@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix skos: . vivo:departmentOrSchool "Science, Faculty of"@en, "Physics and Astronomy, Department of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Lewis, Ryan B."@en ; dcterms:issued "2014-04-17T13:58:21Z"@en, "2014"@en ; vivo:relatedDegree "Doctor of Philosophy - PhD"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description """In this thesis, molecular beam epitaxy (MBE) technology and the MBE growth of GaAsBi are investigated. MBE is a non-equilibrium technique whereby precisely controlled molecular beams are deposited onto a heated substrate at temperatures much lower than for equilibrium growth techniques. A novel closed-cycle cooling setup is implemented to replace liquid nitrogen (LN₂) cooling of the MBE cryo-shroud. The temperature dependence of cryopanel pumping is explored, and GaAs and AlGaAs layers grown using the new cooling setup and with LN₂ cooling of the shroud are characterized. Strong AlGaAs photoluminescence and low impurity concentrations indicate closed-cycle cooling is a promising cost-saving technique for MBE. The relatively unexplored III-V-Bi family of alloys is an exciting frontier of III-V semiconductor alloy exploration. The GaAsBi alloy exhibits many novel properties, including an unparalleled bandgap reduction per change in the size of the crystal lattice, presenting a wide range of potential device applications. A systematic study of the dependence of Bi incorporation on MBE growth conditions is presented. Bi incorporation is found to rapidly increase as the As₂:Ga flux ratio is lowered to 0.5 and saturate for lower flux ratios. This indicates Bi incorporation is sensitive to the surface stoichiometry. A GaAsBi growth model is proposed where Bi from a wetting layer incorporates on surface sites which are terminated by Ga. Low growth temperatures are required as the weak Bi-Ga incorporation bond can be broken thermally, ejecting Bi back to the wetting layer. GaAsBi layers with up to 21.8% Bi, record Bi-content, were grown at temperatures as low as 200C. These layers have up to 2.6% mismatch from the GaAs substrates and show unusually large critical thicknesses for relaxation, a result of the low growth temperature. Optical absorption measurements on pseudomorphic GaAsBi layers with up to 18.7% Bi show the bandgap decreases strongly with increasing Bi-content, reaching 0.5 eV at 18.7% Bi. Si-doped n-GaAsBi layers with up to 4% Bi show the concentration of acceptor states increases rapidly with increasing Bi-content. The acceptor concentration is equal to that of closed Bi3 clusters, suggesting they are the source of deep acceptor states in GaAsBi."""@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/46478?expand=metadata"@en ; skos:note "Molecular Beam Epitaxy Growth Technology and Properties of GaAsBi Alloys by Ryan B. Lewis B.Sc., Dalhousie University, 2006 M.A.Sc., The University of British Columbia, 2008 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Physics) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) April 2014 © Ryan B. Lewis, 2014 ii Abstract In this thesis, molecular beam epitaxy (MBE) technology and the MBE growth of GaAsBi are investigated. MBE is a non-equilibrium technique whereby precisely controlled molecular beams are deposited onto a heated substrate at temperatures much lower than for equilibrium growth techniques. A novel closed-cycle cooling setup is implemented to replace liquid nitrogen (LN2) cooling of the MBE cryo-shroud. The temperature dependence of cryopanel pumping is explored, and GaAs and AlGaAs layers grown using the new cooling setup and with LN2 cooling of the shroud are characterized. Strong AlGaAs photoluminescence and low impurity concentrations indicate closed-cycle cooling is a promising cost-saving technique for MBE. The relatively unexplored III-V-Bi family of alloys is an exciting frontier of III-V semiconductor alloy exploration. The GaAsBi alloy exhibits many novel properties, including an unparalleled bandgap reduction per change in the size of the crystal lattice, presenting a wide range of potential device applications. A systematic study of the dependence of Bi incorporation on MBE growth conditions is presented. Bi incorporation is found to rapidly increase as the As2:Ga flux ratio is lowered to 0.5 and saturate for lower flux ratios. This indicates Bi incorporation is sensitive to the surface stoichiometry. A GaAsBi growth model is proposed where Bi from a wetting layer incorporates on surface sites which are terminated by Ga. Low growth temperatures are required as the weak Bi-Ga incorporation bond can be broken thermally, ejecting Bi back to the wetting layer. GaAsBi layers with up to 21.8% Bi, record Bi-content, were grown at temperatures as low as 200C. These layers have up to 2.6% mismatch from the GaAs substrates and show unusually large critical thicknesses for relaxation, a result of the low growth temperature. Optical absorption measurements on pseudomorphic GaAsBi layers with up to 18.7% Bi show the bandgap decreases strongly with increasing Bi-content, reaching 0.5 eV at 18.7% Bi. Si-doped n-GaAsBi layers with up to 4% Bi show the concentration of acceptor states increases rapidly with increasing Bi-content. The acceptor concentration is equal to that of closed Bi3 clusters, suggesting they are the source of deep acceptor states in GaAsBi. iii Preface The third chapter is the first of the research chapters. For this chapter I was the lead investigator, responsible for most of the experimental design, most of the samples growth, residual gas analyzer data collection, much of the photoluminescence (PL) data collection and the majority of the analysis. T. Tiedje was the research supervisor and responsible for the early development of the experiment concept. J. A. Mackenzie had a major role in the initial experimental setup and D. A. Beaton was involved in the initial experimental setup and sample growth. V. Bahrami-Yekta carried out growths, Hall transport measurements and PL. M. Masnadi-Shirazi also performed Hall transport measurements. K. P. Watkins and P. M. Mooney carried out the C-V and deep level capacitance spectroscopy measurements. M. J. Patel conducted some PL experiments. A portion of this work has been published as R. B. Lewis, J. A. Mackenzie, T. Tiedje, D. A. Beaton, M. Masnadi-Shirazi, V. Bahrami-Yekta, K. P. Watkins and P. M. Mooney. Closed cycle chiller as a low cost alternative to liquid nitrogen in molecular beam epitaxy. J. Vac. Sci. Technol. B, 31:03C116, 2013. I was the leading contributor to the manuscript composition. In addition, a second paper containing a version of the remaining material has published as R. B. Lewis, V. Bahrami-Yekta, M. J. Patel, T. Tiedje and M. Masnadi-Shirazi. Closed-cycle cooling of cryopanels in molecular beam epitaxy. J. Vac. Sci. Technol. B, 32:02C102, 2014. I was responsible for the manuscript composition. A version of chapter 4 has been published as R. B. Lewis, M. Masnadi-Shirazi and T. Tiedje. Growth of high Bi concentration GaAs1-xBix by molecular beam epitaxy. Appl. Phys. Lett., 101(8):082112, 2012. I was the lead investigator for this work, responsible for the experimental design, data collection, analysis and manuscript composition. M. Masnadi-Shirazi assisted in the data collection and T. Tiedje was the research supervisor. For chapter 5, I was the lead investigator, responsible for the majority of the experimental design, growth, data collection and analysis. T. Tiedje was the research supervisor. M. Chicoine performed Rutherford backscattering experiments on my samples. Scanning electron microscopy images were recorded by M. J. Fryer, M. Masnadi-Shirazi and myself. iv For the work described in section 6.1, I prepared the samples, was involved in data analysis and had a leading role in the experimental design. M. Masnadi-Shirazi was the lead investigator, performing the data collection and most of the analysis. In section 6.2, I was the lead investigator, responsible for the experimental design, sample growth, data analysis and some data collection. Hall transport devices were prepared and measured by V. Bahrami-Yekta and M. Masnadi-Shirazi. Capacitance-voltage (C-V) measurements were performed by K. P. Watkins, Zenan Jiang and P. M. Mooney. This thesis focuses on research that was lead by me while I was a PhD student. During this time I also collaborated on several other projects and coauthored other publications that are not mentioned in this preface. These works, which are not the focus of this thesis, are cited in the text when referred to. v Table of Contents Abstract .................................................................................................................................... ii Preface ..................................................................................................................................... iii Table of Contents .................................................................................................................... v List of Tables ......................................................................................................................... vii List of Figures ....................................................................................................................... viii Glossary ............................................................................................................................... xvii Acknowledgements ............................................................................................................ xviii 1 Introduction ....................................................................................................................... 1 1.1 Compound semiconductors for optoelectronic devices ........................................................ 1 1.2 Heterostructure growth techniques ....................................................................................... 4 1.3 Early investigations of III-V-Bi semiconductor alloys ......................................................... 7 1.4 Properties of GaAsBi alloys ................................................................................................ 10 1.5 Molecular beam epitaxy growth of GaAsBi ....................................................................... 15 1.6 The scope of this research ................................................................................................... 17 2 Experimental methods .................................................................................................... 19 2.1 Molecular beam epitaxy growth.......................................................................................... 19 2.1.1 Cooling the MBE cryo-shroud ........................................................................................ 21 2.2 In situ monitoring techniques .............................................................................................. 23 2.2.1 Substrate temperature measurement ............................................................................... 23 2.2.2 Reflection high energy electron diffraction .................................................................... 25 2.2.3 Elastic light scattering..................................................................................................... 29 2.2.4 Beam flux measurement ................................................................................................. 30 2.3 Ex situ characterization ....................................................................................................... 32 2.3.1 High resolution x-ray diffraction .................................................................................... 32 2.3.2 Rutherford backscattering spectroscopy ......................................................................... 36 2.3.3 Scanning electron microscopy ........................................................................................ 36 2.3.4 Optical absorption ........................................................................................................... 37 2.3.5 Electrical transport .......................................................................................................... 38 vi 3 Closed-cycle cooling of the MBE cryo-shroud: a low cost alternative to liquid nitrogen .................................................................................................................................. 39 3.1 Surface adsorption model for H2O ...................................................................................... 42 3.2 Effect of cryopanel temperature on residual gas pressures ................................................. 46 3.3 Effect of shroud temperature on the properties of GaAs layers .......................................... 55 3.4 Properties of AlGaAs layers grown with the closed-cycle cooled shroud .......................... 59 4 The dependence of Bi incorporation in GaAsBi on MBE growth conditions ........... 62 4.1 Bi wetting layer coverage on GaAs .................................................................................... 62 4.2 Lu et al.’s GaAsBi growth study ......................................................................................... 65 4.3 The dependence of Bi-content on MBE growth conditions ................................................ 69 4.4 A new GaAsBi growth model ............................................................................................. 73 4.4.1 The stoichiometry of GaAs surfaces ............................................................................... 75 4.4.2 As and Ga hopping on GaAsBi surfaces ........................................................................ 78 4.4.3 Comparing the GaAsBi growth model with experiment ................................................ 79 4.4.4 Making use of the growth model .................................................................................... 82 5 Structural characterization of GaAsBi layers .............................................................. 83 5.1 Characterization with high-resolution x-ray diffraction...................................................... 83 5.2 Strain relaxation .................................................................................................................. 86 5.2.1 Formation of dislocations ............................................................................................... 86 5.2.2 Observation of strain relaxation in GaAsBi layers ......................................................... 88 5.3 Rutherford backscattering spectroscopy ............................................................................. 93 5.4 Scanning electron microscopy ............................................................................................ 96 5.5 Growth of GaAsBi on InP and GaSb substrates ............................................................... 100 6 Optical and electronic properties of GaAsBi layers .................................................. 105 6.1 The composition dependence of the GaAsBi bandgap ..................................................... 105 6.1.1 Optical absorption spectroscopy ................................................................................... 107 6.2 Electrical properties of n-GaAsBi layers .......................................................................... 112 6.2.1 Compensation of free electrons in n-GaAsBi layers .................................................... 113 6.2.2 Electron Hall mobility in n-GaAsBi ............................................................................. 119 7 Conclusions .................................................................................................................... 122 7.1 Future work ....................................................................................................................... 125 Bibliography ........................................................................................................................ 127 vii List of Tables Table 3.1 Calculated residence times for some binding energies at room temperature, -80C (lowest achievable temperature with our chiller) and -196C (boiling point of LN2). The temperature for a residence time of 104 s is also indicated.......................................................................................................................... 45 Table 3.2 Pumping speeds of active gases relative to the combined speed of the cryopump and ion pump for the respective gases ( , and ) for different cryopanel temperatures. The “total chiller” speed is the pumping speed when the system is configured for operation with the chiller (combined speed of the cryopump, ion pump, -196C TSP and shroud at -78C). The “total LN2” is for the shroud at -196C instead of -78C. The “*” indicates the pumping speed of the shroud at -196C was predicted by multiplying the -196C TSP pumping speeds by the ratio of the H2O pumping speeds for the shroud and TSP at -78C. ............................................................................... 54 Table 3.3 Concentrations of all deep levels seen in p-GaAs and n-GaAs samples grown with the shroud cooled with LN2 and the closed cycle chiller at -70C as determined by DLTS....................................................................................... 57 Table 5.1 Summary of layer properties for the samples from Fig 5.6. “Etch” denotes samples where droplets were removed by an HCl:H2O etch. Thicknesses were determined from pendellösung fringes in (004) HRXRD scans (“**” indicates thickness was determined from growth rate). Samples with “*” were found with RBS to contain multiple layers of varying Bi content. In these cases the composition of the thickest layer was taken. .................................................. 95 Table 5.2 The lattice parameter of some III-V substrates, showing the Bi-content where GaAsBi is lattice matched and the mismatch to GaAs. ................................ 101 viii List of Figures Figure 1.1 Bandgap as a function of lattice constant for many pure group III-V and IV semiconductors (open points) and alloys thereof. II-VI semiconductors are shown as solid points. Figure reproduced from Veeco Instruments Inc. with permission [9]. .................................................................................................. 3 Figure 1.2 Bandgap as a function of lattice constant for bandgap lowering GaAs-based ternary alloys. Solid lines have been verified with experimental data. Broken lines correspond to theoretical results for GaAsBi and TlGaAs and extrapolation of the data fit for GaNAs. Data sources are as follows: GaNAs [32], GaAsBi [46.47], TlGaAs [44.45], InGaAs [48] and GaAsSb [49] ........ 11 Figure 1.3 Bandgap as a function of lattice mismatch to GaAs for ternary alloys with In, Sb, N and Bi. Data sources are as follows: GaNAs [32], GaAsBi [47], InGaAs [48] and GaAsSb [49].. ................................................................................... 12 Figure 1.4 Schematic of the bandstructures of GaNAs and GaAsBi illustrating the conduction (CB), heavy hole (HH), light hole (LH) and split off (SO) bands. The resonance of the N 2s and Bi 6p orbitals with the band edges is indicated as well as the position of localized states. Figure originally published in [53]......................................................................................................................... 13 Figure 2.1 An illustration of the intersection of the Ewald sphere with reciprocal lattice rods of finite width, resulting in elongated streaks in the RHEED pattern. The two diffracted beams for each rod indicate the angular width of the diffracted beams. ............................................................................................................. 26 Figure 2.2 Images of a (21) RHEED reconstruction. The diffraction direction is indicated on the images and the spacing between the lines on the right hand image corresponds to the bulk periodicity. Figure originally published in [79]......................................................................................................................... 27 ix Figure 2.3 Surface reconstruction maps of a) GaAs and b) GaAsBi. The growing GaAsBi surface has an incident Bi beam equivalent pressure (BEP) of 3×10-9 Torr. Surface reconstructions are indicated for the electron beam oriented along [110] and azimuths. The incident Ga BEP was kept constant at 8×10-8 Torr during the experiment, corresponding to a growth rate of 0.1 μm/hr. Figure originally published in [79] ................................................................. 28 Figure 2.4 An illustration of the light scattering setup on the MBE. The detector angle is not representative of the actual configuration used in this work. Figure originally published in [91] ............................................................................. 30 Figure 2.5 A diagram illustrating the geometry of a (004) θ-2θ scan on a sample consisting of a (001) substrate with a compressively strained epilayer. The real space (left) and reciprocal space (right) images are shown ............................ 33 Figure 2.6 An illustration of fully strained (left) and fully relaxed (right) GaAsBi unit cells on a GaAs (001) substrate. ..................................................................... 35 Figure 3.1 A plot illustrating an idealized distribution of H2O desorption sites as a function of desorption energy for a cooled metal surface. Non-pumping, pumping and filled sites are indicated. The filled sites result from the surface being in a steady state with the background H2O partial pressure before cooling. Figure originally published in [112] ................................................. 44 Figure 3.2 Partial pressures of mass 12, 14, 18, 28 and 44 as a function of time during the cooling of the TSP reservoir with LN2 (-196C). The filling of the reservoir began at t = 0 and the two dips in the mass 28 u and 14 u signals are a result of the reservoir being filled in two stages. The shroud was maintained at +70C, with the Ga and As cells at operating temperature (Ga at 921C and As at 345C with the As-cracker at ~1000C). The substrate and all other cells were at 300C. Figure originally published in [112] ...................................... 48 x Figure 3.3 Partial pressures of mass 18, 28 and 44 as a function of time during the cooling of the TSP reservoir to -78C with a dry-ice/ethanol slurry. Filling of the reservoir commenced at t = 0. The shroud was maintained at +70C, with the Ga and As cells at operating temperature (Ga at 921C and As at 345C with the As-cracker at ~1000C). The substrate and all other cells were at 300C. It took about 30 min to completely fill the reservoir. Figure originally published in [112] ........................................................................................... 50 Figure 3.4 Partial pressures as a function of time while the shroud is cooled in steps of 20C from +20C to the lowest achievable temperature of -78C and then warmed back to +20C. The steps in the H2O partial pressure correspond to changes in the shroud temperature. For these experiments, the TSP reservoir was empty and the Ga and As cells were at operating temperature (Ga at 921C and As at 345C with the As-cracker at ~1000C). The substrate and all other cells were at 300C. Figure originally published in [112] ................ 52 Figure 3.5 H2O partial pressure as a function of shroud temperature for the cooling of the shroud. The initial shroud temperature was between +40C and +70C. Figure originally published in [112] ........................................................................... 53 Figure 3.6 Room temperature photoluminescence spectra for n and p-type GaAs samples grown under similar growth conditions with LN2-cooling (-196C) and closed-cycle chiller cooling (-70C) of the shroud. Sample numbers for p- and n-GaAs are r2229 and r2230 for samples grown with LN2, respectively, and r2233 and r2234 for samples grown with the chiller, respectively. Figure originally published in [113] ........................................................................... 58 Figure 3.7 Room temperature PL spectra of AlGaAs layers on GaAs grown with closed-cycle cooling of the shroud. Each spectrum shows emission from the AlGaAs layer and the GaAs buffer. The Al0.49Ga0.51As layer is 550 nm thick and the other layers are ~920 nm thick. The PL measurement conditions are the same for all the samples. Sample numbers are r2436, r2413 and r2442 for the xi samples grown at 555C, 580C and 700C respectively. Figure originally published in [112] ........................................................................................... 60 Figure 4.1 a) Bi coverage (inferred from RHEED measurements) as a function of substrate temperature. The surface was exposed to a Bi BEP of 1.4 10-5 Torr. b) Bi coverage as a function of Bi flux at a substrate temperature of 460C. Here the Bi source temperature was ramped linearly. The coverage from the Langmuir model is plotted on a) and b). Figure originally published in [74]. 64 Figure 4.2 Bi-content as a function of the Bi:As flux ratio, showing experimental data (points) and Lu’s incorporation model (curves). The solid triangles correspond to varying only the Bi flux, while the solid circles and squares vary only the As flux (for two separate Bi fluxes, respectively). For the hollow points, only the substrate temperature was changed. The solid lines are model curves for varying the As flux (with the highest Bi:As ratio of each curve corresponding to the 1:1 As:Ga flux ratio). The broken lines are model curves for varying the Bi flux at different substrate temperatures. The inset shows an illustration of Lu’s incorporation model, illustrating the three processes which are proposed to affect Bi incorporation, as discussed in the text. Figure originally published in [76]. ............................................................................................................. 67 Figure 4.3 Bi-content as a function of As2:Ga BEPR, for samples grown with large Bi fluxes that are expected to result in Bi-saturated surfaces (data points). The curves correspond to the model discussed in the next section and are plotted as a function of the flux ratio on the top scale. The kink in the data at a BEPR of ~2.25 corresponds to an As:Ga atomic ratio of unity. A temperature of 225C was used to draw the curve corresponding to the 220-230C data. Figure originally published in [77] ............................................................................. 70 Figure 4.4 The temperature dependence of Bi-content for samples grown with As2:Ga flux ratios <0.5. Bi:Ga BEPRs were 0.59 ± 0.06 for the solid data points, 0.09 for the open triangles and 6.5 for the open circle. The solid line corresponds to xii the model discussed in the next section, calculated for the growth conditions of the solid-circle points. Figure originally published in [77]. ....................... 71 Figure 4.5 Bi:Ga BEPR dependence of Bi-content for samples grown at 330oC and 1.0 μm/h, with As2:Ga BEPRs between 2.5 and 3.3 (As2:Ga flux ratio between 0.56 and 0.73). The sample corresponding to the open circle has droplets on the surface, while the other samples do not. The curves correspond to the model discussed in the next section. The black curves are calculations of Bi-content as a function of the Bi:Ga flux ratio (indicated on the top scale) for various As2:Ga flux ratios . The blue curves are model calculations of the Bi surface coverage (right hand scale) as a function of Bi:Ga flux ratio ............ 73 Figure 4.6 A model of the semiconductor surface and Bi wetting layer illustrating: 1) incorporation of Bi on a Ga site, 2) thermal ejection of incorporated Bi. ...... 74 Figure 4.7 The calculated Ga surface coverage as a function of the As2:Ga flux ratio for several choices of PAs and PGa. ....................................................................... 77 Figure 5.1 (004) -2 HRXRD scans (solid lines) and dynamical simulations (broken lines) of three GaAsBi films on GaAs. The Bi-contents are 5.8%, 14.2% and 21.8%, and pendellösung fringes indicate thicknesses of 54 nm, 53 nm and 17 nm respectively. Sample numbers are r2313, r2361 and r2267 ...................... 84 Figure 5.2 An RSM of the (224) film and substrate peaks for the 53 nm thick GaAs0.858Bi0.142 sample from Fig. 5.1 (r2361). The upper peak corresponds to the GaAs substrate and the lower peak is the GaAsBi layer. The solid yellow line points to the origin, indicating the line where a 100% relaxed film would lie. The green contours are spaced by factors of 10 in intensity. .................... 85 Figure 5.3 RSMs of the and film and substrate peaks from a 75 nm GaAs0.86Bi0.14 sample (r2345). The sample has an asymmetry in the strain relaxation, with 11% in the direction and 71% in the direction. A film with 100% relaxation would lie on the solid yellow line. ................... 89 xiii Figure 5.4 A plot of film thickness as a function of mismatch to GaAs and Bi-content, for samples where relaxation was (squares) and was not (crosses) detected with HRXRD. Growth temperatures were between 210C and 330C for these samples. The MB critical thickness is also plotted, as are critical thickness values estimated from the work by France et al. for samples grown at 350C (circles) [142] .................................................................................................. 91 Figure 5.5 Diffusely scattered light signal from a rotating wafer, collected during the growth of a p+/n GaAsBi structure (r2344). Near t = 0 the substrate temperature was rapidly increased from 330C to 550C. The appearance of spikes corresponds to the formation of a crosshatch pattern on the wafer surface. ............................................................................................................ 92 Figure 5.6 Strained out-of-plane lattice parameter (blue squares) and corresponding relaxed lattice parameter (black circles) as a function of the RBS Bi-content for GaAsBi films on GaAs. The GaBi lattice parameter was determined from the best fit line assuming Vegard’s law and that Poission’s ratio for GaAsBi is 0.31. RBS measurements were performed by M. Chicoine at Université de Montréal .......................................................................................................... 94 Figure 5.7 RBS signal and SIMNRA simulations for two GaAsBi/GaAs samples. The peak on the right side of each figure corresponds to backscattering from Bi atoms in the GaAsBi layer. The large step to the left is from Ga and As in the layer and substrate. Measurements performed by M. Chicoine at Université de Montréal. ......................................................................................................... 96 Figure 5.8 An EDS image of a bimetallic droplet of Ga (red/top) and Bi (blue/bottom) on a GaAs0.969Bi0.031 surface. The sample was grown with a substrate temperature of 350C under slightly Ga-rich conditions and showed a (21) reconstruction during growth. Map recorded by M. J. Fryer. ................................................. 97 Figure 5.9 An SEM image of bimetallic droplets on a GaAs0.87Bi0.13 surface. The droplet in the centre of the frame is Bi-rich on the top and Ga-rich on the bottom. The xiv sample was grown with a substrate temperature of 250C, low As2:Ga BEPR of 0.85, high Bi:Ga BEPR of 0.50 and showed a (21) RHEED reconstruction during growth .................................................................................................. 98 Figure 5.10 An SEM image of a droplet free GaAs0.942Bi0.05.8 surface. The sample was grown with a substrate temperature of 265C, As2:Ga BEPR of 2.53, Bi:Ga BEPR of 0.29 and showed (21) and (2chevrons) RHEED reconstructions during growth. Image recorded by M. Masnadi-Shirazi ................................. 99 Figure 5.11 (004) -2 HRXRD scans for three GaAsBi layers grown on InP substrates. The sample numbers are r2271 (215C), r2272 (245C) and r2258 (260C). The sample grown at 215C has ~70% relaxation. The composition range of the other films is determined from the (004) scans, assuming the relaxation lies between 100% and 70% ......................................................................... 102 Figure 5.12 An SEM image of the GaAs0.85Bi0.15/InP sample grown at 215C (r2271). The dark parts of the droplets are Bi and the light parts are Ga. Image recorded by M. Masnadi-Shirazi....................................................................................... 103 Figure 6.1 The GaAsBi bandgap as a function of Bi-content and lattice parameter from density functional theory at 0 K [46] and room-temperature PL [47] .......... 107 Figure 6.2 Normalized transmission spectra for GaAsBi/GaAs samples recorded with a) a Ge-detector and b) a PbS detector. The steep rise at ~870 nm corresponds to the GaAs absorption edge. The deviation from the GaAs signal is due to absorption in the GaAsBi layers. Spectra recorded by M. Masnadi-Shirazi 109 Figure 6.3 A plot of the square of the absorption coefficient as a function of energy for several GaAsBi layers. The bandgap was obtained by fitting linear functions between 2 = 2106 cm-2 and 2 = 108 cm-2 and then extrapolating the best fit curves to zero absorption. Data and fits by M. Masnadi-Shirazi .................. 