<<

Properties of thin film III-V/IV alloys and nanostructures

By

Roger Jia

B.S.E. and Engineering University of Michigan, Ann Arbor, 2011

Submitted to the Department of Material Science and Engineering in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy in Materials Science and Engineering at the Massachusetts Institute of Technology

September 2017

© 2017 Massachusetts Institute of Technology. All rights reserved.

Signature of Author:______Department of Materials Science and Engineering July 31, 2017

Certified by:______Eugene A. Fitzgerald Merton C. Flemings-SMA Professor of Materials Science and Engineering Thesis Supervisor

Accepted by:______Donald R. Sadoway John F. Elliott Professor of Materials Chair, Departmental Committee on Graduate Students

2

Properties of thin film III-V/IV semiconductor alloys and nanostructures

By

Roger Jia

Submitted to the Department of Material Science and Engineering on July 31, 2017 in Partial Fulfilment of the Requirements for the Degree of Doctor of Philosophy in Materials Science and Engineering

ABSTRACT

A large amount of research and development has been devoted to engineering materials for the next generation of semiconductor devices with high performance, efficiency, and economic viability. To this end, significant efforts have been made to grow semiconductor thin films with the desired properties onto lattice constants with viable, cost effective substrates. Comparatively less effort has been made to explore III-V/IV heterovalent nanostructures and alloys, which may exhibit properties not available in existing materials. The investigation of these structures, grown using MOCVD, is the goal of this thesis and is motivated by two factors: one, that III-V/IV nanostructures should be good thermoelectrics based on the “phonon glass electron ” concept, and two, that (GaAs)1-x(Ge2)x alloys were observed to exhibit near- infrared room temperature luminescence, a result that can have significant implications for low bandgap optical devices. A survey of various growth conditions was conducted for the growth of the model GaAs/Ge system using MOCVD to gain insight in the involving heterovalent materials and to identify structures suitable for investigation for their thermoelectric and optical properties. A significant decrease in the thermal conductivities of GaAs/Ge nanostructures and alloys relative to bulk GaAs and bulk Ge was observed. This reduction can be attributed to the presence of the heterovalent interfaces. The electron mobilities of the structures were determined to be comparable to bulk Ge, indicating minimal disruption to electron transport by the interfaces. A further reduction in thermal conductivity was observed in an (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy; the alloy had a thermal conductivity of 4.3 W/m-K, comparable to some state-of-the-art . Room temperature photoluminescence measurements of various compositions of (GaAs)1-x(Ge2)x alloys revealed a maximum energy transition of 0.8 eV. This bandgap narrowing is the result of composition fluctuations; the fluctuations create regions of lower bandgap, resulting in a weak dependence on luminescence as a function of Ge composition as well as lower bandgap than the homogeneous alloy with the same composition. As was added to the (GaAs)1-x(Ge2)x alloy, the bandgap increased despite the composition fluctuations. Based on the results from this work III-V/IV nanostructures show promise for thermoelectric and optical applications.

Thesis Supervisor: Eugene A. Fitzgerald Title: Merton C. Flemings-SMA Professor of Materials Science and Engineering

3

4

Acknowledgements

This thesis is the culmination of six years of work, and it could not have been done without the advice, guidance, and aid of the many people that I have met during my time here at MIT. I would like to thank my advisor, Professor Eugene Fitzgerald, for guiding me not only through the actual research project but also through the ups and downs of the PhD process. Particularly with the downs; Gene always knew what to say whenever I came to him with big misgivings on whether I was actually going through the “usual PhD valley of death” and not some struggle due to an inability to grasp how to do research. I would like to thank my thesis committee, Professor Carl Thompson and Professor Jeff Grossman, for providing feedback and evaluating my work. Their advice was very helpful in ultimately creating a more coherent work and final presentation. I would like to thank my collaborators, Lingping Zeng from Professor Gang Chen’s group in the department of mechanical engineering, Dongwook Lee from Professor Yang Shao-Horn’s group in the department of materials science and engineering, and Tony Zhu from Professor Vladimir Bulović’s group in the department of electrical engineering and computer science, for providing me with the thermal conductivity, Seebeck coefficient, and low temperature photoluminescence results in this work. In addition, I would like to thank several other members and alumni of Professor Chen’s group, Kimberly Collins, Maria Luckyanova, and Lee Weinstein, for their help in various early stage experiments. I would like to make a special note of the Fitzgerald research group. The vast majority of the research in the group relies on the operation of the MOCVD reactor. Not even six months of my PhD passed when the reactor began to exhibit significant problems hindering normal operation, and malfunctions continued to occur often enough for the following 3 years to never allow us to feel comfortable. Maintaining reactor operation throughout the years would have been many times more stressful if not for the great camaraderie of the group. I would also like to thank the members for all their help over the years from making TEM samples to research advice and suggestions: Adam Jandl, who was my first point of contact when I started out, Prithu Sharma, Tim Milakovich, Kunal Mukherjee, Ryan Iutzi, Chris Heidelberger, Rushabh Shah, and Mayank Bulsara. I would like to thank Dr. Yong Zhang and Dr. Shiahn Chen of the MIT Center of Materials Science and Engineering for their instruction and assistance with operation of the TEMs. I would like to acknowledge funding support from the Solid-State Solar Thermal Energy Conversion Center, an Energy Frontier Research Center of the U.S. Department of Energy, and the Singapore-MIT Alliance for Research and Technology. Lastly, this work could not be done without the huge support of my family. I hope this work brings joy to them, especially my father, who had to abandon his own PhD ambitions 27 years ago in order to provide for a certain newborn baby.

