.91% Bi  102 50  100  150 200 Temperature, K  250  300  50  100  150 200 Temperature, K  250  300  250  300  102 1.5  T  102  µ, cm2/V s  µ, cm2/V s  T  -1.5  5.5% Bi 101  3.5% Bi 101 50  100  150 200 Temperature, K  250  300  50  100  150 200 Temperature, K  Figure 6.7: Temperature dependence of the hole mobility for GaAs1-x Bix samples with 1.75%, 1.91%, 3.5% and 5.5% bismuth content. Data fits are shown along with the components from each contribution to the model.  With increasing bismuth content, there are increasing contributions from the temperature independent term. For the highest bismuth content samples measured, 4.4% and 5.5%, the hole mobility was found to be entirely limited by the temperature independent term. This is believed to be related to scattering from bismuth clusters, where a sharp increase in their density is to be expected at high bismuth concentration. The scattering crosssection calculated from CBi for the lowest bismuth concentration sample (x = 0.0094) is 7.5x10−15 cm2 . This is in good agreement with the estimated value for the scattering cross-section of an isolated bismuth atom based on the work by Fahy et al., 2.2x10−14 cm2 . The scattering cross-section found experimentally using the value of CBi for the high concentration samples is 3.4x10−14 cm2 , much larger than the values found for the 0.94% GaAs1-x Bix sample. This is an indication that there is scattering from Bi clusters, and/or an increase in hole effective mass[82]. Therefore, I suggest that the temperature independent mobility for the 5.5% GaAs1-x Bix sample is controlled by scattering from bismuth cluster states. Nitrogen dimers also produce a resonant state near the edge of the conduction band and this may also be the case for bismuth clusters and the valence band edge. Low temperature high resolution PL measurements of the GaAs1-x Bix with 0.045% (measured by  73  SIMS) did not show emission from isolated or bismuth-bismuth pairs in the band gap of GaAs1-x Bix [83]. This implies that bismuth dimers may also have a resonant state within the valence band.  Conclusions GaAs samples grown under conventional conditions, as well as under conditions of low growth temperature and low As:Ga flux ratio with and without bismuth surfactant were explored for comparison with the GaAs1-x Bix epi-layers. GaAs grown under conventional conditions gave expected hole mobilities reported elsewhere in the literature[1, 2, 31]. A reduction in hole mobility was observed for samples grown at low temperatures (≃ 300◦ C). A further reduction was found for GaAs grown at low temperatures with a reduced As:Ga flux ratio (≃ 2). When bismuth is used as a surfactant during the growth of low temperature, low As:Ga flux ratio GaAs, the hole mobility is found to return to the expected room temperature value. As significant surface bismuth is present during the growth of GaAs1-x Bix , the reduction in hole mobility in GaAs1-x Bix is attributed to the incorporation of bismuth and not the growth conditions. The hole mobility is measured to decrease with increasing amounts of bismuth up to 5.5%. This reduction is expected as bismuth perturbs the valence band of the host GaAs. The predicted value for the mobility based on the work by Fahy et al. on the scattering due to isolated isovalent impurities gives a good estimate of the measured values for low concentration samples. The reduction of hole mobility in GaAs1-x Bix is measured to be weaker as a fraction of the GaAs mobility than the analogous case of electron mobility in GaNx As1-x films. The hole mobility of GaAs1-x Bix is expected to increase for bismuth concentrations greater than 6% due to the formation of a bismuth impurity band[82]. Temperature dependence of GaAs samples showed various trends. For samples grown under conventional conditions a T−3 dependence was observed at high temperatures, a T−1 dependence was observed at intermediate temperatures and a T1.5 dependence found at low temperatures. The T1.5 is expected for ionized impurity scattering. Neither the T−3 dependence nor the T−1 dependence are associated with any known scattering mechanism(s), though they have been reported elsewhere[1, 2]. GaAs grown under conditions of low temperature and low As:Ga flux ratio, show only a T−1.5 dependence at high temperatures (acoustic phonon scattering) and T1.5 dependence at low temperatures. These samples grown under low temperature low As:Ga ratio conditions where bismuth is used as a surfactant were treated as [Bi]= 0% epi-layers. The temperature dependence of the GaAs1-x Bix hole mobility was fit with a model that included terms for phonon scattering, ionized impurity scattering and a temperatue independent term to account for the effect of bismuth scattering. The temperature independence consistent with theoretical work on the mobility of on dilute nitride GaNx As1-x by Fahy et al. including cluster states. There is an increasing contribution from the bismuth 74  related terms for increasing bismuth content. High bismuth content samples with 4.4% and 5.5% bismuth were found to have mobilities that were constant over the temperature range 75-300 K. This temperature independent mobility is associated with scattering from bismuth clusters.  75  chapter 7:  DEVICE FABRICATION and CHARACTERISTICS This chapter discusses the growth, fabrication and measurements on GaAs, GaNx As1-x and GaAs1-x Bix devices. The optical properties of GaAs1-x Bix were explored using p-i-n LED heterostructures. These results were then compared to measurements on GaAs LED’s and a triple-quantum-well Inx Ga1-x As p-i-n LED device. These results will be discussed first. The latter part of the chapter focuses on the investigation of deep traps in the materials using deep level transient spectroscopy (DLTS). The preferred device design for DLTS measurements is a one-sided abrupt junction, which was realized using a semiconductor/metal interface, i.e., a Schottky junction. These result are complementary to the mobility measurements in the investigation of the GaAs1-x Bix material system. Deep traps present in the material do not act as strong scattering centers and therefore will have very effect little on the carrier mobility. The LED heterostructures also allowed for investigation using DLTS, but the two-sided nature of the junction means that we can not distinguish between electron traps and hole traps without additional information. However, this limited analysis did result in insights into possible defects associated with bismuth incorporation and the growth procedure.  7.1  Light emitting diodes  GaAs1-x Bix layers were included in p-i-n structures to investigate the potential of GaAs1-x Bix as the optically active material in LED’s. Earlier work had shown bulk GaAs1-x Bix to be a strong light emitter[25, 27]. The growth of the GaAs1-x Bix layer is as described in section 5.1 and also covered briefly below. LED’s were fabricated with active regions containing GaAs1-x Bix layers 50 nm thick with up to 5.5% bismuth; GaAs1-x Bix composition is based on fits to XRD spectra. The optical and electrical characterisation was performed primarily by R.B. Lewis. Solid (opaque) metal contacts, as opposed to ring-like contacts, were used. The electroluminescence was collected from the periphery of the contact and photoluminescence measurements were performed in the areas between the metal contacts. These device structures also allowed for the investigation of defects in 76  Ohmic contacts on etched mesas  p-GaAs, 1000nm GaAs, 25nm GaAs1-xBix, 50nm GaAs, 25nm n-GaAs, 1000nm on n+ GaAs substrate  Ohmic contact  Figure 7.1:  Schematic of LED sample showing device structure and contacts.  Device structure is indicated by shading: darkest shaded regions are the highly doped substrate and buffer (n-type) and epi-layer (p-type); the active region shows the undoped GaAs layers (darker) and the GaAs1-x Bix core (lighter). Contacts are shown in gold. Size, spacing and layer thicknesses are not to scale.  GaAs1-x Bix layer through the use of DLTS in collaboration with Dr. P. Mooney at Simon Fraser University, which is discussed later in the chapter.  7.1.1 Device fabrication A p-i-n diode structure was used to test the light emitting capabilities of GaAs1-x Bix for use in LED’s. Device design was done in collaboration with R.B. Lewis, who also performed the electrical and optical characterisation of the devices. The active region of the p-i-n structure consisted of 50 nm of GaAs1-x Bix sandwiched between 25 nm undoped GaAs spacer layers, as shown in Fig. 7.1. The diodes were grown on n+ GaAs substrates. First a 1000 nm thick n-doped GaAs buffer, where the initial 100 nm had a doping gradient from 2x1018 cm−3 to 5x1017 cm−3 , was deposited. The growth was interrupted between the n-doped layer and the undoped layer to adjust growth conditions: substrate temperatures of 300◦ C-350◦ C, arsenic cell temperature of 350◦ C and gallium cell temperature of 850◦ C. The interruptions typically lasted for 10 minutes. The interruption was necessary for the same reasons as described in the GaAs1-x Bix growth procedure (section 5.1) and also to switch the gas source from SiBr4  77  to CBr4 . When a Si cell was included on the growth flange of the chamber this step was no longer needed. The intrinsic region consisted of 50 nm of GaAs1-x Bix with 25 nm of intrinsic GaAs on either side. This entire region was grown at a low growth rate (0.1 µm/h). The two surrounding GaAs regions were grown at standard As2 overpressure, while the GaAs1−x Bix layer was grown with the As2 overpressure lowered to nearly stoichiometric levels. Bismuth flux was incident on the substrate for the entirety of the intrinsic region. No growth interruption was used for the transition from intrinsic to p-doped layers and the growth rate and substrate temperature were ramped back to standard conditions during the growth process. This causes a small region (approximately 25 nm) where the p-doping is nonuniform. A further 1000 nm of p-doped (5x1017 cm−3 ) GaAs was grown followed by 100 nm of highly p-doped (2x1018 cm−3 ) GaAs. The highly doped capping layer was used so that ohmic contacts could be more easily achieved. LED’s with active regions containing Inx Ga1-xAs and GaAs were also grown for comparison, where the active regions of these devices were grown at 580◦C and 300◦C, respectively. The Inx Ga1-xAs LED structure had three Inx Ga1-xAs quantum wells 5 nm thick separated by 15 nm of GaAs in the active region. The device was identical in all other aspects to the GaAs1-x Bix LED’s. The Inx Ga1-xAs quantum layers were grown at the same conditions as the GaAs surrounding layers at a growth temperature of 580◦ C. The three quantum wells had compositions of 18% indium Inx Ga1-x As as determined by XRD. Small circular Ti/Pt/Au dots (0.32 mm diameter) were used for p-type ohmic contacts[71] and Ni/AuGe/Au for n-type ohmic contacts[72]. Each circular p-type contact defined a single device on the wafer. The large n-type contact, located on the back of the wafer was common to all devices. Contacts were deposited through a metal shadow mask using e-beam evaporation. The wafer was annealed at 450◦ C for 20 seconds after deposition to improve the contact conductivity. Using the Ti/Pt/Au contacts as a mask, 300 nm mesas were etched on the surface of the p-doped epi-layer to minimise current spreading. The height of the etched mesas were checked using profilometry. Fig. 7.2 shows the IV characteristics of several devices fabricated on a single wafer for a GaAs1-x Bix LED containing a 50 nm thick active region of 1.8 % bismuth. All the devices show low leakage currents, e.g. 5 µA at 1 V reverse bias. Similar leakage currents were observed for all LED devices. The devices obey the diode equation: I = Is  e  VD/  n VT  −1  (7.1)  where IS is the saturation current, VD is the voltage across the diode, n is the ideality factor and VT is the thermal voltage. A saturation current, IS = 7.1 µA/cm2 and an ideality factor, n= 2.26 are obtained. A series resistance of 100 Ω is found at the highest voltages measured. For radiative electron-hole recombination in the active region of this device (mid-junction), an ideality factor of 2 is expected. It should be noted that large non-radiative recombination in the junction will also lead to a measured ideality factor of 2.  