111 xv Figure 6.4 A plot of the composition and lattice parameter dependence of the GaAsBi bandgap. The absorption results are shown as well as a fit to the absorption data. PL data from Lu et al. [47] and Janotti’s DFT calculation [46] (shifted to match the room temperature GaAs bandgap) are shown for comparison. .... 112 Figure 6.5 Calculated Bi-cluster concentrations assuming Bi randomly occupies group V lattice sites [159]. The open and closed Bi3 concentrations are the sum of multiple configurations ................................................................................. 114 Figure 6.6 A plot of the expected Si-dopant concentration as a function of Bi-content for Si-doped GaAs and GaAsBi layers, indicating whether samples show n-type doping or are resistive/depleted. Whether the samples were doped or restive/depleted was determined from Hall measurements on Si-doped GaAsBi epilayers (closed data points) and from C-V measurements on p+/n devices (open data points). The concentration of closed Bi3 clusters is plotted assuming Bi randomly populates group V lattice sites [159]. ...................... 116 Figure 6.7 Measured As2:Ga BEPRs for resistive/depleted and n-type Si-doped GaAsBi samples. Square date points correspond to C-V measurements on p+/n structures and circle date points correspond to Hall measurements on epilayer samples. The BEPRs of n-type GaAs samples are also shown. All samples were grown with a substrate temperature of 330C. No correlation is found between BEPR and compensation. ............................................................... 117 Figure 6.8 A plot of electron Hall mobility as a function of Bi-content for n-GaAsBi samples. All samples from this work were grown at 330C and 1 μm/h growth rate. Measurements by Kini et al. are shown for comparison [60]. These samples were grown with a substrate temperature of 380C at 2 μm/h. ....... 119 Figure 6.9 The dependence of electron Hall mobility on the Si-dopant concentration for n-GaAsBi and n-GaAs layers. The open square n-GaAs samples were grown at ~550C while all other samples from this work were grown at 330C. For the work of Kini et al. [60], samples were grown at 380C and compensation xvi was neglected. The broken line corresponds to the carrier-concentration dependence of n-GaAs electron mobility [163] ............................................ 121 xvii Glossary AFM atomic force microscopy ALE atomic layer epitaxy BEP beam equivalent pressure BEPR beam equivalent pressure ratio BFM beam flux monitor C-V capacitance-voltage DFT density functional theory DLTS deep level transient spectroscopy DRS diffuse reflectance spectroscopy EDS energy dispersive x-ray spectroscopy HBT heterojunction bipolar transistor HRXRD high resolution x-ray diffraction IR infrared LN2 liquid nitrogen LPE liquid phase epitaxy LS light scattering LT low-temperature grown MB Matthews-Blakeslee MBE molecular beam epitaxy MOVPE metalorganic vapor phase epitaxy PBN pyrolytic boron nitride PL photoluminescence PMT photomultiplier tube RBS Rutherford backscattering spectroscopy RGA residual gas analyzer RHEED reflection high energy electron diffraction RSM reciprocal space map SEM scanning electron microscopy SFU Simon Fraser University SI semi-insulating SIMS secoindary ion mass spectroscopy TEM transmission electron microscopy TSP titanium sublimation pump UBC University of British Columbia UHV ultra high vacuum UVic. University of Victoria VBM valence band maximum xviii Acknowledgements Thanks to my supervisor for sharing his wisdom and distinguishing logic. Thanks to my lab mates for their help and friendship. Thanks to my parents for giving me the confidence and support to follow my curiosity. Thanks to Dayna for the harmony that you bring to my life. 1 Chapter 1 Introduction Developments in the field of compound semiconductors over the last 50 years have had a profound impact on human life. Most notably, devices based on these materials have revolutionized telecommunication, where infrared lasers and detectors are the workhorses that power the internet and radio-frequency amplifiers drive wireless communication, allowing for Wi-Fi, smart phones, GPS and many other devices. In addition, compound semiconductor lasers and detectors are used for optical data storage (CD/DVD/Blu-ray) and diode lasers are used for pointers and pumping solid-state lasers, which are used in medicine and manufacturing, amongst many other applications. Solid state lighting with light emitting diodes is reducing our energy consumption and solar cells, based on compound semiconductors, are poised to play an important role in the production of clean energy. These are some of the most notable applications of compound semiconductors but there are many others, some that are commercially available now and others that have yet to be realized. Each application requires a specific set of material properties, which motivates the aggressive research taking place in this field today. In addition to developing and improving existing device concepts, these materials are used in cutting-edge experiments that give new insight into condensed matter physics, providing the seed for the continued growth of new technology. 1.1 Compound semiconductors for optoelectronic devices In late 1962, in a tremendous achievement, the first semiconductor laser diodes were realized by four independent research groups [1-4]. The devices were very primitive, consisting of p-n junctions made by diffusing p-type dopants into bulk chunks of n-doped GaAs and GaAsP. The laser cavity was created by the cleaved facets of the chunk and lasing was only obtained with large current pulses with the devices submersed in liquid nitrogen. With high optical losses and no carrier confinement, there was much room for improvement. 2 Microfabrication techniques soon made it possible to confine the current laterally to a narrow strip [5], however, carrier confinement in the transverse direction was a more serious problem [6]. The solution, proposed by Kroemer in 1963, was to use a double heterostructure whereby the active region is sandwiched between layers of a larger bandgap [7]. This structure provides transverse carrier confinement and also reduces optical loss, leading to AlGaAs/GaAs double heterostructure lasers operating continuously at room temperature by 1970 [8] . A heterostructure is formed when two or more dissimilar materials are joined together, with the joining interface(s) known as the heterojunction(s). The ability to produce layered structures, where material properties such as bandgap, refractive index, band offsets and transport properties are changed between layers, allows for huge flexibility in device designs. A plot of bandgap as a function of lattice constant for a number of III-V binary semiconductors and ternary alloys thereof, as well as group IV and II-VI semiconductors, is shown in Fig. 1.1. Conveniently most III-V semiconductors have the same zincblende crystal structure, and alloying generally results in a smooth change in material properties with changing composition. Although all wavelengths from ultraviolet (UV) to infrared (IR) can be reached by forming alloys, the variations in lattice constants significantly limit the available bandgaps that can be incorporated into heterostructures, without incurring serious degradation of the material quality. The AlGaAs/GaAs system is almost ideal for making heterostructures. AlAs has less than 0.14% mismatch with GaAs, allowing for thick AlGaAs layers to be grown on GaAs substrates without the formation of strain relaxing dislocations. The large AlAs bandgap of 2.15 eV means AlGaAs layers can be grown with bandgaps between 1.42 and 2.15 eV, though above about 40 % Al the bandgap becomes indirect, which is not desired. This material system has made GaAs a very attractive substrate for optoelectronic devices operating even outside this bandgap range. For these devices, AlAs/GaAs stacks are often used for distributed Bragg reflectors in surface emitting lasers and AlGaAs layers are used for carrier confinement. 3 Figure. 1.1. Bandgap as a function of lattice constant for many pure group III-V and IV semiconductors (open points) and alloys thereof. II-VI semiconductors are shown as solid points. Figure reproduced from Veeco Instruments Inc. with permission [9]. Unlike AlGaAs, for most ternary III-V alloys, the lattice mismatch is severely restrictive. The addition of a fourth alloying element can provide increased flexibility, but at the expense of a more complicated growth process. For example, the InGaP alloy is lattice matched to GaAs only at one composition and bandgap. By adding another element, Al, the quaternary AlGaInP alloy can access a limited wavelength range lattice matched to GaAs. AlGaInP devices are commercially available at 670 nm [6]. Similarly, InGaAsP alloys on the larger InP substrate are used to make lasers for the telecommunications industry, which operate in the 1.3-1.55 μm range. Even still, gaps in the spectrum exist and there is much room for new alloys with better properties to improve existing devices. Transitions between the conduction and valence bands in direct bandgap III-V semiconductors are not the only way of making optoelectronic devices. Indirect bandgap group IV elements Ge and Si are popular detector materials for the near IR. Furthermore, the direct bandgap II-VI alloy HgCdTe has been the most successful detector material in the mid-to-long wavelength IR. Superlattices (periodic structures) can also be used to make 4 optoelectronic devices. In this case, intersubband transitions can allow devices to operate at wavelengths that cannot be easily reached with interband devices. Each device technology has its own unique set of challenges. The focus of this thesis is on direct bandgap III-V alloys. 1.2 Heterostructure growth techniques Highly crystalline layers with low densities of defects and impurities are typically required for good semiconductor device performance. Heterostructure growth techniques must facilitate the deposition of single crystal layers where each subsequent layer is arranged with the same crystal structure and orientation as the preceding layer, a process known as epitaxy. The influence of the substrate crystal structure can allow for the growth of highly strained layers and alloys with crystal structures that do not exist in bulk form. The growth techniques described below typically take place well below the melting point of the bulk materials. For example, molecular beam epitaxy (MBE) growth of GaAs is typically done at approximately half the bulk melting point or lower. This low growth temperature minimizes the number of thermodynamic defects in the grown films and allows for the growth of metastable alloys and structures that are not achievable with equilibrium growth techniques. The earliest method for growing epitaxial semiconductor heterostructures was liquid phase epitaxy (LPE), which emerged in 1963 from RCA laboratories [10]. Then in 1968, the two technologies that would come to dominate the semiconductor heterostructure growth field emerged: molecular beam epitaxy (MBE) [11, 12] and metalorganic vapor phase epitaxy (MOVPE) [13]. These growth techniques are briefly described and compared in this section. When instructive, the growth of GaAs is used as an example. Liquid phase epitaxy: A seed crystal or substrate is brought into contact with a supersaturated liquid, resulting in the epitaxial growth of material on the surface of the substrate. In the case of GaAs growth, the supersaturated liquid is prepared by first heating Ga metal in 5 contact with a piece of GaAs. Some GaAs dissolves into the solution until the concentration of dissolved As reaches saturation. The solution is slightly cooled, producing a Ga-rich liquid that is slightly supersaturated with dissolved As. The substrate, which is at the same temperature as the liquid (typically 600C-900C), is brought in contact with the liquid, resulting in the deposition of epitaxial layers of GaAs on the surface. The substrate can be dipped in the solution before the solution is cooled to dissolve the surface layer of the substrate, reducing surface contamination and resulting in fewer defects in the grown film. Multiple epitaxial layers can be deposited by sequentially dipping the substrate into different liquid sources. One way to grow multiple layers is the “sliding substrate holder” method, where the substrate is pulled on a track where it comes in contact with a series of saturated liquids [8]. For each layer, the time that it spends in contact with the liquid (pulling speed) as well as the temperature and degree of saturation of the liquid, determine the thickness of the grown layer. Growth rates are typically ~2 μm/h. Advantages: Simple, inexpensive and relatively safe. Disadvantages: Only simple layer structures can be grown, can’t grade composition, relatively poor uniformity and thickness control, no in situ characterization of growth process. Metalorganic vapor phase epitaxy: Elements that are to be incorporated into the film are transported to the substrate in the gas phase as part of organic-metallic molecules. These molecules thermally decompose on the heated substrate, depositing the inorganic element. The remaining organic reactants evaporate from the surface and are evacuated from the system. Ultra-pure precursor gases are required to produce quality films and the growth takes place in a vacuum chamber, with a base pressure of ~10-7 Torr. The reaction gases produce chamber pressures between 0.001 and 1 atm. By varying the precursor gases, complicated heterostructures can be easily grown with about 1 nm precision. Trimethylgallium and arsine precursors are commonly used in GaAs growth as sources of Ga and As. For these precursors, between substrate temperatures of 6 ~600C to 850C the growth rate is determined by the flow rates of the precursor gases. Below ~600C the growth rate begins to rapidly decline with decreasing temperature [14]. This is because there is not enough thermal energy at the substrate to decompose the precursor elements. More exotic precursors, which require less thermal energy to decompose, are required for growth at low temperatures [15]. The native oxide layer on the wafer is typically removed from the surface before the onset of the growth by heating the substrate in the presence of either hydrogen or the As-precursor. MOVPE is the most common growth technique used for the production growth of III-V heterostructures. This is due to a higher throughput due to the ease of producing large systems that can accommodate growth on many wafers in a single run, the higher growth rates and shorter maintenance down-time with MOVPE when compared to MBE [16]. Advantages: High growth rates (typically 2-4 μm/h, but can be >20 μm/h [16]), scalability for production, can easily grow complex structures with abrupt and graded interfaces, many sources possible on one system, high quality layers of uniform composition and thickness, less maintenance down-time than MBE (~1% maintenance down time [16]). Disadvantages: Handling large volumes of toxic gases poses safety/environmental concerns and increases operating costs, less availability of in situ characterization experiments due to high gas pressures, complex chemistry, costly source materials, flexibility limited by available precursor species, low temperature growth restricted by precursor decomposition temperature, interfaces less abrupt than with MBE, hydrogen and carbon in the process environment. Molecular beam epitaxy: Molecular beams are directed towards a heated substrate in an ultra-high vacuum (UHV) chamber with a base pressure of ~10-11 Torr. The beams are generated by high-purity elemental evaporation sources, where the source fluxes are controlled by the cell temperatures and/or valves. Gas sources can also be used either in molecular form or excited as a plasma. Different beams can be turned on and off with less than a 7 monolayer of growth by mechanical shutters in front of each source, allowing for multiple layers with abrupt interfaces to be grown. The low pressures involved ensure that the beams only interact on the substrate surface. In most cases, the growth rate is determined kinetically by the fluxes, normally the group III flux in III-V MBE. The simple chemistry and availability of in situ characterization with reflection high energy electron diffraction (RHEED) has made MBE a very popular growth method for research and development of new materials. MBE is used for production as well, especially when very thin and/or abrupt layers are required. One such example is in pseudomorphic high electron mobility transistors (pHEMTs), which are used as low noise amplifiers in wireless phones. For GaAs growth, elemental Ga is evaporated from one effusion source with a single heating filament. An As2 flux is typically used as a source of As, which is generated by first evaporating As4 from solid As and then flowing the As4 through a valved cracker at ~1000C to produce As2. The native oxide layer is removed from the surface before the growth by heating the substrate to above 600C under an As flux, or can be etched off with the Ga flux or a flux of hydrogen. During growth of GaAs the substrate is typically kept below 600C. Advantages: Very abrupt interfaces and highly doped layers possible, availability of in situ characterization, low growth temperatures possible, chemistry more simple and safe than with MOVPE, source materials cheaper than with MOVPE, can easily grow complex structures (although graded interfaces not as easy as with MOVPE). Disadvantages: Low growth rate (~1 μm/h typical, but can be up to 4 μm/h [16]), number of materials limited by the number of cell ports, More system maintenance (~10% maintenance down time [16]) and multiple wafers more challenging than with MOVPE. 1.3 Early investigations of III-V-Bi semiconductor alloys Bismuth, the heaviest non-radioactive element, has received much less attention than its lighter family members as an alloying element in III-V semiconductors. The same is true for 8 thallium, the heaviest stable group III element. Reports on III-V-Bi alloys first surfaced in 1969, for the growth of the ternary InSbBi alloy [17-19]. The work was motivated by the semi-metal character of InBi, which seemed attractive for extending the InSb wavelength to reach the 8-14 μm atmospheric infrared (IR) window. The alloy was first made using bulk growth techniques, where it was found that incorporation of Bi caused the InSb bandgap to shrink by about 36 meV/% Bi [18]. The solubility of InBi in InSb is very low, only 2.6% [19], which restricted the compositions which could be reached with equilibrium growth techniques like Czochralski. To increase the Bi-content, non-equilibrium growth techniques were employed. Using sputtering, InSbBi films with up to 12 % Bi were grown polycrystalline [20], and then single crystal by sputtering with ion bombardment [21]. Absorption measurements indicated that the semiconductor-semimetal transition (zero bandgap) occurs at 11 % Bi [21] . In the 1980s, MBE and MOVPE were used to grow InSbBi, as well as InAsBi and InAsSbBi [22-25]. Photoluminescence (PL) measurements on the InAsBi and InAsSbBi layers showed a strong bandgap reduction with increasing Bi content of 55 meV/% Bi [25]. For MBE and MOVPE growth, the incorporation of Bi was found to be very sensitive to the V:III flux ratio, the growth temperature and the Bi surface coverage [23, 25]. In the case of MOVPE, the low growth temperatures required for Bi incorporation proved problematic for existing precursor species, leading to searches for precursors that decompose at lower temperatures [15]. With both MBE and MOVPE growth, the formation of InBi and InBi2 droplets presented “major problems” for the growth of these alloys [23, 24, 26]. These droplet issues likely detracted interest from these materials and prevented future exploration of new III-V-Bi alloys. Early investigations into the GaAsBi alloy were motivated by two separate concepts: the possibility of creating a laser with a temperature insensitive bandgap [27], and that co-doping with nitrogen could alleviate some of the detrimental effects of nitrogen-cluster formation in GaNAs [28, 29]. A laser with a temperature insensitive emission wavelength would have profound implications for the telecommunications industry. Oe and Asai predicted that the most challenging part of building such a laser would be finding a material with a temperature insensitive bandgap [30]. They proposed that alloying a semimetal (as many Bi-compounds 9 are) with a semiconductor would result in a material with a less temperature-dependent bandgap, suggesting InGaAsBi as a suitable material. The incorporation of N into GaAs gives rise to bound states that form below the conduction band edge. These states are due to the formation of N-clusters, which evolve into an impurity band as the N concentration is increased [31]. This effect is responsible for the giant 200 meV/% bandgap reduction at small N concentrations [32], but also leads to a dramatic degradation of the electron mobility. Mascarenhas et al. predicted that co-doping GaAs with N and Bi would have a number of positive effects [28]. Just as N-states below the conduction band in GaNAs behave like deep acceptor states, Bi was predicted to produce deep donor states above the valence band. In the case of charge co-doping, incorporating p and n-type dopants as pairs has been shown to reduce the dopant activation energy and increase carrier mobility and dopant solubility [28, 33, 34]. Mascarenhas et al. proposed that if Bi and N atoms incorporate as pairs, the resulting strain and charge perturbations on the lattice would be in the form of short-range dipoles, rather than the long-range monopoles which result from isolated impurities. This could result in improved transport properties relative to GaNAs. GaAsBi was first grown by MOVPE in 1998 in Japan by Oe et al. [27] and then by MBE in 2003 at UBC and in Japan [29, 35]. Like with previously studied Bi-containing III-V alloys, difficulties with Bi incorporation necessitated low growth temperatures and As:Ga flux ratios on the brink of having an As-shortage. Problems with metallic surface droplets were also encountered, as well as limits to the amount of Bi that could be incorporated in to the films. Measurements of the temperature dependence of the bandgap by the Japanese group on samples with a few percent Bi indicated that the temperature coefficient of the bandgap energy (Eg/T) was about 1/3 of that of GaAs [27, 36, 37]. However, other reports have found the temperature dependence to be essentially the same as that of GaAs[38, 39]. The reason for this discrepancy may be related to localized states near the top of the valence band associated with Bi alloying [28, 39]. Co-doping of N and Bi in GaAs was found to produce no enhancement in the electron mobility over N-doping alone [40]. Nevertheless, GaAsBi had presented itself as an intriguing new semiconductor alloy with many exciting features to 10 motivate further investigation. These features are discussed in the following section, followed by a more detailed discussion of the MBE growth of GaAsBi. The success of GaAsBi has renewed interest in Bi-containing semiconductors and more recently new III-V-Bi alloys have been investigated. InGaAsBi on InP is under investigation as a way to extend device wavelengths beyond those achievable with InGaAs on InP [41, 42] and GaSbBi has also recently been grown [43]. In July 2010 the first international workshop on Bi-containing semiconductors was held at the University of Michigan, with annual meetings being held thereafter. 1.4 Properties of GaAsBi alloys Alloying GaAs with Bi results in a huge reduction of the GaAs bandgap for only a small amount of Bi incorporation, ~83 meV/% for Bi concentrations up to a few percent [38]. Figure 1.2 shows the bandgap as a function of lattice constant for bandgap-reducing GaAs-based ternary alloys: InGaAs, GaAsSb, GaNAs, TlGaAs and GaAsBi. The solid lines are fits to experimental data, and the broken lines are from first principles calculations for Bi and extrapolation of the data fit for GaNAs. The GaAsBi and GaNAs experimental data are for pseudomorphic films (coherently strained to the substrate), while all the other data corresponds to free standing material. The TlAs bandgap and lattice parameter is from first principles calculations [44] and the TlGaAs bowing parameter is estimated by extrapolating the scaling relationship between bowing parameter and lattice mismatch for other III-V ternary alloys [45]. It is clear that N, Bi and Tl result in a much larger bandgap reduction per change in the lattice constant than the traditional GaAs alloying elements, In and Sb. GaBi has not been synthesized, but is predicted to have a large negative bandgap of -1.45 eV [46], allowing GaAsBi alloys with bandgaps 0 eV< Eg <1.42 eV to be made, in principle. Fig. 1.2 indicates Tl should behave similar to Bi when alloyed with GaAs, however, very few reports on Tl-containing III-V alloys exist. Bandgap measurements on TlGaAs have not been reported, nor have predictions of the relative effect of Tl on the conduction and valence bands. 11 Figure 1.2. Bandgap as a function of lattice constant for bandgap lowering GaAs-based ternary alloys. Solid lines have been verified with experimental data. Broken lines correspond to theoretical results for GaAsBi and TlGaAs and extrapolation of the data fit for GaNAs. Data sources are as follows: GaNAs [32] , GaAsBi [46, 47] , TlGaAs [44, 45] , InGaAs [48] and GaAsSb [49]. The dependence of the bandgap on the lattice mismatch to GaAs is shown in Fig. 1.3 for alloys of N, In, Sb and Bi (from experimental data). Below ~1.18 eV, GaAsBi has the least amount of mismatch from GaAs of any alloy, including GaNAs, which is renowned for its anomalous bandgap bowing. This unmatched bandgap reduction makes GaAsBi very appealing for extending the wavelength of devices (such as light emitters and detectors), that can be fabricated on GaAs substrates as well as substrates with larger lattice parameter like InP, to beyond what traditional alloying elements offer. The potential impact of materials with increased bandgap tunability can be seen when one considers the relatively new \"extended InGaAs detectors\". These devices are fabricated on compliant substrates [50] in order to reach bandgaps below that of InGaAs lattice matched to InP. The compliant substrate consists of a single crystal that it is thinner than the critical thickness of the material 12 which is to be grown. During the growth of the InGaAs layer, the substrate deforms plastically to accommodate the strain in the film. Growth on compliant substrates requires a complicated fabrication process to produce the thin substrate, usually weakly bonded to a host substrate. Figure 1.3. Bandgap as a function of lattice mismatch to GaAs for ternary alloys with In, Sb, N and Bi. Data sources are as follows: GaNAs [32], GaAsBi [47], InGaAs [48] and GaAsSb [49]. The large bandgap reduction in GaAsBi results from a strong perturbation of the host valence band by Bi. The Bi 6p orbitals are predicted to reside near the valence band maximum (VBM) in GaAs [46, 51]. As a result, the large bandgap reduction with Bi incorporation results from an upward motion of the heavy and light hole bands. The position of the split off band changes very little, resulting in a huge increase in the spin orbit splitting energy [52]. This large coupling could be useful in spintronic devices, as well as to suppress auger recombination and inter-valence band absorption at higher Bi concentrations, when the splitting energy exceeds the bandgap (above ~10% Bi) [53, 54]. 13 Although the isolated Bi energy level is believed to reside below the VBM [51], there exists evidence that few-atom clusters that form randomly in the alloy result in bound states above the VBM [55, 56]. These localized states can trap excitons, preventing them from diffusing to non-radiative defects, which could explain the strong PL observed in these low-temperature grown materials. Localized states are also consistent with the observed broad-spectrum PL emission peaks [47]. Figure 1.4 shows a schematic of the band structure of GaAsBi and the analogous GaNAs alloy, illustrating the resonant interaction of N and Bi on the conduction and valence bands, respectively, as well as the position of localized states. Figure 1.4. Schematic of the bandstructures of GaNAs and GaAsBi illustrating the conduction (CB), heavy hole (HH), light hole (LH) and split off (SO) bands. The resonance of the N 2s and Bi 6p orbitals with the band edges is indicated as well as the position of localized states. Figure originally published in [53]. Nominally undoped GaAsBi samples show p-type conductivity, with the hole concentration increasing with increasing Bi content. This suggests the presence of Bi-induced localized acceptor states near the VBM [57]. Due to the perturbation of Bi on the valence band, the hole mobility in p-GaAsBi has been found to decrease with increasing Bi-content [58, 59]. ConductionBandV le ceBa dEnergyEgNN2HHN 2sSOLHCB CBGaNxAs1-xGaAs1-xBixSOLHBi cluster ?Bi 6pHHEgo 14 As for electron transport properties, unlike N-incorporation Bi is not found to strongly affect electron mobility [60, 61]. The fact that Bi reduces the GaAs bandgap by increasing the VBM and has little effect on electron transport and conduction band alignment, makes GaAsBi an ideal base layer for low threshold voltage heterojunction bipolar transistors (HBTs). HBTs are used as power amplifiers in mobile phones and reducing the turn-on voltage could significantly reduce power consumption. A US patent for an HBT design incorporating Bi has been granted [62]. Like annealed low-temperature grown GaAs (LT-GaAs), low temperature grown GaAsBi can have sub-picosecond electron trapping times and high electron mobilities, making these materials useful for the fast photoconductive switches used to generate and detect THz radiation [63]. One advantage that GaAsBi has over LT-GaAs is the reduced bandgap, which can allow for GaAsBi photoconductive switches to be pumped with the wide array of laser sources with longer than 1 μm emission wavelengths. A GaAsBi THz source, with emission bandwidth up to 5 THz and designed to be pumped by sources operating at wavelengths of 1030 nm and shorter, is now commercially available from Ekspla, a Lithuanian optoelectronics company. Despite the low growth temperatures required to obtain Bi incorporation, GaAsBi layers have shown strong room temperature PL [47]. Strong electroluminescence has been demonstrated from GaAsBi light emitting diodes [39] and lasing has been obtained from optically-pumped GaAsBi layers [64]. Very recently, an electrically pumped GaAsBi laser operating at room temperature has been demonstrated [65]. An international patent application has been filed for a light emitting device incorporating GaAsBi as the active material [66]. A separate international patent application has been filed for a GaAsBi light emitting device where the split off energy exceeds the bandgap, suppressing the Auger recombination [67]. These are some of properties of GaAsBi alloys, which provide a wide range of exciting device possibilities. Potential applications for GaAsBi range from IR light sources and detectors, high efficiency solar cells, spintronics, HBTs, THz sources and detectors and thermoelectrics. The recent release of a THz source based on GaAsBi marks the first commercial deployment of this unique material. 