5

Table of Contents

List of Figures ...... 9 List of Tables ...... 15 Chapter 1: Introduction and Background ...... 16 1.1. Motivation ...... 17 1.2. Thesis Objectives and Organization ...... 19 1.3. III-V/IV epitaxy...... 20 1.3.1. Polar on non-polar epitaxy ...... 22 1.3.1.1. Antiphase Domains ...... 23 1.3.2. Non-polar on polar epitaxy ...... 24

1.3.3. Metastable (GaAs)1-x(Ge2)x alloys ...... 26 1.4. Overview of Applications ...... 27 1.4.1. Thermoelectrics ...... 27 1.4.1.1. Thermoelectric figure of merit ...... 28 1.4.1.2. Nanostructuring...... 29 1.4.2. Low bandgap materials lattice matched to GaAs ...... 30 Chapter 2: Materials Growth and Characterization ...... 32 2.1. Film Epitaxy ...... 33 2.2. Structural Characterization ...... 34 2.2.1. Transmission Electron Microscopy (TEM) ...... 34 2.2.2. Electron-Dispersive X-ray Spectroscopy (EDS) ...... 36 2.3. Optical Characterization ...... 37 2.3.1. Photoluminescence (PL) ...... 37 2.3.2. Absorption Spectroscopy ...... 38 2.4. Time-Domain Thermoreflectance (TDTR) ...... 38 2.5. Electrical Characterization ...... 39 2.5.1. Van der Pauw – Hall method ...... 39 2.5.2. Seebeck Measurements...... 41 Chapter 3: Epitaxy of the GaAs/Ge System ...... 42 3.1. Introduction ...... 43 3.2. Growth of GaAs/Ge ...... 44

6

3.2.1. 650°C and 100 Torr growth (samples 1 and 2) ...... 46 3.2.2. 500°C and 100 Torr growth (samples 3 and 4) ...... 47 3.2.3. 650°C and 250 Torr growth (samples 5 – 12) ...... 50 3.2.3.1. Reduction in period thicknesses (samples 9 – 12) ...... 54 3.2.4. 500°C and 250 Torr growth (samples 13 and 14) ...... 55

3.3. Growth of (GaAs)1-x(Ge2)x alloys ...... 56 3.4. Summary ...... 63 Chapter 4: Thermoelectric Properties of III-V/IV Nanostructures ...... 65 4.1. Introduction ...... 66 4.1.1. Structural quality of materials ...... 66 4.2. Design of experiment ...... 67 4.2.1. GaAs/Ge superlattices ...... 67 4.2.1.1. Determination of interface density...... 69 4.2.2. III-V/IV alloy structures ...... 71 4.3. GaAs/Ge structures ...... 72 4.3.1. Thermal conductivity ...... 72 4.3.2. Electrical mobility ...... 75 4.3.3. Seebeck coefficient ...... 79 4.4. Estimation of thermoelectric potential for the GaAs/Ge system...... 79

4.5. Thermoelectric properties of (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 ...... 82

4.5.1. Optimization of (In1-xGaxAs)1-z(Si1-yGey)z ...... 84 4.6. Summary ...... 87 Chapter 5: Optical Properties of III-V/IV Alloy Structures ...... 89 5.1. Introduction ...... 90

5.2. Optical measurements of the (GaAs)1-x(Ge2)x alloys ...... 91

5.3. Optical measurements of (GaAs)1-y(Si1-xGex)y and (In1-xGaxAs)1-z(Si1-yGey)z ...... 95 5.4. Possible explanations for the origin of photoluminescence ...... 99 5.4.1. Bandgap narrowing due to composition fluctuation...... 99 5.4.2. Radiative and non-radiative defects ...... 100 5.4.3. Recombination at interfaces due to spatial separation of photoexcited carriers ...... 101 5.5. Summary ...... 102 Chapter 6: Conclusions and Future Work ...... 104

7

6.1. Summary of results...... 105

6.1.1. MOCVD of GaAs/Ge superlattices and (GaAs)1-x(Ge2)x alloys ...... 105 6.1.2. Thermoelectric properties of III-V/IV structures ...... 106 6.1.3. Optical properties of III-V/IV alloys ...... 107 6.2. Future directions ...... 108 Appendix A ...... 110 References ...... 116

8

List of Figures

Figure 1.1. Bandgap vs. lattice parameter diagram of II-VI, III-V, and group IV . Lines denote ternary compositions (binary in case of SiGe) of the constituent endpoint semiconductors. Line colors represent the conduction band valley lowest in energy; materials with a Γ-valley have direct bandgaps, while materials with X- or L- valleys have indirect bandgaps. Dashed lines indicate estimated compositions based on a linear trend (no experimental data available). Image adapted from Woodall Research Group.21 ...... 19

Figure 1.2. GaAs-Ge phase diagram. Mutual immiscibility is observed in the material system. Phase separation would be expected at all compositions absent kinetic limitations. Adapted from Ansara22...... 21

Figure 1.3. Primary modes of thin film growth a) Volmer-Weber (island growth) b) Frank-van der Merwe (-by-layer) c) Stranski-Krastanov (layer-by-layer up to a critical thickness, followed by island growth).23 The growth mode of a material depends on the interaction of the adatom and surface...... 22

Figure 1.4. Diagram of a GaAs lattice on a (001) group IV substrate with a single atomic step, resulting in the formation of an antiphase boundary (vertical dashed line).25 ...... 24

Figure 1.5. Diagram of a GaAs lattice on a (001) group IV substrate with a double atomic step. No antiphase boundary forms since the lattice on either side of the step has the same orientation.25...... 24

Figure 1.6. Model of the GaAs 풄(ퟒ × ퟒ) surface reconstruction.29 ...... 25

Figure 1.7. Model of the GaAs 풄(ퟖ × ퟐ) surface reconstruction.29 ...... 26

Figure 1.8. (002) dark-field cross-section transmission electron microscopy (XTEM) image of (GaAs)1-x(Ge2)x layers grown by MOCVD, showing pronounced phase separation. Dark regions are Ge-rich.9 ...... 27

Figure 1.9. Conflicting nature of the thermoelectric properties. Axis scale: Thermal conductivity, κ, 0 to 10 W m-1 K-1. Seebeck coefficient, α, 0 to 500 µV K-1. Electrical conductivity, σ, 0 to -1 -1 33 5000 Ω cm . Based on Bi2Te3 data...... 29

Figure 1.10. Dependence of the thermoelectric figure of merit, zT, of various materials as a function of temperature.33 ...... 29

Figure 1.11. Direct bandgap of (GaAs)1-x(Ge2)x with composition. Image adapted from Koiller et al.8 ...... 31

Figure 2.1. Schematic of an MOCVD system.50 ...... 33

Figure 2.2. Diffraction pattern of a GaAs crystal in the [110] zone axis...... 35

9

Figure 2.3. Diagram of a) brightfield imaging and b) off-axis darkfield imaging in TEM.51 ...... 35