78  10-2  Current, A  10-3  device 1 device 2 device 3 device 4  10-4 10-5 10-6 10-7 -1.5  Figure 7.2:  -1  -0.5  0 Potential, V  0.5  1  1.5  Current-Voltage curves of four [Bi] = 1.8% GaAs1-x Bix light  emitting diodes at 300 K. All devices show similar leakage current and forward bias characteristics.  7.1.2 Optical emission from GaAs1-xBix light emitting diodes Temperature dependence of the PL is shown in Fig. 7.3 for a GaAs1-x Bix LED with 1.8% bismuth. A 523 nm pulsed laser was used as the excitation source, more details about the PL measurements are found in section 3.3.3. Peaks are observed for both the GaAs and GaAs1-x Bix at 865 nm and 975 nm, respectively. There is also a wide low energy tail along with the GaAs1-x Bix emission peak which we attribute to shallow states in the band gap near the valence band edge associated with bismuth clusters. Similar shallow states have been observed in GaNx As1-x [6]. The peak wavelength of the GaAs1-x Bix emission at 300 K agrees with the expected shift in band gap energy for 1.8% bismuth incorporation[11]. Both peaks are blue shifted with decreasing temperature. Shown in the inset of Fig. 7.3 is a fit using 2 the Varshni equation, Eg (T ) = Eg (O K) − TαT+β (eqn. 2.12) to the temperature dependence of the peak energy of the emission. The 0 K band gap energies are found to be 1.48 eV and 1.32 eV for the GaAs and GaAs1-x Bix emission, respectively, α = 0.36 meV/K and β = 356 K is used for both fits. The shift from the known 0 K value for the GaAs (1.519 eV) band gap is due to doping in the p-type layer and corresponds to carbon doping of approximately 4x1018 cm−3 [85], which is close to the expected doping in the topmost region of the p-type layer. The shift in the 0 K band gap energy of the GaAs1-x Bix again agrees with the expected change in energy for this bismuth concentration[11]. The relative decrease in intensity of the GaAs emission at low temperatures is attributed to greater confinement of the carriers in the smaller band gap GaAs1-x Bix active region.  79  Intensity, arb. units  GaAs  PL Peak, eV  GaAs0.982Bi0.018  GaAs GaAsBi  1.5 1.4 1.3  100 200 300 Temperature, K  8K 50K 150K 250K 300K 800  900  1000 1100 Wavelength, nm  1200  1300  Figure 7.3: PL spectra for a 1.8% GaAs1-x Bix p-i-n structure over the temperature range 8 K to 300 K. The inset shows the peak emission energies as a function of temperature for both the GaAs and GaAs1-x Bix peaks fit with the Varshni equation (dashed lines). Spectra are offset vertically for clarity. This figure was originally presented in J. Crys. Growth 311:1872 (2008)[84]  Electroluminescence (EL) spectra for an LED with a 50 nm 1.8% bismuth GaAs1-x Bix active region are shown in Fig. 7.4 for various forward bias current densities, along with the room temperature PL for the same sample taken from Fig. 7.3 for comparison. Two clear peaks are once again observed corresponding to GaAs and GaAs1-x Bix , along with the low energy tail. Peak emission is observed at 870 nm (GaAs) and 987 nm (GaAs1-x Bix ). The low energy emission is observed to be relatively more intense compared to the GaAs1-x Bix peak in the EL spectra as compared to the PL. There is no significant wavelength shift observed with increasing current density. The peak intensity is measured to increase superlinearly with current. The radiative recombination rate increases approximately quadratically with carrier density. The EL peak emission wavelength is red-shifted relative to the room temperature PL peak. This is attributed the longer lifetimes of the injected carriers in the case of EL. The longer lifetimes of the carriers allows them to thermalise with the shallow in-gap states and then recombine from these states. The temperature dependence of the 50 A/cm2 EL emission for the same LED with 50 nm 1.8% bismuth GaAs1-x Bix active region is shown in Fig. 7.5. The GaAs emission peak is observed to shift to shorter wavelengths with decreasing temperature in agreement with the PL results and the Varshni formula. There is no shift in the emission wavelength of the GaAs1-x Bix layer over this temperature range. This is believed to result from two competing 80  Intensity, arb. units  RT PL  100A/cm2  75A/cm2  50A/cm2 800  Figure 7.4:  900 1000 1100 Wavelength, nm  1200  EL spectra for the same 1.8% GaAs1-x Bix light emitting diode for  various injection current densities at 300 K; room temperature (300 K) from Fig. 7.3 PL is shown for comparison. Spectra are offset vertically for clarity.  Intensity, arb. units  This figure was originally presented in J. Crys. Growth 311:1872 (2008)[84]  100K 150K 200K 250K 300K 800  Figure 7.5:  900 1000 1100 Wavelength, nm  1200  EL spectra of the same 1.8% GaAs1-x Bix LED using an injection  current density of 50 A/cm2 for temperatures ranging from 100 K to 300 K. This figure was originally presented in J. Crys. Growth 311:1872 (2008)[84]  81  processes: the increase in band gap with decreasing temperature and an increased tendency for the emission to come from lower energy states with decreasing temperature. The fact that the intensity of the GaAs1-x Bix peak and low energy peak increases with decreasing temperature, is in agreement with the previous statement. The intensity of the emission from the in-gap states surpasses the intensity of the GaAs1-x Bix peak at 100 K. A similar increase in the in-gap emission at low temperature has been observed in GaAs1-x Bix bulk epi-layers and GaNx As1-x [6]. Carriers trapped in shallow traps, which recombine radiatively, are increasingly likely to recombine from the trapped state with decreasing temperature. In other words, the holes will spend more time in shallow traps at low sample temperatures. At low temperatures they are less likely to thermalize back to the band edge and are therefore more likely to recombine with electrons from these states. The performance of these devices was compared to a similar p-i-n LED structure with three Inx Ga1-xAs quantum wells, grown as described above. The integrated room temperature EL intensity was 102 × higher than the emission from the GaAs1-x Bix device for the same injection current density. Quantum wells are known to be more efficient light emitters than bulk epi-layers and it is speculated that an active region comprised of three GaAs1-x Bix 5 nm quantum wells would be significantly brighter than a single 50 nm thick GaAs1-x Bix active region design. One possible limiting factor in the light emission of the GaAs1-x Bix LED’s may be a large number of non-radiative recombination centers in the device. The defects are expected to be primarily located in the GaAs1-x Bix layer where the incorporation of bismuth may add defects and the growth conditions are very different from those used for GaAs. Defects may also be present in the low temperature grown GaAs surrounding layers on either side of the GaAs1-x Bix layer. Deep levels present in the material were explored using deep level transient spectroscopy (DLTS), which is discussed in the following section.  7.2  GaNx As1-x and GaAs1-x Bix Schottky diodes  In an effort to better understand the nature of any possible negative effects of incorporation of bismuth, deep levels present in the GaAs1-x Bix were explored. The activation energy, capture cross-section and concentration of deep traps in MBE grown GaAs, GaNx As1-x and GaAs1-x Bix films were measured using DLTS (described in section 3.3.6 and in the appendix). The focus of the DLTS measurements was deep hole traps related to bismuth incorporation. Again, in the analogous case of GaNx As1-x electron traps were investigated in relation to lowered PL intensity with nitrogen incorporation[86–88]. The traps found were shown to limit the PL intensity, as well as the performance of the material in solar cells. It is expected that the incorporation of bismuth may also result in deep traps that could potentially limit the material’s application in devices. As will be shown, Schottky devices fabricated on GaAs1-x Bix epi-layers had performance problems due to the sensitive dependence of the device characteristics on the junction interface. This led to high variability between different GaAs1-x Bix samples and within the devices on the same wafer.  82  This work was done in collaboration with Dr. Pat Mooney’s group at Simon Fraser University. The contributing students were: I. Koslow and A. Royle (GaAs and GaNx As1-x ); Z. Jiang and E. Chen (GaAs, GaAs1-x Bix and LED’s). Work on p-type GaNx As1-x laid the ground work for device design and fabrication of devices for investigation of the GaAs1-x Bix alloys. Measurements on GaAs devices, both Schottky diodes and p-i-n junctions, were used as control samples in order to be able to differentiate traps due to changes in growth conditions as opposed to traps due to bismuth incorporation. It should be noted that this work is not complete and that further investigation of the effect of the growth conditions is needed to fully understand the nature and cause of the deep levels in GaAs1-x Bix . Future directions for this work are discussed in the next chapter.  7.2.1 Device fabrication Samples were grown on either p+ or n+ GaAs substrates depending if the device was to be p-type or n-type, respectively. These substrates have dopant concentrations of approximately 2x1018 cm−3 . Silicon is used as a n-type dopant and zinc for p-type in the substrates used. A highly doped (approximately 1x1018 cm−3 ) buffer layer (> 300 nm thick) is grown on top of the de-oxidized surface. This thick doped buffer layer is required to have the substrate/buffer interface well removed from the depletion region of the Schottky diode. After the growth of an initial highly doped buffer layer, the doping concentration in the buffer is then lowered to the desired level for the device. Typical values were in the range of 2x1016 cm−3 to 2x1017 cm−3 , where doping concentrations in the lower part of this range were preferred. The thickness of the GaAs buffer varied depending on the time required for the CBr4 source to achieve the desired level and the time required to set the growth conditions to those of the epi-layer. Typical thicknesses were 500 − 1000 nm. Once the desired doping level is achieved and the growth conditions for the epi-layer set, a further layer of doped GaAs is grown that is > 100 nm thick in the case of GaAs or GaNx As1-x and > 30 nm thick in the case of GaAs1-x Bix devices. This is to ensure that the doping in the epi-layer of the DLTS device will be uniform. The growth as described here only refers to the doped GaAs buffer layer (see Fig.7.6), the epi-layers are grown as described as in section 4.1 for GaNx As1-x and section 5.1 for GaAs1-x Bix . The GaAs1-x Bix films were typically post growth annealed in-situ to remove any excess surface bismuth. GaAs1-x Bix samples were also grown with and without a thin (≤ 10 nm) similarly doped GaAs capping layer in attempts to reduce leakage currents in the devices. / hr ) in GaNx As1-x films were grown at higher than normal growth rates (approximately 2 µm order to achieve sufficiently low doping requirements. All epi-layer thicknesses were thicker than the expected depletion widths; 1x1016 cm−3 doping gives depletion widths of 210 nm for a built-in potential of 0.7 and a reverse bias voltage of 2 V. The width of the depletion region accessible to DLTS measurements is set by the doping concentration of the epi-layer, as well the Schottky barrier height (which includes the built-in voltage) and applied voltage.  