15 1.5 Molecular beam epitaxy growth of GaAsBi The Ga-Bi phase diagram indicates there are no stable Ga-Bi compounds that exist for any temperature or Ga:Bi ratio [68, 69]. This lack of reactivity between Bi and Ga and the large size of the Bi atom, makes the incorporation of Bi into GaAs challenging. Metastable GaAsBi, like the other III-V-Bi alloys mentioned above, require the use of low temperature growth methods, as well as careful control of growth conditions. For these reasons, MBE is an ideal growth technique as it allows for low growth temperatures, in situ experiments to monitor the growth process, and precise control of temperatures and source fluxes. Although the first GaAsBi films were grown by MOVPE [27], GaAsBi is currently mostly grown by MBE. To generate a Bi flux in the MBE, a standard Ga-type effusion cell loaded with a charge of elemental Bi is typically used. High purity source materials are required to minimize the amount of electrically active impurities in MBE grown films. The purity of the Bi, Ga and As charges in our MBE system are 99.9999%, 99.999999% and 99.999995%, respectively. The Bi effusion cell produces a flux that is a mixture of Bi monomers and dimers [26, 70]. Like Ga, Bi does not wet the surface of a pyrolitic boron nitride (PBN) crucible, so a Ga-type hot-lip cell is the preferred cell type. A hot-lip cell maintains the mouth of the cell at a higher temperature than that of the charge. This prevents the source material from condensing and forming droplets on the cell lip. In the case of Ga, these droplets are known to cause source spitting, which produces topographic defects in grown films [71, 72]. In the growth of GaAsBi, As2 is most often used as the source of As, but As4 has also been used. Ga evaporates as monomers from a standard effusion cell. Under the normal conditions suitable for GaAs growth by MBE, namely substrate temperature 550C -580C with a large As2:Ga flux ratio, Bi does not incorporate into the film. Instead, Bi acts as a surfactant, wetting and then evaporating from the surface. The growth of GaAs, AlGaAs, InGaAs and GaNAs with a Bi surfactant results in vast improvements in material quality, such as smoother surfaces and stronger PL [73-75]. The presence of Bi on the surface is believed to reduce the interface energy between the semiconductor surface and the vacuum, resulting in enhanced adatom diffusion. The 16 desorption energy of Bi from a GaAs surface is found to be the same as from liquid bismuth [74], inviting the interpretation that the surface Bi can be thought of as a liquid. To get Bi to incorporate into GaAs, temperatures below 400C, to as low as 200C, are required, as well as an atomic As:Ga flux ratio close to unity or lower [29, 76, 77]. This results in a narrow growth window for the growth of smooth Bi-containing films, since below the unity atomic As:Ga flux ratio Ga droplets form on the surface. The low growth temperature also means that excess Bi will form Bi droplets rather than evaporate, requiring a controlled Bi flux. Recent reports suggest that replacing the As2 source with As4 can result in reduced sensitivity of Bi incorporation to the As:Ga ratio [78]. Further investigation is still required to assess the efficacy of As4 in the growth of GaAsBi. As Bi incorporation depends on so many growth parameters (As2:Ga flux ratio, substrate temperature and Bi flux), in situ characterization tools have proven invaluable to understanding the growth process and increasing reproducibility. Measurements with RHEED show that at the conditions required for Bi incorporation, namely low substrate temperature, atomic As:Ga flux ratio close to 1 and the presence of a Bi flux, a (21) surface reconstruction appears that is not seen without the Bi flux [79]. The reconstruction is indicative of the surface periodicity and details about RHEED are discussed in chapter 2. In situ light scattering allows for highly sensitive detection of droplet formation and can also be used to accurately calibrate the As2:Ga flux ratio [80] (as can RHEED). Optical bandgap thermometry provides accurate measurement of the substrate temperature, which is particularly useful at the low temperatures needed for GaAsBi growth, where pyrometry is difficult [81]. Developing a better understanding of the physical processes that govern the MBE growth of GaAsBi has been one of the main goals of my research. As such, the MBE growth of GaAsBi is explored in detail in the research chapters of this thesis. To date, the record reported Bi-incorporation stands at 21.8% [77], corresponding to this work. 17 1.6 The scope of this research The GaAsBi alloy has many appealing features that offer potential advantages over more conventional III-V alloys. Challenges associated with the growth have slowed the characterization of this material system and the assessment of whether it has device applications. In addition to suitable material properties, a detailed understanding and high reproducibility of the growth procedure are required before a material can be commercialized. This relatively new III-V alloy has shown very interesting and unique properties, however, there is much work that remains before its properties are fully understood. In this thesis, an investigation of the growth of the GaAsBi alloy by MBE is described. The goal is to obtain an understanding of the physical processes governing GaAsBi growth. The advantages of MBE, outlined in section 1.2, make it a very suitable technique for studying the growth of GaAsBi. These advantages include the simple chemistry, accessibility of low temperature growth conditions, precise control of fluxes and the availability of in situ characterization. A systematic study of the dependence of the Bi-content on MBE growth conditions has led to the development of a physical model for the growth of GaAsBi. This model explains the underlying physics that necessitates the unusual growth conditions to obtain Bi incorporation. In addition, the model acts as a guide for the growth of films for future experiments. This understanding has led to the realization of films with record Bi incorporation of 21.8%, nearly double the previously reported record. Experiments were carried out to assess the structural, optical and electronic properties of these new alloys. The low growth temperature required to obtain high Bi incorporation was found to result in large critical thicknesses for the epitaxial layers, allowing for strained films with bandgaps as low as 0.5 eV to be made on GaAs substrates. These results suggest that GaAsBi could open up a new wavelength range for devices on GaAs and InP substrates. MBE is a versatile tool that is used for both research and production of thin epitaxial layers of semiconductors, metals and insulators. For the reasons discussed above, MBE is particularly well suited for the synthesis and study of new materials. The growth of GaAsBi requires pushing MBE technology to obtain very precise control of the fluxes (the As2:Ga 18 flux ratio in particular), low growth temperatures and use of in situ characterization tools to provide real-time feedback during growth. For production, cost is often the determining factor when choosing an epitaxial growth technique. Therefore, advancing MBE technology should also involve cost reduction. In addition to making MBE more viable for production, lowering operating costs makes MBE a more accessible tool for research and development in the university environment. Liquid nitrogen (LN2) cooling of MBE cryopanels represents a significant operating cost of MBE. To reduce this cost, alternate cooling methods were investigated, with the goal of making MBE more affordable for research and production. This entailed implementing a closed-cycle chiller, operating to as low as -80C, to cool the MBE cryo-shroud. The chiller reduced LN2 consumption by about an order of magnitude. A study of the temperature dependence of cryopanel pumping efficacy in the MBE system was undertaken. In addition, the properties of GaAs layers grown with LN2 and closed-cycle cooling of the shroud, and AlGaAs films grown with closed-cycle cooling of the shroud, were investigated, in order to evaluate efficacy of this new cooling technique. As an MBE grower, I have been privileged to collaborate with many others both within our research team and at other institutions. In this thesis, I have tried to acknowledge the contribution of others to this research, when applicable. Unless otherwise mentioned, the samples were grown and characterized by me. 19 Chapter 2 Experimental methods This chapter describes experimental methods that are relevant to the research discussed in this thesis. Details of the MBE system and the sample growth procedure are discussed, as are the in situ and ex situ experiments that were carried out during growth and on the grown films. 2.1 Molecular beam epitaxy growth Samples were grown in a VG V80H MBE system equipped with a growth chamber, a sample preparation chamber and a sample entry lock. The base pressure of the system is ~110-10 Torr after a being baked to ~180C for several days. The growth chamber has 8 ports for accommodating water cooled effusion cells. As of Nov. 2013, these ports were occupied by standard single-filament effusion cells for Ga, Si and Al; a double-filament cell for Bi that allows independent control of the temperature of the tip of the crucible; a duel-dopant source that contains In in one of the cells (the other is empty); and a two-zone valved cracker for generating As2 (or As4 by reducing the cracking zone temperature). Except for the As source, effusion cell beams are turned on and off with shutters (some actuated pneumatically and others electromagnetically), allowing for abrupt changes in film composition with less than a monolayer of growth. The flux of the cells is controlled by the cell temperature, which is monitored by a thermocouple that is in contact with the cell crucible. For the As source, the flux is controlled by a combination of the bulk source temperature and the cracker valve position. The remaining two cell ports have heatable windows on them that allow for optical access to the sample. In addition, the system has a hydrogen-cracker source and a CBr4 gas-injection system for a source of carbon doping. The growth chamber is pumped by an ion pump, cryopump and a Ti sublimation pump (TSP) with a LN2 cooled trap. The combined pumping speed of the cryopump and the ion pump at 20 the growth chamber is estimated to be ~1000 L/s for H2O and ~800 L/s for air. A cryo-shroud, designed to be cooled with LN2, surrounds the inside of the growth chamber (except for numerous holes that allow access to the substrate by the manipulator, effusion cells, shutters, optical ports, RHEED gun and sample loading system). In addition to removing the heat radiated by the hot effusion cells and substrate heater, the LN2 filled cryo-shroud has the added advantage of acting as a gas-selective cryopump. For water, which has a very low vapor pressure of 10-21 Torr at -196C (the boiling point of LN2), the cooled cryo-shroud and TSP reservoir are believed to offer a combined pumping speed of ~50,000 L/s [82]. The TSP is located above and behind the cryo-shroud with no line of sight to the hot effusion cells or substrate heater. The MBE system is capable of growth on wafers up to 76.2 mm in diameter, however growths were mostly done on ¼ of 50.5 mm wafers and sometimes on full 50.5 mm wafers. Most often growths were carried out on (001)-oriented semi-insulating (SI) GaAs wafers with less than 0.5 offcut and 350±25 μm thickness. Growths were also done on (001) n-type GaAs wafers and a handful of growths were done on substrates with larger lattice parameters: InP and GaSb. After loading the GaAs substrates through the entry lock they were degassed at ~400C for 1-2 h in the preparation chamber, before being loaded into the growth chamber. Before initiating the growth, the native oxide was thermally desorbed by heating the substrates to ~610C for 10 min with an As2 flux incident on the sample surface. Thermal desorption of the native oxide results in surface roughening, which is observed by in situ elastic light scattering (discussed below). Oxide removal is also observed by RHEED (also discussed below). After removing the oxide, GaAs buffer layers 300-500 nm thick were typically grown on the GaAs wafers. The GaAs layers were grown at substrate temperatures of 540-580C under As2:Ga BEPRs of ~8, followed by the growth of the epilayers of interest. For the growth of low-temperature layers, such as GaAsBi, sometimes growth interrupts were used to adjust the substrate and source temperatures, and sometimes they were not. When the temperature of the As source was changed, ample time was needed to allow the source to stabilize (~75 min). Samples were rotated at ~0.5 Hz during growth to increase uniformity. 21 2.1.1 Cooling the MBE cryo-shroud In order to reduce the cost associated with cooling the MBE cryo-shroud with LN2 and to investigate the effect of shroud temperature on the pumping of active gas species, the MBE system was reconfigured to be cooled with a chiller operating at ~-80C with a polysiloxane recirculating fluid. This is a new concept that has a potential broad impact on the MBE community. Growths were carried out with the system configured for both LN2 cooling and for cooling with the chiller. During operation of the MBE with LN2 cooling, the shroud was gravity-fed through vacuum-insulated lines from a phase separator that was in turn fed by another vacuum insulated line from a pressurized 5,000 L external storage tank. During growth, the MBE system used ~25 L/h of LN2 at the phase separator. With a latent heat of vaporization of 162 kJ/L for LN2, this corresponds to 1.1 kW of cooling power. The TSP reservoir was filled from the phase separator about an hour before growth was initiated and liquid remained for more than 8 hours. In our experience, when averaged over a year the MBE used ~50 L of LN2 for every hour of operation. This estimate includes evaporation from the storage tank and loss in transferring to the phase separator. In Victoria, the cost of LN2 is ~$0.45 /L, so it costs about $20 /h for LN2 to run the MBE. Large production MBEs consume LN2 at rates up to 500 L/h, representing a significant fraction of the total operating cost [16]. For growth with the chiller-cooled shroud, LN2 was replaced with Dow Chemical Syltherm XLT heat transfer fluid, which is composed of dimethyl polysiloxane with an average molecular weight of 317 u and a freezing point of -111C. The fluid is cooled with an ultra-low temperature re-circulating chiller, RC311 made by FTS Systems. This water-cooled chiller is rated at 6 kW electrical input power and has an operating temperature range from -90C to +75C, with 1.2 kW of cooling capacity at -70C and 0.5 kW at -80C. It is also equipped with a 1.5 kW heater. The syltherm XLT fluid has an operating temperature range of -100C to 260C and a heat capacity of 1.6 kJ kg-1K-1 at -80C. Initially, when the system was at UBC, the chiller was connected to the MBE with the vacuum-insulated lines that were used for LN2. After moving the system to UVic., these were switched to 5/8 copper tubing connected to the MBE using 1.5 sanitary fittings with silicone gaskets. The copper tube was 22 insulated with 2-layers of Armaflex flexible foam insulation. Copper tubing with the Armaflex insulation is more flexible, easier to install and much less expensive than vacuum insulated lines. During operation of the MBE with the silicone fluid cooled shroud, the TSP reservoir was filled once with LN2 about an hour before the growth was initiated. Running the MBE with the As, Si and Ga cells at operating temperature and the substrate at ~600C, the lowest achievable temperature of the heat transfer fluid in the chiller reservoir was -78C, indicating that ~0.5 kW of heat was being extracted from the MBE system in this state. With these conditions, the temperatures of the exterior of the copper tubes at the inlet and outlet to the shroud were measured to be -71C and -67C, respectively, with a thermocouple. The estimated cooling capacity of the LN2 used during growth (1.1 kW) is twice the cooling provided by the chiller at -80C (0.5 kW) under similar growth conditions. The difference in heat load may be due to thermal losses in the phase separator and in the connections between the transfer lines and the MBE system when operating with LN2. The MBE system with the chiller setup can be baked with the copper lines left connected to the MBE. The insulation is removed from the copper tubing inside the bake box and the silicone fluid is blown out of the cooling lines with air by disconnecting the lines at the chiller. Removing the fluid from the lines is necessary to thermally isolate the hot fluid inside the bake box. During the bake the lines are left disconnected at the chiller to prevent pressure from accumulating in the shroud. The effect of the shroud and TSP temperature on the residual gas partial pressures inside the chamber was investigated with a quadrupole residual gas analyzer (RGA) from Stanford Research Systems, capable of detecting gases with masses up to 200 u. The RGA was located just outside the shroud with no line of sight to the substrate heater or effusion cells. In a separate measurement, to test the effect of RGA location on the measured pressures, the RGA was located inside the shroud by insertion through a viewport. Similar values of gas partial pressures were measured with the RGA in both locations. GaAs samples, undoped, n-type and p-type were grown with the MBE system immediately before and after switching from LN2 cooling to the new chiller. Samples were grown by D. A. Beaton and myself. Measurements on these films were performed to test whether the 23 change in the cooling system affected the electronic properties of the films. For the nominally undoped samples, resistivity measurements were performed in order to determine the residual doping level. On the doped samples, the density of deep level defects in the films was characterized by deep level transient spectroscopy (DLTS). The resistivity measurements were carried out by V. Bahrami-Yekta and M. Masnadi-Shirazi and the DLTS measurement were done by K. P. Watkins and P. M. Mooney at Simon Fraser University (SFU). Photoluminescence measurements were performed on the GaAs layers by me. In addition, AlGaAs layers, both undoped and n-doped with Si, were grown with the MBE system configured for cooling with the chiller. These layers were grown by V. Bahrami-Yekta and myself. Photoluminescence and transport measurements were performed on the AlGaAs layers by M. Masnadi-Shirazi, M. Patel, V. Bahrami-Yekta and myself. 2.2 In situ monitoring techniques One of the strengths of MBE for research and development is the accessibility of the sample for in situ characterization during growth. Our MBE system is equipped with several important tools: a retractable ion gauge beam flux monitor (BFM); a quadrupole residual gas analyzer (RGA) from Stanford Research Systems, capable of analyzing gases with molecular masses up to 200 u that also facilitates easy leak detection; a reflection high energy electron diffraction (RHEED) setup; an optical bandgap thermometry setup for measuring substrate temperature; and an elastic light scattering (LS) setup for monitoring sample roughness. All were integral to the research described in this thesis. 2.2.1 Substrate temperature measurement Substrate temperature measurement in MBE is nontrivial, as bringing a temperature probe into physical contact with the substrate is difficult and the temperature range is often too low for pyrometry (especially for GaAsBi growth). In our system, the substrate thermocouple is suspended in the space between the pyrolytic boron nitride (PBN) heater and the back of the 24 sample holder. As a result, the thermocouple reading during growth typically differs from the actual wafer temperature by several hundred C. For SI-GaAs, n-GaAs and n-InP wafers, substrate temperature was inferred by measuring the optical absorption edge of the wafers, using diffuse reflectance spectroscopy (DRS). For the DRS measurement, the filament of a halogen bulb is imaged on the substrate through a heatable window on one of the cell ports (25 from the substrate normal). The diffusely scattered light is then collected through a viewport at normal incidence, focused into a large core (550 μm diameter) optical fiber cable and sent to a Control Development spectrometer with an InGaAs array detector. This diffuse light, scattered from the unpolished back surface of the wafer or from the rough PBN plate behind the wafer, has made two passes through the substrate. To eliminate the radiation from the substrate heater, the difference between the spectra with the halogen light on and with the light shuttered is taken. Fitting the wavelength and width of the absorption edge allows for the substrate temperature to be inferred with an accuracy of ~2C. For more information, the reader is referred to the work by Johnson et al. [81, 83]. The bandgap of GaSb is too small to be measured with our DRS setup. It has been shown that the dependence of the substrate temperature on the substrate current follows a power law for the resistive heating of Si and Ge wafers in vacuum [84, 85]. We also observe this behaviour for SI-GaAs and n-GaAs wafers heated with the substrate heater in the MBE, with the temperature of the SI and n-GaAs wafers following the same power of the heater voltage, differing only by a constant factor. Assuming the same exponent applies for GaSb wafers and by noting (with RHEED) the substrate heater voltage where the native oxide desorbs (known to happen at ~500C for GaSb), the expression T(C)  30.2(V sub) 0.735 was obtained. One caveat is that for GaAs, it is found that the heat load from the cells and variations in surface roughness (which changes the amount of heater radiation absorbed) can cause large (~50C) variations in the substrate temperature. This variation is not taken into account in the GaSb temperature estimate, which is based on the heater power alone. 25 2.2.2 Reflection high energy electron diffraction Reflection high energy electron diffraction has been used extensively as an in situ structural analysis tool since the early days of MBE. A typical setup consists of an electron gun that is incident on the wafer surface at grazing incidence (<3) with a phosphor screen placed in the reflected beam path. Remarkably, a huge amount of insight into the growth process can be gleaned from this simple setup. RHEED is highly surface-sensitive, as the grazing electron beam only penetrates one or two monolayers into the substrate. Therefore, from the perspective of the RHEED beam, the crystal is two-dimensional. A consequence of this is that the reciprocal lattice consists of an array of reciprocal “rods” elongated in the out-of-plane direction, rather than points. For experiments described in this thesis, a 20 keV STAIB electron gun was operated at 15 keV. The image of the reflected beam on the phosphor screen was recorded with a video camera interfaced with a computer. EE2010 software was used for displaying and recording images and allows for measurement of various properties of the images, such as intensity oscillations of the spots during growth. For electrons at 15 keV the de Broglie wavelength is 0.010 nm, 57 smaller than the GaAs lattice spacing. In reciprocal space, this means the radius of the Ewald sphere is much greater than the spacing of reciprocal lattice points, resulting in small diffracted angles. This allows for a large piece of reciprocal space to be imaged on the phosphor screen at one time. For elastic scattering, energy and momentum conservation gives the requirements of Eq. 2.1 for constructive interference, where and are the incident and final wave vectors and is a vector of the reciprocal lattice. Figure 2.1 shows the diffraction of an incident wave from reciprocal lattice rods. Diffraction occurs in directions where the Ewald sphere intersects the rods, causing short streaks to appear on the RHEED screen. The length of the streak is related to the thickness of the 26 reciprocal rods, which is inversely proportional to the lateral coherence of the sample surface [86]. Figure 2.1. An illustration of the intersection of the Ewald sphere with reciprocal lattice rods of finite width, resulting in elongated streaks in the RHEED pattern. The two diffracted beams for each rod indicate the angular width of the diffracted beams. The periodicity of the epilayer surface often differs from that of the bulk. When this happens, the resulting surface is known as a “surface reconstruction”, a change in surface structure created in order to minimize surface energy associated with dangling bonds. For (001) zincblend crystals, the convention is to describe the surface reconstruction by the periodicity along the orthogonal and [110] directions. Figure 2.2 shows an images of a (21) surface reconstruction, the preferred reconstruction for the MBE growth of GaAsBi. Here the 2 denotes that the periodicity in the direction is twice the bulk, and the 1 denotes bulk periodicity in the [110] direction. 27 Figure 2.2. Images of a (21) RHEED reconstruction. The diffraction direction is indicated on the images and the spacing between the lines on the right hand image corresponds to the bulk periodicity. Figure originally published in [79]. Figure 2.3 shows the surface reconstruction phase map for GaAs and GaAsBi growth, illustrating the growth conditions where the (21) reconstruction exists. Samples grown under the (21) reconstruction show increased Bi incorporation and stronger PL [79]. In Fig. 2.3, an As2:Ga flux ratio of 0.5 is equal to an atomic As:Ga ratio of unity. The As2:Ga flux ratio was calculated from the measured As2:Ga beam equivalent pressure ratio (BEPR) as measured by a retractable ionization gauge according to [87], as discussed in section 2.2.4. Diffraction from the amorphous native oxide initially covering the GaAs wafer, results in a uniform haze on the RHEED screen. Heating the wafer to remove the oxide exposes a 3-dimensional rough crystalline surface, resulting in a “spotty” RHEED pattern. This is due to the 3-dimensional nature of the surface shortening the reciprocal lattice rods into points. The growth of a GaAs buffer layer smoothes the surface and causes the RHEED spots to elongate into very long streaks, like those seen in Fig. 2.2. Although streaks are often associated with smooth surfaces, they result from a small lateral coherence length of the surface, which widens the reciprocal lattice rods. This is associated with the nucleation and coalescence of small islands on the surface and also surface reconstruction antiphase domains [86]. If 28 growth results in the formation of facets on the surface, diffraction from the facets causes chevron-shaped formations to appear in the RHEED pattern. Figure 2.3. Surface reconstruction maps of a) GaAs and b) GaAsBi. The growing GaAsBi surface has an incident Bi beam equivalent pressure (BEP) of 3×10-9 Torr. Surface reconstructions are indicated for the electron beam oriented along [110] and azimuths. The incident Ga BEP was kept constant at 8×10-8 Torr during the experiment, corresponding to a growth rate of 0.1 μm/hr. Figure originally published in [79]. When the growth of a smooth annealed surface is initiated, the intensity of the RHEED spots is observed to oscillate with a periodicity corresponding to the growth time for a single monolayer. This allows the growth rate to be measured using RHEED. It has been suggested that the intensity oscillations are associated with the change in step-edge density on the surface as a new monolayer nucleates and then fills in [88]. For the work described in this thesis, RHEED was mostly used to monitor the surface reconstruction during GaAsBi growth, as well as to detect the removal of the native oxide from the wafers. 29 2.2.3 Elastic light scattering In situ elastic light scattering is a non-evasive technique that allows for real-time monitoring of the film surface morphology during growth. Light scattering was used to detect the thermal desorption of the substrate native oxide, monitor surface roughness evolution during growth, detect the formation of Ga and Bi droplets and detect strain relaxation. It was found that LS is particularly useful for determining the 1:1 As:Ga atomic flux ratio, as the onset of surface roughness associated with Ga-rich growth is easily detected. The early detection of this roughening with LS allows the surface morphology to be recovered before Ga-droplets form on the surface, after which the surface cannot be recovered. The amount of light that is scattered diffusely from a surface is proportional to the power spectral density (PSD) of the surface, which is related to the roughness. The spatial frequency on the surface that the scattering is sensitive to depends on the wavelength of light () and the incident and scattered angles, as given by Eq. 2.2. In this equation i and s are the incident and scattered angles relative to the surface normal and s is the angle out of the plane of incidence (relative to the projection of the reflected beam on the substrate surface). In our setup, 488 nm light from an Ar-ion laser was used for the light source, and i, s and s were 55, 25 and 90, respectively. This configuration gives a spatial frequency of 11.9 μm-1, corresponding to a length scale of = 528 nm.   A diagram of the light scattering setup is shown in Fig. 2.4. Laser light at 488 nm, from a Lexel 3500 Ar-ion laser, is chopped at a frequency of 2 kHz and then coupled to a fiber and sent to the MBE chamber. The light is focused on the sample, through a modified shutter port, with a spot size on the substrate of ~5 mm. The specular beam is allowed to exit the chamber through another modified shutter port to minimize the amount of scattered light inside the chamber. This is critical as the sample has a mirror-like finish, which only scatters a small amount of light. The detector assembly is mounted on a cell port. A collecting lens, iris and line filter are aligned in front of a photomultiplier tube (PMT), such that only the 30 laser light scattered from a region of the sample is detected. The PMT is connected to an SR810 lock-in amplifier that interfaces with a PC via LabView. For further reading, the reader is referred to [89-91]. Figure 2.4. An illustration of the light scattering setup on the MBE. The detector angle is not representative of the actual configuration used in this work. Figure originally published in [91]. 2.2.4 Beam flux measurement Precise knowledge of the MBE source fluxes is critical in the growth of GaAsBi. Beam equivalent pressures (BEPs) were measured with a retractable nude Bayard-Alpert-type ion gauge on a daily basis when GaAsBi was grown. The measurements were often taken just before initiating the growth to accurately set the As2:Ga flux ratio, and almost always after the growth was finished. For As-rich growth of GaAs, the growth rate is determined by the Ga flux, since Ga evaporation is negligible under normal growth conditions [92]. By measuring a single Ga BEP using the beam flux monitor (BFM) and then measuring the corresponding GaAs growth rate at that BEP, a relationship between the measured Ga BEP and the absolute Ga flux can be obtained. The growth rate is determined by x-ray diffraction 31 (section 2.3.1) measurements on a test structure or alternatively by the period of RHEED oscillations during growth. In this thesis, the BEP refers to the reading (in units of pressure) given by the BFM, which is factory calibrated for N2 at 300 K. Flux ratios can be inferred from BEPRs using Eq. 2.3 [87], where Fx and BEPx is the flux (number of molecules/area/time) and BEP of species x and ηx, Tx and Mx are the relative ionization efficiency, absolute temperature and molecular mass of x. Equation 2.3 is a result of the BFM measuring molecular density of the beam. Preobrazhenskii et al. measured ionization efficiencies for Ga and As2 and As4 of 1.74, 4.0 and 6.8, respectively [87]. Bi is known to evaporate as a mixture of monomers and dimers [26, 70], but there are no reports of the ionization efficiencies of these species. From Eq. 2.3 and the absolute Ga flux, the fluxes of other species (eg. As2) can be determined from the BEPR. For As2 at 1000C (cracker temperature) and Ga at 950C, Eq. 2.3 indicates that an As2:Ga flux ratio of 0.5 (atomic ratio of 1) occurs at a BEPR of 1.7. From light scattering and RHEED, this unity flux ratio was found at a BEPR of 2.2  0.2. A higher BEPR would result if the As sticking-coefficient were less than unity, however, systematic errors arising from different flux measurement geometries could also offset the measured BEPRs between different MBE systems. To effectively measure As2:Ga flux ratios and communicate them between MBE systems, standard practice is to rely on measurements of the 1:1 atomic flux ratio, as determined by RHEED or light scattering. To relate the Bi:Ga BEPR and flux ratio, profilometry measurements on a Bi-metal film, grown with the Bi-cell at 600C, were performed. This yielded the relationship FBi/FGa = (0.51 ± 0.05)(Bi:Ga BEPR), assuming the density of the film was equal to the bulk density. A lower film density would result in a constant of proportionality less than 0.51. The Bi-incorporation data discussed in chapter 4 suggests FBi/FGa~(0.323  0.008)(Bi:Ga BEPR) for a Ga cell at 941C and the Bi cell between 450C and 525C. 32 2.3 Ex situ characterization This section describes ex situ experiments carried out to characterize the structural, optical and electronic properties of grown films. 2.3.1 High resolution x-ray diffraction The most important technique for characterizing the structure of epitaxial films is high resolution x-ray diffraction (HRXRD). HRXRD gives a measure of the long range order, which can allow for determination of properties such as layer composition, thicknesses, strain and relaxation. Two HRXRD systems were used for the research described in this thesis, a PANalytical X’Pert Pro Materials Research Diffractometer at UBC and a new Bruker D8 Discover at UVic. The systems are quite similar, with differences only in technical specifications, software and degree of automation. They generate x-rays by accelerating electrons at 40-45 kV towards a Cu target in a sealed vacuum tube. The x-rays exit the tube through a Be window, where monochromating optics produce a highly collimated beam of k1 x-rays (wavelength 0.154051 nm). The heart of both systems is a high-resolution two-circle goniometer. The sample stage, which has 3-rotational and 3-translational degrees of motion, is mounted to the θ circle and the detector is mounted to the 2θ circle. The majority of scans were conducted with the Bruker system, for which the goniometer has a resolution of 0.0001. The incident beam optics consist of a Göbel mirror and a Ge crystal channel-cut for two asymmetric (004) reflections, resulting in a typical beam divergence of less than 16 arcsec. The detector assembly can be adjusted by software so that the scintillation counter detects x-rays through either a pair of motorized slits or a 3-bounce analyzer crystal. The condition for constructive interference from a set of atomic planes is given by Bragg’s law, where n is an integer,  is the wavelength of the x-rays, d is the plane spacing and θ is the angle from the plane.  