Figure 3.1. (002) Darkfield XTEM image of GaAs/Ge superlattices grown at 650°C and 100 Torr with substrate oriented in (a) the exact (001) direction (sample 1) and (b) (001) 6° offcut towards the <111>A direction (sample 2). Antiphase boundaries (APBs) in the GaAs layers were confirmed through observing contrast reversal in adjacent domains in (002) darkfield and (00-2) darkfield images (not shown)...... 46

Figure 3.2. XTEM images of GaAs/Ge superlattices grown at 500°C and 100 Torr. (a) (002) darkfield and (c) (00-2) darkfield images of sample 3, grown using a substrate oriented in the exact (001) direction. (b) (002) darkfield and (d) (00-2) darkfield images of sample 4, grown using a substrate oriented (001) 6° offcut towards the <111>A direction. APBs were determined from observing contrast reversal in adjacent domains in the (002) darkfield and (00-2) darkfield images. Stacking faults (SF) are observed in sample 3, and are also present in sample 4 (not shown)...... 49

Figure 3.3. Dependence of Ge growth rate on total system pressure. Growth was done at 750°C on GaAs substrates with orientations (001) 3° offcut towards <110>, (110), and As terminated (111).60 ...... 51

Figure 3.4. (002) Darkfield XTEM images of GaAs/Ge superlattices grown at 650°C and 250 Torr using the N2 anneal surface preparation method. (a) Sample 5, grown on a substrate oriented in the exact (001) direction. (b) Sample 6, grown on a substrate oriented (001) 6° off towards the <111>A direction. APBs in the GaAs layers were confirmed through observing contrast reversal in adjacent domains in (002) darkfield and (00-2) darkfield images (not shown)...... 52

Figure 3.5. (002) Darkfield XTEM images of GaAs/Ge superlattices grown at 650°C and 250 Torr using the TMGa pulse surface preparation method. (a) Sample 7, grown on a substrate oriented in the exact (001) direction. (b) Sample 8, grown on a substrate oriented (001) 6° off towards the <111>A direction. APBs in the GaAs layers were confirmed through observing contrast reversal in adjacent domains in (002) darkfield and (00-2) darkfield images (not shown)...... 53

Figure 3.6. (002) Darkfield XTEM images of GaAs/Ge superlattices grown at 650°C and 250 Torr. Sample 9 (a) and sample 11 (b) were grown on substrates oriented in the exact (001) direction, while sample 10 (c) and sample 12 (d) were grown on substrates oriented (001) 6° offcut towards the <111>A direction. N2 anneal was used in (a) and (c) while TMGa pulsing was used in (b) and (d). Precursor flows and flow times during deposition of each GaAs and Ge layer were the same for all samples. APBs in the GaAs layers were confirmed through observing contrast reversal in adjacent domains in (002) darkfield and (00-2) darkfield images (not shown)...... 55

Figure 3.7. (002) Darkfield XTEM images of GaAs/Ge superlattices grown at 500°C and 250 Torr. (a) substrate oriented in the exact (001) direction (sample 13). (b) substrate oriented (001) 6° offcut towards the <111>A direction (sample 14). Stacking faults are observed in both samples...... 56

10

Figure 3.8. (002) Darkfield XTEM images of (GaAs)0.73(Ge2)0.27 alloys grown at 650°C and 100 Torr. a) substrate oriented in the exact (001) direction (sample 15). b) substrate oriented (001) 6° off towards the <111>A direction (sample 16)...... 59

Figure 3.9. (002) Darkfield XTEM images of (GaAs)0.72(Ge2)0.28 alloys grown at 650°C and 250 Torr. a) substrate oriented in the exact (001) direction (sample 17). b) substrate oriented (001) 6° off towards the <111>A direction (sample 18)...... 60

Figure 3.10. a) (002) Darkfield XTEM image of a (GaAs)0.77(Ge2)0.23 alloy grown at 650°C and 250 Torr. with substrate oriented (001) 6° off towards the <111>B direction (sample 19). b) magnified image of box in a) ...... 61

Figure 3.11. a) (002) Darkfield XTEM image of a (GaAs)0.83(Ge2)0.17 alloy grown at 575°C and 250 Torr with substrate oriented in the exact (001) direction (sample 20). b) (002) Darkfield XTEM image of a (GaAs)0.71(Ge2)0.29 alloy grown at 700°C and 250 Torr with substrate oriented in the exact (001) direction (sample 21)...... 61

Figure 3.12. (002) Darkfield XTEM image of a GaAs – (GaAs)0.77(Ge2)0.23 – GaAs structure grown at 650°C and 250 Torr with substrate oriented in the exact (001) direction (sample 22).. 63

Figure 4.1. (002) Darkfield XTEM image of GaAs/Ge on-axis samples with interface densities of (a) 0.02 nm-1, (b) 0.037 nm-1, (c) 0.03 nm-1, (d) 0.109 nm-1, (e) 0.12 nm-1, and (f) 0.106 nm-1. Bright bands/regions correspond to GaAs while dark bands/regions correspond to Ge. Antiphase boundaries (APBs) can be seen as dark vertical lines in the GaAs regions. See section 4.2.1.1 for discussion on the interface density...... 68

Figure 4.2. (002) Darkfield XTEM image of GaAs/Ge superlattice offcut samples with interface densities of (a) 0.022 nm-1, (b) 0.039 nm-1, (c) 0.038 nm-1, (d) 0.101 nm-1, (e) 0.105 nm-1, and (f) 0.068 nm-1. Bright bands/regions correspond to GaAs while dark bands/regions correspond to Ge. Antiphase boundaries (APBs) can be seen as dark vertical lines in the GaAs regions. See section 4.2.1.1 for discussion on the interface density...... 69

Figure 4.3. Threshold (right) of an (002) darkfield TEM image (left). An image processor can be used to calculate the perimeter of the dark regions. This perimeter is then summed with the perimeter calculated for the reversed image (result: 2x interface length + original image border). Total interface length is then extracted by subtracting the original image border and dividing the result by two...... 70

Figure 4.4. (002) Darkfield XTEM image of (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy grown on (001) exact GaAs. Defects and phase separation were not observed in the structure...... 71

Figure 4.5. Representative measured phase signal along with the best model fit based on the thermal diffusion model. The dashed curves represent the model fits by adjusting the best fit thermal conductivity by ±10% to show the sensitivity of the measurement to the underlying thermal conductivity...... 72