83  Schottky contacts doped epi-layer doped GaAs buffer SI GaAs substrate  backside contact  Figure 7.6:  Schematic of DLTS sample showing device structure and contacts.  The device structure is indicated by shading: the highly doped substrate is darkest; the doping gradient layer is shown in a lighter shade; and the low doped epi-layer is shown in the lightest shade. The back and surface contacts Cr/Au are shown in gold. Size, spacing and layer thicknesses are not to scale.  The Schottky diode devices used for DLTS were fabricated using a shadow mask technique with Cr/Au or Al/Au metal contacts. Small circular (0.32 mm diameter) contacts were deposited onto the as-grown epi-layer by electron beam deposition. The front contact was 25 nm Cr and then 100 nm of Au. The Schottky contact is made at the Al or Cr semiconductor interface. The low work functions of Cr and Al make them suitable for Schottky contacts to p-type GaAs[2]. The Au upper part of the contact pad is necessary for wire bonding the device in the DLTS apparatus. The device area is defined by the size of the contact (nominal areas of 0.125 mm2 were used). A much larger contact of 15 nm of Cr followed by 50 nm of Au, deposited onto the back surface of the substrate comprised the second contact; due to its large size and the high doping of the substrate, this contact is non-rectifying. A schematic diagram of the device is shown in Fig. 7.6. Shown in Fig. 7.7 are example IV data representative of the DLTS devices from a GaAs1-x Bix Schottky device with Al/Au frontside contacts before the device was wire bonded. The IV characteristics were checked before and after wire bonding the devices to ensure that the wire bonding process did not change the device performance. Typical p-type GaNx As1-x DLTS devices, using only Cr as the Schottky contact, had leakage currents on the order of 0.3 − 30 µA for −1.0V applied bias. GaAs Schottky devices had leakage currents of 2 − 20 µA for −1.0V. Leakage currents are given in the summary tables in the Appendix. High bismuth concentration GaAs1-x Bix (x > 3%) Schottky devices were found to be very leaky, as shown in Fig. 7.7. In an effort to reduce this leakage current, thin (≃ 10 nm) 84  Current, A  10-3  10-4  10-5  device 2,2 device 3,4 device 4,2 device 2,4 device 4,4  10-6 -5.0  Figure 7.7:  -4.0  -3.0  -2.0 -1.0 Potential, V  0.0  1.0  Current-Voltage data for several DLTS p-type Schottky devices  fabricated on one wafer with a [Bi]=4.1% GaAs1-x Bix epi-layer. These particular devices used Al as the Schottky contact, which showed a lower leakage current compared to Cr. Solid lines show devices with low leakage currents and dotted lines show leaky devices.  doped GaAs capping layers were included in the device heterostructure and Al/Au Schottky contacts were also tried. Only devices where both techniques, GaAs capping layer and Al/Au contacts, were applied had sufficiently low leakage currents to be suitable for analysis by DLTS. High leakage currents are believed to be related to poor surface morphology and surface metallic droplets. Fig. 7.8 shows an example of the capacitance-voltage relationships for DLTS devices. This data set is representative of all DLTS devices measured. From the slope of /1 C2 as a function of voltage it can be seen that the epi-layer is uniformly doped over the probed depth (refer to eqn. D.3), 1.9 x1017 cm−3 p-type for the devices shown. Variations from device to device are due to slight differences in the area of the contacts, which define the capacitor area. The built-in voltage is found from the x-intercept of the same data. Vbi = 0.97 V for the devices shown. For the same wafer where Cr/Au Schottky contacts were used Vbi was measured to be 0.23 V. This decreased built-in potential resulted in leakier devices.  85  1.6  200  1.4  175 150  1 0.8  125  0.6  100  Capcitance (pF)  1/C2 (1020 F-2)  1.2  0.4 75  0.2 0  -3  -2  -1  0  50  Potential (V)  Figure 7.8: Capacitance-Voltage data for several DLTS p-type Schottky devices fabricated on a single wafer with a [Bi]=4.1% GaAs1-x Bix epi-layer. These devices 1 2 shows that the epi-layer had Al Schottky contacts. A linear relationship for / C is uniformly doped.  7.2.2 Deep level transient spectroscopy measurements Fig. 7.9 illustrates schematically the results of the DLTS measurements on GaAs, GaNx As1-x and GaAs1-x Bix Schottky devices. Trap energies are shown relative to the band edges and concentrations are averaged over several devices. Tables showing a more complete summary of the data (including leakage currents, built-in potentials and trap capture crosssections) are included in the Appendix. Each trap is denoted by type (electron, E or hole, H) and the material in which it was observed (G for GaAs, N for GaNAs and B for GaAsBi), e.g. a hole trap observed in GaAs is labelled HG-#, where the number is arbitrarily assigned. In the case of the LED’s as the type of trap can not be determined, traps are labelled as GL or BL for GaAs LED and GaAsBi LED respectively. Again the trap number is assigned arbitrarily.  i) GaAs A hole trap (HG-1) with EA = 0.57 eV was found in all p-type GaAs Schottky devices for growth temperatures of 560◦ C and 450◦C. Two other hole traps with EA of 0.73 eV (HG-2) and 0.17 eV (HG-3) were also observed, but only in some samples. A p-type GaAs Schottky diode grown at 300◦ C also showed the 0.57 eV trap and also had associated with it a new 86  1017  Concentration, (cm-3)  mid-gap  p-GaAs, 580C p-GaAs, 450C p-GaAs, 310C n-GaAs, 550C n-GaAs, 390C  HG-X  1016  HG-1  1015  EG-4  HG-3  EG-3 EG-2  1014  EG-1  HG-2 1013 EG-5 12  10  Ev  Ec  Energy, (eV)  1017  Concentration, (cm-3)  mid-gap 1016 HN-1 HN-2  15  10  HG-1  p-GaNAs, 0.17% p-GaNAs, 0.29% p-GaNAs, 0.36% p-GaNAs, 0.48% p-GaNAs, 1.8%  1014 1013 1012 Ev  Ec  Energy, (eV)  1017  Concentration, (cm-3)  mid-gap 1016 HB-2  p-GaAsBi, 4.1% p-GaAsBi, 3.0% p-GaAsBi, <0.2%  HB-1 HB-3  1015  HB-4 HG-1  1014 1013 1012 Ev  Figure 7.9:  Energy, (eV)  Ec  Summary of DLTS measurements of defect activation energy and  concentration for GaAs, GaNx As1-x and GaAs1-x Bix Schottky diodes. Hole traps are shown by solid lines and electron traps are shown by dashed lines. The 0.17% GaNx As1-x samples (black) was grown using bismuth surfactant.  87  trap at 0.9 ± 0.1 eV (HG-X). This trap was found with a concentration of approximately 1x1016 cm−3 , but the measured peak height (related to trap concentration) was found to strongly vary with measurement rate window. This fact and the anomalously large activation energy suggest that the cross-section is likely strongly dependent on temperature. Hole traps similar in activation energy and capture cross-sections to HG-1 are reported elsewhere and attributed to either an arsenic anti-site (AsGa ) EL2-type defect[86] or Fe impurities[89]. Fe in GaAs is reported to give defects with activation energy 0.54 eV and a cross-section of 3.4x10−16 cm2 [90] and other hole traps associated with Fe are reported to have EA = 0.52 eV and cross-sections of 0.3-3.4x10−16 cm2 [89]. The AsGa EL2-type hole trap is reported to have EA = 0.52-0.54 eV, but is not observed in epitaxial GaAs[91]. The AsGa trap is also easily observed in n-type GaAs at 0.75 eV below the CBM[92], but was not observed in any n-GaAs samples here. It is therefore most likely that this hole trap is related to an Fe impurity. No electron traps were found in n-GaAs grown at or above 450◦C in which solid Si or SiBr4 were used as the dopant source, where the detection limit for trap concentration is approximately 1x1012 cm−3 . Two n-type Schottky GaAs devices were grown at 390◦ C and 300◦ C. The device grown at 390◦ C showed several new electron traps and details for these traps are given in Table D.1 in the Appendix. The electron traps at EG-1, EG-2 and EG-3 are typically found in bulk GaAs[92]. These traps are believed to be related to either Ga vacancies or As interstitials, as the materials are grown under arsenic-rich conditions. A further defect was observed in the DLTS spectrum of the n-GaAs Schottky device grown at 390◦ C, but it could not be properly analysed due to proximity to the 0.54 eV trap. The n-type GaAs Schottky device grown at 300◦ C was found not to be strongly rectifying and therefore unsuitable for DLTS measurements. As the device was grown under the same conditions as previous devices, the doping concentration is expected to be very similar. The non-rectifying behaviour is possibly due to a high density of traps with energies near the middle of the band gap.  ii) GaNx As1-x As was noted earlier the incorporation of small amounts of nitrogen in GaAs greatly reduces the PL intensity. The use of bismuth as a surfactant has been found to mitigate the dencrease in the PL intensity. Deep levels present in the material will act as nonradiative recombination centers which will lower PL intensity. Several GaNx As1-x epi-layers with x = 0.18% − 0.48% were explored and three new hole traps found. The work on these GaNxAs1-x devices also laid the ground work for the device fabrication on GaAs1-x Bix epi-layers. Fig. 7.9b shows schematically the traps found in the GaNx As1-x p-type Schottky devices. Three new hole traps were found to be associated with the incorporation of nitrogen: HN-1, 88  HN-2 and HN-3. Trap details can be found in Table D.2 in the Appendix. There is an observed increase in the density of HN-1 with increased nitrogen content, but no clear trend was found in HN-2. Of the hole traps found only HN-1 and HN-2 were consistently observed in all GaNxAs1-x devices measured. Hole trap HN-3 could not be reliably measured, and was found in only a few devices on a single wafer. The DLTS signal from HG-3 was relatively weak and noisy, and due to overlap with HG-1 its activation energy could not be accurately measured. There is a notable increase in hole trap concentrations when bismuth has been used as a surfactant during the growth of the epi-layer. This is surprising, as samples grown with bismuth as a surfactant have shown improved photoluminescence intensity[23], and as such a lower defect density would have been expected. Hole traps similar to HN-1 have been observed in other GaNx As1-x films grown by MOCVD with EA = 0.21 eV[86][91]. A further similar hole trap was observed in electron irradiated GaAs with EA = 0.20 eV and a cross-section of 1.2x10−15 cm2 . This would suggest that the defect was related to the use of the plasma source, however this hole trap is not observed in all cases where an rf plasma is used in the growth of GaNx As1-x [93]. Furthermore, as this hole trap is not always detected in GaNx As1-x epi-layers, it is not likely related to nitrogen. Hole traps with activation energy 0.35 eV similar to HN-2, are often reported in relation to GaNx As1-x [93, 94], as well as GaAs[89]. No source has been identified for the defect. Defects of activation energies of 0.33 and 0.32 eV have also been reported in GaNx As1-x [91][86]. In both cases the defects were removed by annealing the sample. In the measurements performed on these GaNx As1-x samples, higher values for the activation energy of HN-2 were observed in samples with increased nitrogen content. Though the valence band is not affected by the incorporation of nitrogen, the fact that the films are strained could account for this small change of activation energy. Since no similar increase in activation energy is observed in HN-1, this is further confirmation that the HN-1 hole trap may not be related to nitrogen.  