33 For thin epitaxial films, where the substrate and film plane spacing is similar, it is often more useful to write Bragg’s law in differential form, as shown in Eq. 2.5. Here θB is the Bragg angle and is the strain between the layers. Often it is useful to work in reciprocal space, where Bragg’s law takes on the form given in Eq. 2.1. In reciprocal space, atomic planes form the reciprocal lattice, where the distance from the origin to a particular point corresponds to the inverse of the plane spacing. The most common scan, performed on every single GaAsBi film grown, was a (004) θ-2θ scan. Here the sample and detector are rotated together to scan along the [001] direction of reciprocal space near the (004) peaks. An illustration of a (004) scan geometry on a sample consisting of a (001) GaAs substrate and compressively strained GaAsBi epilayer is shown in Fig. 2.5, in both real space and reciprocal space. Figure 2.5. A diagram illustrating the geometry of a (004) θ-2θ scan on a sample consisting of a (001) substrate with a compressively strained epilayer. The real space (left) and reciprocal space (right) images are shown. 34 Bi-incorporation increases the size of the GaAs lattice. The lattice mismatch between film and substrate ( ) is given by Eq. 2.6, where af and as are the film and substrate relaxed lattice parameters. In the pseudomorphic case, the relation between and the (004) strain is given by Eq. 2.7, where  is Poisson’s ratio for the film.   For a pseudomorphic film, measuring the out of plane strain allows determination of the film composition if the relation between lattice parameter and composition is known. A GaBi relaxed lattice parameter of 6.330.06 Å was previously obtained by extrapolating HRXRD and Rutherford backscattering spectroscopy (RBS) measurements on GaAsBi samples with up to a few % Bi, assuming Vegard’s law applies and that Poisson’s ratio for GaAsBi is equal to that of GaAs ( ) [29]. For quick reference, this results in a spacing of ~300 arcsec between the film and substrate (004) peaks for each percent Bi incorporated, for a pseudomorphic layer. In Fig. 2.5, the in-plane lattice constant of the film and substrate is the same, indicating the film is pseudomorphic. In reciprocal space, this is indicated by off-axis peaks (eg. (224)) having the same in-plane value as the substrate. Figure 2.6 illustrates fully strained and fully relaxed unit cells on a substrate. Relaxation of a compressively strained film causes the in-plane lattice constant to enlarge, while the out-of-plane lattice constant shrinks. The tilt angle of off-axis planes also changes. For films where relaxation is present, the composition cannot be determined solely from a (004) θ-2θ scan as knowledge of the in-plane lattice constant is also required. 35 Figure 2.6. An illustration of fully strained (left) and fully relaxed (right) GaAsBi unit cells on a GaAs (001) substrate. The degree of relaxation is determined by Eq. 2.8, where ain and aout are the in-plane and out-of-plane lattice constants of the film and is the substrate lattice constant. On samples where relaxation was suspected, (224) reciprocal space maps (RSMs) were carried out to determine ain. For high-quality layers, (004) θ-2θ scans typically showed interference oscillations around the substrate and film peaks. These are known as pendellösung fringes, which arise as interference between the diffracted beams from the top layer(s) and substrate. The presence of pendellösung fringes indicates a film with smooth interfaces, good composition uniformity and minimal relaxation. The thickness of the top layer can be determined from the period of the oscillations, using Eq. 2.9, where t is the layer thickness,  is the wavelength of the x-rays, p is the angular spacing of the oscillations and B is the Bragg angle [93].  36 In practice, composition and layer thickness was determined by dynamically simulating measured (004) scans, using Leptos and RADS Mercury. 2.3.2 Rutherford backscattering spectroscopy Determining the Bi-content of GaAsBi films from HRXRD requires knowledge of the relation between the lattice parameter and the Bi-content. This relationship has been measured by Tixier et al. [29], by RBS measurements of the Bi-content and HRXRD measurements of the lattice parameter, for GaAsBi films with up to a few % Bi. Given that layers with up to 22 % Bi have now been realized, it was important to characterize these films by RBS in order to measure the relationship between Bi-content and lattice parameter in this new non-dilute composition range. RBS is an excellent method for determination of Bi concentrations due to the large mass of Bi relative to Ga and As. Consequently, alpha particles that backscatter from Bi atoms have considerably higher energy from those scattered from As and Ga atoms, and thus are easy to distinguish. RBS measurements were carried out by Martin Chicoine at the Université de Montréal using a 2 MeV beam alpha particles. The detector was placed at a scattering angle of 170 and the sample was tilted to ~7 from the surface normal to minimize channeling effects. Concentrations and thicknesses were obtained by simulating the RBS spectra using SIMNRA. 2.3.3 Scanning electron microscopy To obtain high resolution images of GaAsBi surfaces and Ga-Bi droplets, samples were imaged with a scanning electron microscope (SEM). Images were recorded using a Hitachi S-4800 field emission SEM at UVic.’s advanced microscopy facility. The SEM operates with an accelerating voltage of 0.5 to 30 kV and has a resolution as small as 1 nm. The system is equipped with a secondary electron detector, a ring-type backscatter detector and a Bruker 37 Quantax system for energy-dispersive x-ray spectroscopy (EDS). Images presented in this thesis were collected by M. Masnadi-Shirazi, M.J. Fryer and myself. 2.3.4 Optical absorption Optical transmission measurements were performed on GaAsBi/GaAs films to determine their bandgap. Measurements were carried out by M. Masnadi-Shirazi at UVic. My contribution was in providing some initial motivation for these experiments, growing the GaAsBi/GaAs samples, as well as aiding in the experimental setup and data analysis. Light from a halogen bulb was chopped and then focused on the entry slit of an Oriel Cornerstone 260 monochromator (0.26 m focal length, on loan from L. Chrostowski). The monochromator has two gratings: one with 1200 lines/mm with an operating range of 450-1400 nm; the other with 300 lines/mm with an operating range of 1100-4800 nm. An appropriate long-pass filter was placed near the exit of the monochromator, followed by the GaAsBi/GaAs sample and finally the detector. Un-cooled Ge (800-1750 nm) and PbS (1000-2900 nm) photodetectors were used. The Detector signal was sent to an SR510 lock-in and a LabView program was used to interface with the monochroator and the lock-in. Samples with Ga-Bi droplet coverage were etched with an HCl:H2O solution (1:4 ratio) to remove the droplets. This greatly reduced the amount of light being scattered from the sample surface. Since the bandgap of the GaAsBi layer is lower than that of GaAs, absorption from the thick substrate was not a problem in the region of the GaAsBi bandgap. Once normalized for the optical throughput of the system, the transmission spectra of a GaAsBi/GaAs sample is the product of the transmission spectra of the layer and the substrate. Dividing the transmission spectra of the GaAsBi/GaAs samples by the transmission spectrum from a GaAs substrate, the transmission spectra of the GaAsBi layers (TGaAsBi) can be isolated, as indicated by Eq. 2.10 (neglecting multiple reflections). HereGaAsBi() is the absorption coefficient of the GaAsBi layer at wavelength , and dGaAsBi is the thickness of the GaAsBi layer.     38 2.3.5 Electrical transport In order to investigate the effect of Bi incorporation on n-type transport properties, as well as to calibrate n and p-type dopant sources, electrical transport measurements were performed on selected samples. Post-growth fabrication and characterization were generously carried out by V. Bahrami-Yekta and M. Masnadi-Shirazi. Samples were grown by me. In the case of the n-doped GaAsBi samples, the study was directed and the data analyzed by me. Carrier concentration and mobility were measured using the van der Pauw method, which is described in detail elsewhere [94, 95]. Samples consisted of square 7  7 mm pieces with ohmic contacts on the four corners. Contacts were deposited by e-beam evaporation, Ni/AuGe/Au for n-type [96, 97] and Ti/Pt/Au for p-type [98], then annealed to 430C for 30 s to decrease the contact resistance. Hall measurements were carried out in a magnetic field of 0.27 Tesla perpendicular to the sample surface. The n-doped GaAsBi layers measured in this thesis were grown at 330C and it is not expected that they are significantly affected by annealing at 430C. A recent GaAsBi annealing study showed that annealing GaAsBi layers grown at 220C to as low as 500C can introduce inhomogeneities in the Bi-content, while GaAsBi layers grown at 315C are stable when annealed to above 600C [99]. It was suggested that the reorganization of Bi atoms may be facilitated by native defects, such as Ga vacancies, which are more abundant for lower growth temperatures. 39 Chapter 3 Closed-cycle cooling of the MBE cryo-shroud: a low cost alternative to liquid nitrogen A significant financial burden to MBE growth programs is the cost associated with LN2 [16]. From the 1970s onward, standard MBE practice has been to line the inside of the growth chamber with LN2-filled cryopanels. These panels remove the heat radiated by the hot cells and substrate, capture the excess material being evaporated from the sources, and adsorb or condense many gases in the process environment. Although cryopanels are normally cooled with LN2, the effect of operating the cryopanels at higher temperature on system performance has not been previously investigated, even though this would have a large impact on operating costs. In this chapter, a novel cooling system is implemented where a recirculating polysiloxane fluid, cooled to as low as -80C by a chiller, is used as an alternative to LN2. The temperature dependence of cryopanel pumping is investigated, with H2O cryo-adsorption being analyzed in detail. In addition, GaAs samples grown with the shroud cooled to -70C with the chiller and with the shroud cooled to -196C with LN2 are characterized, as are AlGaAs layers grown with the chiller setup. The goal of this work is to make materials growth by MBE more affordable and accessible both for research and production. The typical cryopanel arrangement results in direct line-of-sight exposure of the LN2 filled cryo-shroud to the thermal radiation from the hot effusion cells and substrate heater. This leads to high LN2 consumption and contributes to the high cost of operating an MBE system. Raising the temperature of the coolant in the shroud naturally facilitates energy savings, as the coefficient of performance (COP) of a Carnot refrigerator is 5.4 higher when operating at -80C than at -196C (boiling point of LN2). The COP of a Carnot refrigerator is given by Eq. 3.1, where TC and TH are the temperatures of the cold and hot reservoirs. 40 If liquid nitrogen costs 0.45 $/L (cost in Victoria) and its latent heat of vaporization is 162 kJ/L, then it costs about $10 to remove a kWh of heat from the MBE by boiling liquid nitrogen. In our case the wastage from storage and transfer approximately doubles the consumption. In contrast, our chiller is rated at an input power of 6 kW and has a cooling capacity of 500 W at -80C and 1200 W at -70C. If electricity costs $0.1 per kWh then it costs $1.20 and $0.5 to remove a kWh of heat with our chiller at -80C and -70C, respectively. For good semiconductor device performance it is important to minimize the incorporation of electrically active impurities. Therefore, cleanliness of the inside of the growth chamber, minimizing the partial pressure of residual reactive gases, such as H2O, O2, CO and CO2, and the use of high-purity source materials, are of critical importance. Improvements in MBE material quality in the 1980s, reflected by low-temperature electron mobility as well as GaAs/AlGaAs laser threshold currents, can be attributed to various advancements in MBE technology that improved system cleanliness [100-103]. These include innovations such as implementing a load-lock chamber and baking the system after exposure to air, surrounding the inside of the growth chamber with an LN2-filled cryo-shroud and improvements in source material purity. In addition, our V80H and other modern MBE systems have a sample preparation chamber. This allows the substrate to be baked before loading into the growth chamber, so that no unbaked surfaces are exposed to the growth chamber. The vapor pressure of H2O at the boiling point of LN2 (-196C) is ~10-21 Torr. The typical H2O partial pressure in a baked MBE chamber is much higher, 10-11-10-10 Torr, therefore at LN2 temperature arbitrary amounts of H2O can be pumped on LN2 cooled cryopanels through cryo-condensation. At -143C, the vapor pressure of H2O is 10-10 Torr. Above this temperature in a chamber with a H2O partial pressure of 10-10 Torr, a surface covered with many monolayers of H2O (bulk conditions) would act as a source of H2O rather than a pump. Thus, to pump large amounts of H2O while maintaining a partial pressure below 10-10 Torr, the shroud must be cooled below -143C. Water is particularity troublesome in the MBE as it reacts strongly with species such as Al and also bonds relatively strongly to the chamber walls. This is one of the reasons for baking 41 the MBE chamber after exposure to air: to increase the desorption rate of H2O from the walls, so it can be pumped from the chamber. LN2 paneling has been used in MBE systems since the early 1970s [104, 105]. Early systems did not have load-locks, so venting the growth chamber to atmosphere was required to change the substrate between growths [103]. Even the first systems with load-locks did not have preparation chambers and the system design still resulted in exposure of unbaked surfaces to the growth chamber during deposition [102]. These configurations would have resulted in higher H2O partial pressures than with modern MBE systems, necessitating large pumping capacities for H2O. This raises the question as to whether a different, higher temperature coolant, instead of LN2 might be adequate for cryo-shrouds in modern MBE systems, which also typically have better pumping. H2O and other gases adhere more strongly to metal surfaces than to their own condensate. This allows for pumping by cryo-adsorption [106] at higher temperatures than what would be expected based on the vapor pressure, so long as the surface coverage remains relatively low. At -196C, the vapor pressures of CO, O2, CH4 and CO2, are still relatively high: 500, 200, 10, and 10-8 Torr, respectively, so arbitrary amounts of these gases cannot be pumped by cryo-condensation even at LN2 temperature. As a result, maintaining continuous LN2 cooling of the MBE cryopanels likely results in oxygen and the carbon containing gases not being pumped once the metal surface of the cryopanel is saturated. Warming the cryopanels between MBE growths may improve the pumping of these species, by keeping their coverage on the cryopanels low. However this has the disadvantages of the stress from thermal cycling and dumping the cryo-condensed H2O back into the chamber. Recently, MBE system manufacturers Veeco and Riber began offering split shrouds with a separate cooling loop for the part of the shroud close to the effusion cells. This part can be cooled with water or another closed cycle coolant to minimize LN2 costs. In addition, Eiko Corporation offers several small water-cooled MBE systems, which can be baked with internal lamp heaters between uses to maintain system cleanliness. An unpublished report on the Eiko web page [107] describes GaAs/AlGaAs HEMT structures with low temperature mobilities of 1106 cm2V-1s-1 grown in an Eiko water cooled MBE system. The authors note 42 however, that for cooling the shroud with water at 5C the system is ineffective at pumping As from the process environment. 3.1 Surface adsorption model for H2O The pumpdown of UHV vacuum chambers after being exposed to air is dominated by H2O desorption from the chamber walls. The desorption rate of water from stainless steel is found to follow a t- dependence, where t is the pumping time and 1 [108, 109]. A power law time dependence results from an exponential distribution of binding energies for H2O adsorption sites [108], in the energy range between about 0.7 eV and 1.1 eV for relevant experimental time scales. When this distribution is in equilibrium, the surface coverage as a function of pressure (p) at constant temperature is proportional to pc, where c = T/Tc and kTc is the characteristic energy for the distribution of binding sites [108]. This form for the adsorption isotherm is known as a Freundlich isotherm [110] and the corresponding desorption rate follows a t- dependence with  = 1+c = 1+T/Tc. In the limit that the density of adsorption sites is a constant as a function of binding energy (Tc>>T) a t-1 dependence results. The composition of the cryopanel surfaces in the MBE system vary. The composition of the outside surface of the shroud will be stainless steel, while the inside surface facing the effusion cells is coated with arsenic. Similarly for the titanium sublimation pump, the inner surface will be covered with TiO2 and the outer surface is expected to be stainless steel. The strong adhesion of H2O to the walls of a vacuum chamber is an asset for cryopanel pumping. In MBE systems, after a 200C bakeout the level of H2O contamination on the walls is <<1 monolayer and the empty sites are available for pumping by cryo-adsorption. The adsorption residence time  is given by Eq. 3.2, where o is the attempt time (~10-13 s), U is the binding energy of the site and T is the temperature. If an uncooled surface has reached a steady state with the background H2O 43 pressure in the chamber, adsorption sites with  longer than the time between H2O adsorption events (~105 s for a background H2O pressure of 10-11 Torr) are mostly filled. Sites with shorter lifetimes are mostly empty. Reducing the temperature of surfaces in the MBE increases the residence time of adsorbed gas molecules, increasing the steady state H2O coverage for a given partial pressure. Once the residence time of a site exceeds the process time p, the length of time that the system is operated with the surface cooled, it is considered to be a pumping site since an adsorbed H2O molecule will remain for the duration of the MBE growth process. We assume typical process times are 3-6 hr (p = 1-2104 s). If the H2O partial pressure in the chamber is 10-12-10-10 Torr then it takes ~106-104 s for a monolayer to be incident on the surface. This is the minimum time for the surface to become saturated and the low coverage assumption to hold. For sticking coefficients less than unity the time for a monolayer to adsorb would be longer. Multiple monolayers of H2O can also adhere strongly [111]. We consider cryo-adsorption on a surface where the density of surface sites falls exponentially with increasing desorption energy U. The density of sites as a function of desorption energy is given by, Here go is a constant and kTc is the characteristic energy for the exponential distribution. The density of surface sites has a maximum at the binding energy Um and decreases at lower energy. The functional form of the decrease in density for U1107 /square, indicating the layers were fully depleted. Assuming that the Fermi-level of the layer is pinned at mid-gap at the front and back surface of the epilayer, a background dopant concentration less than 21015 cm-3 would cause the film to be fully depleted. p-type and n-type GaAs epilayers 660-710 nm thick and doped at ~11017 cm-3 were grown on highly doped p and n-type GaAs buffer layers, respectively. These were in turn grown on doped GaAs (100) substrates of the same type. Samples were grown with the shroud cooled with LN2 and to -70C with the chiller. These growths were carried out by D.A. Beaton. Schottky diodes were made by depositing 400 μm diameter circular Al contacts on the p-type layers and Cr/Au contacts to the n-layers. In both cases a large-area Cr/Au contact was deposited on the back of the wafer, forming a low resistance back contact common for all devices. Capacitance-voltage (C-V) measurements were performed on the Schottky diodes at room temperature to determine the built in potential and doping concentration. Doping concentrations were (1.00.1)1017 cm-3 in the pair of samples grown using LN2 and (1.60.1)1017 cm-3 in the pair grown using the chiller. The built-in voltage was 0.67±0.01 eV and 0.94±0.04 eV for the p- and n-type Schottky diodes, respectively. The concentrations of deep level defects in the grown films were assessed by deep-level transient spectroscopy (DLTS) measurements on the Schottky diodes using a SULA Technologies Deep Level Spectrometer. The C-V and DLTS measurements were carried out by P. M. Mooney and K. P. Watkins at SFU. Table 3.3 summarizes all the traps found in the p-type and n-type samples. Hole and electron traps are labeled as “H” and “E” respectively, followed by the thermal emission activation energy of the trap in eV. The trap activation energy is for hole emission to the valence band maximum in p-type samples and for electron emission to the conduction band minimum in n-type samples. For the p-GaAs samples, three traps are seen in the sample grown with LN2 in the shroud, and these traps plus an additional trap E(0.71) are seen in the sample grown with the closed cycle chiller. In the n-GaAs samples, no traps are seen in the sample grown with the chiller and only a weak trap E(0.62), likely the M6 trap reported by Lang et al. [118], is seen in the sample grown with LN2. 57 Table 3.3. Concentrations of all deep levels seen in p-GaAs and n-GaAs samples grown with the shroud cooled with LN2 and the closed cycle chiller at -70C as determined by DLTS. Trap NT (cm-3) LN2 shroud NT (cm-3) -70C shroud H(0.25 eV) 11014 21014 H(0.41 eV) 31014 51014 H(0.54 eV) 81014 21015 H(0.71 eV) --- 21015 E(0.62 eV) 31014 --- All trap concentrations were found to be 21015 cm-3. For the p-GaAs samples, concentrations are about a factor of two higher in the sample grown with the chiller. Trap H(0.54) is always seen in p-type samples grown in this MBE system and is believed to be related to Fe impurities, while traps H(0.25) and H(0.41) may be due to Cu [119]. Possibly these impurities are present in the source materials. Concentrations of trap H(0.54) have fluctuated between 41014 cm-3 and 51015 cm-3 over 6 years of growths with LN2 cooling in our MBE system, so a factor of two is not deemed to be significant. The origin of H(0.71), the hole trap only seen in the p-GaAs sample grown with the closed cycle cooling setup, is unknown. This trap is unlikely related to oxygen impurities as the dominant oxygen related traps are electron traps [120] and the n-GaAs sample, grown with the closed cycle chiller, showed no traps at all (<51013 cm-3). For more information on the DLTS results, the reader is referred to the paper by Lewis et al. [113]. Photoluminescence (PL) spectra were collected from the p and n-type GaAs layers with an optical spectrometer equipped with an LN2 cooled InGaAs array detector. The samples were excited by 20 ns pulses at 523 nm from a frequency doubled Nd:YLF laser. Room 58 temperature PL spectra for the p and n-type GaAs layers grown with the two different cooling systems are shown in Fig. 3.5. No difference in the emission intensity is seen for the p-GaAs samples and the small intensity difference between the two n-type samples is judged not to be significant. The surface treatment for both sets of samples is the same and consists of extended exposure to lab air. Figure 3.6. Room temperature photoluminescence spectra for n and p-type GaAs samples grown under similar growth conditions with LN2-cooling (-196C) and closed-cycle chiller cooling (-70C) of the shroud. Sample numbers for p- and n-GaAs are r2229 and r2230 for samples grown with LN2, respectively, and r2233 and r2234 for samples grown with the chiller, respectively. Figure originally published in [113] 59 3.4 Properties of AlGaAs layers grown with the closed-cycle cooled shroud It is well known that oxygen and carbon containing gas species react more strongly with and produce more detrimental effects on Al-containing layers than GaAs [120, 121]. Therefore it is of much interest to characterize AlGaAs layers grown with the new cooling arrangements. Before installing the closed-cycle shroud cooling setup, AlGaAs growth had not been carried out with our MBE system in several years. As such, direct comparison of the effect of the new cooling system on AlGaAs quality is not possible. Nevertheless, several AlGaAs layers, both nominally undoped and n-doped with Si were grown with the chiller setup to assess the quality of these layers. Figure 3.7 shows room temperature PL from several AlGaAs layers of varying Al-content on GaAs. The PL intensities are relative to the emission of a p+GaAs wafer, which is a strong emitter. Each spectrum shows two emission peaks, one at short wavelengths corresponding to the AlGaAs layer and the other corresponding to the GaAs buffer layer. The AlGaAs layers are 550-920 nm thick and were grown at growth temperatures between 555C and 700C. The two spectra corresponding to 700C are from the same layer. Above ~650C Ga desorption is significant [92], and the non-uniform temperature profile across the wafer caused variation in composition and layer thickness in this sample. Near the centre of the wafer, which is presumably the hottest part, the layer contained 49% Al and was 550 nm thick, while near the edge the layer was found to contain the 29% Al and was 920 nm thick. The weaker PL from the Al0.49Ga0.51As layer is likely a result of this layer having an indirect bandgap, while the other layers have direct bandgaps. Compositions were inferred from AlGaAs emission wavelengths. Photoluminescence from AlGaAs layers can be very weak due to non-radiative recombination associated with deep level defects associated with oxygen [121-123]. As a result, PL from AlGaAs is known to decrease with increasing H2O, CO and CO2 contamination levels. The PL intensity also increases with increasing growth temperature, as the incorporation of these species is thought to be lower at higher temperatures [121, 122]. 60 The strong emission from the AlGaAs layers in Fig. 3.7 is taken as an indication of good quality AlGaAs. Figure 3.7. Room temperature PL spectra of AlGaAs layers on GaAs grown with closed-cycle cooling of the shroud. Each spectrum shows emission from the AlGaAs layer and the GaAs buffer. The Al0.49Ga0.51As layer is 550 nm thick and the other layers are ~920 nm thick. The PL measurement conditions are the same for all the samples. Sample numbers are r2436, r2413 and r2442 for the samples grown at 555C, 580C and 700C respectively. Figure originally published in [112]. A 600 nm thick Al0.22Ga0.78As layer, n-doped with Si, was grown at 580C to assess the transport properties of the layer. Hall measurements indicate the layer has a free electron concentration of 1.61018 cm-3 and electron mobility of 1020 cm2V-2s-1. The carrier concentration was the expected value based on calibrations of the Si flux for n-GaAs and the measured electron mobility is in agreement with literature values [124]. Hall measurements were also performed on the nominally undoped Al0.31Ga0.69As sample grown at 555C (PL shown in Fig. 3.7) to assess background doping. This 920 nm thick sample showed a high sheet resistance of 7108 /square, indicating the film is depleted. The conductivity was p- 61 type with a hole mobility of 192 cm2V-1s-1, in agreement with literature values in this Al concentration range [124]. Assuming the Fermi level is pinned at mid-gap at the front and back surfaces of the layer, a p-type doping concentration of 41015 cm-3 or less would cause the layer to be depleted. The observation of strong PL and the expected electron mobility of AlGaAs layers provides compelling evidence that good quality AlGaAs layers can be grown with the closed-cycle-cooled shroud. To further evaluate the quality of the AlGaAs layers, low temperature measurements could be undertaken. A frequently used factor of merit for AlGaAs quality is the intensity ratio of bound exciton and donor to acceptor recombination at 4 K, also, observation of free exciton emission is evidence of very low defect levels [123]. We have measured PL at temperatures down to ~11 K on the samples from Fig. 3.7, however, we could only resolve a single peak. Perhaps the most important test of the quality of the AlGaAs will be how the new cooling configuration affects the performance of devices that contain these layers. We expect the increase in the oxygen-content of AlGaAs grown with the closed-cycle chiller to be modest and in line with total increase in the partial pressure of oxygen-containing species in the chamber. Aluminum has the lowest electronegativity of the group III and V elements. As a result, it reacts the strongest with oxygen containing gases. Aluminum-free III-V alloys, such as GaAsBi should be less affected by the increased CO and CO2 levels present in the chiller-cooled MBE chamber. Furthermore, the group V elements all have higher electronegativities than Ga, and so changing the anion should not increase the sensitivity to oxygen containing species from that of GaAs. Elements with lower electronegativities than Al, such as Mg and Mn, may be more adversely affected by the increased shroud temperature. 62 Chapter 4 The dependence of Bi incorporation in GaAsBi on MBE growth conditions Understanding the underlying physics of an MBE growth process can facilitate the precise and reliable growth of films with desired compositions and material properties. Such an understanding is particularly important for GaAsBi, where the Bi incorporation is not determined by a simple ratio of the source fluxes under most growth conditions. The first two sections of this chapter describe previous investigations, where important knowledge pertaining to the GaAsBi growth process was obtained. This includes a growth study and a proposed growth model by Lu et al. [76], which I contributed to during my M.A.Sc. research. Later, a systematic investigation of the dependence of Bi content on MBE growth conditions is presented and a new growth model is proposed [77], offering a new interpretation of the physical processes governing GaAsBi growth. This model, which differs from that of Lu et al., provides quantitative agreement over a much wider range of growth conditions. 4.1 Bi wetting layer coverage on GaAs Using RHEED, Young et al. have studied the adsorption and desorption of Bi from a GaAs surface [74, 125]. It had already been reported, for the growth of InGaAs/GaAs structures, that the presence of a sufficiently large Bi flux causes the RHEED reconstruction to change from the normal (24) pattern to a (13) pattern [75]. Young et al. showed that on a GaAs surface, heated to above 400C with the co-deposition of As2 and Bi fluxes, the intensity of the specular reflection of the RHEED pattern increases with increasing Bi flux when the flux is in an appropriate range. For sufficiently large Bi flux, the RHEED intensity is found to saturate. Like the usual interpretation that the peak of a RHEED intensity oscillation corresponds to the completion of the growth of one monolayer, the saturation of the RHEED intensity was interpreted as corresponding to a full monolayer of Bi on the surface. Figure 63 4.1 shows plots of the Bi coverage (inferred from the RHEED intensity) as a function of substrate temperature and Bi BEP for a GaAs surface exposed to fluxes of As2 and Bi. The presence of As2 was to ensure that the GaAs surface remained As-terminated, as Bi desorption is expected to be different for As and Ga-terminated surfaces. It is likely that Bi is more strongly bound to a surface that is Ga-terminated. The fact that the RHEED intensity saturated and there was no evidence for the formation of Bi droplets, led to the hypothesis that the coverage could be modeled with a Langmuir adsorption isotherm. For the Langmuir model, the bonding of the adsorbate atoms to themselves is assumed to be much weaker than the bonding of the adsorbate to the surface. In this case, adsorption occurs only on the bare GaAs surface and saturates at a single monolayer. The Langmuir isotherm can be derived kinetically by noting that in equilibrium adsorption and desorption rates are equal. The rate of Bi adsorption is assumed to be proportional to the impinging Bi flux times the fraction of surface that is bare, while the desorption rate is proportional to the Bi coverage. This leads to the expression in Eq. 4.1 for the Bi coverage (θBi), where FBi is the impinging Bi flux on the surface and b is the product of the Bi site area and the residence time of a Bi atom on the surface. The expression for b is given by Eq. 4.2, where o is the Bi site area and o/2 is the vibrational frequency of the Bi atom on a surface site, assumed to be 1012 s-1 [74, 126]. Uo is the binding energy for Bi in the dilute limit where the surface Bi atoms do not interact with each other. The term Bi takes into account lateral interaction of adjacent Bi atoms in a mean field way. This term results in a steeper transition from low coverage to full coverage, but does not shift where the transition occurs as a function of temperature and pressure. The Bi coverage from a Langmuir isotherm model also plotted in Fig. 4.1, where the Langmuir isotherm is fit with Uo = 1.8±0.4 eV,  = 0.12 eV and o = 0.2 nm2 [74]. 