11

Figure 4.6. Thermal conductivities and the corresponding interface density of the on-axis samples. The alloy sample is shown on the right. Thermal conductivities of bulk GaAs (triangle) and Ge (*) are shown at 0 interface density for comparison...... 73

Figure 4.7. Thermal conductivities and the corresponding interface density of the offcut samples. The alloy sample is shown on the right. Thermal conductivities of bulk GaAs (triangle) and Ge (*) are shown at 0 interface density for comparison...... 74

Figure 4.8. Electron mobility of bulk GaAs.81 The samples in Table 4.1 are plotted for comparison...... 78

Figure 4.9. Electron mobility of bulk Ge. 1. Low temperature measurement (77K) 2. Room temperature measurement (300K).82 The samples in Table 4.1 are plotted for comparison...... 78

Figure 4.10. Dependence of thermal resistivity of In1-xGaxAs on composition. Experimental values shown as a solid line. Dotted line denotes a theoretical fit.87 ...... 83

88 Figure 4.11. Dependence of thermal conductivity of Si1-xGex on composition...... 83

90 Figure 4.12. (Solid curves) Dependence of electron mobility of In1-xGaxAs on composition. 1 – n=3×1015 cm-3; 2 – n=4×1016 cm-3; 3 – n=2.3×1017 cm-3 ...... 86

91 Figure 4.13. Dependence of electron mobility of Si1-xGex on composition...... 86

Figure 5.1. Room temperature photoluminescence spectra of (GaAs)1-x(Ge2)x alloy samples with Ge compositions less than 20%. The Luminescence peaks of all samples at approximately 1550nm. Luminescence intensity decreases with decreasing Ge composition...... 92

Figure 5.2 Absorption coefficient dependence on excitation energy for the (GaAs)0.83(Ge2)0.17, (GaAs)0.88(Ge2)0.12, and (GaAs)0.94(Ge2)0.06 alloys. The absorption edge for each sample appears to be around 0.8 eV. (Note: A measurement artifact is responsible for the apparent absorption between 0.6 – 0.8 eV.) ...... 92

Figure 5.3. 77K photoluminescence spectrum of (GaAs)0.81(Ge2)0.19. Peak wavelength at approximately 1495nm...... 93

Figure 5.4. Room temperature photoluminescence spectra of the (GaAs)0.81(Ge2)0.19, (GaAs)0.77(Ge2)0.23, and (GaAs)0.71(Ge2)0.29 alloy samples. Luminescence peak for (GaAs)0.81(Ge2)0.19 at approximately 1550nm (0.8 eV) is consistent with the results obtained in Fig. 5.1. Luminescence peaks for (GaAs)0.77(Ge2)0.23 and (GaAs)0.71(Ge2)0.29 at approximately 1720nm (0.72 eV) and 1760nm (0.7 eV), respectively...... 94

Figure 5.5. 77K photoluminescence spectrum of (GaAs)0.71(Ge2)0.29, with multiple luminescence peaks resolved. These peaks likely correspond to different striated regions in the structure...... 94

Figure 5.6. Comparison of the photoluminescence spectra at room temperature for (GaAs)0.8(Si0.1Ge0.9)0.2 and (GaAs)0.81(Ge2)0.19. The addition of Si causes a blueshift in the peak luminescence...... 96

12

Figure 5.7. Comparison of the photoluminescence spectra at 77K for (GaAs)0.8(Si0.1Ge0.9)0.2 and (GaAs)0.81(Ge2)0.19. The blueshift in the peak luminescence is significantly larger at low temperature...... 96

Figure 5.8. Photoluminescence spectra at room temperature for (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 and (GaAs)0.81(Ge2)0.19...... 97

Figure 5.9. Photoluminescence spectra at 77K for (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 (actual signal × 10) and (GaAs)0.81(Ge2)0.19...... 97

Figure 5.10. 77K photoluminescence spectrum of the (In0.15Ga0.85As)0.8(Si0.1Ge0.9)0.2 alloy...... 98

Figure 5.11. 77K photoluminescence spectrum of the (In0.15Ga0.85As)0.8(Si0.1Ge0.9)0.2 alloy. Peaks are observed at 1495nm (0.83 eV), 1710nm (0.73), 1860nm (0.67), and 1880nm (0.66). The significant degree of phase separation observed in the morphology likely causes the numerous luminescence peaks...... 98

Figure 5.12. Possible band diagram schematic of the (GaAs)1-x(Ge2)x alloys. Based on our results in section 4.3.2, we estimate that the conduction band of (GaAs)1-x(Ge2)x at all compositions is approximately aligned. Due to the composition fluctuations in the samples, regions with higher Ge compositions than the sample average, and thus smaller bandgaps, may exist. This causes a narrowing of the bandgap. We suspect that by reducing the Ge composition (bottom), these high Ge composition regions are still present, but simply reduced in density. . 102

Figure A1. (002) Darkfield XTEM image of the (GaAs)0.81(Ge2)0.19 alloy. The sample was grown by codeposition of GaAs and Ge at 650°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX...... 110

Figure A2. (002) Darkfield XTEM image of the (GaAs)0.88(Ge2)0.12 alloy. The sample was grown by codeposition of GaAs and Ge at 575°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX...... 111

Figure A3. (002) Darkfield XTEM image of the (GaAs)0.94(Ge2)0.06 alloy. The sample was grown by codeposition of GaAs and Ge at 575°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX...... 112

Figure A4. (002) Darkfield XTEM image of the (GaAs)0.77(Ge2)0.23 alloy. The sample was grown by codeposition of GaAs and Ge at 650°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX...... 113

13

Figure A5. (002) Darkfield XTEM image of the (GaAs)0.8(Si0.1Ge0.9)0.2 alloy. The sample was grown by codeposition of GaAs and SiGe at 650°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX...... 114

Figure A6. (002) Darkfield XTEM image of the (In0.15Ga0.85As)0.8(Si0.1Ge0.9)0.2 alloy. The sample was grown by codeposition of InGaAs and SiGe at 650°C and 250 Torr on a semi- insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX...... 115

14

List of Tables

Table 3.1. Summary of grown structures and epitaxy conditions. Thicknesses of the GaAs and Ge layers were determined using TEM. For samples with non-uniform layers, approximate average thicknesses are given. For samples 3, 4, 13, and 14, thicknesses given do not reflect the thickness of the first Ge layer deposited; Ge growth rates are not determined due to difficulty in separating out incubation time...... 45