iii) GaAs1-x Bix As discussed in the GaAs1-x Bix growth section, the ideal growth parameters are difficult to realize and as a result metallic surface droplets form and the surface roughens over time. The preferred devices to perform DLTS measurements required GaAs1-x Bix epi-layer thicknesses greater than 300 nm. Using the modified growth procedure (section 5.1), Schottky devices were fabricated on GaAs1-x Bix epi-layers, however poor surface morphology prevented an in-depth study of the GaAs1-x Bix Schottky devices. One GaAs1-x Bix Schottky device containing only a small fraction of bismuth was measured along with two high concentration (x 3%) samples. These latter two GaAs1-x Bix Schottky devices were found to be very leaky, with typical leakage currents of 0.2-2.0 mA for a reverse bias voltage of 1.0 V (see Fig. 7.7). DLTS spectra from the GaAs1-x Bix p-type Schottky device where < 0.2% (based on XRD) showed two new hole traps: HB-1 and HB-2. 89  Peaks in the DLTS spectra from the leaky bulk high bismuth concentration GaAs1-x Bix Schottky devices were very broad and analysis based on a distribution of traps could not be performed. Analysis was performed assuming the peaks observed were due to individual traps with a single EA : HB-3 and HB-4. The HB-3 trap found in one device is relatively shallow (approximately 0.12 eV away from VBM) and would typically not be measured using this technique. Furthermore the value found for its cross-section (0.6-33x10−20 cm−2 ) is unphysically small.  iv) Light emitting diodes The p-i-n heterostructures of the GaAs1-x Bix LED’s were also suitable for DLTS measurements. However, because these junctions are two-sided it is not possible to discern whether the observed defects are hole traps or electron traps. The device structure also limited the analysis in regards to the location of the defects, as the depletion width of the devices was found to be larger than the intrinsic region. Therefore, observed traps may be present in the GaAs1-x Bix layer, in the undoped surrounding GaAs layers, throughout the low temperature grown intrinsic region, in the n- or p-type GaAs or in all of the above layers. The traps found from the DLTS measurements on the LEDs are shown in Fig. 7.10. Comparisons to the above GaAs Schottky devices and LEDs where the active region does not include a GaAs1-x Bix layer allowed for further insight into the defects observed. Simulations of the band structure of the devices under various bias conditions performed by Z. Jiang allowed for some insight on trap location based on trap activation energy. The shaded region in Fig. 7.10 represents the activation energies of traps that may be observed if they exist in the intrinsic region of the LED. The GaAs p-i-n structure with the active region grown at 580◦ C showed only one trap with a forward bias filling pulse of 0.8 V at the familiar energy of 0.57 eV (GL-1). This is the activation energy of the HG-1 hole trap observed in all p-GaAs Schottky devices. The trap observed in the GaAs p-i-n heterostructure also has a similar cross-section of 7.0x10−16 cm2 . Based on its activation energy and measurement parameters the simulations of the device placed this trap in the intrinsic region of the LED. As it was measured to be a minority trap, this suggests that the intrinsic region is nominally n-type. This is unexpected as MBE grown GaAs is typically p-type due to background carbon in the chamber. However, residual silicon in the chamber may n-type dope the intrinsic region. The lower growth temperature (300◦ C) GaAs p-i-n diode shows two further hole traps (HL-2 and HL-3), one of which is also a minority carrier trap. These traps were found for a forward bias filling pulse of 0 V when the observation of a minority traps is unexpected. This anomalous device behaviour was attributed to a large increase in traps. Three p-i-n GaAs1-x Bix LED heterostructures were measured using DLTS containing 1.4%, 1.8% and 4.7% bismuth. Several forward bias filling pulses were used. Many new traps were found. Both the 1.4% and 4.7% GaAs1-x Bix p-i-n show a trap at 0.56 eV (BL-1). This trap 90  1016  1016  GL-2 10  GL-3  15  GL-3  1015  1014  1014  GL-1  1013  1013  1012  1012 Ev  Energy, (eV)  1017  Concentration, (cm-3)  Concentration, (cm-3)  1017  GaAs LED, 580C GaAs LED, 300C  mid-gap  mid-gap  Ec  Energy, (eV)  1017  GaAsBi LED, 4.7% GaAsBi LED, 1.8% GaAsBi LED, 1.4%  1016  BL-3 BL-4, 5,6, 7, 8 BL-1  1015  1016  BL-3 BL-2  BL-2  BL-1  8, 7, 6, 5  4 1015  1014  1014  1013  1013  1012  1012 Ev  Energy, (eV)  Figure 7.10:  mid-gap  mid-gap  Energy, (eV)  Ec  Summary of DLTS measurements of defect activation energy and  concentration for GaAs and GaAs1-x Bix p-i-n diodes. As most traps could not be determined to be either electron or hole traps, holes traps (solid lines) corresponding to the measured defects are shown on the left and electron traps on the right (dashed lines). Activation energy is given relative to the band edge. Traps present with the shaded region are present only in the intrinsic region of the p-i-n heterostructure. Traps GL-1 and GL-2 are believed to hole traps based on measurements on GaAs Schottky diodes.  91  Concentration, (cm-3)  Concentration, (cm-3)  1017  corresponds to a shoulder observed in the n-GaAs Schottky device grown at 390◦C. It is possible that this trap is an electron trap observed in GaAs[92].  Conclusions GaAs1-x Bix has been incorporated into the active region of a p-i-n heterostructure and shown to prodeuce light emission in simple light emission devices. Light intensity from the device is 100× weaker than similar devices more closely resembling commercially used products. Electroluminescent emission from GaAs1-x Bix layers in the devices was measured to be temperature insensitive, while the Varshni relation is obeyed for the GaAs peak and all optically excited emission. Both PL and EL show emission from shallow in-gap states, where the in-gap emission is stronger in the case of EL due to the longer lifetimes of the injected carriers. This effect also caused a red-shift of the peak energy in the EL spectra. These shallow states are believed to be related to bismuth clusters, similar to those observed in GaNx As1-x alloys. Deep levels present in the layers of the device structure were explored using DLTS and several defect states were found. Based on the small number of deep traps found in GaAs samples indicates that the source materials in use are of very high purity and that the growth chamber is itself also very clean. Recall that the detection limit for the measurements discussed above is on the order of 1x1012 cm−3 . The hole trap found in nearly all p-type Schottky devices has been identified as being related to an Fe impurity and not the an AsGa anti-site. The Fe incorporation is believed to be due to background contamination present in the growth chamber, possibly from the stainless steel gas lines of the CBr4 source itself. There is an increase in the number of electron traps found with decreasing temperature. No electron traps are observed in samples grown at 580◦ C, whereas devices grown at 300◦C were no longer rectifying due to the defect concentration. It is possible that potential impurities in the chamber are incorporated only at lowered substrate temperatures. Of the two hole traps found in the GaNxAs1-x Schottky devices, only one is believed to be nitrogen related and the other related to the growth conditions. An increase (approximately 10×) in the trap densities of the GaNx As1-x is observed in samples where bismuth has been used as a surfactant during growth. This is an unexpected result, as the use of bismuth surfactant has been shown to increase photoluminescent intensity and reduce shallow gap states[6, 23]. High leakage current in the GaAs1-x Bix Schottky diodes, due to the formation of metallic droplets on the surface, limited the analysis of the deep levels in GaAs1-x Bix by DLTS. These preliminary results obtained show traps not previously observed in either GaAs or GaNx As1-x . Further work on the growth conditions of GaAs1-x Bix is needed in order to fabricate devices with lower leakage currents. Based on the results of the high growth temperature GaAs p-i-n, the intrinsic region is  92  nominally n-type, the minority trap corresponds to the Fe impurity hole trap found in GaAs Schottky diodes. The nominal doping in the intrinsic region may be due to latent silicon in the growth chamber after the growth of the n-type layer. Lower activation traps (< 0.36 eV) are only observed in devices where a growth interrupt was included in the heterostructure. Devices where no interrupt is used show no traps at these energies. This suggests that these shallow traps are introduced into the device at the interruption of the growth. While the preliminary DLTS work described above on GaAs1-x Bix , along with the measurements on GaAs and GaNx As1-x , resulted in a few insights more work is required on the growth of GaAs1-x Bix epi-layers before further work can be done on GaAs1-x Bix Schottky diodes. It is possible to use abrupt p-n junctions where the doping in one side is much higher than the other. This in effect would act as a one-sided junction and remove the problem of producing contacts on the GaAs1-x Bix epi-layer directly. The GaAs1-x Bix layer need only be on one side of the junction and the top most layer could be grown as thick as necessary for a smooth, clean surface to deposit contacts.  93  chapter 8:  CONCLUSIONS This thesis focused on the growth and characterisation of GaNx As1-x and GaAs1-x Bix alloys. These two material systems are often classed together as special cases of III-V ternary alloys due to the dramatic effects of the material properties due to small incorporated amounts as a result of resonant states within the bands of GaAs. Nitrogen incorporation results in a resonant state within the conduction band of GaAs and bismuth produces a state within the valence band. GaNx As1-x at the time of the undertaking of this thesis had been extensively explored and the work performed sheds light on some unanswered questions. On the other hand, GaAs1-x Bix was relatively unexplored, but treated as an analogous material system and results from work on GaNx As1-x are often used as a road map for work on GaAs1-x Bix . Therefore the aim of the work in this thesis was to explore the effect of bismuth incorporation on the hole transport. The focus of the work on GaNx As1-x was the effect of the operation conditions of the rf plasma active nitrogen source on the optoelectrical efficiency of the GaNx As1-x epi-layers. By changing the relative ratio of active atomic or molecular nitrogen the preferred species for improved luminescence was explored. In the case of GaAs1-x Bix , where the resonant state lies on the valence band of GaAs, the focus of the work was on the effect of bismuth incorporation on the electronic transport of holes. Incorporation of nitrogen has been shown to greatly reduce the mobility of electrons. Temperature dependence of the measured hole mobility gave insight into the nature of the scatterers present due to addition of bismuth. Before the transport measurements could be performed, much work was necessary on the growth procedure of GaAs1-x Bix to produce sufficiently thick doped epi-layers. The rf plasma active nitrogen source used in the growth of GaNx As1-x can be operated in one of two modes: on-resonance, which favours active atomic nitrogen; and off-resonance, which favours excited nitrogen dimers. It was also found that over time the plasma discharge tube becomes contaminated. Growth where the plasma discharge tube is