64 Figure 4.1 a) Bi coverage (inferred from RHEED measurements) as a function of substrate temperature. The surface was exposed to a Bi BEP of 1.4 10-5 Torr. b) Bi coverage as a function of Bi flux at a substrate temperature of 460C. Here the Bi source temperature was ramped linearly. The coverage from the Langmuir model is plotted on a) and b). Figure originally published in [74]. 65 The desorption energy of Uo = 1.8±0.4 eV is equal to the Bi self desorption energy of 1.7 eV [74], which implies multiple layer accumulation cannot be neglected. Multiple layer adsorption models, like the model by Brunauer, Emmett and Teller (BET) [126, 127], may be more applicable for Bi adsorption than the Langmuir model. The BET model allows an arbitrary number of monolayers to accumulate. A special desorption energy is used for the first monolayer with all subsequent monolayers having the self desorption energy. The RHEED intensity saturating for sufficiently large Bi fluxes and low temperatures may not indicate that the Bi coverage saturates at a single monolayer, but only that the surface is fully covered with Bi. If Bi does not accumulate layer-by-layer then there is no reason to expect RHEED oscillations as multiple layers accumulate. Metallic Bi droplets can form on the surface at GaAsBi growth temperatures, indicating Bi accumulation beyond a single monolayer at these conditions. For normal GaAsBi growth, conditions resulting in more than a monolayer of Bi on the surface should be avoided. So, despite potential deviations at high Bi coverage, the Langmuir model can still be useful at GaAsBi growth conditions. A simple interpretation is that after the completion of the Bi monolayer, the formation of Bi droplets or multiple layers is likely. 4.2 Lu et al.’s GaAsBi growth study Lu et al. [76] systematically explored the effect of varying the Bi flux, As flux and growth temperature on Bi incorporation. Figure 4.2 shows the measured Bi-content plotted against the atomic Bi:As flux ratio. The curves correspond to a growth model, which is discussed below. In one data set (solid triangles), the Bi flux was varied over a wide range with the other parameters fixed. This shows that for low Bi fluxes the incorporation increases with increasing Bi flux, and at high Bi fluxes the incorporation is independent of the Bi flux, presumably because the surface is saturated with Bi. In another data set (open symbols), the growth temperature was varied between 270C and 320C with the other parameters held fixed, showing the Bi-content increases as the temperature is decreased. Finally, the effect of varying the As2 flux was explored for two different Bi fluxes (solid squares and circles) in a fairly narrow range, showing the Bi-content increases as the As:Ga ratio is lowered close to 66 the 1:1 flux ratio. The authors were able to grow GaAsBi films with up to 10% Bi, which was the record incorporation at the time. A growth model was proposed by Lu et al. [76], which is the first attempt to explain the growth conditions dependence of Bi incorporation in terms of physical processes. This model provided quantitative agreement with the authors’ experimental data. As is shown in the later sections of this chapter, the As2:Ga ratio, which was not studied in detail by Lu et al., is the most sensitive parameter in the growth of GaAsBi. The treatment of the As2:Ga dependence of the Bi incorporation is the main shortcoming of Lu’s model. As was shown by Young et al. [74], and discussed above, Bi has a strong tendency to surface segregate and form a wetting layer on GaAs. Lu et al. assumed this wetting layer to be the source of Bi for incorporation, which sits on the GaAsBi semiconductor surface. The semiconductor surface is assumed to be group-V-terminated, as the atomic As flux is typically greater than the Ga flux, even for the growth of GaAsBi. An illustration of this model is shown in the inset of Fig. 4.2, where three processes proposed to affect the incorporation of Bi into the semiconductor surface layer are indicated: 1) a Ga adatom inserts between an incorporated As and the Bi surfactant, forming an As-Ga-Bi bond. This is the Bi incorporation step, which is assumed to be proportional to the product of the Bi coverage (θBi), the Ga flux (FGa) and the fraction of the surface that is As-terminated (1-x). 2) A Ga adatom inserts between an incorporated Bi and the Bi wetting layer, forming a Bi-Ga-Bi bond. This process was neglected as it was assumed that the large strain resulting from having two Bi atoms so close would be prohibitive. 3) An incoming As atom displaces an incorporated Bi. The rate of this process is proportional to the As flux (FAs). The process involves breaking a Ga-Bi bond and consequently is assumed to be thermally activated. Equation 4.3 was proposed for the rate of change of the Bi concentration (x) of the semiconductor surface, where processes 1) and 3) correspond to the two terms on the right hand side of the equation. For steady state growth, x is assumed to be equal to the bulk Bi-content. In Eq. 4.3, U1 is the activation energy for As to displace Bi, and a is a constant that accounts for the relative cross sections for the two processes. 67 Figure 4.2. Bi-content as a function of the Bi:As flux ratio, showing experimental data (points) and Lu’s incorporation model (curves). The solid triangles correspond to varying only the Bi flux, while the solid circles and squares vary only the As flux (for two separate Bi fluxes, respectively). For the hollow points, only the substrate temperature was changed. The solid lines are model curves for varying the As flux (with the highest Bi:As ratio of each curve corresponding to the 1:1 As:Ga flux ratio). The broken lines are model curves for varying the Bi flux at different substrate temperatures. The inset shows an illustration of Lu’s incorporation model, illustrating the three processes which are proposed to affect Bi incorporation, as discussed in the text. Figure originally published in [76]. Solving Eq. 4.3 in the steady state gives Eq. 4.4. A modified Langmuir isotherm, Eq. 4.5, is used for the Bi coverage (Bi). Equation 4.5 neglects the lateral interaction of Bi atoms, for simplicity. Also, the Bi incorporation rate (xFGa) has been subtracted from the incident flux 68 (FBi). The bo and Uo terms have the same interpretation as in Eq. 4.2, but they were used as free parameters to fit the incorporation data. The model curves in Fig. 4.2 are obtained for Uo = 1.3 eV, bo = 8.510-11 nm2s, U1 = 0.8 eV, a = 2.5108 [76]. Lu’s model was able to correctly account for the experimental data in Fig. 4.2: for low Bi:Ga ratios the incorporation is proportional to the Bi:Ga ratio, and at high Bi:Ga ratios the incorporation is independent of the Bi flux, as the surface is saturated with Bi. If the Bi flux is sufficiently large to maintain a Bi wetting layer, the model predicts that the incorporation should increase as the substrate temperature or the As2:Ga ratio is reduced, until a Bi-content of 100 % is reached. The As:Ga flux ratio dependence has the form 1/(cFAs/FGa+1), where c is a constant. This saturates at 100% as the As:Ga ratio approaches zero. As will be shown in the next section, this prediction disagrees with recent experimental data. Below an As:Ga atomic ratio of unity, the assumption that the growing surface is group-V-terminated is obviously invalid. Therefore, the Lu model cannot be expected to apply in this range. The fact that an As:Ga atomic flux ratio close to unity is required to achieve Bi incorporation, begs the question of whether the surface can ever be assumed to be group-V terminated for the growth of GaAsBi. This conundrum, along with new data, has led to a new model being proposed. The model provides quantitative agreement over a much wider range of growth conditions, and has allowed us to reach new levels of Bi incorporation, up to 21.8% Bi. This takes the GaAsBi alloy, undeniably, well beyond the “dilute” range. 69 4.3 The dependence of Bi-content on MBE growth conditions In this section, a systematic study of the dependence of Bi-content on the As2:Ga BEPR, the Bi:Ga BEPR and the growth temperature is presented. Figure 4.3 shows the As2:Ga BEPR dependence of Bi incorporation. Sets of samples were grown at substrate temperatures 220-230C, 265C and 330C, with Bi:Ga BEPRs of 0.47, 0.35 and 0.09, respectively. The Bi fluxes are expected to result in surfaces that are saturated with Bi. The 330C samples were grown at a growth rate of 1.0 μm/h while the other samples were grown at 0.13 μm/h. The composition of these films was determined with high resolution x-ray diffraction and selected HRXRD scans are presented in chapter 5. Interestingly, the three data sets show similar behavior. Below an As2:Ga BEPR of ~2.25 the Bi incorporation saturates. Further lowering of the As2:Ga BEPR does not increase the Bi incorporation. For BEPRs between 2.25 and 3.6, the Bi-content decreases strongly with increasing As2:Ga BEPR. Above an As2:Ga BEPR of 4.5, no Bi incorporation was detected with x-ray diffraction (<0.1% detection limit). Intriguingly, the As2:Ga BEPR of 2.25 corresponds to the unity atomic As:Ga ratio, as measured for GaAs growth and discussed in section 2.24. The uncertainty in the As2:Ga BEPRs is ~10%. Growths at substrate temperatures of 265C and 330C with As2:Ga BEPRs > 4.5 showed no Bi and (13) RHEED pattern, while the Bi containing samples showed (23), (21) or (2chevrons) RHEED patterns at these temperatures. Samples grown at 220-230C with As2:Ga BEPRs < 3.6 showed (21), (11) and spotty RHEED patterns, while the sample grown with As2:Ga BEPR above 4.5 showed no observable RHEED pattern. The solid lines correspond to the incorporation model discussed below. The As2:Ga dependence shown in Fig. 4.3 cannot be explained with Lu’s model. Below the BEPR of 2.25 the incorporation is flat, rather than following the predicted 1/(cBEPR+1). Above a BEPR of 2.25, the incorporation falls much faster than 1/(cBEPR+1). As mentioned before, it should not come as a surprise that the Lu model fails for Ga-rich growth conditions, due to the assumption of a group-V-terminated surface. The sharp change in the Bi incorporation at the unity atomic As:Ga ratio is suggestive that the incorporation actually depends on the stoichiometry of the growing surface. 70 Figure 4.3. Bi-content as a function of As2:Ga BEPR, for samples grown with large Bi fluxes that are expected to result in Bi-saturated surfaces (data points). The curves correspond to the model discussed in the next section and are plotted as a function of the flux ratio on the top scale. The kink in the data at a BEPR of ~2.25 corresponds to an As:Ga atomic ratio of unity. A temperature of 225C was used to draw the curve corresponding to the 220-230C data. Figure originally published in [77]. A plot of the Bi-content as a function of the growth temperature is shown in Fig. 4.4. Samples were grown with As2:Ga BEPR in the range 0.81-1.7, below where the Bi incorporation has saturated, as indicated in Fig. 4.3. Ga fluxes would have resulted in 0.13 µ/h and 1.0 µ/h growth rates, had the low As2:Ga ratio not resulted in Ga droplets. The solid data points were grown with a large Bi:Ga BEPR of 0.59 ± 0.06. The open triangles had a Bi:Ga BEPR of 0.09 and the open circle had a Bi:Ga BEPR of 6.5. These conditions are expected to yield Bi incorporation, which is maximized with respect to As2:Ga and Bi:Ga BEPR, however, they also guarantee Ga and Bi droplets on the surface. Bi incorporation was 71 found to increase with decreasing temperature, with the lowest growth temperature of 200C resulting in 21.8% Bi. An exponential fit to the data below 270C gives an activation energy of 0.25±0.01 eV. Comparing the samples grown at 1.0 μm/h and 330C (open triangles) with the 0.13 μm/h data suggest the Bi-content depends on the growth rate in this temperature range (for constant As2:Ga and Bi:Ga BEPRs), while at 225C the Bi-content appears to be independent of the growth rate. Possible reasons for this discrepancy are discussed in the next section. The solid line in the figure corresponds to the incorporation model discussed below. Figure 4.4. The temperature dependence of Bi-content for samples grown with As2:Ga flux ratios <0.5. Bi:Ga BEPRs were 0.59 ± 0.06 for the solid data points, 0.09 for the open triangles and 6.5 for the open circle. The solid line corresponds to the model discussed in the next section, calculated for the growth conditions of the solid-circle points. Figure originally published in [77]. 72 A plot of the Bi-content as a function of the Bi:Ga BEPR is shown in Fig. 4.5 for samples grown at 330C and a growth rate of 1.0 μm/h. The highest Bi:Ga BEPR sample was grown with an As2:Ga BEPR of 3.3 (As2:Ga flux ratio of 0.73), while the other samples had BEPRs varying between 2.5 and 3.3 (As2:Ga flux ratios of 0.56-0.73). Optical microscopy confirmed that all samples were droplet free, except the hollow point sample, which had droplets on the surface. At low Bi flux, when the Bi wetting layer coverage is low, the Bi-content is proportional to the Bi:Ga BEPR, consistent with Ptak et al. [128]. In this unsaturated range, the Bi-content appears to be independent of small fluctuations in the As2:Ga flux ratio. Figure 4.3 indicates that in this same flux range, for Bi-saturated surfaces, these fluctuations would results in Bi-content fluctuations. At high Bi:Ga ratios, the Bi-content is independent of the Bi:Ga ratio, but depends on the other growth conditions, as indicated by Fig. 4.3 and Fig 4.4. As the Bi flux increases, eventually the surface becomes saturated with Bi, after which, the maximum incorporation is independent of the Bi:Ga BEPR. It is expected that Bi droplets start to appear as the Bi coverage approaches unity. The fact that only the highest Bi:Ga BEPR sample had droplets is consistent with this interpretation. The relationship between Bi:Ga BEPR and the atomic flux ratio is determined from the linear portion of the data in Fig. 4.5, assuming that at low Bi:Ga ratios the Bi-content is equal to the flux ratio. The resulting relationship is 1.5 times less than what was determined by profilometry measurements on a masked Bi metal film, as discussed in section 2.2.4. The linear portion of the data in Fig. 4.5 yields the relation FBi/FGa = (0.323  0.008)(Bi:Ga BEPR). This relationship is used in the growth model discussed below. The curves shown on Fig. 4.5 correspond to the Bi-content and Bi wetting layer coverage from the model discussed below. 73 Figure 4.5. Bi:Ga BEPR dependence of Bi-content for samples grown at 330oC and 1.0 μm/h, with As2:Ga BEPRs between 2.5 and 3.3 (As2:Ga flux ratio between 0.56 and 0.73). The sample corresponding to the open circle has droplets on the surface, while the other samples do not. The curves correspond to the model discussed in the next section. The black curves are calculations of Bi-content as a function of the Bi:Ga flux ratio (indicated on the top scale) for various As2:Ga flux ratios . The blue curves are model calculations of the Bi surface coverage (right hand scale) as a function of Bi:Ga flux ratio. 4.4 A new GaAsBi growth model For the growth model proposed by Lu et al. [76], the dependence of Bi content on the As flux results from As displacing incorporated Bi. The observation that the Bi content is independent of As2:Ga ratio below the unity atomic flux ratio and then decreases much faster than the As2:Ga BEPR cannot be explained entirely by incoming As2 displacing incorporated Bi. It is proposed instead that the Bi incorporation depends on the Ga:As surface coverage ratio of the growing film. Careful control of the As2:Ga BEPR is required in GaAsBi growth since the Ga coverage is very sensitive to the As2:Ga flux ratio near 0.5. With this insight, a 74 new model for Bi incorporation has been developed. It is assumed that any site of the semiconductor surface can be terminated with either an incorporated As, Ga or Bi atom. A Ga-terminated site can equivalently be thought of as a group V vacancy. In addition, a Bi wetting layer is allowed to exist on top of the semiconductor surface. This wetting layer acts as the source of Bi atoms for incorporation into the GaAsBi surface layer. Figure 4.6 illustrates a GaAsBi surface, indicating two processes that affect the net incorporation rate of Bi atoms at the crystal surface, . Here is the number of Bi atoms per unit area in the layer. Process 1) corresponds to the incorporation of a Bi atom from the wetting layer on a Ga site. The rate of this process is assumed to be proportional to the product of the Bi wetting layer coverage and the Ga surface coverage (GaBi). Process 2) is the thermal ejection of an incorporated Bi atom from an incorporation site (bonded to Ga) back into the wetting layer, which is proportional to the film Bi-content and characterized by the activation energy U1. Figure 4.6. A model of the semiconductor surface and Bi wetting layer illustrating: 1) incorporation of Bi on a Ga site, 2) thermal ejection of incorporated Bi. For steady state film growth, is equal to the Bi concentration x times the Ga incorporation rate . For growth that is free of Ga droplets, the Ga incorporation rate is simply the Ga flux, FGa. The net rate of Bi incorporation at the crystal surface is given by Eq. 4.6, where the two terms on the right hand side of this equation correspond to the two processes indicated in Fig. 4.6. The factors a1 and a2 are rate constants. 75 The modified Langmuir isotherm of Lu et al. [76] (Eq. 4.5) is applied here for Bi. Rather than fitting the incorporation energy (Uo) and rate constant (bo), the values obtained from Young’s RHEED experiments (discussed in section 4.1) are used, namely U0 = 1.8 eV and bo = 2o/o, where o = 0.2 nm2 and o/2 = 1012 s-1. The As2:Ga flux ratio dependence of the Ga surface coverage Ga is calculated in the following sections. Experimentally, the Bi-content is not found to decrease with increasing growth rate, for fixed Bi, As2:Ga flux ratio and growth temperature. This indicates the rate of thermal ejection ( in Eq. 4.6) is much larger than the net rate of Bi incorporation . The physical interpretation is that the incorporated Bi on the surface and the Bi in the wetting layer are nearly in thermal equilibrium, such that Bi atoms are rapidly exchanging between the incorporation sites and the wetting layer. The steady state Bi-content can then be determined by balancing the incorporation term a1GaBi and the thermal ejection term as these are large compared to . We note that neglecting the net incorporation term is not necessary to solve Eq. 4.6, rather it is based on the experimental observation in Fig. 4.4. 4.4.1 The stoichiometry of GaAs surfaces Extensive work has been done on modeling various dynamics of GaAs growth. Surface stoichiometry during GaAs growth has been modeled using Monte Carlo [129, 130]. Also, the stoichiometry of GaAs surfaces exposed to beams of As2 has been modeled analytically [131]. To obtain an analytical expression for the stoichiometry during GaAs growth, a model is proposed. This model allows hopping of Ga and As adatoms on the GaAs surface, assuming a solid-on-solid configuration (no vacancies or overhangs). The assumptions are analogous to those used to calculate surface stoichiometry and other features of GaAs growth using Monte Carlo simulations [129, 130]. The GaAs surface is composed of surface sites that can be populated by either As or Ga. Incident As2 molecules are assumed to dissociate into two As adatoms that diffuse on the surface. Incident Ga atoms also diffuse on the surface and reactions between the free Ga and 76 As adatoms are neglected. The As adatoms will permanently attach if they land on a site that is populated by Ga, converting that site into an As site and so increasing the As surface coverage (As). If As lands on an As site, it will either evaporate with probability PAs or hop to a new site with probability 1-PAs. Thus the As adatom searches the surface for Ga sites by random hopping events, and upon each hop there is a probability that the adatom evaporates and is lost from the surface. For an incident As atom, the probability that it will find a Ga site before evaporating is given by a geometric series. The effect of an incident As flux on As is given by the first term in Eq. 4.7, where the term contains the aforementioned geometric series. Equation 4.7 is the rate of change of the As surface coverage, As. In this equation FGa is the Ga flux and FAs is the atomic As flux, which is twice the As2 flux. Ga adatoms undergo a similar hoping process, attaching only when landing on As sites. Ga evaporation is negligible at the temperature range of interest for GaAs and GaAsBi growth [92], however, there is a small probability (PGa) for each hop that the Ga atom will be lost to droplet formation. This becomes important at low As2:Ga ratios, where Ga atoms undergo many hopping events in search of As sites. The effect of an incident Ga flux on As is given by the second term in Eq. 4.7, which also contains a geometric series for Ga hopping. These geometric series’ are easily evaluated and have been summed in the second line of Eq. 4.6. The As coverage (As) is the fraction of the surface sites that are populated by As. A steady state solution for the corresponding Ga coverage, Ga, is obtained from Eq. 4.7, noting the surface normalization condition, As+Ga = 1. The Ga coverage (Ga) is plotted as a function of As2:Ga flux ratio in Fig. 4.7 for several different values of PAs and PGa. Increasing PAs results in larger Ga coverage for As2:Ga flux ratios above 0.5. Choosing PGa to be very small, results in Ga 1 for As2:Ga ratios less than 0.5. Good agreement with the As2:Ga dependence of the Bi-content in Fig. 4.3 is obtained for PAs and PGa values of 0.12 and ~0.001. The choice of PGa is not important and only has to be small enough to result in G1 77 for Ga-rich conditions. The evaporation probability per hop, PAs, is the ratio of the evaporation rate and the hopping rate. These are two thermally activated processes, so it is expected that PAs is also temperature dependent and of the form , where Ee and Ed are the energy barriers for As evaporation and diffusion and ko is a constant. Figure 4.7. The calculated Ga surface coverage as a function of the As2:Ga flux ratio for several choices of PAs and PGa. Monte Carlo simulations by Kaspi and Barnet [129] allowed Ga and As adatoms to diffuse on the surface by randomly hopping between sites. In this case, a fixed number of hops were used for Ga and As adatoms. If they did not find incorporation sites within the allowed number of hops, As adatoms would desorb and Ga atoms would be captured by (or nucleate) droplets. This is very similar to our model, except we allow finite evaporation and droplet probabilities for each hop, and solve for the coverage analytically. Fitting to measurements of the surface stoichiometry at the RHEED transition from the As-rich (24) to the Ga-rich 78 (42) at 450C, Kaspi and Barnet [129] obtained good agreement when As adatoms were allowed to take 18 hops before desorbing and Ga adatoms were allowed to take between 50 and 100 hops [129]. This corresponds to PAs and PGa values of 0.06 and 0.01-0.02, respectively. These values are comparable to those obtained by fitting to the Bi incorporation data, which was obtained at lower temperatures and under different surface reconstructions. 4.4.2 As and Ga hopping on GaAsBi surfaces In this section, our As and Ga hopping model is applied to the GaAsBi surface. With the presence of Bi, a surface site can be populated with either As, Ga or an incorporated Bi atom. In this case the normalization condition of the crystal surface layer is Ga +As +xs = 1, where xs is the fraction of the surface sites populated by incorporated Bi atoms, which is assumed to be equal to the bulk Bi-content (x) in the steady state. The Bi wetting layer, which is assumed to exist on top of the crystal surface in the Bi incorporation model, is assumed not to effect As and Ga adatom hopping. With Bi on the surface, As and Ga adatoms may land on incorporated Bi sites. For As adatoms, there are two extreme possibilities: Bi sites behave like As sites, meaning As does not displace Bi. In this case As either evaporates or hops to a new site; or the sites behave like Ga sites, in which case the Bi is always displaced by the As adatoms. The analogy with the model by Lu et al. [76] is that in the extreme cases, the activation energy for As displacement of Bi is either infinitely large or zero. The choice of whether Bi is displaced by As or not has little effect on the incorporation model that will be presented in the next section. For simplicity and to reduce confusion as to which process is responsible for the As2:Ga dependence of Bi incorporation, it is assumed that As does not displace Bi. As Bi and Ga tend to react weakly, it is expected that when a Ga atom lands on an incorporated Bi site, the probability that it will attach is <<1. From the perspective of a Ga atom the Bi site looks like a Ga site. This assumption is required to reproduce the observation that the Bi-content saturates at low As2:Ga flux ratios. We note that a finite probability for Ga sticking to Bi is required for film growth. 79 The rate of change of As is given in Eq. 4.8 for the case where As does not displace Bi. In the other case, where As always displaces Bi, the expression for the rate of change of As is the same as in Eq. 4.7, where there was no Bi. We note that in this case the resulting Ga coverage still differs from that of the Bi-free case, due to the surface normalization condition. The As2:Ga dependence of the Ga surface coverage Ga on a GaAsBi surface is obtained by solving Eq. 4.8 in the steady, substituting As = 1-Ga -x and finally rearranging for Ga. 4.4.3 Comparing the GaAsBi growth model with experiment A solution for the steady state Bi-content is obtained by substituting the expressions for Bi (Eq. 4.5) and Ga (obtained from Eq. 4.8) into Eq. 4.6 and solving for x. As mentioned above, the net Bi incorporation term in Eq. 4.6 is neglected. This general solution is very cumbersome and so is not shown. The curves in Figs 4.3, 4.4 and 4.5 are obtained by setting PAs = 0.12, PGa = 0.001, U1 = 0.28 eV and a2/a1 = 3300. As mentioned in the previous section, it is expected that the As evaporation probability (PAs) should be temperature dependent. However, a constant value was adequate to fit the data presented in the Fig. 4.3. The model accurately fits the experimental data in Fig. 4.3, indicating that the As2:Ga flux ratio dependence of the Bi-content tracks with the Ga surface coverage Ga. Below a flux ratio of 0.5, the As-coverage As on the surface is negligible and the Bi-content is independent of the flux ratio. Just above the unity atomic As:Ga flux ratio, the rapid decrease in the Bi-content is due to the rapid decrease in Ga as the surface transitions to the As-rich regime. A discrepancy exists for the highest As2:Ga ratio samples in Fig. 4.3, where the model overestimates the Bi-content. The explanation for this could be associated with the reconstruction changing to (13) [79]. It is possible that the Bi incorporation site, which 80 tracks with the Ga coverage on the other surfaces, is not present on the (13) surface. The Bi fluxes from the data used to fit the model curves in Fig. 4.3, were high enough to maintain Bi-saturated surfaces (B1) for samples grown at 220-230C and 265C. Thus the Bi incorporation is saturated with respect to the Bi flux in these conditions, and as such, is primarily dependent on the growth temperature and the As2:Ga ratio. For the samples grown at 330C with the two lowest As2:Ga ratios, the Bi flux was only adequate to produce B  1 after accounting for the fact that at such low As2:Ga ratios, approximately half the Ga flux is lost to droplet formation. This increases the effective Bi:Ga flux ratio. The 330C model curve is drawn assuming a Bi-saturated surface (B = 1). The small Bi thermal ejection energy U1 = 0.28 eV indicates incorporated Bi atoms are weakly bound. This is consistent with Bi behaving as a surfactant, which implies a weak interaction with the semiconductor surface. The model curve in Fig. 4.4 was calculated for the growth conditions of the solid circle data points in the figure (Bi:Ga BEPR = 0.59, As2:Ga flux ratio <0.5 and 0.13 μm/h growth rate). The 0.28 eV ejection energy results in the near-constant slope of the model curve below ~350C. The fall-off of the model for T>350C is due to the onset of Bi evaporation, and consequent loss of Bi coverage. The thermal ejection energy U1 = 0.28 eV was obtained by fitting the data in Fig. 4.3 and Fig. 4.4 by eye. The 4 data points in Fig. 4.4, grown at 0.13 μm/h at substrate temperatures above 300C, were excluded from the fit as they were inconsistent with the majority of the data in these figures. Increasing the Bi flux 11 (open circle) at 350C did not increase the Bi-content at the low growth rate, indicating that the surface is saturated with Bi even for the lower Bi flux data points at this temperature. The reason why the 4 samples, grown with 0.13 μm/h growth rate at substrate temperatures above 300C, have lower Bi-content than predicted by the model, is unknown. Increasing the growth rate can be considered analogous to decreasing the growth temperature, as MBE growth dynamics typically scale as the ratio of the flux and the thermally activated surface diffusion rate. This could explain why the samples grown at 330C and 1 μm/h do not show the same drop in Bi-content as the samples grown at 0.13 μm/h. It is possible that at low growth rate and high substrate temperature, a new process that is not accounted for in the model is beginning to dominate. These samples 81 were grown under conditions resulting in excess Ga and Bi surface coverage. One possibility is an enhanced capture of Bi by droplets on the Ga and Bi covered surface at high substrate temperatures and low growth rates, as these conditions increase the migration of Bi adatoms. Further investigation is required to understand this discrepancy. Figure 4.5 shows model calculations of the Bi-content (black lines) as a function of the Bi:Ga flux ratio for a growth temperature of 330C and a 1.0 μm/h growth rate. The curves are calculated for As2:Ga flux ratios 0.50, 0.60 and 0.73. The Bi surface coverage Bi from the model is also plotted for two As2:Ga flux ratios (blue lines), indicating the Bi incorporation follows Bi. As shown with the Bi plots, the point where the surface saturates with Bi depends on the other growth conditions, such as the As2:Ga flux ratio. It is expected that Bi droplets start to appear as the Bi coverage approaches unity. The composition of the highest Bi:Ga ratio sample is fit best with the As2:Ga ratio of 0.60, despite being grown with the measured ratio of 0.73. In the incorporation model, the As evaporation probability (PAs) was taken as a constant value, which was fit to data mostly at lower growth temperatures. At 330C the As evaporation probability is likely higher. This would result in a higher Ga surface coverage Ga for a given As2:Ga flux ratio, which can explain why a lower As2:Ga ratio is required to fit the data in Fig. 4.5. The model curves in Fig. 4.5 show the Bi incorporation increases linearly with the Bi:Ga BEPR when Bi is unsaturated. Consistent with experimental data, the Bi-content is insensitive to small fluctuations in the As2:Ga flux ratio (i.e. in Ga) when the surface is not saturated with Bi. This differs from the Bi-saturated case shown in Fig. 4.3, which shows that in this range of As2:Ga ratios, small fluctuations in the As2:Ga flux ratio cause the Bi-content to also fluctuate. The stabilization that occurs on the unsaturated surfaces is a result of the fluctuations in Ga being offset by compensating fluctuations in Bi. This insensitivity to the As2:Ga ratio, along with the lack of Bi droplets, motivate keeping Bi < 1 during GaAsBi growth. Although only a single saturated point is shown in Fig. 4.5, the data shown in Figs 4.3 and 4.4 were all grown with Bi-saturated surfaces, where the Bi-content is primarily dependent on the As2:Ga ratio and the growth temperature. This data could not be included on Fig 4.5 as the other growth parameters were different. 