Table 3.2. Summary of epitaxy conditions of the (GaAs)1-x(Ge2)x alloys. Compositions were determined using EDX. Sample 22 – “double heterostructure” with 200nm GaAs and GaAs substrate cladding the alloy film...... 57

Table 4.1. Thermoelectric transport properties of several offcut superlattice samples and the GaAs/Ge alloy structures...... 76

Table 4.2. Thermoelectric properties of (GaAs)0.77(Ge2)0.23 and (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy structures...... 82

Table 5.1. Alloy structures investigated. Compositions were determined using EDX...... 90

15

Chapter 1: Introduction and Background

16

1.1. Motivation

A heterovalent structure is composed of materials AX and BY, where A and B have different valences, as well as X and Y. In semiconductors, heterovalent structures largely fall under the

II-VI/III-V material system or the III-V/IV material system. Many of these structures, such as

GaP/Si,1,2 GaAs/Ge,3–9 ZnTe/GaSb,10–12 and ZnSe/GaAs,13 have attracted significant interest for photovoltaic and optoelectronic applications due to the ability to access multiple different bandgaps on the same lattice constant. An even wider range of bandgaps can be accessed by alloying to obtain ternary compositions (binary for the group IV system), as seen in Fig. 1.1.

Additionally, the III-V/IV material system is one of the primary options to consider for continued progress within the semiconductor industry as current devices approach limits set by the inherent properties of silicon. III-V semiconductors in general have superior electrical and optical properties than silicon; thus, III-V/IV structures can potentially allow for better and new devices while still maintaining use of the silicon fabrication infrastructure to keep costs low.

We can consider the heterovalent structures discussed above as “integration level” structures; that is, the materials are within a single device structure but each material has its own function and for the most part exhibits its own properties. For example, GaAsP-SiGe structures have been widely investigated for the possibility of creating high efficiency dual junction solar cells on silicon substrates.14–16 In the structure, the GaAsP and SiGe layers each have a separate purpose (high gap cell and low gap cell, respectively). Fewer studies have been conducted on heterovalent nanostructures, superlattices, and alloys – structures in which the materials are combined and investigated as a single, new class of material. These “material level” structures from the III-V/IV system are the focus of this thesis.

17

There is limited knowledge of the material properties of III-V/IV superlattices; however, current material trends in the thermoelectric field suggest that such structures should be effective in those applications. Specifically, nanostructured materials have been shown to exhibit significantly lower thermal conductivities, which is desirable for thermoelectrics, due to increased interfacial phonon . (A discussion on thermoelectrics can be found later in this chapter.) Most investigations have largely been focused on structures and materials with isovalent interfaces; however, heterovalent interfaces should in theory provide even stronger interfacial phonon scattering. III-V/IV alloys are also potentially interesting structures to study for thermoelectrics; although nanostructured materials have been observed to exhibit better properties than the alloys of the same constituents, most thermoelectric devices in active use still consist of alloy materials, likely due to easier fabrication and more consistent results.

In addition to their potential as thermoelectric materials, III-V/IV alloys may be potential materials for low bandgap optoelectronic applications. The bandgaps of metastable (GaAs)1- x(Ge2)x alloys are one of the few properties that have been investigated previously; however, the studies used growth techniques that are not optimal for device fabrication.4,17–19 A later study using metalorganic chemical vapor deposition (MOCVD) was unable to produce structures resembling metastable single phase alloys9 and only detected weak photoluminescence signals at extremely low (5K) temperatures.20 If metastable single phase III-V/IV alloys can be obtained using a conventional epitaxy technique like MOCVD and exhibit low bandgaps, they can potentially allow for low bandgap devices while still maintaining a lattice constant close to or even matching that of GaAs.

18

Figure 1.1. Bandgap vs. lattice parameter diagram of II-VI, III-V, and group IV semiconductors. Lines denote ternary compositions (binary in case of SiGe) of the constituent endpoint semiconductors. Line colors represent the conduction band valley lowest in energy; materials with a Γ-valley have direct bandgaps, while materials with X- or L- valleys have indirect bandgaps. Dashed lines indicate estimated compositions based on a linear trend (no experimental data available). Image adapted from Woodall Research Group.21

1.2. Thesis Objectives and Organization

The objectives of this thesis are as follows:

1. Investigate the effects of various growth parameters on the morphology of GaAs/Ge

superlattices and (GaAs)1-x(Ge2)x alloys. The aim of this investigation is primarily to

identify suitable structures for the investigations detailed in the following objectives. The

results for this objective are presented in chapter 3.

2. Investigate the thermoelectric properties of GaAs/Ge superlattices and (GaAs)1-x(Ge2)x

alloys that were identified in objective 1. In particular, we aim to understand the effect of

heterovalent interfaces on those properties. We also investigate the thermoelectric

19

properties of (In1-yGayAs)1-x(Si1-zGez)x alloy structures. This structure is a more

representative sample of what the III-V/IV material system may offer to the

thermoelectric field. The results for this objective are presented in chapter 4.

3. Investigate the optical properties of suitable (GaAs)1-x(Ge2)x alloys identified in objective

1, as well as the optical properties of several (In1-yGayAs)1-x(Si1-zGez)x alloys, including

the structures from objective 2. These results are presented in chapter 5. A large focus of

the chapter will be to understand the origins of the photoluminescence signals obtained

from the structures.

The remainder of this chapter reviews the literature and background relevant to this thesis.

Chapter 2 discusses the methods used in the investigations. Chapter 6 summarizes the results of this thesis and discusses possible directions for future work.

1.3. III-V/IV epitaxy

Epitaxy involving both a III-V material and a group IV material will necessitate the formation of

III-IV and IV-V bonds, which are generally thermodynamically unfavorable in comparison to

III-V and IV-IV bonds. Additionally, III-V materials are polar while group IV materials are nonpolar. Thus, if the growth process is thermodynamically limited (i.e. growth kinetics are sufficient as to not result in metastability), processes such as phase separation or Volmer-Weber growth (island growth mode, preferred mode when adatom- is higher than adatom-adatom energy) can be expected to play a large role in the formation of the final morphologies. This section will review the literature on III-V/IV structures and the effects of

20 controlling the growth kinetics on the morphologies, with a focus on the lattice matched

GaAs/Ge system.