contaminated resulted in increased emission from shallow in-gap states associated with clusters of nitrogen atoms. This was especially true for off-resonance operation. The increase of clustering of nitrogen is expected due to the higher ratio of excited dimers to active atomic nitrogen. No contaminants (C, Si, B, or O) were detected to be incorporated in the grown epi-layers 94  using SIMS with a detection limit of 1x1016 cm−3 . Growth of GaNx As1-x using a cleaned plasma discharge tube greatly reduces the luminescence from the in-gap states and increased the amount of incorporated nitrogen with the same foreline pressures for both modes of operation. As the contamination of the plasma discharge tube will occur over the period of any growth campaign, growth using the plasma source in the on-resonance mode is preferred to reduce low energy emission from shallow in-gap states related nitrogen clusters. As the contamination of the plasma discharge is unavoidable and higher probability of incorporated clusters of nitrogen atoms in the material with off-resonance operation of the plasma source, it is concluded that excited atomic nitrogen is the preferred species for dilute GaNx As1-x . Use of the rf plasma source in the on-resonance mode results in a higher ratio of excited atomic nitrogen. Starting from the growth procedure used previously by growers in the research group, the growth process window of GaAs1-x Bix was explored. The significant changes included: a reduction of the growth rate to compensate for the lower adatom diffusion at the reduced growth temperatures; reduction of the arsenic source pre-valve pressure to allow for better dynamic range on the source valve; and the inclusion of a growth interrupt before the growth of the GaAs1-x Bix epi-layer. The last step reduced the amount of GaAs buffer grown at low temperatures and allowed the surface to smooth under arsenic overpressure. Using RHEED to monitor the surface of the growing epi-layers, it was found that growth under conditions where a 2×1 surface reconstruction is observed produces better GaAs1-x Bix films. These epi-layers have electrically active carbon doping and more uniform compositions. The latter is based on both XRD and SIMS measurements. It was observed that there exists a short time delay before the observation of the 2×1 surface reconstruction and a further delay before bismuth is incorporated. These time delays are correlated, the length of the delay for bismuth incorporation is approximately twice as long as the delay to the surface reconstruction transition. These delays are attributed to the time required for sufficient surface bismuth to build up, where twice as much surface bismuth is necessary for incorporation as for the surface reconstruction transition. The hole mobility of GaAs samples grown under conventional conditions were measured to assess the effect of the growth conditions. Samples grown under conventional conditions gave expected hole mobilities based on empirical models in the literature. Hole mobility was observed to be reduced for samples grown at low temperatures (≃ 300◦ C) and further reduced growth at low temperatures with a reduced As:Ga ratio (≃ 2). Samples where bismuth surfactant is used during the growth of low temperature, low As:Ga GaAs had hole mobilities at the expected values. As significant levels of bismuth are present during the growth of GaAs1-x Bix epi-layers, any reduction in the hole mobility in GaAs1-x Bix epi-layers was attributed solely to the incorporation of bismuth. The hole mobility in GaAs1-x Bix is measured to decrease with increasing amounts of bismuth up to 5.5%. This reduction is expected as bismuth perturbs the valence band of the host GaAs, analogous to the case of the effect of nitrogen incorporation on the electron mobility. The reduction of hole mobility in GaAs1-x Bix is not as severe as is the reduction 95  of the electron mobility in GaNx As1-x . The heavier holes will be less effectively scattered than the conduction band electrons. A resonant state with energy further shifted from the valence band edge, compared to the position of the isolated nitrogen resonant state in the conduction band, would also less effectively scatter charge carriers. It has been found that a bismuth impurity band forms at concentrations above 6% and as a result the hole mobility should increase with increasing concentrations above this amount. With increasing bismuth content alloy scattering may begin to play a larger role in the mobility. This has been suppressed in similar material systems by using growth conditions that result in an ordered alloy, where this ordering has been observed in GaAs1-x Bix epi-layers with concentrations of approximately 10%. Further work on the growth conditions of GaAs1-x Bix will allow for the growth of high concentration films. The temperature dependence of the hole mobility was used to quantify the various scattering mechanisms. A temperature independent term was used to account for the effect of bismuth incorporation. This term was included into a simple model which also had terms for phonon scattering and ionized impurity scattering. There was an increasing contribution from the bismuth related term for increasing bismuth content. Samples with 4.4% and 5.5% bismuth were found to have mobilities that were constant over the temperature range 75300 K. Based on the fitted coefficients from the model it was concluded that clusters of bismuth atoms play a significant role in the scattering of charge carriers. A calculation of the scattering cross-section from a GaAs1-x Bix (x = 0.0094) was in good agreement with a theoretical value based on work on GaNx As1-x electron mobility. Doped GaAs and GaNx As1-x epi-layers were used to fabricate Schottky diodes and deep levels in the material grown in the MBE reactor were explored using DLTS. The initial work on GaNx As1-x devices laid the groundwork for the fabrication of devices on GaAs1-x Bix epi-layers. The measurements on the GaAs layers showed the reactor chamber and the sources used to be very clean and pure. Only one trap was consistently found in GaAs epi-layers grown using the standard conditions of Tsubstrate ≃ 580◦ C and a As:Ga flux ratio of 5 or higher. This defect is an Fe impurity and the gas lines of the CBr4 dopant source is suspected to be the likely source. An increase in deep traps was found for decreasing growth temperatures below 400◦ C and p-type devices where the epi-layer was grown at 300◦ C were found to be non-rectifying due to the density of deep traps. Incorporation of nitrogen resulted in two further hole traps, where only one was related to the incorporated nitrogen. Devices fabricated from a sample with trace amounts of bismuth (< 0.2%) showed two further defects related to bismuth incorporation. Poor surface morphology in high bismuth concentration GaAs1-x Bix epi-layers limited DLTS analysis, though a broad DLTS spectrum was observed indicating a number of defects with similar activation energies. Further work on the growth of GaAs1-x Bix is required before a more in-depth DLTS study can be undertaken. A more complete exploration of the effects of reduced temperature and As:Ga flux ratio will also be necessary to fully understand the results from GaAs1-x Bix DLTS measurements. Light emitting p-i-n heterojunctions were also explored using DLTS, however the two-sided nature of the devices made it difficult to determine if the defects were hole or electron traps. 96  It was found that many traps were incorporated into the devices at the interface of the n-type and intrinsic regions where a growth interruption is used in the growth procedure at this point. Several other defects found in these devices were attributed to either the growth conditions used or bismuth incorporation. As strong light emission was measured in these devices, it is believed that many of the traps found in the DLTS measurements are present only in the surrounding undoped GaAs layers of the intrinsic region. A similar device containing Inx Ga1-x As quantum wells was found to be approximately 100× more intense than the GaAs1-x Bix p-i-n devices. The electroluminescence of the GaAs1-x Bix layer in the p-i-n LED heterostructure was found to be temperature insensitive, whereas the optically excited (PL) emission followed the expected Varshni relation. This is consistent with measurements on GaAs1-x Bix epi-layers. The temperature insensitivity of GaAs1-x Bix epi-layers has been observed by other research groups and attributed to the combination of a semiconductor with a semi-metal.  97  Bibliography [1] S. Adachi. Physical Properties of III-V Semiconductor Compounds. Wiley, 1992. [2] S. Adachi. Properties of Group-IV, III-V and II-VI Semiconductors. Wiley, 2005. [3] A. Janotti, S.-H. Wei, and S.B. Zhang. Theoretical Study of the Effects of Isovalent Coalloying of Bi and N in GaAs. Phys. Rev. B, 65:115203, 2002. [4] H. A. McKay, R. M. Feenstra, T. Schmidtling, and U. W. Pohl. Arrangement of nitrogen atoms in GaAsN alloys determined by scanning tunneling microscopy. Appl. Phys. Lett., 78:82, 2001. [5] P.R.C. Kent, L. Bellaiche, and Alex Zunger. Pseudopotential theory of dilute III-V nitrides. Semicond. Sci. Tech., 17:851, 2002. [6] D.A. Beaton. Temperature Dependence of Photoluminescence of the Dilute Nitride Semiconductor GaNx As1−x . Master’s thesis, University of British Columbia, 2003. [7] D.L. Young, J.F. Geisz, and T.J. Coutts. Nitrogen-induced Decrease of the Electron Effective mass in GaAs1−x Nx Thin Films Measured by Thermomagnetic Transport Phenomena. Appl. Phys. Lett., 82:1236, 2003. [8] S.R. Kurtz, A.A. Allerman, C.H. Seager, R.M. Sieg, and E.D. Jones. Minority carrier diffusion, defects, and localization in InGaAsN, with 2% nitrogen. Appl. Phys. Lett., 77:400, 2000. [9] E.V.K. Rao, A. Ougazzaden, Y. Le Bellego, and M. Juhel. Optical properties of low band gap GaAs1−x Nx layers: Influence of post-growth treatments. Appl. Phys. Lett., 72:1409, 1998. [10] S.R. Kurtz, A.A. Allerman, E.D. Jones, J.M. Gee, J.J. Banas, and B.E. Hammons. InGaAsN solar cells with 1.0 eV band gap, lattice matched to GaAs. Appl. Phys. Lett., 74:729, 1999. [11] S. Francoeur, M.J. Seong, A. Mascarenhas, S. Tixier, M. Adamcyk, and T. Tiedje. Bandgap of GaAs1−x Bix , 0<x<3.6%. Appl. Phys. Lett., 82:3874, 2003. [12] W. Shan, W. Walukiwiecz, J.W. Ager, E.E. Haller, J.F. Geisz, D.J. Fiedman, J.M. Olson, and S.R. Kurtz. Band Anticrossing in GaInNAs Alloys. Phys. Rev. Lett., 82:1221, 1999.  [13] S. Fahy and E.P. O’Reilly. Intrinsic Limits on Electron Mobility in Dilute Nitride Semiconductors. Appl. Phys. Lett., page 3731, 2003. [14] E. Strohm. Optical and Electrical Properties of Dilute GaNx As1−x Alloys. Master’s thesis, University of British Columbia, 2001. [15] D.G. Cooke, E.C. Young, F.A. Hegmann, and T. Tiedje. Electron Mobility in Dilute GaAs Bismide and Nitride Alloys Measured by Time-resolved Terahertz Spectroscopy. Appl. Phys. Lett., 89:122103, 2006. [16] K. Oe and H. Okamoto. New Semiconductor Alloy GaAs1−x Bix Grown by Metal Organic Vapour Phase Epitaxy. Jpn. J. Appl. Phys., 37:L1283, 1998. [17] Hiroshi Okamoto and Kunishige Oe. Growth of metastable alloy inasbi by low-pressure movpe. J. J. Appl. Phys., 37:1608, 1998. [18] Q. Du, J. Alperin, and W.I. Wang. Molecular beam epitaxial growth of GaInSbBi for infrared detector applications. Journal of Crystal Growth, 175-176:849, 1997. [19] B. Joukoff and A. M. Jean-Louis. Growth of InSb1−x Bix Single Crystals by Czochralski Method. J. Cryst. Growth, 12:169, 1971. [20] K. Oe. Metalorganic Vapor Phase Epitaxial Growth of Metastable GaAs1−x Bix Alloy. J. Cryst. Growth, 237-239:1481, 2002. [21] H. Shen, S.H. Pan, Z. Hang, J. Leng, F.H. Pollack, J.M. Woodall, and R.N. Sacks. Photoreflectance of GaAs and Ga0.82 Al0.18 As at elevated temperatures up to 600C. Appl. Phys. Lett., 53:1080, 1988. [22] M.R. Pillai, S.S Kim, S.T. Ho, and S.A. Barnett. Growth of Inx Ga1−x As/GaAs Heterostructures Using Bi as a Surfactant. J. Vac. Sci. Tech. B, 18:1232, 2000. [23] S. Tixier, M. Adamcyk, E.C. Young, J.H. Schmid, and T. Tiedje. Surfactant Enhanced Growth of GaNAs and InGaNAs Using Bismuth. J. Crys. Growth, 251:449, 2003. [24] S. Tixier, M. Adamcyk, T. Tiedje, S. Francoeur, A. Mascarenhas, P. Wei, and F. Schiettekatte. Molecular Beam Epitaxy Growth of GaAsBi. Appl. Phys. Lett., 82:2245, 2003. [25] Erin Christinia Young. GaNAs and GaAsBi: Structural and Electronic Properties of two Resonant State Semiconductor Alloys. PhD thesis, University of British Columbia, Canada, December 2006. [26] N. Zangenberg, D.A. Beaton, T. Tiedje, S. Tixier, M. Adamcyk, R. Kumaran, J.A. MacKenzie, E. Nodwell, E.C. Young, and G.I. Sproule. Molecular Beam Epitaxy Growth of the Dilute Nitride GaAs1−x Nx with a Helical Resonator Plasma Source. J. Vac. Sci. Technol. A, 25:850, 2007. [27] X. Lu, D. A. Beaton, R. B. Lewis, T. Tiedje, and Yong Zhang. Composition Dependence of Photoluminescence of GaAs1−x Bix Alloys. Appl. Phys. Lett., 95:041903, 2009.  [28] R.N. Kini, L. Bhusal, A.J. Ptak, R. France, and A. Mascarenhas. Electron Hall Mobility in GaAsBi. J. Appl. Phys., 106:043705, 2009. [29] Charles Kittel. Introduction to Solid State Physics. John Wiley and Sons, New York, 7th edition, 1996. [30] S.M. Sze. Physics of Semiconductor Devices. John Wiley and Sons, 1981. [31] M. Balkanski and R.F. Wallis. Semiconductor Physics and Applications. Oxford, 2000. [32] Y.P. Varshni. Temperature Dependence of the Energy Gap in Semiconductors. Physics, 334:149, 1967. [33] B. Monemar and L. Samuelson. Optical transitions via the deep O donor in GaP. I. Phonon interaction in low-temperature spectra. Phys. Rev. B, 18:809, 1978. [34] K.P. O’Donnell and X. Chen. Temperature-dependence of Semiconductor Band-gaps. Appl. Phys. Lett., 58:2924, 1991. [35] H.P. Hjalmarson, P. Vogl, D.J. Wolford, and J.D. Dow. Theory of Substitutional Deep Traps in Covalent Semiconductors. Phys. Rev. Lett., 44:810, 1980. [36] L. Vegard. Die Konstitution der Mischkristalle und die Raumfullung der Atome. Z. Physik, 5:17, 1921. [37] R.E. Nahory, M.A. Pollack, W.D. Johnston Jr., and R.L. Barns. Band Gap versus Composition and Demonstration of Vegard’s law for In1−x Gax Asy P1−y Lattice Matched to InP. Appl. Phys. Lett., 33:659, 1978. [38] S. Tixier, S.E. Webster, E.C. Young, T. Tiedje, S. Francoeur, A. Mascarenhas, P. Wei, and F. Schiettekatte. Band Gaps of the Dilute Quaternary Alloys GaNx As1−x−y Biy and Ga1−y Iny Nx As1−x . Appl. Phys. Lett., 86:112113, 2005. [39] J.W. Matthews and A.E. Blakeslee. Defects in Epitaxial Multilayers 1. Misfit Dislocations. J. Cryst. Growth, 27:118, 1974. [40] E.C. Young, A.N. Koveshnikov, S. Tixier, K.L. Kavanagh, and T. Tiedje. Strain relaxation by ¡100¿ misfit dislocations in dilute nitride InGaAsN/GaAs quantum wells. Phys. Stat. Sol. A, 202:2849, 2005. [41] G. Pozina, I. Ivanov, B. Monemar, J.V. Thordson, and T.G. Anderson. Properties of Molecular-beam Epitaxy-grown GaNAs from Optical Spectroscopy. J. Appl. Phys., 84:3830, 1998. [42] I. Vurgaftman, J.R. Meyer, and L.R. Ram-Mohan. Band parameters for III-V Compound Semiconductors and their Alloys. J. Appl. Phys., 89:5815, 2001. [43] S.-H. Wei and A. Zunger. Giant and Composition-Dependent Optical Bowing Coefficient in GaAsN Alloys. Phys. Rev. Lett., 76:664, 1996.  [44] U. Tisch, E. Finkman, and J. Salzman. The Anomalous Bandgap Bowing in GaAsN. Appl. Phys. Lett., 81:463, 2002. [45] X. Liu, M.E. Pistol, L. Samuelson, S. Schwetlick, and W. Seifert. Luminescence in GaAs. Appl. Phys. Lett., 56:1451, 1990.  Nitrogen Pair  [46] K. Alberi, J. Wu, W. Walukiewicz, K. M. Yu, O. D. Dubon, S. P. Watkins, C. X. Wang, X. Liu, Y.-J. Cho, and J. Furdyna. Valence-band anticrossing in mismatched III-V semiconductor alloys. Phys. Rev. B, 75:045203, 2007. [47] W. Shan, W. Walukiewicz, K.M. Yu, J.W. Ager, E.E. Haller, J.F. Geisz, D.J. Friedman, J.M. Olson, S.R. Kurtz, H.P. Xin, and C.W. Tu. Band anticrossing in III-N-V alloys. Pkys. Stat. Sol. B, 223:75, 2001. [48] K. Alberi, O.D. Dubon, W. Walukiewicz, K. M. Yu, K. Bertulis, and A. Krotkus. Valence band anticrossing in GaBix As1−x . Appl. Phys. Lett., 91, 2007. [49] B. Fluegel, A. Mascarenhas, A.J. Ptak, S. Tixier, E.C. Young, and T. Tiedje. E+ transition in GaAs1−x Nx and GaAs1−x Bix due to isoelectronic-impurity-induced perturbation of the conduction band. Phys. Rev. B, 76:155209, 2007. [50] N.W. Ashcroft and N.D. Mermin. Solid State Physics. Saunders College Publishing, New York, USA, 1976. [51] Raymond Serway. Physics for Scientists and Engineers. Thomson Brooks/Cole, 4th edition, 1996. [52] E.H.C. Parker. The Technology and Physics of Molecular Beam Epitaxy. Plenum Press, New York, 1985. [53] T.A. Flaim and P.D. Ownby. Observations on Bayard-Alpert Ion Gauge Sensitivities to Various Gases. J. Vac. Sci. and Tech., 8:661, 1971. [54] V.V. Preobrazhenskii, M.A. Putyato, O.P. Pchelyakov, and B.R. Semyagin. Experimental Determination of the Incorporation Factor of As4 During Molecular Beam Epitaxy of GaAs. J. Cryst. Growth, 201/202:170, 1999. [55] M.R. Melloch, J.M. Woodall, E.S. Harmon, N. Otsuka, F.H. Pollak, D.D. Nolte, R.M. Feenstra, and M.A. Lutz. Low-Temperature-Grown III-V Materials. Annu. Rev. Mater. Sci., 25:547, 1995. [56] S.R. Johnson, C. Lavoie, and T. Tiedje. Semiconductor Substrate Temperature Measurement by Diffuse Reflectance Spectroscopy in Molecular Beam Epitaxy. J. Vac. Sci. Techol., 11:1007, 1993. [57] S.R. Johnson and T. Tiedje. Effect of Substrate Thickness, Back Surface Texture, Reflectivity, and Thin Film Interference on Optical Band-gap Thermometry. J. Cryst. Growth, 175/176:273, 1997.  [58] L. Dweritz and R. Hey. Reconstruction and Defect Structure of Vicinal GaAs(001) and Alx Ga1−x As(001) Surfaces during MBE Growth. Surface Science, 236:15, 1990. [59] Martin B. Adamcyk. Epitaxial Growth of Dilute Nitride-Arsenide Compound Semiconductors by Molecular Beam Epitaxy. PhD thesis, University of British Columbia, 2002. [60] M. Masnadi Shirazi. Surface Reconstructions during growth of GaAs1−x Bix Alloys. Master’s thesis, University of British Columbia, 2010. [61] Michael Brian Whitwick. Surface Evolution during Gallium Arsenide Homoepitaxy with Molecular Beam Epitaxy. PhD thesis, University of British Columbia, Canada, December 2009. [62] Anders Ballestad. Epitaxial Growth Dynamics in Gallium Arsenide. University of British Columbia, 2005.  PhD thesis,  [63] S. R. Johnson. Optical Bandgap Thermometry in Molecular Beam Epitaxy. PhD thesis, University of British Columbia, 1995. [64] S.R. Johnson and T. Tiedje. Temperature Dependence of the Urbach Edge in GaAs. J. Appl. Phys., 78:5609, 1995. [65] M. Beaudoin, A.J.G. deVries, S. R. Johnson, H. Laman, and T. Tiedje. Optical Absorption Edge of Semi-insulating GaAs and InP at High Temperatures. Appl. Phys. Lett., 70:3540, JUN 1997. [66] Ayahiko Ichimiya and Philip I Cohen. Refelction High Energy Electron Diffraction. Cambridge University Press, New York, 2004. [67] E.C. Young, M.B. Whitwick, T. Tiedje, and D.A. Beaton. Bismuth Incorporation in GaAs1−x Bix Grown by Molecular Beam Epitaxy with In-situ Light Scattering. Phys. Stat. Sol. C, 4:1707, 2007. [68] L.T. van der Pauw. A Method for Measuring the Resistivity and Hall Coefficient on Lamellae of Arbitrary Shape. Philips Res. Rep., 13, 1958. [69] E.C. Young, S. Tixier, and T. Tiedje. Bismuth Surfactant Growth of the Dilute Nitride GaNx As1−x . J. Cryst. Growth, 279:316, 2005. [70] M. Shirazi-Masnadi, D. A. Beaton, R. B. Lewis, and T. Tiedje. Surface phase diagram of MBE-grown GaAs1-x Bix (001). to be published, 2011. [71] G. Stareev. Formation of Extremely Low Resistance Ti/Pt/Au Ohmic Contacts to p-GaAs. Appl. Phys. Lett., 62:2801, 1993. [72] R. Williams. Modern GaAs Processing Methods. Artech House, 1990.  [73] T. Tiedje, E. C. Young, and A. Mascarenhas. Growth and properties of the dilute bismide semiconductor GaAs1−x Bix a complementary alloy to the dilute nitrides. Int’l J. NanoTech., 5:963, 2008. [74] R.C. Newman. The upper limits of useful n- and p-type doping in GaAs and AlAs. Materials Science and Engineering B, 66:39, 1999. [75] Cheong, Byoung-Ho and Chang, K. J. Compensation and diffusion mechanisms of carbon dopants in gaas. Phys. Rev. B, 49:17436, 1994. [76] S. Fahy, A. Lindsay, and E.P. O’Reilly. Intrinsic Limits on Electron Mobility in Disordered Dilute Nitride Semiconductor Alloys. IEE Proceedings - Optoelectronics, 151:352, 2004. [77] G.B. Stringfellow and G.S. Chen. Atomic Ordering in III/V Semiconductor Alloys. J. Vac. Sci. Tech. B, 9:2182, 1991. [78] Y. Nakata and O. Ueda and T. Fujii. High two-dimensional electron gas mobility enhanced by ordering in ingaas/n-inalas heterostructures grown on (110)-oriented inp substrates by molecular beam epitaxy. J. J. Appl. Phys., 30:L249, 1991. [79] K. Volz. Structural characteristics of dilute nitride & bismide iii/v semiconductors. Presentation at 2010 Bismuth Containg Semiconductors Workshop, Ann Arbor, MI. [80] C. Erginsoy. Neutral Impurity Scattering in Semiconductors. Phys. Rev., 79:1013, 1950. [81] O.F. Sankey, J.D. Dow, and K. Hess. Theory of resonant scattering in semiconductors due to impurity central-cell potentials. Appl. Phys. Lett., page 664, 1982. [82] G. Pettinari, A. Polimeni, J.H. Blokland, R. Trotta, P.C.M. Christianen, M. Capizzi, J.C. Maan, X. Lu, E.C. Young, and T. Tiedje. Unusual Compositional Dependence of the Exciton Reduced Mass in GaAs1−x Bix (x=0-10%) Investigated by Magnetophotoluminescence Spectroscopy. to be published, xx:xx, 2010. [83] A. Mascarenhas, R. Kini, Yong Zhang, R. France, and A. Ptak. Comparison of the Dilute Bismide and Nitride alloys GaAsBi and GaAsN. Phys. Stat. Sol. B, 246:504, 2009. [84] R.B. Lewis, D.A. Beaton, X. Lu, and T. Tiedje. GaAs1−x Bix Light Emitting Diodes. J. Crys. Grow., 311:1872, 2009. [85] Shinho Cho and Eun Kyu Kim. Band Gap Narrowing Effects in Carbon Doped GaAs by Photoluminescence Spectroscopy. In TENCON 99. Proceedings of the IEEE Region 10 Conference, volume 2, page 1099, 1999. [86] So Tanaka, A. Moto, M. Takahashi, T. Tanabe, and S. Takagishi. Spatial Distribution of Deep Level Traps in GaNAs Crystals. J. Crys. Growth, 221:467, 2000. [87] S. Kurtz, S. Johnston, and H.M. Branz. Capacitance-spectroscopy identification of a key defect in N-degraded GalnNAs solar cells. Appl. Phys. Lett., 86, 2005.  [88] S.W. Johnston and S.R. Kurtz. Comparison of a dominant electron trap in n-type and p-type GaNAs using deep-level transient spectroscopy. J. Vac. Sci. Tech. A, 24:1252, 2006. [89] A. Mitonneau, G.M. Martin, and A. Mircea. Hole Traps in Bulk and Epitaxial GaAs Crystals. Electronics Letters, 13:666, 1977. [90] M. Kleverman, P. Omling, L-˚ A . Ledebo, and H.G. Grimmeiss. Electrical Properties of Fe in GaAs. J. Appl. Phys., 54:814, 1983. [91] Qiang Gao, Hark Hoe Tan, Chennupati Jagadish, and Prakash N. K. Deenapanray. Defect Evolution in Annealed p-type GaAsN Epilayers Grown by Metalorganic Chemical Vapour Deposition. Jpn. J. Appl. Phys., 42:6827, 2003. [92] G.M. Martin, A. Mitonneau, and A. Mircea. Electron Traps in Bulk and Epitaxial GaAs Crystals. Electronics Letters, 13:191, 1977. [93] P. Krispin, S.G. Spruytte, J.S. Harris, and K.H. Ploog. Origin and Annealing of Deeplevel Defects in p-type GaAs/Ga(As,N)/GaAs Heterostructures grown by Molecular Beam Epitaxy. J. Appl. Phys., 89:6294, 2001. [94] R.J. Kaplar, A.R. Arehart, S.A. Ringel, A.A. Allerman, R.M. Sieg, and S.R. Kurtz. Deep Levels and their impact on Generation Current in Sn-doped InGaAsN. J. Appl. Phys., 90:3405, 2001. [95] James Ward Brown and Ruel V. Churchhill. Complex Variables and Applications. McGraw-Hill Inc., New York, 6th edition, 1996. [96] Dieter K. Schroder. Semiconductor Material and Device Characterization. WileyInterscience, New York, 2nd edition, 1998.  