82 4.4.4 Making use of the growth model The new incorporation model explains the sensitivity of Bi incorporation to the As2:Ga flux ratio, substrate temperature and Bi:Ga flux ratio. Generally, the Bi-content depends on all three of these parameters. For As2:Ga flux ratios less than 0.5 and for high Bi fluxes, the Bi incorporation is saturated with respect to these parameters. Unfortunately, these conditions also result in excess Ga and Bi droplets, which is not desirable for obtaining smooth films. The preferred parameter to limit the incorporation is certainly the Bi:Ga ratio, which is easily controlled by the Bi and Ga cell temperatures. When the surface is not saturated with Bi, the Bi-content is approximately linear with the Bi:Ga ratio and fluctuations in the As2:Ga ratio are buffered by fluctuation in the Bi surface coverage Bi. Fluctuations in the Bi:Ga ratio will only result in linear fluctuations in the Bi-content. In contrast, fluctuations in the As2:Ga ratio, could result in highly non-linear variations in the Bi-content if the As2:Ga ratio is used to limit the Bi incorporation. To avoid droplets, care must be taken to ensure excess Ga or Bi is not present on the surface. If the incorporation is not limited by the Bi:Ga ratio (i.e. is limited by As2:Ga or temperature), then it is likely that there is excess Bi on the surface. Therefore, the preferred growth method is to choose an As2:Ga ratio and temperature that can allow for more than the desired Bi incorporation, then, regulate the Bi-content down with the Bi:Ga ratio. Nevertheless, the steep dependence of Bi incorporation on the As2:Ga ratio, depicted in Fig. 4.3, illustrates the difficulty in obtaining droplet free high Bi-content films. Alternative growth techniques such as atomic layer epitaxy (ALE) may be useful for obtaining high Bi-contents, while avoiding the buildup of excess Ga and Bi on the surface. For ALE, molecular species are typically deposited sequentially in monolayer or sub-monolayer quantities. By pulsing the As2 flux during growth, Ga-rich surfaces could be created briefly between pulses, with the As pulse periodically removing the excess Ga from the surface. This is one way to produce a Ga rich surface that is free of droplets and could increase the process latitude for growing smooth high Bi-content GaAsBi films. 83 Chapter 5 Structural characterization of GaAsBi layers Achieving a high level of structural quality and reproducibility has been a challenge in the MBE growth of GaAsBi. In this chapter, the structural characterization of GaAsBi layers is discussed. The insight gained from our growth model has allowed films with a high degree of crystalline perfection to be grown. The low growth temperature for GaAsBi is found to inhibit strain relaxation, allowing for highly-strained pseudomorphic films to be grown on GaAs substrates. In addition to growth on GaAs, GaAsBi films were grown under tensile strain on InP and GaSb substrates. This was done to determine whether the large compressive strain, resulting from high Bi-content GaAsBi on GaAs, inhibits Bi incorporation. 5.1 Characterization with high-resolution x-ray diffraction Figure 5.1 shows (004) -2 HRXRD scans of three GaAsBi epilayers on GaAs substrates. The sharp peaks at 0 correspond to the GaAs substrate and the weaker peaks are the GaAsBi layers. The angular separations between the substrate and film peaks give Bi-contents of 5.8%, 14.2% and 21.8% Bi, respectively, for pseudomorphic films. All scans show pendellösung fringes, indicating a high degree of film uniformity and interface perfection. The spacing of these fringes indicates film thicknesses of 54 nm, 53 nm and 17 nm, for the 5.8%, 14.2% and 21.8% Bi films, respectively. Dynamical simulations using these parameters are shown as the broken lines on the figure. The growth temperature, As2:Ga BEPR and observed RHEED patterns for the samples from Fig. 5.1 are: 265C, 2.53, (21) and (2chevrons) for 5.8% Bi; 250C, 2.25, (21) for 14.2% Bi; 200C, 0.85, (11) for 21.8% Bi. The sample with 21.8% Bi was grown well into the Ga-rich growth regime with a large Bi flux. This resulted in a high density of Ga-Bi droplets on the surface. The other samples were grown with atomic As:Ga ratios close to unity (BEPR of 84 2.25 is believed to result in an atomic As:Ga atomic ratio of 1). Despite having a much smoother surface than the 21.8% Bi sample, sub-μm size droplets could be detected on the 14.2% Bi sample under 100 magnification with an optical microscope. No droplets were seen on the sample with 5.8% Bi with the optical microscope or with SEM (section 5.4). As of Nov. 2013, the film with 21.8% Bi currently holds the world record for highest Bi-content [77]. This is nearly double previous reported incorporation records [76, 128, 132]. Figure 5.1. (004) -2 HRXRD scans (solid lines) and dynamical simulations (broken lines) of three GaAsBi films on GaAs. The Bi-contents are 5.8%, 14.2% and 21.8%, and pendellösung fringes indicate thicknesses of 54 nm, 53 nm and 17 nm respectively. Sample numbers are r2313, r2361 and r2267. The GaAsBi layers with 14.2% and 21.8% Bi have a huge amount of mismatch to the GaAs substrate: 1.7% and 2.6%, respectively. Nevertheless, these films were found to be fully strained by reciprocal space mapping of the {224} peaks. Figure 5.2 shows an RSM around the (224) film and substrate peaks for the 14.2% Bi sample. The upper peak is from the GaAs substrate and the lower peak is from the GaAsBi film. The in-plane component of the film 85 peak (qx) exactly matches the substrate, indicating the film is fully strained (relaxation uncertainty 1%). Pendellösung fringes are also seen between the film and substrate peaks. If relaxation were detected, this would need to be included in the simulation of the (004) scan to correctly extract the Bi-content. A solid yellow line is directed from the substrate peak toward the origin. For fully relaxed films, off axis planes have no tilt relative to the substrate planes. As a result, a fully relaxed film peak would lie on a line directed from the substrate peak to the origin. A map of the peaks for this sample showed the film to be pseudomorphically strained in the orthogonal direction. Figure 5.2. An RSM of the (224) film and substrate peaks for the 53 nm thick GaAs0.858Bi0.142 sample from Fig. 5.1 (r2361). The upper peak corresponds to the GaAs substrate and the lower peak is the GaAsBi layer. The solid yellow line points to the origin, indicating the line where a 100% relaxed film would lie. The green contours are spaced by factors of 10 in intensity. In the process of mapping out the growth condition dependence of Bi-incorporation, many attempts were made to grow GaAsBi with high Bi-content (>10% Bi). Often growth 86 conditions resulting in large excesses of Ga and Bi on the sample surface were used. In some cases, attempts were made to grow thick films under these conditions, ignoring surface degradation. It was found that, despite following a growth procedure expected to result in thick layers, only very thin layers were detected in the HRXRD scans. It is possible that too much excess Ga and Bi on the surface effectively stops the layer growth. This effect is not taken into account by the GaAsBi growth model. It is observed that not only is careful control of the As2:Ga BEPR and Bi:Ga BEPR necessary to minimize droplets at high Bi-content, it is also necessary to allow for the growth of thick layers. 5.2 Strain relaxation In this section, dislocation formation in III-V semiconductors is discussed, as is strain relaxation in GaAsBi films. Evidence of a critical layer thickness for relaxation of GaAsBi/GaAs films is observed, but at greater thicknesses than predicted by energy minimizing arguments. This is believed to result primarily from the low growth temperature for GaAsBi. 5.2.1 Formation of dislocations In zincblende crystals, strain-relieving dislocations typically form by glide of {111} slip planes [133]. These dislocations have Burgers vectors of the form , which lie at 60 angles from the dislocation lines ( and lines on a (001) surface). Therefore the Burgers vectors have components both parallel and perpendicular to the dislocation lines, and so these dislocations are of mixed edge and screw character. The atomically dense {111} planes have the largest spacing, resulting in the lowest Peierls stress and activation energy for glide [133]. The critical thickness, for which a layer will no longer grow epitaxially, was famously predicted by Matthews and Blakeslee (MB) [134]. They predicted that dislocations would form if it lowered the total energy of the layer. The critical thickness is determined with a 87 bulk-energy balance approach, by equating the elastic energy released from the glide of a dislocation to the energy needed to produce the dislocation segment. This allows for predicting the thickness where dislocations should begin forming, as well as the residual strain in layers beyond that thickness. The critical layer thickness (hc) of a single exposed layer is given by Eq. 5.1. Here b is the length of the Burgers vector (4.00 Å for GaAs), f is the mismatch,  is Poisson’s ratio,  is the angle between the dislocation line and the Burgers vector (60), and  is the angle between the Burgers vector and the direction of relaxation (60) [134].  is a parameter representing the contribution of the dislocation core energy, which has a value of 4 for most semiconductors [135].    MB’s predicted critical thickness was found to agree with TEM measurements on GaAsP/GaAs layers grown at 750C [134]. However, the energy minimization arguments were insufficient for estimating the amount of residual strain in the layers after relaxation. The theory predicted when already present dislocation would glide, but residual strain in the layers was much greater than expected. This is due to the large energy barrier, relative to kBT, to nucleate new dislocations, as well as the interaction between dislocations. Furthermore, dislocation glide velocity is a thermally activated process, having an Arrhenius temperature dependence. At low growth temperatures, the glide velocity can slow to a point where the time to glide the length of the film exceeds the growth time [136-138]. This results in increasing critical thicknesses as the growth temperature decreases and growth rate increases. In GaAs-alloys, dislocations with lines along are normally observed to have higher glide velocities than dislocations with lines along [139]. The dislocations with lines are known as  -type and have cores located on As sites, while lines are -type and have cores located on Ga sites [140]. This leads to different activation energies for nucleation and glide in the two directions. France et al. [141] have investigated the growth condition dependence of strain relaxation in MBE grown InGaAs/GaAs layers with in situ wafer curvature, HRXRD and transmission electron microscopy (TEM). They found that strain relaxation, especially in the direction, is highly sensitive to V:III flux ratio and 88 substrate temperature during growth. Lowering the V:III flux ratio and/or increasing the growth temperature was found to result in smaller critical thicknesses. Furthermore, the direction in which relaxation happens first is dependent on the growth conditions. It was suggested that this sensitivity to V:III ratio could be related to changes in surface morphology. 5.2.2 Observation of strain relaxation in GaAsBi layers There have been few reports on dislocation formation in GaAsBi alloys. Publications from our group have noted that GaAsBi layers can be grown well beyond the critical thickness predicted by MB [58, 77]. France et al. reported measuring relaxation in GaAsBi layers with in situ wafer curvature [142], where they compared GaAsBi and InGaAs layers grown at 350C with the same V:III flux ratio and amount of mismatch. The GaAsBi layers were found to begin relaxing at a lower thickness than InGaAs, however, thick GaAsBi layers had more residual strain than thick InGaAs layers. This was attributed to a lower energy to nucleate dislocations in GaAsBi, but a higher energy to encourage glide or multiplication. Figure 5.3 shows and RSMs for a 75 nm thick GaAs0.86Bi0.14 epilayer on a GaAs substrate. The sample was grown with a growth temperature of 250C. For each map, the peak at the top corresponds to the substrate and the lower peak corresponds to the GaAs0.86Bi0.14 layer. In the map, the in-plane components of the substrate and film peaks are equal, to within the 1% measurement uncertainty, indicating no significant strain relaxation in the direction. The GaAs0.86Bi0.14 peak has a lower in-plane value than the substrate peak, indicating 71% relaxation in the direction. This sample is thicker than the sample from Fig. 5.2. The asymmetry in the film relaxation shown in Fig. 5.3 is consistent with that typically observed for InGaAs/GaAs samples, which normally show more relaxation in the direction. In contrast, France et al. measured relaxation in two GaAsBi films and found no significant asymmetry in the amount of relaxation [142]. France et al.’s earlier work on InGaAs indicated that relaxation is highly sensitive to growth conditions [141]. It is therefore 89 not surprising that our GaAs0.86Bi0.14 sample, which was grown at different conditions than those of France et al., does show asymmetric relaxation. The predicted critical thickness for a GaAs0.86Bi0.14 layer on GaAs is ~5 nm according to MB. Figure 5.3. RSMs of the and film and substrate peaks from a 75 nm GaAs0.86Bi0.14 sample (r2345). The sample has an asymmetry in the strain relaxation, with 11% in the direction and 71% in the direction. A film with 100% relaxation would lie on the solid yellow line. The GaAs wafer specifications indicate a typical etch-pit density < 5000 cm-2. Assuming a density of threading dislocation of 5000 cm-2, and allowing each one to glide the full length of the ~2 cm sample, results in a density of dislocation lines of 104 cm-1. The mismatch of GaAs0.86Bi0.14 to GaAs is 1.7% and the projection of the Burgers vector in the direction of relaxation is 2 Å. This means that to relieve 100 % of the strain, a misfit dislocation every 12 nm is required. Therefore, glide of substrate dislocations alone can only produce 1% strain relaxation, less than the 71% observed in the direction. This suggests that new dislocations are being nucleated/multiplied, perhaps at the edge of the wafer or on imperfections. In highly strained InGaAs/GaAs layers, strain relaxation is also known to 90 occur through changes in the surface morphology, namely through the formation of 3D islands [143]. Surface morphology changes could also play a role in GaAsBi growth. Strain relaxation was measured in only a few GaAsBi films with HRXRD and only along one direction was significant relaxation detected. Only for the above sample (r2345) was it confirmed that the relaxation direction was along [110], as the other samples had been previously diced such that the crystal directions could not be identified. Figure 5.4 shows a plot of film thickness as a function of mismatch to GaAs and Bi-content, for GaAsBi/GaAs samples where RSMs were collected along both the and directions. The figure indicates for which samples relaxation was and was not detected. The critical thickness for two GaAsBi layers is also estimated from the wafer curvature measurements by France et al. [142]. For reference, the MB critical thickness calculation is shown on the figure. The point at 0.0051 mismatch has overlapping relaxed and strained symbols. This is because the {224} maps showed unusually large broadening of the film peak in the direction of relaxation, even though no bulk relaxation was detected. This was interpreted as the onset of relaxation in the film. The data in Fig. 5.4 indicates a GaAsBi critical thickness about an order of magnitude greater than that predicted by bulk-energy minimization. This is likely a result of the low growth temperature for these layers of 210C-330C. Evidence that the low growth temperatures are responsible for the large critical thicknesses was seen in the growth of a p+/n GaAsBi structure (n-GaAs/n-GaAsBi/p+GaAsBi/p+GaAs). This structure was grown for characterization of deep level defects in the n-GaAsBi layers by DLTS, in collaboration with P. M. Mooney’s group at SFU. After the growth of ~500 nm of n- and p+ GaAs0.972Bi0.028, a p+GaAs cap layer was grown. At the beginning of which, the substrate temperature was ramped from 330C to above 500C in about one minute. During the temperature increase, a crosshatch pattern of dislocations was observed to form on the surface of the wafer. Figure 5.5 shows the in situ diffusely-scattered light signal collected from the rotating substrate, during the growth of the p+/n structure. The substrate temperature was rapidly increased from 330C to 550C near t = 0. The appearance of periodic spikes in the scattered light signal corresponds to scattering from a crosshatch pattern of dislocations on the surface, as the sample is continuously rotated. The intensity of 91 the spikes oscillates with a period of two, suggesting there is a stronger crosshatch pattern along one crystal direction, consistent with the presence of asymmetric relaxation. AFM measurements confirmed that the crosshatch pattern was denser along one direction. Relaxation was detected in the GaAsBi layers with HRXRD, as was tensile strain in the GaAs cap. These results confirm that the low growth temperature of GaAsBi inhibits the formation of strain relieving dislocations. Figure 5.4. A plot of film thickness as a function of mismatch to GaAs and Bi-content, for samples where relaxation was (squares) and was not (crosses) detected with HRXRD. Growth temperatures were between 210C and 330C for these samples. The MB critical thickness is also plotted, as are critical thickness values estimated from the work by France et al. for samples grown at 350C (circles) [142]. The low GaAsBi growth temperature allows layers to be grown well beyond the critical thickness that would be expected from energy minimization. Low growth temperatures can also prevent dislocation formation in other materials, such as InGaAs. However, this may not be practical, due to degradation of other material properties. The fact that GaAsBi layers 92 grown below 400C show strong light emission, suggests that GaAsBi may have a unique ability to take advantage of the increased critical thickness at lower growth temperatures. However, low growth temperatures present challenges for incorporating GaAsBi layers in to devices structures, as the ideal growth temperatures for most III-V semiconductors are well above that of GaAsBi. For many structures, the growth of overlayers and device fabrication steps would require heating GaAsBi above its growth temperature. Even in this case, the low GaAsBi growth temperature can still provide critical thickness benefits. The critical thickness of a capped layer is 2 that of an uncapped epilayer [134]. The factor of 2 results from the formation of a dislocation in a capped layer producing two dislocation lines (one on the top interface and one on the bottom), rather than a single dislocation line on the lower interface for an uncapped layer. This increases the energy of the dislocation by 2. Alternatively, the benefits of growth with a Bi surfactant [73-75] could allow high quality capping layers to be grown below their normal growth temperatures. Figure 5.5. Diffusely scattered light signal from a rotating wafer, collected during the growth of a p+/n GaAsBi structure (r2344). Near t = 0 the substrate temperature was rapidly increased from 330C to 550C. The appearance of spikes corresponds to the formation of a crosshatch pattern on the wafer surface. 93 5.3 Rutherford backscattering spectroscopy Tixier et al. have characterized GaAsBi films with up to 3.1% Bi using RBS and HRXRD [29]. Similar experiments have been carried out by Takehara et al. on films with up to 4.8 % Bi [144]. In both cases, the lattice constant was found to increase linearly with Bi-content, obeying Vegard’s law. Extrapolation of these results (assuming Poisson’s ratio  = 0.31 for all GaAsBi compositions) yield estimates of the GaBi relaxed lattice parameter: 6.330.06 Å from Tixier et al. and ~6.23 Å from Takehara et al. RBS and HRXRD measurements on films with up to 10.8 % Bi by Pacebutas et al. show agreement with the earlier results [132, 145]. Channeling measurements on GaNAsBi films with less than 2% Bi have also been performed [146]. These measurements indicate the Bi incorporation is fully substitutional, while conversely, the N substitutional fraction was found to be ~71%. With the recent realization of GaAsBi layers with over 20% Bi, it is important to revisit these experiments in order to measure the relationship between Bi-content and lattice parameter in this new composition range. Figure 5.6 shows the strained out-of-plane lattice parameter of several GaAsBi layers on GaAs as a function of Bi-content (blue squares). The lattice constant was measured from (004) HRXRD scans and the Bi-content was determined by RBS. In samples where relaxation was suspected, (224) RSMs were collected from and directions. No significant relaxation was detected in any of these scans, indicating all the layers are pseudomorphic. With this information, the relaxed lattice parameter of these films is calculated assuming  = 0.31 (black circles in Fig. 5.6). Fitting a line to this data and extrapolating to 100 % Bi yields a GaBi relaxed lattice constant of 6.370.05 Å, in agreement with the earlier work by Tixier et al [29]. Sample properties for the layers from Fig. 5.6 are listed in Table 5.1. The layers with higher Bi-contents had some Ga-Bi surface droplet coverage. To assess whether the droplets affected the RBS signal, pieces of the two most heavily droplet-covered samples were measured after etching the droplets from the surface with a HCl:H2O etch. The results were compared with un-etched pieces from the same samples. For the two samples, etching showed opposite effects on the RBS Bi-content, namely, etching increased the measured Bi- 94 content in one case and decreased it in the other. This variation is within the uncertainty of the RBS measurements ( 6% Bi for the ~30 nm thick high Bi-content layers). This indicates the droplet coverage did not significantly affect the RBS-measured Bi-content. Figure 5.6. Strained out-of-plane lattice parameter (blue squares) and corresponding relaxed lattice parameter (black circles) as a function of the RBS Bi-content for GaAsBi films on GaAs. The GaBi lattice parameter was determined from the best fit line assuming Vegard’s law and that Poission’s ratio for GaAsBi is 0.31. RBS measurements were performed by M. Chicoine at Université de Montréal. 95 Table 5.1. Summary of layer properties for the samples from Fig 5.6. “Etch” denotes samples where droplets were removed by an HCl:H2O etch. Thicknesses were determined from pendellösung fringes in (004) HRXRD scans (“**” indicates thickness was determined from growth rate). Samples with “*” were found with RBS to contain multiple layers of varying Bi content. In these cases the composition of the thickest layer was taken. sample # RBS [Bi] (%) thickness (nm) T (C) d/d 2250 etch 9.6 24 225 0.0207 2299 8.0* 124** 225 0.0214 2304 4.8* 267** 215 0.0100 2313 7.2 54 265 0.0136 2341 16.0 48** 225 0.0423 2341 etch 19.4 48** 225 0.0423 r2361 16.6 53 250 0.0320 r2378 18.6 33 230 0.0393 r2378 etch 12.4 30 230 0.0393 r2381 4.0 255 330 0.0083 The RBS spectra for r2381 and r2361 are shown in Fig. 5.7. To minimize channeling, the samples were tilted so the beam was incident at 7 to the surface normal. The measured data (red points), as well as the simulated spectra, are shown. The peak near channel 500 corresponds to backscattering from Bi atoms from the GaAsBi layer. The height of this peak indicates the Bi-content while the width gives the layer thickness. The sample on the left has a GaAsBi layer 255 nm thick, while the sample on the right is only 53 nm thick. For thin samples that do not result in flat-top peaks, the Bi-content is determined by fitting the shape 96 of the peak. As a result, the uncertainty in the Bi-content is increased. The uncertainty in the Bi-content of the thick samples, such as r2381, is less than  1%, while the 30 nm thick high Bi-content samples have  6% uncertainty. Figure 5.7. RBS signal and SIMNRA simulations for two GaAsBi/GaAs samples. The peak on the right side of each figure corresponds to backscattering from Bi atoms in the GaAsBi layer. The large step to the left is from Ga and As in the layer and substrate. Measurements performed by M. Chicoine at Université de Montréal. 5.4 Scanning electron microscopy It was mentioned in the first chapter, that the Ga-Bi phase diagram has no solid Ga-Bi compounds at any Ga:Bi ratio or temperature [68, 69]. The bond strengths of Ga-Ga, Bi-Bi and Ga-Bi diatomic molecules are 13821 kJ/mol, 200.47.5 kJ/mol and 15917 kJ/mol, respectively [147]. This indicates that although Ga-Bi bonding is stronger than Ga-Ga bonding, the total bond energy is greater if Ga and Bi bond to themselves rather than each other. This is due to the comparatively strong Bi-Bi bond. A weak Ga-Bi reactivity was also required in our GaAsBi growth model, to obtain agreement with experimental data. 97 Figure 5.8 shows an EDS map of a Ga-Bi bimetallic droplet on a GaAs0.969Bi0.031 surface. The droplet has two distinct parts, one composed of Bi (or a Bi-rich metal) and the other composed of Ga (or a Ga-rich metal). The phase segregation maximizes Ga-Ga and Bi-Bi bonding. The film was grown at 350C, above the melting points of both Ga (29.8C) and Bi (271.3C). Above the Bi melting point, Ga and Bi form a homogeneous liquid for all compositions. Therefore, the droplet likely phase segregated after the growth was stopped and the sample cooled. The Ga-Bi phase diagram also contains regions where Bi-rich and Ga-rich liquids coexist. Figure 5.8. An EDS image of a bimetallic droplet of Ga (red/top) and Bi (blue/bottom) on a GaAs0.969Bi0.031 surface. The sample was grown with a substrate temperature of 350C under slightly Ga-rich conditions and showed a (21) reconstruction during growth. Map recorded by M. J. Fryer. Figure 5.9 shows an SEM image of several Ga-Bi composite droplets, this time on a GaAs0.87Bi0.13 surface. EDS measurements indicate that the droplet in the centre of the frame is Bi-rich on top and Ga-rich on the bottom. Facets are visible on the Bi-rich part, but no features are seen on the Ga-rich part, which may still be liquid. The sample was grown with a substrate temperature of 250C, so Bi was likely solid during the growth. The As2:Ga BEPR was very low (0.85) and the Bi:Ga BEPR was 0.50, which resulted in large amounts of excess Ga and Bi on the surface during growth. 98 Figure 5.9. An SEM image of bimetallic droplets on a GaAs0.87Bi0.13 surface. The droplet in the centre of the frame is Bi-rich on the top and Ga-rich on the bottom. The sample was grown with a substrate temperature of 250C, low As2:Ga BEPR of 0.85, high Bi:Ga BEPR of 0.50 and showed a (21) RHEED reconstruction during growth. Figure 5.10 shows an SEM image of a droplet-free GaAs0.942Bi0.058 surface. This is the highest Bi-content layer that I achieved that did not have droplets. Many droplet-free samples were grown with less than 5% Bi while no great effort was made to grow droplet-free films with higher Bi-contents. This should be possible, although higher Bi-contents require ever more careful control of the As2:Ga BEPR. This sample was growth with a substrate temperature of 265C and a As2:Ga BEPR of 2.53, near stoichiometry. The Bi:Ga BEPR of 0.29 should have resulted in excess surface Bi, so it is surprising that no droplets appear on this sample. The small features on the surface appear to be pin holes. These features may be due to initial contamination on the wafer or Bi droplets in their infancy that evaporated, possibly after the growth was stopped. 99 Figure 5.10. An SEM image of a droplet free GaAs0.942Bi0.05.8 surface. The sample was grown with a substrate temperature of 265C, As2:Ga BEPR of 2.53, Bi:Ga BEPR of 0.29 and showed (21) and (2chevrons) RHEED reconstructions during growth. Image recorded by M. Masnadi-Shirazi. Vardar and co-workers have recently investigated droplet formation in the MBE growth of GaAsBi [148]. For As-rich growth of GaAsBi at 350C, no droplets were observed to form. When the As2:Ga ratio was reduced to within 10% of stoichiometry, Ga droplets formed on the surface. The stoichiometric As:Ga ratio was determined by the point where the RHEED degrades and droplets form for GaAs growth at 350C. Below stoichiometry, composite Ga-Bi droplets were seen. The interpretation was that for As-rich growth at these conditions, Bi behaves as a surfactant; with excess Bi evaporating from the surface rather than forming droplets. The formation of Ga droplets within 10% of stoichiometry was interpreted as a Bi “anti-surfactant effect”. The stoichiometric As2:Ga flux ratio and Bi anti-surfactant-induced droplet formation was found to depend on growth rate. A simple explanation for the observations of Vardar et al., is that the presence of Bi on the surface increases the As-evaporation probability, thereby shifting the threshold for Ga droplet formation to higher As:Ga flux ratios. Furthermore, according to our growth model, incorporated Bi on the surface increases both the As evaporation rate and the Ga droplet capture rate, for a given As2:Ga flux ratio. This is due to As not displacing Bi and Ga having 100 a low probability of reacting with Bi. The result is that Ga droplets will begin to form at a higher As2:Ga flux ratio. 