Figure 1.2. GaAs-Ge phase diagram. Mutual immiscibility is observed in the material system. Phase separation would be expected at all compositions absent kinetic limitations. Adapted from Ansara22

21

Figure 1.3. Primary modes of thin film growth a) Volmer-Weber (island growth) b) Frank-van der Merwe (layer-by-layer) c) Stranski-Krastanov (layer-by-layer up to a critical thickness, followed by island growth).23 The growth mode of a material depends on the interaction energies of the adatom and surface.

1.3.1. Polar on non-polar epitaxy

Elemental group IV in the bulk have four bonds, bonding in a tetrahedral structure. In the absence of surface reconstruction, surface atoms would have two dangling bonds, resulting in a high surface free energy. While surface reconstruction reduces the number of dangling bonds, it is still sufficiently high in energy that it is energetically favorable for a deposited III-V material to initially form a monolayer rather than a 3-dimensional cluster. In vapor phase epitaxy of

GaAs on Ge where the surface is allowed to equilibrate, a monolayer of As forms, resulting in a very stable As-terminated Ge surface. Subsequent deposition on this low energy As-terminated surface favors 3-dimensional clustering and island growth over layer growth.24 By controlling the kinetics, however, this process can be altered to promote more “2D-like” growth. In this case, control involves allowing sufficient kinetics for adatoms to find local minimum energy states but not global minimums. Typically, this control is done by changing temperature; at lower

22 temperatures, the adatom surface diffusion is decreased, increasing the likelihood that adatoms bond to the surface before it can migrate to an adatom cluster.

1.3.1.1. Antiphase Domains

An additional complication in the epitaxy of III-V materials on a group IV surface is the possible formation of antiphase domains. The zincblende of III-V materials is composed of two interpenetrating face-centered cubic (fcc) lattices arranged such that nearest neighbor atoms are in a tetrahedral formation, with one fcc sublattice comprised of the group III element and the other comprised of the group V element. If the two elements exchange position, the resulting crystal is 90° rotated with respect to the original arrangement. This particular distinction in crystallographic orientation is absent in the diamond cubic group IV material since the two fcc sublattices have the same element.

The surface of (001) Si and Ge substrates typically exhibit single atomic layer steps throughout

(Fig. 1.4).25 During deposition of GaAs or other III-V material, this step can result in a plane of

III-III and V-V bonds that extends up towards the growing surface. This plane of “wrong” bonds is known as an antiphase boundary (APB), separating domains that are rotated 90° to each other.

The local area near APBs will have a different electronic structure compared to the GaAs crystal; thus, APBs can be expected to negatively affect electrical transport.26 Suppression of APBs is generally done by using (001) substrates offcut by a few degrees; studies have shown that by annealing an offcut substrate, the single atomic layer steps on the surface reconstruct into a double step formation (Fig. 1.5).25,27,28 With a double step surface, the domains on either side of the step have the same orientation, resulting in an APB-free structure.

23

Figure 1.4. Diagram of a GaAs lattice on a (001) group IV substrate with a single atomic step, resulting in the formation of an antiphase boundary (vertical dashed line).25

Figure 1.5. Diagram of a GaAs lattice on a (001) group IV substrate with a double atomic step. No antiphase boundary forms since the lattice on either side of the step has the same orientation.25

1.3.2. Non-polar on polar epitaxy

The surface of III-V materials is more complex than that of Si or Ge due to the presence of two different elements. Depending on the stoichiometry (ratio of group V element to group III) of the surface, different surface reconstructions can occur that can impact the growth mechanisms of a group IV overlayer.29,30 In GaAs epitaxy, a high surface As-to-Ga ratio results in the 푐(4 ×

24

4) surface reconstruction, with several As layers on the surface. Bai et al. found that Ge growth on this surface resulted in a highly pitted layer, with no Ge growth at the pits. In situ studies of

Ge deposition on GaAs by molecular beam epitaxy (MBE) indicated that the epitaxy began with the formation of a low energy Ga-Ge dimerized surface.31 This suggests that deposition of Ge on an As-rich surface may involve diffusion of Ge through the As layers to bond to Ga; the excess

As then form clusters that locally prevent Ge growth. By controlling the growth conditions to promote a surface with a higher ratio of Ga to As (e.g. 푐(8 × 2) reconstruction), Ge growth was shown to be free of the pits.30 The strong tendency to form the stable Ga-Ge dimerized surface allows proper Ge initiation on GaAs; however, it also plays a role in the thermodynamic unfavorability of continued 2D growth,31 akin to that seen for GaAs growth on the As-passivated

Ge surface in the previous section.

Figure 1.6. Model of the GaAs 풄(ퟒ × ퟒ) surface reconstruction.29

25

Figure 1.7. Model of the GaAs 풄(ퟖ × ퟐ) surface reconstruction.29

1.3.3. Metastable (GaAs)1-x(Ge2)x alloys

The growth of metastable III-V/IV alloys by ion-beam was investigated by several groups in the 1980s.4,17–19 An important factor in achieving the metastable phases of the material systems is the use of low-energy ion bombardment of the growing film, which alters the adatom incorporation and diffusion. More importantly, it was found that ion bombardment preferentially sputters precipitates forming near the surface, thus inhibiting the formation of a second phase.

Bombardment energies were sufficiently low as to only affect the surface; the underlying bulk material was unaffected. Phase separation in the bulk is limited due to the low bulk self diffusion. Thus, by inhibiting phase separation at the surface, a single metastable phase was achieved.19

26

Investigations in the growth of metastable alloys using other techniques have been limited and often unsuccessful. Banerjee et al. observed phase separation of (GaAs)1-x(Ge2)x alloys grown by molecular beam epitaxy at 550°C.32 In a later study at lower temperatures, a single metastable phase appeared to be achieved; the authors suggested that low surface mobilities of the atoms as a result of the lower temperature may have forced single phase growth.7 Norman et al. investigated the growth of (GaAs)1-x(Ge2)x layers by metalorganic chemical vapor deposition; however, they observed pronounced phase separation associated with surface faceting.9

Figure 1.8. (002) dark-field cross-section transmission electron microscopy (XTEM) image of (GaAs)1-x(Ge2)x layers grown by MOCVD, showing pronounced phase separation. Dark regions are Ge-rich.9

1.4. Overview of Applications

1.4.1. Thermoelectrics

Thermoelectric energy conversion technology is a solid-state technique that can directly convert a heat flux into useful electrical power or vice versa. It has attracted significant research interest for various applications due to its mechanical stability, reliability, and scalability. Currently, thermoelectrics are limited by their low efficiency; the technology is primarily used in niche

27 applications where their advantages outweigh the importance of energy efficiency, such as deep space exploration.