Appendix A:  Wafer Conventions  (a) US flats  Figure A.1:  (b) EU flats  Correct, as labelled wafer identifications for both US and EU flats  for (100) substrates.  As we have been ordering substrates mainly from Wafer Tech. and as they typically have only wafers with EU flats in stock, we have been ordering what is available. The image above correctly identifies the crystallographic directions on the flats for both wafer conventions. Please note that often we refer to (001) substrates, in which case the US flats are on [1¯10] and [110] (major and minor), EU flats are then [¯1¯10] and [¯110] (major and minor). The minor flat direction in the EU convention is equivalent to the major flat direction in the US convention. The image on the growth log sheets is not necessarily correct for substrate used for any particular growth, therefore care must be taken to correctly identify the wafer convention used for any given growth.  105  Appendix B:  Sheet Resistance from Contacts on the Periphery of a Thin Film In order to find the mobility of a thin film, both the carrier concentration and sheet resistance must be independently measured. The carrier concentration can be readily measured from the transverse potential that builds up to counter the Lorentz force, while the sheet resistance, RS , is found from the following relation[68]:  e  −πRP S,QR RS  +e  −πRP Q,RS RS  =1  (B.1)  where RP S,QR is the resistance measured VIPQRS (current enters at contact P, exits at contact S and the voltage is measured across contacts Q and R). Similarly for RP Q,RS . This relation is referred to as the van der Pauw equation The derivation of the above relation begins by considering a thin film (see Fig. B.1) where current is applied to contact, P. This current will flow away radially to infinity. If we now also consider the current to exit via another contact, S we can define the current density at a point r by: I (B.2) J= πrd where d is the thickness of the film. From this we can define the electric field strength at point r as: ρI E = ρJ = (B.3) πrd where ρ is the resistivity of the material. We can now find the potential difference between any two points on the thin film, for convenience we choose points on the edge aligned with the entrance and exit points of the current, P and S R  VQ − VR =  Edr = Q  ρI πd  R Q  ρI dr = − ln r πd  ac (b + c)(a + b)  (B.4)  Which yields for RP S,QR: RP S,QR =  −RS ln π 106  ac (b + c)(a + b)  (B.5)  Figure B.1: Semi-infinite conducting thin film used in the derivation of the van der Pauw equation. Conformal mapping can be used to transform this semi-infinite sheet into any other arbitrary, simply connected shape. The contacts on the real axis correspond to contacts on the edge of the transformed thin film. Contacts here labelled P, Q, R and S correspond to the contacts numbered 1 through 4 on the samples measured in this thesis.  where I have replaced  ρ d  with RS . Similarly for RP Q,RS we have: RP Q,RS =  −RS ln π  (a + b)(b + c) b(a + b + c)  (B.6)  when the two equations above are added together we get the van der Pauw relation, eqn. B.1. Conformal mapping is used to transform the above geometry into any well-connected (no holes) thin film of arbitrary shape. The transformation used to map the upper half plane (UHP) to a square is: w = f (z) = B + A  (z − x1 )  −α1 π  (z − x2 )  −α2 π  (z − x3 )  −α3 π  dz  (B.7)  where the integral is over the UHP, the xn ’s are points from the real axis in z space, and the αn ’s are the exterior angles of the line segments ( π2 for the case given above). This is a use of the Schwartz-Cristoffel transformation with transforms the UHP into an n-sided polygon using n − 1 terms in the integral[95].  107  Appendix C:  Hall Measurement Wiring Schematic 4P/4T switch a b c d  Vacuum feed-thru  a b c d  A 1 4  2 3  B  Res/Hall  a b c d  C  a b c d  D  Figure C.1:  V+ r h  V-  r h  I+ I-  Wiring schematic used in the Hall measurement apparatus. The  4P/4T switch cyclically changes the configuration of the voltage and current at the sample. Swapping the contacts for the current source and voltage output allows for the current/voltage to be sourced/measured on non-adjacent contacts for Hall measurements. Care must be taken to correctly record the measurements for the calculation of the sheet resistance and carrier concentration.  108  Appendix D:  Deep Level Transient Spectroscopy Measurement Technique DLTS is a capacitance transient technique that is used to detect trapped charges within a junction. Focus of this discussion will be on the one-sided (Schottky) junctions where information about the doping concentration can also be found along with the device built-in potential, Vbi . The two-sided (p-i-n) junctions do not allow for a measurement of the doping concentration, nor it is possible to determine if defects are electron or hole traps. Diagrams for the two types of junctions are shown in Fig. D.1. The junctions form a depletion width, W , which can be treated as a parallel plate capacitor, where the capacitance, C, of a Schottky junction is given by[96]: C=−  dW dQs = qANA dV dV  (D.1)  where Qs is the free charge near the edge of the depletion region, V is the applied voltage across the junction, q is the charge of an electron, A is the area of the capacitor and NA is the doping concentration. This is equivalent to: C=  ǫs ǫo A W  (D.2)  where ǫs is the dielectric constant of the material (ǫs for GaAs is 12.9). If we differentiate the above and substitute it into eqn. D.1, we can solve for the doping concentration: NA = −  2 C3 = −2 dC 2 qǫs ǫo A dV qǫs ǫo A2 d(CdV )  (D.3)  From this it can be seen that the doping concentration can be found from a measurement of the capacitance-voltage characteristics of the Schottky device. The slope of /1 C2 as a function of V assuming a linear relationship. A linear relationship indicates a uniform doping concentration over the probed region.  109  Ec qVbi  p-type  Ec  EF  W Ev  n-type  EF  Ev  (a) p-i-n junction  Figure D.1:  qΦm  qVbi p-type  metal  (b) Schottky junction  Band diagram schematics for a) p-i-n and b) metal-semiconductor  (Schottky) junctions used in DLTS measurements. The Schottky junction is shown for a p-doped semiconductor. Both the built-in potential, Vbi , and the depletion width, W, are shown for both cases.  The built-in potential, Vbi , of the junction may also be found from the measurement of the device capacitance-voltage characteristics, where we have[96]: Vbi − V =  NA W 2 q 2ǫs ǫo  (D.4)  In combination with eqn. D.2 the above can be written as; 1 2 (Vbi − V ) = C2 qA2 NA ǫs ǫo It can be seen that Vbi can be found from the x-intercept of the plot of applied voltage.  (D.5) 1 C2  as a function of  To perform a DLTS measurement the junction is first held in reverse bias and then a short forward biasing voltage pulse is applied. The brief forward biasing fills the depletion region with carriers and some of these carriers then become trapped by defects in the depletion region. After the forward biasing pulse, carriers are then emitted back to their respective bands, the conduction band for electrons and the valence band for holes. Shown schematically in Fig. D.2. This results in time varying capacitance, which is what is measured in DLTS. Assuming that the forward biasing pulse was sufficiently long to completely fill all the trap states in the depletion region, the occupation of a trap, nT , is given by: nT (t) = NT e −ent  (D.6)  where NT is the trap density and en is the emission rate of an electron trap (ep is used for holes) given by: σn vn Nc -∆ E / kB T en = e (D.7) g 110  a  b  CBM EF traps  depletion width  c  CBM EF traps  depletion width  V=Vrev  depletion width  V=Vpulse  VBM  V=Vrev  VBM  VBM  d  e  Vpulse  0V  CBM EF traps  Cpulse time  Ceq ∆Co  Vrev  Crev 0  2  4  6  8  time 10 12 14 16  Figure D.2: Schematic diagram of DLTS procedure: a) junction is held in reverse bias; b) forward biasing pulse applied, traps in depletion region are filled; c) junction returned to reverse bias bias, trapped carriers are emitted back to the band; d) voltage v. time for a DLTS measurement; e) capacitance v. time for a DLTS measurement showing a capacitance transient.  where σn is the capture cross section for electrons, vn is the average velocity of the electrons, Nc is the effective density of states in the conduction band and ∆E is the energy difference from the trap to the conduction band edge, Ec − ET . The above is also true for hole traps and the valence band edge. It is not necessarily true that the capture cross section of a trap is temperature independent, and as such an additional factor for the thermal activation of the trap, Eσ , is added to ∆E. In which case we have: σ → σ(T ) = σ e  -Eσ/ kB T  (D.8)  where Eσ is the thermal activation energy of the capture cross-section. ∆E is now written as: ∆E = Ec − ET − Eσ (D.9) A DLTS measurement no longer gives an accurate measure of the trap activation energy.  The presence of the trapped charges in the capacitance of the depletion region. As the trapped electrons (or holes) are thermally emitted back to the conduction (or valence) band, the capacitance returns to its equilibrium value. The measured capacitance decays as follows: ∆C ∝ e −ent  (D.10)  It is this change in capacitance that is being measured in DLTS. The capacitance of the junction is measured twice after the application of the forward biasing filling pulse at times T1 and T2 . These two times define the rate window of the 111  t1  t2 T5  Temperature  Capacitance  T4  T3  T2  T1  Figure D.3:  ∆Capacitance  Illustration of a DLTS measurement of capacitance transients for  several temperatures and the resultant DLTS spectrum.  measurement, t2 − t1 . The DLTS signal as a function of temperature is referred to as a DLTS spectrum. For a given trap activation energy and rate window emission rate from the trap will give a peak in the DLTS spectrum. The temperature of the peak emission rate is: ln( t/2 t1 ) en (Tpeak ) = t2 − t1  (D.11)  At temperatures below Tpeak the trapped charges escape slowly from the defect and the resultant change in the measured capacitance over the rate window is very small. For temperatures above Tpeak , trapped charges easily escape from the defects and the capacitance transient is nearly fully decayed before the initial measurement of the capacitance at t1 . Again the measured difference in the capacitance is very small. At temperatures near Tpeak , larger values are measured in the difference of the capacitance. This is shown schematically in Fig. D.3 A series of DLTS spectra range and the relationship This is typically plotted as than one trap contributing can be difficult. Fig. D.4  are measured for various rate windows over a large temperature of the peak temperature and the rate window determines ∆E. an Arrhenius plot of e/n T 2 versus /1 T . There is typically more to the capacitance transient and as a result analysis of the data  112  (a) n-GaAs DLTS Spectra  (b) DLTS Arrhenius plot  Figure D.4: Example of a) DLTS spectra and b) the corresponding Arrhenius plot from a n-GaAs Schottky diode. Four electron traps were observed and the details given in table D.1. Plots taken from the M.Sc. thesis of Zenan Jiang.  