5.5 Growth of GaAsBi on InP and GaSb substrates The growth of high Bi-content GaAsBi on GaAs substrates results in a large amount of compressive strain in the film. For example, the mismatch between GaAs0.78Bi0.22 and GaAs is 2.6%. It was postulated that this high strain may inhibit the incorporation of Bi. It is generally considered that atoms deposited on a substrate attempt to arrange themselves in a way that minimizes the total free energy of the system. Important elements of the total free energy are the chemical energy (forming stronger bonds being favored) and the strain energy. The exchange of As and Sb atoms at InAs/AlSb and InAs/GaSb interfaces, which prevents the interfaces from being atomically sharp, is believed to results from free energy minimization. In the case of InAs/AlSb, the free energy is reduced by the formation of stronger bonds (Al-As and In-Sb bonds are stronger than In-As and Al-Sb bonds) and a reduction of the interface strain energy [149, 150]. For the growth of InGaNAs/GaAs quantum wells, it has been claimed that strain directly affects N-incorporation [151]. In order to determine whether strain was inhibiting Bi incorporation in GaAsBi layers grown on GaAs, GaAsBi was grown on substrates of larger lattice parameters. Table 5.2 shows the lattice parameter of some III-V substrates, indicating the Bi-content required for GaAsBi to be lattice matched. The mismatch to GaAs for these lattice parameters is also given. The growth of GaAsBi, with less Bi than that required to lattice match, results in films under tensile strain. In this case, increasing the Bi-content would lower the strain energy of the film. This could act as a driving force for Bi incorporation. The growth of GaAsBi on larger lattice constants is also of interest from a device perspective. GaAsBi lattice matched to InP is expected to have a bandgap near 0.1 eV (predicted by extrapolating absorption data from chapter 5), while alloys lattice matched to 101 InAs and GaSb are expected to have negative bandgaps. GaAsBi on InP could open up a new wavelength range on InP that is not currently available. Table 5.2. The lattice parameter of some III-V substrates, showing the Bi-content where GaAsBi is lattice matched and the mismatch to GaAs. substrate material lattice parameter (Å) [Bi] to match (%) mismatch to GaAs (%) GaAs 5.653 0 0 InP 5.869 32 3.8 InAs 6.058 60 7.2 GaSb 6.096 65 7.8 GaAsBi layers were grown on InP and GaSb substrates. Here only the InP results are discussed, as a large uncertainty in the substrate temperature for GaSb made the data difficult to interpret. For growth on InP, the oxide was observed with RHEED to desorb at ~530C (DRS temperature) when the substrates were heated under an As flux. GaAsBi layers were deposited directly on the InP substrates, without the growth of a buffer layer. The layers were grown using Ga and Bi rich conditions, namely low As2:Ga BEPRs ~0.85 and Bi:Ga BEPRs of 0.6-0.7. As with growth on GaAs, these conditions resulted in Ga-Bi droplets on the surface. Figure 5.11 shows HRXRD -2 scans of the (004) film and substrate peaks for GaAsBi/InP samples grown at three different temperatures. Film thicknesses are estimated to be 12-20 nm from the growth time. The InP substrate peak is the sharp peak at 0 and the film peaks are located to the right of the substrate peaks, indicating tensile strain in the GaAsBi layers. The broad peaks and lack of interference oscillations indicate poor film quality. Reciprocal space mapping of the and peaks for the sample grown at 215C showed a wide 102 distribution in the amount of relaxation of the film. The full width a half maximum (FWHM) of the film peak ranged between fully strained and beyond 100% relaxed. The average relaxation in both directions was ~ 70%. Simulation of the (004) -2 scan of this sample, assuming 70% relaxation, indicated a Bi-content of ~15%. RSMs were not recorded for the samples grown at 245C and 260C. These samples were grown with higher temperature and have more strain, so it is reasonable to assume the average amount of relaxation is 70%. The composition range indicated on Fig. 5.11, for the samples grown at 245C and 260C, was obtained by simulating the -2 scans assuming the relaxation lies within 100% and 70%. Figure 5.11. (004) -2 HRXRD scans for three GaAsBi layers grown on InP substrates. The sample numbers are r2271 (215C), r2272 (245C) and r2258 (260C). The sample grown at 215C has ~70% relaxation. The composition range of the other films is determined from the (004) scans, assuming the relaxation lies between 100% and 70%. The GaAs0.85Bi0.15 sample has 2.0% mismatch from the InP substrate, the same as a GaAs0.84Bi0.16 layer from GaAs. This tensile strained GaAs0.85Bi0.15/InP film shows much 103 more relaxation than compressive films with similar mismatch and thickness on GaAs. This observation is consistent with Marée et al. [152], who showed that diamond-type and zincblende-type layers show more strain relaxation under tension than compression. The effect arises due to the dissociation of the 60 mixed dislocations into 90 and 30 partials. Marée et al. showed that 90 partials form first under tensile strain and 30 partials forms first under compressive strain, resulting in easier dislocation nucleation in tensile strained films [152]. Figure 5.12 shows an SEM image of the GaAs0.85Bi0.15/InP sample grown at 215C. As for samples grown under these conditions (Bi and Ga rich) on GaAs, a high density of composite Ga-Bi droplets exists on the surface. The presence of excess Bi and Ga on the surface indicates the Bi incorporation was not limited by the As2:Ga flux ratio or the Bi flux. Figure 5.12. An SEM image of the GaAs0.85Bi0.15/InP sample grown at 215C (r2271). The dark parts of the droplets are Bi and the light parts are Ga. Image recorded by M. Masnadi-Shirazi. Consistent with GaAsBi growth on GaAs, as the substrate temperature is lowered, more Bi is observed to incorporate. When compared to growths on GaAs at the same conditions, less Bi was incorporated for the growths on InP. This is unexpected and may be related to experimental errors, such as offsets in the temperature reading for InP and the difficulty of 104 estimating the Bi-content from the GaAsBi/InP HRXRD data. Also, no buffer layer was grown on the InP substrates and the resulting rough starting surface may have had an effect on the Bi incorporation. Nevertheless, growth under tensile strain did not increase the amount of Bi incorporated. This suggests strain does not significantly contribute to the difficulty in getting Bi to incorporate. 105 Chapter 6 Optical and electronic properties of GaAsBi layers The optical and electronic behaviour of GaAsBi alloys have not been fully explored. With the recent realization of GaAsBi layers with up to 21.8% Bi, nearly double the highest previously reported concentrations [76, 128, 132], one of the most pressing questions is: what is the bandgap energy of these new alloys? In this chapter, absorption measurements on GaAsBi alloys with up to 18.7% Bi are presented. These measurements reveal the composition dependence of the fundamental bandgap energy in this previously unexplored composition range. Transport measurements on n-GaAsBi layers with up to 4% Bi reveal a high concentration of compensating acceptors in the layers. The composition dependence of the compensation suggests that closed Bi3 clusters are responsible for the compensation of n-type carriers. 6.1 The composition dependence of the GaAsBi bandgap The composition dependence of the bandgap of III-V ternary alloys is typically well fit with a quadratic function, so long as there is no transition in the nature of the bandgap, say from direct to indirect. The standard bandgap bowing equation for binary alloys is given by Eq. 6.1, for the case of GaAsBi. Here, EGaBi is the GaBi bandgap, EGaAs is the GaAs bandgap and b is the bowing parameter. Bowing parameters are positive, indicating that bandgaps are less than what would be predicted by linear interpolation between the endpoints. For most III-V alloys, a constant bowing parameter is sufficient to fit experimental data. However, for measurements on highly mismatched alloys, such as GaNAs, GaAsBi and InAlN; bowing parameters are found to be strongly composition-dependent [32, 47, 153]. 106 This is due to the impurity-like nature of highly mismatched alloying elements, which result in a strong perturbation on the host band structure for dilute incorporation amounts. First principles band structure calculations for GaAsBi, by Janotti et al. [46], predict a GaBi bandgap of -1.45 eV and a constant bowing parameter b = 2.1 eV. The bandgap for films with less than 11% Bi has previously been measured experimentally by PL [47], and optical absorption and photomodulated reflectance spectroscopy [132]. For the PL results, a constant bowing parameter was found insufficient to fit the experimental data. A bowing parameter that decreased monotonically with increasing Bi-content, like what had been used for In content in InAlN [153], was used to fit the PL data. The composition-dependent bowing parameter is given by Eq. 6.2. Good agreement was obtained with the PL data for EGaBi = -0.36 eV,  = 9.5 and  = 10.4. A plot of the GaAsBi bandgap as a function of Bi-content, showing Lu’s room temperature PL results and Janotti’s 0 K DFT result, is shown in Fig 6.1. The PL data in Fig. 6.1 are from pseudomorphic films, while the DFT calculations are for free standing material. The bandgap of III-V semiconductors increases when a compressive strain is applied, mostly due to an upward motion of the conduction band. The change in bandgap for a given strain on the lattice is given by Eq. 6.3, where a is the deformation potential constant (~-8.5 eV for GaAs) and xx is the strain in the x direction [154]. For the pseudomorphic case,   where  is Poisson’s ratio. The deformation potential constant has not been measured for GaAsBi alloys, but it is likely similar to that for GaAs. In this case, the strain in a pseudomorphic GaAs0.9Bi0.1/GaAs film would cause the bandgap to increase by ~0.11 eV, 107 relative to a free standing film. Accounting for strain would increase the separation between the DFT and PL results in Fig. 6.1. Figure 6.1. The GaAsBi bandgap as a function of Bi-content and lattice parameter from density functional theory at 0 K [46] and room-temperature PL [47]. 6.1.1 Optical absorption spectroscopy Optical absorption is a simple technique that can yield a great deal of information about the material band structure. By varying the wavelength of the incident light, all possible electron transitions from filled to empty states can be excited. For the recently-grown high Bi-content GaAsBi samples, which were grown at very low temperatures, PL intensity is likely to be very weak. In this case, absorption spectroscopy is a well suited technique for determining 108 the bandgap. Furthermore, at the time these measurements were taken, our lab did not possess a liquid-nitrogen-cooled detector extending to wavelengths beyond 1.6 μm. For the absorption measurements, the transmitted light intensity is on the order of the incident intensity for energies below the bandgap. In this case, high-sensitivity cooled detectors are not required. The transmission spectra presented below were recorded by M. Masnadi-Shirazi. As was mentioned in chapter 2, my role was in providing some initial motivation for these experiments, growing the GaAsBi/GaAs samples, and aiding in the experimental setup and data analysis. Figure 6.2 shows selected transmission spectra for several GaAsBi/GaAs samples. The spectra have been normalized for the throughput of the optical system. The spectra in a) were collected using a Ge detector and the spectra in b) were collected using a PbS detector. The Ge detector is more sensitive, but has a shorter cut-off wavelength (1750 nm for Ge vs. 2900 nm for PbS). On both plots, the transmission spectrum of a GaAs wafer is shown for comparison. The steep rise in the transmitted signal at ~870 nm corresponds to the GaAs absorption edge. For shorter wavelengths, the incident light is nearly all absorbed by the thick GaAs substrate. For photon energies below the GaAs bandgap but above the GaAsBi bandgaps, the thin GaAsBi layers are only partially absorbing. In spectra b), the noise around 2600 nm results from the very low sensitivity of the PbS detector in this range. The undulations in b) may be due to speckle or other interference effects. As was mentioned in the chapter 2, dividing the GaAsBi/GaAs transmission spectra by the GaAs spectrum isolates the absorption from the GaAsBi layer. Neglecting multiple reflections, the absorption coefficient of the GaAsBi layer is given by Eq. 6.5, where d is the layer thickness, and Tsample() and TGaAs() are the transmittance of the GaAsBi/GaAs sample and the GaAs substrate at wavelength .  For direct bandgap semiconductors, parabolic band theory predicts an abrupt absorption edge of the form of Eq. 6.6 [155] near the band edge. Here h is the photon energy, Eg is the bandgap and A* is a constant. 109 Figure 6.2. Normalized transmission spectra for GaAsBi/GaAs samples recorded with a) a Ge-detector and b) a PbS detector. The steep rise at ~870 nm corresponds to the GaAs absorption edge. The deviation from the GaAs signal is due to absorption in the GaAsBi layers. Spectra recorded by M. Masnadi-Shirazi. 110 Experimentally, rather than no absorption below the bandgap, an exponential band tail is observed. This is known as the Urbach edge [156], resulting from tail states near the conduction and valence band edges that arise due to impurities and disorder in the lattice [155]. In GaAs at 300 K, free exciton absorption results in a small bump in the absorption spectrum ~4 meV below bandgap [157, 158]. Nevertheless, a square-root dependence of the absorption edge is observed for direct semiconductors in the region just above the bandgap. A plot of 2 as a function of photon energy is shown in Fig. 6.3 for selected GaAsBi layers. Most of the data are well fit with straight lines, indicating direct bandgaps. Band tails are visible at low absorption, resulting in deviations from the linear fits. The bandgaps are obtained by fitting straight lines to the absorption curves from 2 = 2106 cm-2 (below this the band tails dominate) to 2 = 108 cm-2, and then extrapolating these linear fits to zero. The sample with 9.7% Bi (r2299) showed composition variation in the RBS spectra. This is the cause of the slope variations for this sample in Fig. 6.3. Composition variation is the likely cause for variations in the other curves as well. At low absorption levels, the signal corresponds to the highest Bi-content portion of the film. If there is composition variation, this layer is thinner than the total layer thickness, which is used in calculating the absorption coefficient. This results in an underestimate of the absorption coefficient, and thus, a less steep slope in this region. Furthermore, if an incorrect thickness is assumed for a uniform layer, this will result in an incorrect slope. However, this will not affect the extrapolated bandgap. Figure 6.4 shows a plot of the GaAsBi bandgap as a function of Bi-content and lattice parameter, showing the absorption data, Lu’s PL data [47] and Janotti’s DFT results [46]. The DFT curve has been shifted downward by 0.1 eV to match the room temperature GaAs bandgap energy of 1.42 eV. Using the composition dependent bowing parameter from Eq. 6.2, a best fit to the absorption data is obtained for EGaBi = -2.04 eV,  = 5.03 and  = 8.97. This fit is shown as the solid line in Fig. 6.4. 111 Figure 6.3. A plot of the square of the absorption coefficient as a function of energy for several GaAsBi layers. The bandgap was obtained by fitting linear functions between 2 = 2106 cm-2 and 2 = 108 cm-2 and then extrapolating the best fit curves to zero absorption. Data and fits by M. Masnadi-Shirazi. The bandgap of thin GaAsBi layers has been measured with optical absorption for Bi compositions up to 18.7%. The absorption data is in good agreement with the PL data from Lu et al. [47] for up to ~10% Bi. The sample with highest Bi-content, 18.7%, showed a bandgap of 0.50  0.07 eV. This sample is 36 nm thick and pseudomorphically strained to the GaAs substrate. This illustrates the potential of GaAsBi, for extending the wavelength range of devices operating on GaAs, to well beyond what conventional alloys offer. A crucial next step will be the demonstration of light emission from these high Bi-content layers. 112 Figure 6.4. A plot of the composition and lattice parameter dependence of the GaAsBi bandgap. The absorption results are shown as well as a fit to the absorption data. PL data from Lu et al. [47] and Janotti’s DFT calculation [46] (shifted to match the room temperature GaAs bandgap) are shown for comparison. 6.2 Electrical properties of n-GaAsBi layers Bismuth incorporation in GaAs mainly affects the valence band, due to a resonance of the Bi 6p orbitals with the GaAs VBM [46, 51]. As a result, Bi-incorporation is found to significantly degrade the Hall mobility of holes in p-type GaAsBi [58, 59]. This degradation is much less severe than the effect of N-incorporation on electron mobility in GaNAs [58]. In contrast to N-incorporation, Hall measurements [60] and time-resolved THz spectroscopy [61] showed Bi-incorporation does not significantly affect electron mobility. Electron transport properties will have an impact on the potential device applications of GaAsBi. For example: high speed devices like HBTs and THz sources/detectors require high electron 113 mobility, long minority carrier diffusion lengths are needed for photodetectors and solar cells, and short electron trapping times are needed for THz sources and detectors. GaAsBi grown at very low temperatures has shown the sub-picosecond electron trapping times required for THz devices [63]. 6.2.1 Compensation of free electrons in n-GaAsBi layers p-type conductivity in nominally undoped GaAsBi layers has been observed by Pettinari et al. [57]. The conductivity was attributed to acceptor states near the VBM associated with Bi-clusters and the concentration of free holes was found to increase exponentially with the Bi-content of the layers. For a sample with 10.6% Bi, the temperature dependence of the carrier concentration showed an activation energy of 27 meV and a concentration of 2.41017 cm-3 for the acceptor level. This concentration is much less than the expected concentration of low-order Bi clusters. As multiple acceptor energies are possible and deep acceptors sites may not be fully ionized, the total concentration of acceptor states cannot be reliably determined from measurements of p-type carrier concentration. On the other hand, measuring the compensation of free electrons in n-doped GaAsBi can allow the total number of acceptors states in the gap to be determined. Formation probabilities of low-order clusters (monomers, dimers and trimers) have been calculated by Kreitman and Barnett for cubic and hexagonal lattices, assuming clusters form randomly in the lattice [159]. In the case of zincblende GaAsBi, Bi clusters are assumed to form on the group V face-centred cubic (fcc) sub-lattice. Figure 6.5 shows the composition dependence of the concentration of isolated Bi atoms, Bi2 dimers, and open and closed Bi3 trimers. The cluster concentration is calculated by assuming all group V sites have an equal probability of being occupied by a Bi site (random cluster formation) [159]. A Bi2 cluster has two Bi atoms as nearest neighbours on the group V sub-lattice, with all the other nearest neighbour sites (on the sub-lattice) being occupied by As. The open Bi3 cluster concentration is the sum of the linear and bent cluster configurations. Although there is only one closed trimer cluster configuration in a simple fcc lattice, there are two in a zincblende crystal. In one case, the three Bi atoms are all bonded to a single Ga atom that is slightly raised from the 114 midpoint between the Bi atoms. In the other case, the Ga-Bi bonding is in a ring with only two Bi atoms bonded to each Ga atom on the ring and no Ga in the centre. The two closed Bi3 cluster types have equal formation probabilities and their sum is plotted in Fig. 6.5. For low Bi-contents (less than a few %), cluster concentrations are proportional to the Bi-content raised to the power of the cluster size (1 for monomers, 2 for dimmers, etc). Higher-order clusters were not calculated but will have lower concentrations. Figure 6.5. Calculated Bi-cluster concentrations assuming Bi randomly occupies group V lattice sites [159]. The open and closed Bi3 cluster concentrations are the sum of multiple configurations. Compensation in n-GaAsBi was investigated after it was noticed that GaAsBi layers, doped with Si in an attempt to achieve n-type doping, were often found to be highly resistive and/or depleted. The compensation was first noticed for GaAsBi p+/n device structures. These were grown for characterization of electronic defects in the n-GaAsBi layers by DLTS, in collaboration with P. M. Mooney’s group at SFU. For the p+/n structures, only the free carrier concentrations determined from capacitance-voltage (C-V) profiling are discussed 115 here. The C-V and DLTS measurements were carried out by K. P. Watkins, Z. N. Jiang and P. M. Mooney at SFU and the results of that study are discussed elsewhere [160]. To further explore the issues with doping, Hall and resistivity measurements were performed on GaAsBi and GaAs epilayers. These layers were n-doped with Si and grown 250 nm to 675 nm thick on semi-insulating GaAs substrates. Hall measurements and post-growth fabrication on these samples were carried out by M. Masnadi-Shirazi and V. Bahrami-Yekta. For the p+/n structures and the epilayer samples, the Bi-containing layers were grown at 330C with As2:Ga BEPRs between 2.8 and 4.2 at a growth rate of ~1 μm/h. The Bi-content was regulated by the Bi flux, but on occasion lower than expected Bi-contents resulted, indicating other growth parameters were limiting the Bi incorporation. All samples had smooth, droplet free surfaces. To n-dope GaAs and GaAsBi layers, a Si dopant flux was achieved using a standard single filament effusion cell loaded with pieces of a Si wafer. The Si content in the GaAs layers was assuming to equal the free electron concentration, as measured by Hall and C-V. The Si content in the GaAsBi layers is estimated from the relation between the Si source temperature and the measured free carrier concentration in Si-doped GaAs calibration samples. These calibration samples, which were grown over the same timeframe as the GaAsBi layers, showed free electron concentrations ranging from 71016 cm-3 to 81018 cm-3. The free electron concentration followed an Arrhenius relationship with the cell temperature, given by [n] = 6.681032e-46500/T cm-3, where T is in Kelvin. The average deviation of the n-GaAs calibration samples from the fit was 30% of the free carrier concentration. This is taken as the uncertainty in the Si-concentration of the GaAsBi layers. The Si-doped GaAsBi layers show a large amount of compensation. Figure 6.6 shows a phase diagram plot of Si concentration as a function of Bi-content for Si-doped GaAsBi and GaAs layers. The black data points (circles and squares) indicate samples that show n-type conductivity (Hall) or had ND>NA and were not depleted (C-V), while the star data points correspond to highly resistive/depleted samples. The closed data points correspond to Hall measurements on epilayers and the open data points are from C-V measurements on p+/n samples. Some of the resistive samples show weak p-type conductivity. The plot shows the threshold Si-content where the samples become conductive is strongly dependent on the Bi 116 content. The density of closed Bi3 clusters from Fig. 6.5 is also plotted on Fig. 6.6. The cluster concentration is in good agreement with the Si-content where the layers transition from being resistive to conductive. This suggests closed Bi3 clusters may produce acceptor states in the gap, which may be the source of the electron compensation. Figure 6.6. A plot of the expected Si-dopant concentration as a function of Bi-content for Si-doped GaAs and GaAsBi layers, indicating whether samples show n-type doping or are resistive/depleted. Whether the samples were doped or restive/depleted was determined from Hall measurements on Si-doped GaAsBi epilayers (closed data points) and from C-V measurements on p+/n devices (open data points). The concentration of closed Bi3 clusters is plotted assuming Bi randomly populates group V lattice sites [159]. Secondary ion mass spectroscopy (SIMS) measurements have been performed on different Si-doped GaAsBi samples, grown by D.A. Beaton at UBC and measured by S. Moisa at NRC Ottawa [95]. These measurements show that the incorporation of Bi does not diminish Si incorporation, ruling this out as an explanation for the lack of doping. 117 Silicon is known to be an amphoteric dopant in MBE-grown GaAs, doping n-type for As:Ga flux ratios  stoichiometry and p-type for Ga-rich growth [161]. Presumably, on a Ga-terminated surface Si bonds to Ga, occupying As sites where it is a p-type dopant, while on an As-terminated surface Si bonds to As, occupying Ga sites. Figure 6.7 shows the As2:Ga BEPRs of the resistive/depleted and n-type Si-doped GaAsBi samples. The uncertainty in the BEPRs is ~20%. All samples were grown under slightly As-rich conditions and no correlation between the BEPR (or the RHEED reconstruction) and compensation is found. The BEPRs of Si-doped GaAs samples grown at 330C are also indicated, all of which showed n-type conductivity. This suggests that the low As2:Ga flux ratios are not responsible for the observed compensation in the GaAsBi samples. However, given the close proximity to Ga-rich growth conditions for GaAsBi growth, further investigation should be undertaken to ensure that the amphoteric nature of Si, which could possibly be affected by the presence of Bi, is not responsible for the observed compensation. It is worth noting that Si on As sites forms a shallow acceptor in GaAs. Therefore, if the majority of the Si incorporates on As sites, highly p-type films should result. Figure 6.7. Measured As2:Ga BEPRs for resistive/depleted and n-type Si-doped GaAsBi samples. Square date points correspond to C-V measurements on p+/n structures and circle date points correspond to Hall measurements on epilayer samples. The BEPRs of n-type GaAs samples are also shown. All samples were grown with a substrate temperature of 330C. No correlation is found between BEPR and compensation. Compensation of electrons in GaAsBi has not been previously reported, and in general, little work has been done on doped GaAsBi films. Electron Hall transport measurements on Si- 118 doped GaAsBi layers have previously been reported by Kini et al. [60] for Bi-concentrations up to 2.5%. Their data shows the concentration of free electrons to decrease from 1.01018 cm-3 to 0.51018 cm-3 as the Bi-content increases from 0 to 2.5%, although the authors did not mention whether a constant Si-flux was used. This decrease suggests compensation is present in these samples as well. C-V and DLTS measurements on our GaAsBi p+/n structures, which contain Bi-contents up to 1.1%, indicate trap levels in the 1016 cm-3 range [160]. The trap concentrations, which are lower than the observed compensation levels above, were not found to scale with the Bi content. The study surveyed traps in the top half of the bandgap, except for the region within ~100 meV of the CBM. Fuyuki et al. characterized deep levels in Be-doped p-GaAsBi layers with 0 to 3.4% Bi by DLTS [162]. These experimental conditions resulted in a survey of hole traps in the lower half of the bandgap with energies greater than ~170 meV above the VBM. This study yielded total trap concentrations in the 1015 cm-3 range, much lower than the compensation levels observed above. These suggest that the acceptors may reside nearer to the band edges. To explore the regions closer to the band edges, DLTS characterization to lower temperatures (below the 80 K and 100 K for the above studies, respectively) is of interest. Raman spectroscopy could also be used to search for clusters or other defect complexes in GaAsBi. The high level of compensation in Si-doped GaAsBi suggests the presence of Bi-induced acceptor states. For future growths of n-GaAsBi, it will be critically important to accommodate for these states. The amount of compensation correlates well with the concentration of closed Bi3 clusters. Higher-order Bi clusters may also produce acceptor states in the bandgap, however these clusters will have lower concentrations. The compensation measurements are insensitive to states which produce hole traps in the bandgap, however, unlike for n-type doping, problems achieving p-type doping in GaAsBi have not been encountered. The observation that the compensation follows the concentration of the closed Bi3 clusters suggests they may be the lowest-order clusters that produce acceptor states in the bandgap. Further study is needed to verify the origin of this compensation. 119 6.2.2 Electron Hall mobility in n-GaAsBi Figure 6.8 shows the dependence of the electron Hall mobility on Bi-content for conductive n-GaAsBi epilayers. A general decrease in the mobility is seen with increasing Bi-content. The results of Kini et al. are shown for comparison [60]. At higher Bi concentrations, the electron mobility is lower than that of Kini et al. This may be due to the higher Si-content in our higher Bi-content samples, or other variations in the growth conditions, as discussed below. Figure 6.8. A plot of electron Hall mobility as a function of Bi-content for n-GaAsBi samples. All samples from this work were grown at 330C and 1 μm/h growth rate. Measurements by Kini et al. are shown for comparison [60]. These samples were grown with a substrate temperature of 380C at 2 μm/h. Figure 6.9 shows the dependence of electron mobility on the Si-dopant concentration for GaAs and GaAsBi samples. For the GaAsBi samples, the Si-content was estimated from the n-GaAs calibration curve. For the n-GaAs and Kini et al.’s samples, the measured free carrier 120 concentration is used for the Si-content. An empirical model for the carrier concentration dependence of n-GaAs electron mobility is shown as the broken line [163]. GaAs mobility decreases with increasing doping concentration, due to increased scattering from ionized impurities. The electron mobility of the GaAs layers grown at ~550C and 330C (LT n-GaAs) are only marginally lower than empirical model, while the GaAsBi layers fall well below the model line. The set of GaAsBi samples with Si concentration ~21017 cm-3 contained less than 0.2% Bi and were grown with As2:Ga BEPRs 3.5, 3.8 and 4.2 with all other growth conditions fixed. The mobility of these samples decreases with increasing As2:Ga BEPR, from 2200 cm2V-1s-1 to 1700 cm2V-1s-1. The LT n-GaAs samples correspond to the two samples indicated in Fig. 6.7. Both samples show carrier concentrations close to the expected concentration of 1.31018 cm-3, with the lower BEPR sample showing slightly lower carrier concentration and mobility. Further investigation is required to determine the origin of the low mobility in the highly doped GaAsBi layers. Electron Hall mobility in n-GaAsBi is significantly lower than for n-GaAs. The mobility decreases with increasing Bi-content, although part of this reduction may be associated with the increased Si concentrations used to overcome compensation. Electron mobility values were generally lower than values from Kini et al. [60], possibly due to the lower growth temperature, higher dopant concentrations or differences in the As2:Ga BEPR used during growth. 121 Figure 6.9. The dependence of electron Hall mobility on the Si-dopant concentration for n-GaAsBi and n-GaAs layers. The open square n-GaAs samples were grown at ~550C while all other samples from this work were grown at 330C. For the work of Kini et al. [60], samples were grown at 380C and compensation was neglected. The broken line corresponds to the carrier-concentration dependence of n-GaAs electron mobility [163]. 122 Chapter 7 Conclusions III-V-Bi alloys are an exciting frontier of III-V semiconductor alloy exploration, potentially offering new device possibilities. Historically, investigation of these materials has been held back by the low solubility of Bi in the III-V hosts and consequent growth challenges. The highly mismatched GaAsBi alloy shows many novel properties which are interesting for a wide range of device applications, including infrared light emitters and detectors, solar cells, HBTs, THz sources and detectors and spintronics. Perhaps the most alluring feature of this alloy is the enormous effect that Bi incorporation has on the bandgap. This allows bandgaps of 1.2 eV and less to be reached with less strain to GaAs than for any other known GaAs alloying element, potentially opening up a new wavelength range for devices on GaAs substrates. In addition, GaAsBi lattice matched to InP is expected to have a bandgap of ~0.1 eV, which could allow for new device possibilities on InP substrates, which, like GaAs, are relatively inexpensive. The recent realization of GaAsBi-based LEDs, lasers and THz devices illustrates the potential of this new material and the interest that it is starting to attract. Molecular beam epitaxy (MBE) is a well suited growth technique for GaAsBi as it allows for precise control over the growth conditions, in situ characterization and low growth temperatures, which helps facilitate the growth of metastable alloys. Presented in this thesis is a systematic study of the dependence of Bi incorporation on MBE growth conditions, along with a new model for Bi incorporation in GaAsBi. This work elucidates the underlying physics of the GaAsBi growth process, which was previously poorly understood, and explains the high sensitivity of Bi incorporation on the growth conditions. This new insight clarifies how the growth of this material should be undertaken to maximize composition reproducibility and uniformity, and to avoid Ga and Bi surface droplets. When a sufficient Bi flux is supplied, Bi incorporation is found to rapidly increase as the As2:Ga flux ratio is lowered to 0.5 and to saturate for lower flux ratios. This result is explained by Bi incorporation being sensitive to the stoichiometry of the growing surface, specifically the 123 fraction of the surface which is Ga-terminated. The As2:Ga flux ratio dependence of the surface stoichiometry is modeled by taking into account As hopping and evaporation and Ga hopping and droplet formation. In addition to GaAs, this stoichiometry model should apply to the growth of other III-V semiconductors. For the Bi incorporation model, Bi from a wetting layer incorporates on surface sites which are terminated by Ga. This weak Bi-Ga incorporation bond can be broken by thermal energy, ejecting incorporated Bi atoms back to the Bi wetting layer. An activation energy for Bi ejection of 0.28 eV is found by fitting to the experimental data. These processes are found to account for the growth conditions dependence of the Bi incorporation, providing quantitative agreement with experimental data over an extensive range of growth conditions. This is a striking improvement over a previous growth model [76], which did not properly account for the As2:Ga flux ratio dependence of the Bi incorporation. It is found that avoiding surface droplets and composition fluctuations resulting from fluctuations in the growth conditions, is best achieved by limiting the Bi-content with the Bi flux. With this clear understanding of GaAsBi growth, a detailed assessment of GaAsBi material properties and the effect of the growth conditions on those properties can now be undertaken. Highly strained GaAsBi films with up to 21.8% Bi were grown on GaAs, nearly double the previously reported incorporation records [76, 128, 132]. X-ray diffraction measurements show these epitaxial layers have a high degree of crystalline perfection, despite being grown at temperatures as low as 200C and having up to 2.6% mismatch to the GaAs substrate. The low growth temperature is found to result in unusually large critical thicknesses for relaxation, likely a result of dislocation formation and glide being inhibited at the low growth temperatures. In order to investigate whether the compressive strain resulting from GaAsBi growth on GaAs inhibits Bi incorporation, GaAsBi layers were grown for the first time under tensile strain on InP substrates. No enhancement in the Bi incorporation was found for growth under tensile strain. Optical adsorption experiments were carried out on GaAsBi layers with up to 18.7% Bi in order to determine the relation between the composition and the bandgap in this new composition range. A 36 nm thick 18.7% Bi layer that is pseudomorphically strained to GaAs 124 shows a bandgap of 0.50  0.07 eV. To our knowledge, no other alloy with such a low bandgap can be grown pseudomorphically on GaAs. Electrical transport measurements on Si-doped GaAsBi layers reveal a large concentration of compensating acceptors in the GaAsBi bandgap. Compensation in n-GaAsBi has not been previously reported as doped GaAsBi has been little investigated. The amount of compensation increases rapidly with increasing Bi-content, exceeding 1018 cm-3 by 2% Bi. The dependence of the acceptor concentration on the Bi-content is in quantitative agreement with that of closed Bi3 clusters, calculated assuming Bi randomly populates group V lattice sites. This suggests closed Bi3 clusters may be the source of the acceptor state. Further measurements are required to determine where the acceptor states reside within the bandgap. The electron Hall mobility in n-GaAsBi layers was found to decrease with increasing Bi content, at a faster rate than that previously reported [60]. The MBE growth of GaAsBi requires pushing MBE technology in terms of precise flux control, low temperature growth and use of in situ characterization tools. In the same spirit of MBE innovation, a novel closed-cycle cooling system was implemented to cool the MBE cryo-shroud, which is normally cooled with liquid nitrogen (LN2). Liquid nitrogen is the single largest MBE operating cost and the new cooling system reduces LN2 consumption in a VG V80H MBE system by an order of magnitude. A study of the temperature dependence of cryopanel pumping efficacy in the MBE system is presented. Cryopanels cooled with LN2 (-196C) are found to provide effective pumping of H2O, CO, CO2 and As4 at the levels in the MBE system. For H2O and As4 the vapor pressures at -196C are very low and pumping is by cryo-condensation, while CO and CO2 have relatively high vapor pressures so pumping is only possible by cryo-adsorption. At -78C, the operating temperature of our closed-cycle chiller, H2O and As4 are pumped effectively while CO and CO2 are not. The H2O pumping speed increases exponentially with decreasing temperature, which can be explained by an exponential distribution of binding energies for H2O adsorption sites on the cryopanel surface. Below ~-40C the As4 pumping speed saturates, while the H2O pumping is predicted to saturate at ~-95C. GaAs layers grown with the shroud cooled with LN2 and with the chiller show similar concentrations of deep levels. In addition, AlGaAs layers grown with the chiller show strong photoluminescence, expected electron mobility and background doping 125 levels less than 41015 cm-3. These results indicate closed-cycle cooling is a promising cost-reduction technique for MBE. It is expected that this new technique will be widely adopted by the MBE community. 