1.4.1.1. Thermoelectric figure of merit

The efficiency of a thermoelectric material is governed by the dimensionless thermoelectric

휎푆2 figure of merit, 푧푇 = 푇, where 휎 is the electrical conductivity, 푆 is the Seebeck coefficient, 휅푒+휅푙

휅푒 is the electronic thermal conductivity, 휅푙 is the lattice thermal conductivity, and T is the material’s average operating temperature. Generally, a larger figure of merit yields higher energy conversion efficiency. Desirable thermoelectric materials would therefore have high electrical conductivities and Seebeck coefficients but low thermal conductivities. One of the difficulties in realizing such a material is the interrelation between these properties. High

Seebeck coefficients are often seen in materials with low carrier concentration and thus, low electrical conductivity. In addition, materials with high electrical conductivities, such as metals, also have high thermal conductivities. Furthermore, these properties change with temperature, resulting in a limited temperature range in which a thermoelectric material would be ideal.

28

Figure 1.9. Conflicting nature of the thermoelectric properties. Axis scale: Thermal conductivity, κ, 0 to 10 W m-1 K-1. -1 -1 -1 33 Seebeck coefficient, α, 0 to 500 µV K . Electrical conductivity, σ, 0 to 5000 Ω cm . Based on Bi2Te3 data.

Figure 1.10. Dependence of the thermoelectric figure of merit, zT, of various materials as a function of temperature.33

1.4.1.2. Nanostructuring

Nanostructured materials, such as thin films, , and superlattices, first gained notice in the thermoelectric field in the 1990s, when theoretical studies suggested significant improvement in the Seebeck coefficient through quantum confinement of carriers.34,35 Although experimental studies did not demonstrate this improvement in the Seebeck coefficient, most did observe improvement in the thermoelectric figure of merit due to a decrease in the observed thermal

29 conductivity. This is understood to be from increased disruption to phonon transport at material interfaces and grain boundaries that significantly shortens the effective phonon mean free paths,

35–44 leading to a significant reduction in the lattice thermal conductivity 휅푙.

1.4.2. Low bandgap materials lattice matched to GaAs

There is significant interest in finding low bandgap materials close to the GaAs lattice constant for photovoltaic and photonic applications. Many devices are built upon a GaAs or Ge substrate; thus, a low bandgap material with a similar lattice constant would avoid issues associated with lattice mismatch. A bandgap of 1 eV, in particular, is highly desired for multi-junction photovoltaic cells as it would be significantly more efficient in collecting light with energies in the 1.0-1.4 eV range than the 0.67 eV bandgap Ge. Most research on 1 eV bandgap materials have focused on complex quaternary and quinternary III-V materials or attempt to work with

45–47 lattice mismatched In0.3Ga0.7As and employ a graded composition buffer layer. Smaller direct bandgaps on GaAs are interesting for mid- and long-wave optical detection applications.

Typically, InAs- and InSb-based materials are used to create these small bandgaps, but the large lattice mismatch to GaAs forces these narrow bandgap materials to be formed on expensive, small substrates like InP, InAs, and GaSb. Small direct bandgaps in materials with a smaller lattice constant would allow these more advanced IR detectors to be ported to large, inexpensive substrates like GaAs or Ge; various groups have investigated strained Ge for this purpose.48,49

Metastable (GaAs)1-x(Ge2)x alloys have been observed to exhibit direct bandgaps ranging from

0.6 eV to 1.4 eV depending on the composition (Fig. 1.11).5,17 These alloys are latticed matched to GaAs, making them potentially ideal for the applications mentioned above. A noteworthy

30 feature of the system is the large, negative asymmetrical V-shaped bowing of the bandgap as a function of composition. This has been attributed to an order-disorder transition at approximately 30% Ge composition. Alloys with lower Ge compositions have the zincblende crystal structure of GaAs, with the Ge atoms randomly substituting on both Ga and As lattice sites. Alloys with higher Ge compositions have the diamond cubic structure of Ge, and Ga and

As may sit on both of the fcc sublattices.

8 Figure 1.11. Direct bandgap of (GaAs)1-x(Ge2)x with composition. Image adapted from Koiller et al.

31

Chapter 2: Materials Growth and Characterization

32

2.1. Film Epitaxy

Metalorganic chemical vapor deposition (MOCVD) is a widely used technique for thin film semiconductor synthesis. A schematic of an MOCVD system is shown in Fig. 2.1. H2 carrier gas brings alkyl and precursors to the substrate in the reaction chamber. The substrate is heated to a high temperature to enable pyrolysis of the precursors and deposition of the required materials.

Figure 2.1. Schematic of an MOCVD system.50

In this work, epitaxy was conducted using a Thomas Swan/Aixtron close-coupled showerhead

MOCVD reactor custom built to allow deposition of both III-V semiconductor films (e.g. GaAs,

InGaAs) and group IV semiconductor films (Si, Ge). Nitrogen was used as the carrier gas for all experiments. The group III precursors used were trimethylgallium (TMGa), trimethylaluminum

33

(TMAl), and trimethylindium (TMIn). The hydride precursors include arsine (AsH3) for the group V material, as well as germane (GeH4) and silane (SiH4) for the group IV materials.

2.2. Structural Characterization

2.2.1. Transmission Electron Microscopy (TEM)

Transmission electron microscopy (TEM) is a powerful technique used to characterize the microstructure of materials. A beam of high energy electrons is transmitted through a prepared, thin sample; the interaction of the electrons with the atoms in the sample as it passes through forms an image, which is magnified and focused onto a fluorescent screen or CCD camera. The contrast observed in the image is the result of several interaction mechanisms including absorption, inelastic scattering, and elastic scattering.