113  D.1  Tables of DLTS measurement results  What follows is a series of summarising table of the DLTS measurements performed by Dr. P. Mooney’s research group at Simon Fraser University on both Schottky and p-i-n junctions. Trap activation energy is given relative the correct band edge: CBM for electrons and VBM for holes. Each trap is denoted by type (electron, E or hole, H) and the material in which it was observed (G for GaAs, N for GaNAs and B for GaAsBi), e.g. a hole trap observed in GaAs is labelled HG-#, where the number is arbitrarily assigned. In the case of the LED’s as the type of trap can not be determined, traps are labelled as GL or BL for GaAs LED and GaAsBi LED respectively. Again the trap number is assigned arbitrarily. In the case of the p-i-n junctions the trap concentrations and capture cross-sections are given for both electron and hole traps, as the type of trap can not be determined in these two-sided devices.  114  Sample  NA  Vb  Ileak at -1V  (cm )  (V)  (µA)  name  4.8 x1016  0.54  1.8-8.5  HG-1  1-3.6 x1014  HG-2  1-4.7 x1014  HG-1  38-45 x1014  −3  p-GaAs Tg = 560◦C  4.5 x1016  p-GaAs  0.46  14-26  Tg = 450◦C  traps NT (cm−3 )  HG-3 2.5-3.8 x1013  n-GaAs  1.6x1016  0.69  1.1  EG-5  2.2 x1012  4.1x1016  0.78  1.1  EG-1  1.1 x1014  EG-2  1.8 x1014  EG-3  3.1 x1014  EG-4  3.4 x1014  Tg = 550◦C n-GaAs 115  Tg = 390◦C  Trap EA (eV)  HG-1 0.52-0.57  σ (cm2 ) 0.3-3.7x10−14  HG-2 HG-3  EG-1  EG-2  EG-3  EG-4  EG-5  0.73  0.17  0.14  0.38  0.54  0.66  0.38  -  -  5.1x10−17  2.4x10−14  8.8x10−13  1.8x10−14  2.4x10−15  Table D.1: Summary of DLTS results for p-type and n-type GaAs Schottky devices. Measurements were performed by I. Koslow, A. Royle, Z. Jiang and E. Chen. Doping concentration and built-in potential are averaged over several devices on the same wafer. Trap activation energy is given below the table.  Sample GaNx As1-x  NA  Vb  Ileak at -1V  (cm−3 )  (V)  (µA)  7.4 x1016  0.58  0.3-4.3  traps name  NT (cm−3 )  HG-1 4.9-6.4 x1013 HN-1  1.1-1.6 x1014  HN-2  7.5-12 x1013  HG-1  2.1 x1013  HN-1  8.0 x1014  HN-2  7.6 x1014  HG-1  6-7 x1013  HN-1  1.5-2.0 x1014  HN-2  1.3-1.8 x1014  HN-3  6.5 x1013  HG-1  2-3.7 x1013  HN-1  1.7-2.5 x1014  HN-2  1.2-1.7 x1014  HG-1  7.0 x1014  [N] = 0.17%  HN-1  2.5 x1015  Bi surfactant  HN-2  8.2 x1014  x = 0.18% GaNx As1-x  1.4 x1016  0.53∗  0.61  x = 0.29% GaNx As1-x  7.8 x1016  0.6  0.5-1.1  x = 0.36%  GaNx As1-x  7.3 x1016  0.47  27-33  x = 0.48% GaNx As1-x  1.0 x1016  Trap  Table D.2:  9.3∗  1.17  HN-1  HN-2  HN-3  EA (eV)  0.22  0.33-0.35  ??  σ (cm2 )  2.4x10−14  8.8x10−13  -  Summary of DLTS results for p-type GaNx As1-x Schottky devices.  Measurements performed by I. Koslow and A. Royle. All GaNx As1-x samples were all grown at a 450◦ C with As:Ga ≃ 8. Doping concentration and built-in potential are averages over several devices on the same wafer. Trap activation energy is given  below the table.  116  sample GaAs1-x Bix  NA  Vb  Ileak at -1V  (cm−3 )  (V)  (µA)  name  NT (cm−3 )  6.4x1016  0.46  0.1-5  HB-1  1.9 x1014  HB-2  2.1 x1014  HG-1  1.0 x1014  HB-3  1.5 x1014  < 0.2% GaAs1-x Bix  1.3x1017  0.33-0.55  8.5-37.2  traps  2.7 x1014  3.0%  2.7 x1014 6.8 x1014 GaAs1-x Bix  2.1x1017  0.98-1.01  2.7-3.4  HB-4  4.6 x1014 2.7 x1014  4.1% Trap  HB-1  HB-2  HB-3  HB-4  EA (eV)  0.31  0.30  0.09-0.15  0.36  σ (cm2 )  8x10−15  8x10−14  3.3x10−19 -5.8x10−21  1.2-5x10−18  Table D.3:  Summary of bulk GaAs1-x Bix DLTS results performed by Z. Jiang.  The low concentration GaAs1-x Bix sample ws grown at 300◦ Cusing the ioriginal growth procedure and the two high concetration GaAs1-x Bix samples were grown at 350◦ Cusing the modified growth procedure.  DLTS spectra from the high  concentration samples were broad and analysis is based on a single trap. Doping concentration and built-in potential are averages over several devices on the same wafer. Trap activation energy is given below the table.  117  LED active region  trap  NT electrons  NT holes  (cm−3 )  (cm−3 )  GaAs, Tg = 580◦ C  GL-1∗  GaAs, Tg = 300◦ C  GL-2∗  2.3 x1015  2.1 x1015  GL-3  1.0 x1015  9.5 x1014  -  1.4 x1013  Trap  GL-1  GL-2  GL-3  EA (eV)  0.57  0.52  0.29  σe (cm2 )  1.5x10−14  2.0x10−16  8.9x10−17  σh (cm2 )  7.0x10−16  1.0x10−17  4.5x10−18  Table D.4: Summary of the DLTS results p-i-n junction LED’s with 100 nm thick GaAs active regions grown at 580◦ Cand 300◦ Cusing 0 V forward bias filling pulse performed by Z. Jiang. Traps marked by an ∗ were observed as minority traps. Trap densities and capture cross-section are given for both electron and hole traps.  118  LED active region  trap  NT electrons  NT holes  (cm−3 )  (cm−3 )  BL-4  1.5 x1014  9.9 x1013  [Bi] = 1.4%  BL-6  3.9 x1014  2.6 x1014  BL-7  1.3 x1014  8.4 x1013  BL-1  3.4 x1014  1.7 x1014  BL-3  9.9 x1014  5.1 x1014  GaAs1-x Bix , 50 nm  BL-5  2.4 x1014  1.5 x1014  [Bi] = 1.8%  BL-7  4.3 x1014  2.3 x1014  BL-8  1.5 x1015  8.2 x1014  BL-3  6.7 x1015  2.9 x1015  BL-6∗  1.2 x1014  7.6 x1013  GaAs1-x Bix , 50 nm  BL-1  2.6 x1015  2.4 x1015  [Bi] = 4.7%  BL-2  2.0 x1015  1.8 x1015  119  GaAs1-x Bix , 50 nm  Trap  BL-1  BL-2  BL-3  BL-4  BL-5  BL-6  BL-7  BL-8  EA (eV)  0.56  0.73  0.63  0.22  0.29  0.33  0.37  0.41  σe (cm2 )  0.2-5x10−14  7.3x10−13  1.8-2.8x10−14  9.5x10−16  4.7x10−14  8.2x10−14  0.7-2.2x10−18  8.4x10−15  σh (cm2 )  2.5-8.9x10−18  3.7x10−14  0.9-1.4x10−15  4.7x10−17  2.3x10−15  4.1x10−15  0.4-1.1x10−18  4.2x10−16  Table D.5: Summary of the DLTS results p-i-n junction LED’s with GaAs and GaAs1-x Bix active regions for 0 V forward bias filling pulse performed by Z. Jiang. Traps marked by an for both electron and hole traps.  ∗  were observed as minority traps. Trap densities are given  Appendix E:  Helical Resonator Nitrogen Plasma Source The plasma source is located in a standard effusion cell port in an elemental source molecular beam epitaxy system. The helix itself consists of a sixteen-turn gold-plated, self-supporting OFHC copper coil grounded at one end and open at the other end, inside a coaxial tantalum shield tube with a cap on the open end. A 10 mm diameter pyrolytic boron nitride (PBN) discharge tube is located in the center of the helix. The diameter for the discharge tube was selected following experiments with different diameter tubes, in order to maximise active atomic nitrogen generation, as observed by emission lines in the discharge. The resonator is driven by a wideband oscillator and power amplifier through a three-turn coupling loop as shown in Fig. E.1, with 3.9 m of coaxial line between the power amplifier and the source. Its first two unloaded (absence of a plasma discharge) resonances are located approximately at 63 MHz and 180 MHz. Bombardment with energetic ions is known to degrade the electronic properties of semiconductors. The exit of the discharge tube is equipped with a three-stage PBN baffle (see inset of Fig. E.1) so that there is no line of sight to the plasma and all gas species exiting the discharge zone experience several wall collisions before entering the growth chamber. The baffle is found to reduce the ion current by 2 − 4 orders of magnitude compared to the source without the baffle installed. Helical resonators have the property that at resonance the input impedance at an intermediate tap position is purely real and is adjustable to any value between zero and some maximum value determined by the losses in the resonator. This allows for matching of the input impedance of the helical resonator to the output impedance of the rf generator by adjusting a single tuning parameter[59]. Furthermore as the source is a resonant device, high fields are present in the unloaded resonator and the plasma discharge can start spontaneously under a wide range of operating pressures. This means that quantum well structures, which require rapid cycling of the active nitrogen flux can be grown simply by turning the plasma source on and off. This eliminates the need for shutters, a carrier gas or growth interruptions  120  Figure E.1: Schematic diagram of the rf plasma source designed and built at UBC used to produce active nitrogen species for GaNx As1-x growth. Shown in the inset is the output baffle used to reduce ion current from plasma to substrate. this figure originally was presented in J. Vac. Sci. Technol. A 25, 850 (2007)[26]  for the growth of GaNxAs1-x . This ability to quickly turn on and off the plasma is one of the major advantages of this set-up over other commercially available plasma sources. High purity N2 gas is introduced through a leak valve into the PBN discharge tube inside the resonator. The 5 9’s grade N2 gas is passed through an in-line purifier (SGT SuperClean filters, triple filter part no. F0301). A PBN plug with a small hole was placed at the back end of the discharge tube. This constriction prevents the plasma from extending back into the stainless steel gas feed tube. From the conductance of the baffle on the end of the discharge tube we estimate the pressure in the discharge tube to be about 0.03 mbar under typical growth conditions, with a background nitrogen pressure of 2x10−6 mbar in the growth chamber. The flux is expected to depend on the cosine of the angle from the axis of the source. In some commercial plasma sources the exit aperture consists of an array of small holes oriented so as to maximise the uniformity of the flux at the substrate. The small holes provide a line of sight to the plasma which may result in a higher flux of energetic ions than in the baffled source. Shown in Fog. E.2 is the forward and reflected rf power (PF and PR) are monitored with a directional watt-meter and the difference is the net power to the load, PL = P F − P R[26]. Large evenly spaced oscillations are observed in the reflected power as a function of frequency due to standing waves in the coaxial line between the generator and the resonator due to impedance mismatch. From measurements of the frequency and power required to start and maintain the plasma, the helix resonance is believed to be at 149 MHz with the plasma on. This mode of operation is denoted as on-resonance, at this frequency the reflected power is minimised and the forward power is maximised. For on-resonance operation, it is believed that large longitudinal field inside the discharge tube. Away from the helix resonance, the 121  Figure E.2: Frequency dependence of the forward (△), reflected ( ) and power to the plasma ( ) of the rf plasma nitrogen source. The vertical lines indicated the two modes of operation used: on-resonance at 147 MHz and off-resonance at 187 MHz. this figure originally was presented in J. Vac. Sci. Technol. A 25, 850 (2007)[26]  highest net power to the load PL occurs at 187 MHz. This defines another high power operating mode, which we refer to as off-resonance. The resonances in the transmission line coupling the generator to the plasma source make it possible to optimise coupling to the plasma source at a frequency that does not match the helix resonance so that the helix can also be driven efficiently at a non-resonant frequency. It is speculated that the electric field distribution in the discharge for the non-resonant operation is different than at the helix resonance. For example the electric field may be primarily a circulating (transverse) electric field in the off-resonance condition whereas on-resonance the field is primarily longitudinal. Operation of the plasma source under the two mode results in different ratios of the constituent species: active atomic N and excited dimers, N2∗ . This was determined from the spectrum of the plasma source as shown in Fig. E.3. The atomic emission lines are stronger in the case of on-resonance operation.  122  Figure E.3: Frequency spectra for operation of the plasma source in on-resonance and off-resonance mode. On-resonance operation shows stronger emission lines from atomic nitrogen. this figure originally was presented in J. Vac. Sci. Technol. A 25, 850 (2007)[26]  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0071670/manifest

Comment

Related Items