7.1 Future work The growth study presented in this thesis illustrates the difficulty of obtaining high Bi-content GaAsBi layers while avoiding Ga and Bi metallic droplets on the surface. Applying innovative MBE growth techniques could help achieve the Ga-rich surfaces required for high Bi-content without permitting excess Ga to accumulate on the growth surface. One such possibility is pulsing the As2 flux during growth, either by shuttering the primary As2 source or having a secondary source that is shuttered. This could allow Ga-rich surfaces to be created briefly between pulses, with the As pulse periodically removing the excess Ga from the surface, preventing buildup. GaAsBi with 32% Bi is lattice matched to InP. This alloy is expected to have a bandgap of ~0.1 eV, which could open up exciting device possibilities, motivating efforts to strive for even higher Bi-contents. The effect of growth conditions and annealing on the optical and electronic properties of GaAsBi has not been systematically investigated. For example, it is not known whether growth with the As2:Ga flux ratio as close as possible to stoichiometry results in better or worse films than growth with higher flux ratios. Addressing these questions is the next step in optimizing the GaAsBi growth process. Low temperature DLTS experiments on p-type and n-type GaAsBi layers are an ideal way to search for traps related to the observed free electron compensation in n-GaAsBi. Raman spectroscopy and x-ray absorption could be used to search for Bi clusters and other defect complexes. Theoretical calculations of the effect of Bi clustering on the band structure are of much interest. If clusters do form deep acceptor states this will have consequences for devices, as the deep levels will be recombination centres. The band offsets to GaAs are also important for devices. These have yet to be measured. 126 Given the recent success of InGaNAs as a small bandgap material that is lattice matched to GaAs, the growth of GaNAsBi is of interest. This material, which was one or the original motivations for studying GaAsBi, has the potential to provide more bandgap reduction for a given amount of N incorporation than for In. This should allow longer wavelengths to be reached than with InGaNAs before N-clusters degrade the material performance to a point where it is no longer suitable for devices. Finally, the efficacy of closed-cycle cooling of the MBE cryo-shroud could be further evaluated by comparing the concentrations of deep levels in AlGaAs growth with LN2 and with the chiller and performing low temperature PL measurements on AlGaAs layers. 127 Bibliography [1] R. N. Hall, R. O. Carlson, T. J. Soltys, G. E. Fenner and J. D. Kingsley. Coherent light emission from GaAs junctions. Phys. Rev. Lett., 9(9):366-368, 1962. [2] M. I. Nathan, W. P. Dumke, G. Burns, F. H. Dill and G. Lasher. Stimulated emission of radiation from GaAs p-n junctions. Appl. Phys. Lett., 1(3):62-64, 1962. [3] N. Holonyak and S. F. Bevacqua. Coherent (visible) light emission from Ga(As1-xPx) junctions. Appl. Phys. Lett., 1(4):82-83, 1962. [4] T. Quist, R. Rediker, R. Keyes, W. Krag, B. Lax, A. McWhorter and H. Zeigler. Semiconductor maser of GaAs. Appl. Phys. Lett., 1(4):91-92, 1962. [5] J. Dyment. Hermite-Gaussian mode patterns in GaAs junction lasers. Appl. Phys. Lett., 10(3):84-86, 1967. [6] J. J. Coleman. The development of the semiconductor laser diode after the first demonstration in 1962. Semiconductor Science and Technology, 27(9):090207, 2012. [7] H. Kroemer. A proposed class of heterojunction injection lasers. Proc. IEEE, 51(12):1782-1783, 1963. [8] I. Hayashi, M. B. Panish, P. W. Foy and S. Sumski. Junction lasers which operate continuously at room temperature. Appl. Phys. Lett., 17(3):109-111, 1970. [9] Veeco Instruments Inc., URL http://www.veeco.com/promos/lattice-parameter-and-bandgap-data.aspx. [10] H. Nelson. Epitaxial growth from the liquid state and its application to the fabrication of tunnel and laser diodes,\" RCA Review, 24(4):603-615, 1963. [11] J. Arthur. Interaction of Ga and As2 molecular beams with GaAs surfaces. J. Appl. Phys.,39(8):4032-4034, 1968. [12] A. Cho. Morphology of epitaxial growth of GaAs by a molecular beam method: the observation of surface structures. J. Appl. Phys., 41(7):2780-2786, 1970. [13] H. M. Manasevi and W. I. Simpson. Use of metal-organics in preparation of semiconductor materials. J. Electrochem. Soc., 116(12):1725-1732, 1969. [14] D. Reep and S. Ghandhi. Deposition of GaAs epitaxial layers by organometallic CVD. J. Electrochem. Soc., 130(3):675-680, 1983. 128 [15] C. H. Chen, G. B. Stringfellow, D. C. Gordon, D. W. Brown and B. A. Vaartstra. Tertiarybutyldimethylantimony - a new Sb source for low-temperature organometallic vapor-phase epitaxial-growth of InSb. Appl. Phys. Lett., 61(2):204-206, 1992. [16] M. O'Steen, E. Readinger, M. Doran and D. Hanser. Systems and technology for production-scale molecular beam epitaxy, in Molecular Beam Epitaxy: From Research to Mass Production, 1st ed. M. Henini, Ed. Elsevier Science, Boston, USA, 2012, p.657. [17] A. M. Jeanloui and C. Hamon. Properties of InSb1-xBix alloys .I. electrical measurements. Physica Status Solidi, 34(1):329, 1969. [18] A. M. Jeanloui, B. Ayrault and J. Vargas. Properties of InSb1-xBix alloys .2. optical absorption. Physica Status Solidi, 34(1):341, 1969. [19] B. Joukoff and A. M. Jeanloui. Growth of InSb1-xBix single-crystals by Czochralski method. J. Cryst. Growth, 12(2):169, 1972. [20] J. L. Zilko and J. E. Greene. Growth of metastable InSb1-xBix thin-films by multitarget sputtering. Appl. Phys. Lett., 33(3):254-256, 1978. [21] J. L. Zilko and J. E. Greene. Growth and phase-stability of epitaxial metastable InSb1-xBix films on GaAs .1. crystal-growth. J. Appl. Phys., 51(3):1549-1559, 1980. [22] K. Oe, S. Ando and K. Sugiyama. InSb1-xBix films grown by molecular-beam epitaxy. Jpn. J. Appl. Phys., 20(4):L303-L306, 1981. [23] A. Noreika, W. Takei, M. Francombe and C. Wood. Indium antimonide-bismuth compositions grown by molecular-beam epitaxy. J. Appl. Phys., 53(7):4932-4937, 1982. [24] T. Humphreys, P. Chiang, S. Bedair and N. Parikh. Metalorganic chemical vapor-deposition and characterization of the In-As-Sb-Bi material system for infrared detection. Appl. Phys. Lett., 53(2):142-144, 1988. [25] K. Y. Ma, Z. M. Fang, D. H. Jaw, R. M. Cohen, G. B. Stringfellow, W. P. Kosar and D. W. Brown. Organometallic vapor-phase epitaxial-growth and characterization of InAsBi and InAsSbBi. Appl. Phys. Lett., 55(23):2420-2422, 1989. [26] R. F. C. Farrow. The stabilization of metastable phases by epitaxy. J. Vac. Sci. Technol., B, 1(2):222-228, 1983. [27] K. Oe and H. Okamoto. New semiconductor alloy GaAs1-xBix grown by metal organic vapor phase epitaxy. Jpn. J. Appl. Phys., 37(11A):L1283-L1285, 1998. 129 [28] A. Mascarenhas, Y. Zhang, J. Verley and M. Seong. Overcoming limitations in semiconductor alloy design. Superlattices Microstruct., 29(6):395-404, 2001. [29] S. Tixier, M. Adamcyk, T. Tiedje, S. Francoeur, A. Mascarenhas, P. Wei and F. Schiettekatte. Molecular beam epitaxy growth of GaAs1-xBix. Appl. Phys. Lett., 82(14):2245-2247, 2003. [30] K. Oe and H. Asai. Proposal on a temperature-insensitive wavelength semiconductor laser. IEICE Trans. Electron., E79C(12):1751-1759, 1996. [31] Y. Zhang, A. Mascarenhas, J. Geisz, H. Xin and C. Tu. Discrete and continuous spectrum of nitrogen-induced bound states in heavily doped GaAs1-xNx. Phys. Rev. B, 63(8):085205, 2001. [32] U. Tisch, E. Finkman and J. Salzman. The anomalous bandgap bowing in GaAsN. Appl. Phys. Lett., 81(3):463-465, 2002. [33] O. Brandt, H. Yang, H. Kostial and K. Ploog. High p-type conductivity in cubic GaN/GaAs(113)A by using Be as the acceptor and O as the codopant. Appl. Phys. Lett., 69(18):2707-2709, 1996. [34] K. Ploog and O. Brandt. Doping of group III nitrides. J. Vac. Sci. Technol. A, 16(3):1609-1614, 1998. [35] M. Yoshimoto, S. Murata, A. Chayahara, Y. Horino, J. Saraie and K. Oe. Metastable GaAsBi alloy grown by molecular beam epitaxy. Jpn. J. Appl. Phys., 42(10B):L1235-L1237, 2003. [36] K. Oe. Characteristics of semiconductor alloy GaAs1-xBix. Jpn. J. Appl. Phys., 41(5A):2801-2806, 2002. [37] J. Yoshida, T. Kita, O. Wada and K. Oe. Temperature dependence of GaAs1-xBix band gap studied by photoreflectance spectroscopy. Jpn. J. Appl. Phys., 42(2A):371-374, 2003. [38] S. Francoeur, M. J. Seong, A. Mascarenhas, S. Tixier, M. Adamcyk and T. Tiedje. Band gap of GaAs1-xBix, 0 < x < 3.6%. Appl. Phys. Lett., 82(22):3874-3876, 2003. [39] R. B. Lewis, D. A. Beaton, X. Lu and T. Tiedje. GaAs1-xBix light emitting diodes. J. Cryst. Growth, 311(7):1872-1875, 2009. [40] D. G. Cooke, F. A. Hegmann, E. C. Young and T. Tiedje. Electron mobility in dilute GaAs bismide and nitride alloys measured by time-resolved terahertz spectroscopy. Appl. Phys. Lett., 89(12):122103, 2006. 130 [41] G. Feng, M. Yoshimoto, K. Oe, A. Chayahara and Y. Horino. New III-V semiconductor InGaAsBi alloy grown by molecular beam epitaxy. Jpn. J. Appl. Phys., 44(37):L1161-L1163, 2005. [42] J. P. Petropoulos, Y. Zhong and J. M. O. Zide. Optical and electrical characterization of InGaBiAs for use as a mid-infrared optoelectronic material. Appl. Phys. Lett., 99(3):031110, 2011. [43] S. K. Das, T. D. Das, S. Dhar, M. de la Mare and A. Krier. Near infrared photoluminescence observed in dilute GaSbBi alloys grown by liquid phase epitaxy. Infrared Phys. Technol., 55(1):156-160, 2012. [44] M. Vanschilfgaarde, A. Chen, S. Krishnamurthy and A. Sher. InTlP - a proposed infrared detector material. Appl. Phys. Lett., vol. 65(21):2714-2716, 1994. [45] S. P. Svensson and F. J. Crowne. A proposed TlGaAs/AlGaAs pseudomorphic heterostructure field effect transistor. J. Appl. Phys., 83(5):2599-2603, 1998. [46] A. Janotti, S. Wei and S. Zhang. Theoretical study of the effects of isovalent coalloying of Bi and N in GaAs. Phys. Rev. B, 65(11):115203, 2002. [47] X. Lu, D. A. Beaton, R. B. Lewis, T. Tiedje and Y. Zhang. Composition dependence of photoluminescence of GaAs1-xBix alloys. Appl. Phys. Lett., 95(4):041903, 2009. [48] R. Moon, G. Antypas and L. James. Bandgap and lattice-constant of GaInAsP as a function of alloy composition. J. Electron Mater, 3(3):635-644, 1974. [49] G. Liu, S. Chuang and S. Park. Optical gain of strained GaAsSb/GaAs quantum-well lasers: a self-consistent approach. J. Appl. Phys., 88(10):5554-5561, 2000. [50] K. Vanhollebeke, I. Moerman, P. Van Daele and P. Demeester. Extended wavelength InGaAs detectors grown by metal-organic vapor phase epitaxy (MOVPE) on compliant substrates. P. Soc. Photo.-Opt. Ins., 4169:325-336, 2000. [51] Y. Zhang, A. Mascarenhas and L. Wang. Similar and dissimilar aspects of III-V semiconductors containing Bi versus N. Phys. Rev. B, 71(15):155201, 2005. [52] B. Fluegel, S. Francoeur, A. Mascarenhas, S. Tixier, E. C. Young and T. Tiedje. Giant spin-orbit bowing in GaAs1-xBix. Phys. Rev. Lett., 97(6):067205, 2006. [53] T. Tiedje, E. C. Young and A. Mascarenhas. Growth and properties of the dilute bismide semiconductor GaAs1-xBix a complementary alloy to the dilute nitrides. Int. J. Nanotechnol., 5:963-983, 2008. [54] S. J. Sweeney. Bismide-alloys for higher efficiency infrared semiconductor lasers. 22nd IEEE International Semiconductor Laser Conference, Kyoto, Japan, 2010. 131 [55] S. Francoeur, S. Tixier, E. Young, T. Tiedje and A. Mascarenhas. Bi isoelectronic impurities in GaAs. Phys. Rev. B, 77(8):085209, 2008. [56] S. Imhof, A. Thraenhardt, A. Chernikov, M. Koch, N. S. Koester, K. Kolata, S. Chatterjee, S. W. Koch, X. Lu, S. R. Johnson, D. A. Beaton, T. Tiedje and O. Rubel. Clustering effects in Ga(AsBi). Appl. Phys. Lett., 96(13):131115, 2010. [57] G. Pettinari, A. Patane, A. Polimeni, M. Capizzi, X. Lu and T. Tiedje. Bi-induced p-type conductivity in nominally undoped Ga(AsBi). Appl. Phys. Lett., 100(9):092109, 2012. [58] D. A. Beaton, R. B. Lewis, M. Masnadi-Shirazi and T. Tiedje. Temperature dependence of hole mobility in GaAs1-xBix alloys. J. Appl. Phys., 108(8):083708, 2010. [59] R. N. Kini, A. J. Ptak, B. Fluegel, R. France, R. C. Reedy and A. Mascarenhas. Effect of Bi alloying on the hole transport in the dilute bismide alloy GaAs1-xBix. Phys. Rev. B, 83(7):075307, 2011. [60] R. N. Kini, L. Bhusal, A. J. Ptak, R. France and A. Mascarenhas. Electron Hall mobility in GaAsBi. J. Appl. Phys., 106(4):043705, 2009. [61] D. G. Cooke, F. A. Hegmann, E. C. Young and T. Tiedje. Electron mobility in dilute GaAs bismide and nitride alloys measured by time-resolved terahertz spectroscopy. Appl. Phys. Lett., 89(12):122103, 2006. [62] I. Hase. Heterojunction bipolar transistor with a base layer that contains bismuth. US patent no. 7,009,225, 2006. [63] K. Bertulis, A. Krotkus, G. Aleksejenko, V. Pacebutas, R. Adomavicius, G. Molis and S. Marcinkevicius. GaBiAs: A material for optoelectronic terahertz devices. Appl. Phys. Lett., 88(20):201112, 2006. [64] Y. Tominaga, K. Oe and M. Yoshimoto. Low temperature dependence of oscillation wavelength in GaAs1-xBix laser by photo-pumping. Appl. Phys. Express, 3(6):062201, 2010. [65] P. Ludewig, N. Knaub, N. Hossain, S. Reinhard, L. Nattermann, I. P. Marko, S. R. Jin, K. Hild, S. Chatterjee, W. Stolz, S. J. Sweeney and K. Volz. Electrical injection Ga(AsBi)/(AlGa)As single quantum well laser. Appl. Phys. Lett., 102(24):242115, 2013. [66] D. A. Beaton, R. B. Lewis, X. Lu and T. Tiedje. Gallium arsenide semiconductor material incorporating bismuth and process for epitaxial growth. CA patent application no. 2008/002245, 2009. 132 [67] S. J. Sweeney. Light emitting semiconductor device. GB patent application no. 2010/001249, 2010. [68] B. Predel, Z. Physik. Chem. Frankf. Neue Folge, 24:206, 1960. [69] P. Huber, O. Shpyrko, P. Pershan, B. Ocko, E. DiMasi and M. Deutsch. Tetra point wetting at the free surface of liquid Ga-Bi. Phys. Rev. Lett., 89(3):035502, 2002. [70] A. Kawazu, T. Otsuki and G. Tominaga, Growth of bismuth layers on Si(100) surfaces. Jpn. J. Appl. Phys., 20(3):553-560, 1981. [71] D. Schlom, W. Lee, T. Ma and J. Harris. Reduction of gallium-related oval defects. J. Vac. Sci. Technol. B, 7(2):296-298, 1989. [72] R. Sacks, P. Colombo, G. Patterson and K. Stair. Improved MBE-grown GaAs using a novel, high-capacity Ga effusion cell. J. Cryst. Growth, 175/176:66-71, 1997. [73] S. Tixier, M. Adamcyk, E. C. Young, J. H. Schmid and T. Tiedje. Surfactant enhanced growth of GaNAs and InGaNAs using bismuth. J. Cryst. Growth, 251:449-454, 2003. [74] E. Young, S. Tixier and T. Tiedje. Bismuth surfactant growth of the dilute nitride GaNxAs1-x. J. Cryst. Growth, 279:316-320, 2005. [75] M. R. Pillai, S. S. Kim, S. T. Ho and S. A. Barnett. Growth of InxGa1-xAs/GaAs heterostructures using Bi as a surfactant. J. Vac. Sci. Technol. B, 18(3):1232-1236, 2000. [76] X. Lu, D. A. Beaton, R. B. Lewis, T. Tiedje and M. B. Whitwick. Effect of molecular beam epitaxy growth conditions on the Bi content of GaAs1-xBix. Appl. Phys. Lett., 92(19):192110, 2008. [77] R. B. Lewis, M. Masnadi-Shirazi and T. Tiedje. Growth of high Bi concentration GaAs1-xBix by molecular beam epitaxy. Appl. Phys. Lett., 101(8):082112, 2012. [78] F. Bastiman, A. R. B. Mohmad, J. S. Ng, J. P. R. David and S. J. Sweeney. Non-stoichiometric GaAsBi/GaAs (100) molecular beam epitaxy growth. J. Cryst. Growth, 338(1):57-61, 2012. [79] M. Masnadi-Shirazi, D. A. Beaton, R. B. Lewis, X. Lu and T. Tiedje. Surface reconstructions during growth of GaAs1-xBix alloys by molecular beam epitaxy. J. Cryst. Growth, 338(1):80-84, 2012. [80] E. C. Young, M. B. Whitwick, T. Tiedje and D. A. Beaton. Bismuth incorporation in GaAs1-xBix grown by molecular beam epitaxy with in-situ light scattering. Phys. Status Solidi C, 4(5):1707-1710, 2007. 133 [81] S. Johnson, C. Lavoie, T. Tiedje and J. Mackenzie. Semiconductor substrate-temperature measurement by diffuse reflectance spectroscopy in molecular-beam epitaxy. J. Vac. Sci. Technol. B, 11(3):1007-1010, 1993. [82] G. J. Davies and D. Williams. III-V MBE growth systems, in The Technology and Physics of Molecular Beam Epitaxy, E. H. C. Parker, Ed. Plenum Press, New York, USA, 1985, p. 15. [83] S. Johnson and T. Tiedje. Effect of substrate thickness, back surface texture, reflectivity, and thin film interference on optical band-gap thermometry. J. Cryst. Growth, 175:273-280, 1997. [84] T. Ichikawa and S. Ino. Structural study of Sn-induced superstructures on Ge(111) surfaces by RHEED. Surf. Sci., 105:395-428, 1981. [85] S. Hasegawa, Y. Nagai, T. Oonishi and S. Ino. Hysteresis in phase-transitions at clean and Au-covered Si(111) surfaces. Phys. Rev. B, 47(15):9903-9906, 1993. [86] A. Ichimiya and C. Cohen. Reflection High Energy Electron Diffraction. Camebridge, Cambridge, UK, 2004. [87] V. Preobrazhenskii, M. Putyato, O. Pchelyakov and B. Semyagin. Experimental determination of the incorporation factor of As4 during molecular beam epitaxy of GaAs. J. Cryst. Growth, 201:170-173, 1999. [88] T. Shitara, D. Vvedensky, M. Wilby, J. Zhang, J. Neave and B. Joyce. Step-density variations and reflection high-energy electron-diffraction intensity oscillations during epitaxial-growth on vicinal GaAs(001). Phys. Rev. B, 46(11)6815-6824, 1992. [89] T. Pinnington, Y. Levy, J. MacKenzie and T. Tiedje. Real-time monitoring of InAs/GaAs quantum dot growth using ultraviolet light scattering. Phys. Rev. B, 60(23):15901-15909, 1999. [90] T. H. Pinnington. Surface morphology dynamics in strained-layer epitaxy. PhD thesis, University of British Columbia, 1999. [91] M. B. Whitwick. Surface evolution during gallium arsenide homoepitaxy with molecular beam epitaxy. PhD thesis, University of British Columbia, 2009. [92] J. Ralston, G. W. Wicks and L. F. Eastman. Reflection high-energy electron-diffraction intensity oscillation study of Ga desorption from molecular-beam epitaxially grown AlxGa1-xAs. J. Vac. Sci. Technol. B, 4(2):594-597, 1986. [93] B. K. Tanner and D. K. Bowen. X-Ray metrology in semiconductor manufacturing. CRC/Taylor & Francis, Boca Raton, USA, 2006. 134 [94] The National Institute of Standards and Technology (NIST). Resistivity and Hall measurements. URL http://www.nist.gov/eeel/semiconductor/hall.cfm. [95] D. A. Beaton. Electronic states and transport in GaNxAs1-x and GaAs1-xBix. PhD thesis, University of British Columbia, 2011. [96] R. Williams, Modern GaAs processing methods. Artech House, Boston, USA, 1990. [97] W. Edwards, A. Torrens and W. Hartman. Specific contact resistance of ohmic contacts to gallium-arsenide. Solid State Electron., 15(4):387-392, 1972. [98] G. Stareev. Formation of extremely low-resistance Ti/Pt/Au ohmic contacts to p-GaAs. Appl. Phys. Lett., 62(22):2801-2803, 1993. [99] J. Puustinen, M. Wu, E. Luna, A. Schramm, P. Laukkanen, M. Laitinen, T. Sajavaara and M. Guina. Variation of lattice constant and cluster formation in GaAsBi. J. Appl. Phys., 114:243504, 2013. [100] L. Pfeiffer and K. W. West. The role of MBE in recent quantum Hall effect physics discoveries. Phys. E, 20(1-2):57-64, 2003. [101] A. Y. Cho. Recent developments in molecular-beam epitaxy (MBE). Journal of Vacuum Science & Technology, vol. 16, pp. 275-284, 1979, 1979. [102] A. Y. Cho, H. C. Casey, C. Radice and P. W. Foy. Influence of growth-conditions on the threshold current-density of double-heterostructure lasers prepared by molecular-beam epitaxy. Electron. Lett., 16(2):72-74, 1980. [103] J. W. Robinson and M. Ilegems. Vacuum interlock system for molecular-beam epitaxy. Rev. Sci. Instrum., 49(2):205-207, 1978. [104] A. Cho. Epitaxial growth of gallium phosphide on cleaved and polished (111) calcium fluoride. J. Appl. Phys., 41(2):782, 1970. [105] A. Cho. Film deposition by molecular-beam techniques. J. Vac. Sci. Technol., 8(5):S31, 1971. [106] J. P. Hobson. Cryopumping. J. Vac. Sci. Technol., 10(1):73-79, 1973. [107] Eiko Corporation, T. Takamatsu, S. Takagi and G. Kido. High mobility HEMT growth with water-cooling MBE system. (unpublished), URL http://www.1974eiko.co.jp/pdf/report_3.pdf. [108] P. Redhead. Modeling the pump-down of a reversibly adsorbed phase .1. monolayer and submonolayer initial coverage. J. Vac. Sci. Technol. A, 13(2):467-475, 1995. 135 [109] I. Chun, B. L. Cho and S. M. Chung. Outgassing rate characteristic of a stainless-steel extreme high vacuum system. J. Vac. Sci. Technol. A, 14(4):2636-2640, 1996. [110] R. Sips. On the structure of a catalyst surface. J. Chem. Phys., 16(5):490-495, 1948. [111] R. Elsey. Outgassing of vacuum materials .1. theory and practice of vacuum science and technology in schools and colleges. Vacuum, 25(7):299-306, 1975. [112] R. B. Lewis, V. Bahrami-Yekta, M. J. Patel, T. Tiedje and M. Masnadi-Shirazi. Closed-cycle cooling of cryopanels in molecular beam epitaxy. J. Vac. Sci. Technol. B, 32:02C102, 2014. [113] R. B. Lewis, J. A. Mackenzie, T. Tiedje, D. A. Beaton, M. Masnadi-Shirazi, V. Bahrami-Yekta, K. P. Watkins and P. M. Mooney. Closed cycle chiller as a low cost alternative to liquid nitrogen in molecular beam epitaxy. J. Vac. Sci. Technol. B, 31:03C116, 2013. [114] J. F. O'Hanlon. A User's Guide to Vacuum Technology. Wiley-Interscience, New York, USA, 1980. [115] K. Prior, G. Davies and R. Heckingbottom. The thermodynamics of oxygen incorporation into semiconductor-III-V compounds and alloys in MBE. J. Cryst. Growth, 66(1):55-62, 1984. [116] G. B. Stringfellow, R. Stall and W. Koschel. Carbon in molecular-beam epitaxial GaAs. Appl. Phys. Lett., 38(3):156-157, 1981. [117] B. Lee, S. Bose, M. Kim, G. Stillman and A. Calawa. Spectroscopic studies of the influence of intentionally increased CO partial-pressure on the incorporation of residual-C impurities in molecular-beam epitaxy GaAs. Appl. Phys. Lett., 53(24):2438-2440, 1988. [118] D. V. Lang, A. Y. Cho, A. C. Gossard, M. Ilegems and W. Wiegmann. Study of electron traps in n-GaAs grown by molecular-beam epitaxy. J. Appl. Phys., 47(6):2558-2564, 1976. [119] D. V. Lang and R. A. Logan. Study of deep levels in GaAs by capacitance spectroscopy. J. Electron. Mater., 4(5):1053-1066, 1975. [120] M. Skowronski. Oxygen in gallium arsenide, in Deep Centers in Semiconductors A State-of-the-Art Approach, 2nd Ed. S. T. Pantelides, Ed. USA, Gordon and Breach Science Publishers, 1992. [121] G. Wicks, W. I. Wang, C. E. C. Wood, L. F. Eastman and L. Rathbun. Photoluminescence of AlxGa1-xAs grown by molecular-beam epitaxy. J. Appl. Phys., 52(9):5792-5796, 1981. 136 [122] K. Akimoto, M. Kamada, K. Taira, M. Arai and N. Watanabe. Photoluminescence killer center in AlGaAs grown by molecular-beam epitaxy. J. Appl. Phys., 59(8):2833-2836, 1986. [123] L. Pavesi and M. Guzzi. Photoluminescence of AlxGa1-xAs alloys. J. Appl. Phys., 75(10):4779-4842, 1994. [124] M. Ilegems. Properties of III-V layers, in The Technology and Physics of Molecular Beam Epitaxy, E. H. C. Parker, Ed. Plenum Press, New York, USA, 1985, p. 83. [125] E. C. Young. GaNAs and GaAsBi: Structural and Electronic Properties of Two Resonant State Semiconductor Alloys, PhD thesis, University of British Columbia, 2006. [126] A. W. Adamson. Physical Chemistry of Surfaces, 4th Ed. John Wiley and Sons, New York, USA, 1982. [127] S. Brunauer, P. H. Emmett and E. Teller. Adsorption of gases in multimolecular layers. J. Am. Chem. Soc., 60:309-319, 1938. [128] A. J. Ptak, R. France, D. A. Beaton, K. Alberi, J. Simon, A. Mascarenhas and C. S. Jiang. Kinetically limited growth of GaAsBi by molecular-beam epitaxy. J. Cryst. Growth, 338(1):107-110, 2012. [129] R. Kaspi and S. Barnett. Analysis of GaAs(001) surface stoichiometry using Monte-Carlo simulations. Surf. Sci., 241(1-2):146-156, 1991. [130] C. Heyn and M. Harsdorff. Simulation of GaAs growth and surface recovery with respect to gallium and arsenic surface kinetics. Phys. Rev. B, 55(11):7034-7038, 1997. [131] J. R. Arthur. Surface stoichiometry and structure of GaAs. Surf. Sci., 43(2):449-461, 1974. [132] V. Pacebutas, K. Bertulis, G. Aleksejenko and A. Krotkus. Molecular-beam-epitaxy grown GaBiAs for terahertz optoelectronic applications. J. Mater. Sci. -Mater. Electron., 20(S1):363-366, 2009. [133] J. P. Hirth and J. Lothe. Theory of Dislocations, 2nd Ed. Wiley Inter-science, New York, USA, 1982. [134] J. W. Matthews and A. E. Blakeslee. Defects in epitaxial multilayers .1. misfit dislocations. J. Cryst. Growth, 27:118-125, 1974. [135] D. Hwang. Strain relaxation in lattice-mismatched epitaxy. Mater. Chem. Phys., 40(4):291-297, 1995. 137 [136] L. B. Freund and S. Suresh. Thin Film Materials: Stress, Defect Formation and Surface Evolution. Cambridge University Press, New York, USA, 2003. [137] G. L. Price. Critical-thickness and growth-mode transitions in highly strained InxGa1-xAs Films. Phys. Rev. Lett., 66(4):469-472, 1991. [138] B. Dodson and J. Tsao. Relaxation of strained-layer semiconductor structures via plastic-flow. Appl. Phys. Lett., 51(17):1325-1327, 1987. [139] K. Kavanagh, M. Capano, L. Hobbs, J. Barbour, P. Maree, W. Schaff, J. Mayer, D. Pettit, J. Woodall, J. Stroscio and R. Feenstra. Asymmetries in dislocation densities, surface-morphology, and strain of GaInAs/GaAs single heterolayers. J. Appl. Phys., 64(10):4843-4852, 1988. [140] I. Yonenaga and K. Sumino. Behavior of dislocations in GaAs revealed by etch pit technique and x-ray topography. J. Cryst. Growth, 126(1):19-29, 1993. [141] R. France, A. J. Ptak, C. S. Jiang and S. P. Ahrenkiel. Control of asymmetric strain relaxation in InGaAs grown by molecular-beam epitaxy. J. Appl. Phys., 107(10):103530, 2010. [142] R. France, C. S. Jiang and A. J. Ptak. In situ strain relaxation comparison between GaAsBi and GaInAs grown by molecular-beam epitaxy. Appl. Phys. Lett., 98(10):101908, 2011. [143] C. W. Snyder, B. G. Orr, D. Kessler and L. M. Sander. Effect of strain on surface-morphology in highly strained InGaAs films. Phys. Rev. Lett., 66(23):3032-3035, 1991. [144] Y. Takehara, M. Yoshimoto, W. Huang, J. Saraie, K. Oe, A. Chayahara and Y. Horino. Lattice distortion of GaAsBi alloy grown on GaAs by molecular beam epitaxy. Jpn. J. Appl. Phys., 45(1A):67-69, 2006. [145] V. Pacebutas, K. Bertulis, L. Dapkus, G. Aleksejenko, A. Krotkus, K. M. Yu and W. Walukiewicz. Characterization of low-temperature molecular-beam-epitaxy grown GaBiAs layers. Semicond. Sci. Technol., 22(7):819-823, 2007. [146] P. Wei, S. Tixier, M. Chicoine, S. Francoeur, A. Mascarenhas, T. Tiedje and F. Schiettekatte. Ion beam characterization of GaAs(1-x-y)NxBiy epitaxial layers. Nucl. Instrum. Methods Phys. Res., 219:671-675, 2004. [147] J. A. Kerr. Strengths of chemical bonds, in CRC Handbook of Chemistry and Physics, 73rd Ed. David R. Lide, Ed. CRC Press, Florida, USA, 1992. 138 [148] G. Vardar, S. W. Paleg, M. V. Warren, M. Kang, S. Jeon and R. S. Goldman. Mechanisms of droplet formation and Bi incorporation during molecular beam epitaxy of GaAsBi. Appl. Phys. Lett., 102(4):042106, 2013. [149] J. Schmitz, J. Wagner, F. Fuchs, N. Herres, P. Koidl and J. Ralston. Optical and structural investigations of intermixing reactions at the interfaces of InAs/AlSb and InAs/GaSb quantum-wells grown by molecular-beam epitaxy. J. Cryst. Growth, 150(1-4):858-862, 1995. [150] Q. Xie, J. Van Nostrand, J. Brown and C. Stutz. Arsenic for antimony exchange on GaSb, its impacts on surface morphology, and interface structure. J. Appl. Phys., 86(1):329-337, 1999. [151] A. Albo, C. Cytermann, G. Bahir and D. Fekete. Strain-induced nitrogen incorporation in atomic layer epitaxy growth of InAsN/GaAs quantum wells using metal organic chemical vapor deposition. Appl. Phys. Lett., 95(5):051102, 2009. [152] P. M. J. Marée, J. C. Barbour, J. F. van der Veen, K. L. Kavanagh, C. W. T. Bulle‐Lieuwma and M. P. A. Viegers. Generation of misfit dislocations in semiconductors. J. Appl. Phys., 62(11):4413-4420, 1987. [153] E. Iliopoulos, A. Adikimenakis, C. Giesen, M. Heuken and A. Georgakilas. Energy bandgap bowing of InAlN alloys studied by spectroscopic ellipsometry. Appl. Phys. Lett., 92(19):191907, 2008. [154] I. Vurgaftman, J. Meyer and L. Ram-Mohan. Band parameters for III-V compound semiconductors and their alloys. J. Appl. Phys., 89(11):5815-5875, 2001. [155] J. I. Pankove, Optical Processes in Semiconductors. Prentice-Hall, New Jersey, USA, 1971. [156] F. Urbach. The long-wavelength edge of photographic sensitivity and of the electronic absorption of solids. Phys. Rev., 92(5):1324-1324, 1953. [157] M. Sturge. Optical absorption of gallium arsenide between 0.6 and 2.75 eV. Phys. Rev., 127(3):768, 1962. [158] D. Sell. Resolved free-exciton transitions in the optical-absorption spectrum of GaAs. Phys. Rev. B, 6(10):3750, 1972. [159] M. M. Kreitman and D. L. Barnett. Probability tables for clusters of foreign atoms in simple lattices assuming next-nearest-neighbor interactions. J. Chem. Phys., 43(2):364, 1965. [160] P. M. Mooney, K. P. Watkins, Z. Jiang, A. F. Basile, R. B. Lewis, V. Bahrami-Yekta, M. Masnadi-Shirazi, D. A. Beaton and T. Tiedje. Deep level defects in n-type 139 GaAsBi and GaAs grown at low temperatures. J. Appl. Phys., 113(13):133708, 2013. [161] D. Zhang, K. Radhakrishnan and S. Yoon. The effect of As/Ga flux ratio on Si-doped GaAs-layers grown by molecular-beam epitaxy. J. Cryst. Growth, 135(3-4):441-446, 1994. [162] T. Fuyuki, S. Kashiyama, Y. Tominaga, K. Oe and M. Yoshimoto. Deep-hole traps in p-type GaAs1-xBix grown by molecular beam epitaxy. Jpn. J. Appl. Phys., 50(8):080203, 2011. [163] M. Sotoodeh, A. Khalid and A. Rezazadeh. Empirical low-field mobility model for III-V compounds applicable in device simulation codes. J. Appl. Phys., 87(6):2890-2900, 2000. "@en ; edm:hasType "Thesis/Dissertation"@en ; vivo:dateIssued "2014-05"@en ; edm:isShownAt "10.14288/1.0103418"@en ; dcterms:language "eng"@en ; ns0:degreeDiscipline "Physics"@en ; edm:provider "Vancouver : University of British Columbia Library"@en ; dcterms:publisher "University of British Columbia"@en ; dcterms:rights "Attribution-NonCommercial-NoDerivs 2.5 Canada"@en ; ns0:rightsURI "http://creativecommons.org/licenses/by-nc-nd/2.5/ca/"@en ; ns0:scholarLevel "Graduate"@en ; dcterms:title "Molecular beam epitaxy growth technology and properties of GaAsBi alloys"@en ; dcterms:type "Text"@en ; ns0:identifierURI "http://hdl.handle.net/2429/46478"@en .