In crystalline materials, elastic scattering of the electrons results in discrete Bragg reflections, collectively a diffraction pattern (see example in Fig. 2.2), at the back focal plane of the objective lens. By using an objective aperture to include only one particular Bragg a darkfield image is obtained (Fig. 2.3b) containing diffraction contrast pertaining to that reflection.

A complementary brightfield image can be obtained by tilting the sample to maximize the intensity of only that Bragg reflection while minimizing all other reflections (this is known as the two-beam condition, in which only the incident beam and one diffracted beam are strong), and subsequently passing only the incident beam through the objective aperture (Fig. 2.3a). The two- beam condition can also be used in darkfield imaging to increase the observed contrast.

Diffraction contrast from the 002 reflection is particularly important for samples containing primarily GaAs and Ge. GaAs and Ge have very similar densities, lattice constants, and atomic

34 numbers, resulting in very low mass and strain contrast. Contrast between the two materials is primarily obtained using the 002 reflection; this reflection is forbidden in the diamond cubic Ge but allowed in zincblende GaAs.

Figure 2.2. Diffraction pattern of a GaAs crystal in the [110] zone axis.

Figure 2.3. Diagram of a) brightfield imaging and b) off-axis darkfield imaging in TEM.51

35

In this work, sample cross-sections were imaged using a JOEL 2011 transmission electron microscope with a LaB6 filament operated at 200 kV. A double-tilt holder was used to allow samples to be tilted to the proper diffraction orientations. Samples were prepared as follows: two pieces (~ 3mm × 5mm) are cleaved from the and glued together with the film side facing each other. Silicon cladding pieces are glued on the backs of the samples. The “sandwich” structure is mechanically polished to ~10μm using a series SiC sandpapers with increasingly finer grits and a diamond polish finish. The polished sample is attached to a copper grid and ion milled in a Fischione 1010 ion mill. The milling conditions used are accelerating voltage of 3.5 kV, beam current of 4 mA, and a milling angle of 15° for the first step to form a hole in the middle of the sample; a second step follows with accelerating voltage of 2 kV, beam current of 3 mA, and a milling angle of 13° to enlarge the hole. Images of the sample are taken from the electron transparent regions of the film near the hole.

2.2.2. Electron-Dispersive X-ray Spectroscopy (EDS)

Electron-dispersive x-ray spectroscopy (EDS) is a technique used to analyze the compositions of samples. A high energy beam of electrons or x-rays is focused onto a sample, exciting inner shell electrons within the atoms of the sample from their ground state. Relaxation of an electron in a higher energy shell to the lower energy shell generates an x-ray with energy equal to the energy level difference. These energy level differences are characteristic to each element; thus the x-rays can be analyzed to determine the composition. In this work, film compositions were estimated using a JOEL 2010 field emission transmission electron microscope equipped with an

INCA EDS system. Samples were prepared in the same manner as discussed in Section 2.2.1.

36

2.3. Optical Characterization

2.3.1. Photoluminescence (PL)

Photoluminescence is a technique that can be used to measure the bandgap of a high quality semiconducting material. A excitation source with photon energy higher than the bandgap of the material will generate electron-hole pairs upon absorption of the photons. The electrons and holes then undergo relaxation to the conduction band minimum and valence band maximum, respectively. In the absence of trap states within the bandgap (these may be present when there are defects in the material), the electrons and holes at the band edges recombine, emitting photons with energy equivalent to the bandgap.

In this work, room temperature PL was conducted using a 514.5nm emission continuous-wave argon ion laser operating at 150mW output power. The beam spot size is approximately 250µm in diameter. Luminescence from the samples is passed through a monochromator capable of

1nm resolution and into an appropriate detector. Detectors include an InGaAs detector with good responsitivity between 1000-1650nm and a strained InGaAs detector with good responsitivity between 1500-2550nm. The setup also includes a chopper and lock-in amplifier to improve the signal-to-noise ratio.

Low temperature (77K) PL was conducted using micro-photoluminescence setup with a cryostat compartment. The excitation source was a 1060nm emission continuous-wave diode laser with

90mW maximum output power. A 1064nm longpass dichroic filter was used to separate the excitation beam from the sample luminescence. The detector used was an IR spectrometer with

37 an InGaAs photodiode; good responsivity was obtained between 1300-2050nm. A chopper and lock-in amplifier was used to improve the signal-to-noise ratio.

2.3.2. Absorption Spectroscopy

Absorption spectroscopy is a technique that can be used to determine the absorption of a sample as a function of wavelength. In the measurement, a broad spectrum radiation source is directed through a monochromator towards the sample, and the intensity of the transmitted light is detected. The setup is then configured to measure the reflection; the absorption can be extracted from the measured transmission and reflection spectra. In this work, absorption measurements were conducted using a PerkinElmer Lambda 1050 UV/VIS/NIR spectrophotometer fitted with a

150mm integrating sphere accessory for the reflectance measurements. The operating wavelength range is 175nm – 3300nm.

2.4. Time-Domain Thermoreflectance (TDTR)

Time-domain thermoreflectance (TDTR) is a technique used to measure the thermal properties of materials. The technique uses a pump-probe setup, in which a pump laser heats a sample and induces a localized . This stress causes a change in the reflectance, which is measured by a time-delayed probe laser. The thermal conductivity of the sample can then be extracted by fitting the measured temporal reflectance signal with the standard multi-layer heat diffusion model.52 Further discussion of the technique can be found in the relevant references.52,53

38

In this work, thermal conductivities were measured using an ultrafast TDTR setup at room temperature. A thin (~100 nm) metal layer was coated onto the sample surface using electron beam deposition to act as an optical-thermal transducer in the measurement. The used pump and probe spot sizes in the measurement were 45 m and 11 m, respectively. The pump beam was modulated by an electro-optic modulator and a range of modulation frequencies from 3 MHz to

12 MHz were used to characterize the thermal conductivity. Note that since the pump laser diameter was much larger than the thermal penetration depth (~ 550 nm – 1.15 um depending on the used modulation frequency), the measurement was insensitive to the in-plane thermal transport. The thickness of the metal film was accurately determined through TDTR measurement on a dummy sapphire substrate with known thermal conductivity that was co- placed in the deposition chamber with other GaAs/Ge superlattice samples.54

2.5. Electrical Characterization

2.5.1. Van der Pauw – Hall method

The Van der Pauw – Hall technique can be used to determine the carrier concentration and carrier mobility of semiconductor thin films with known thickness.55 The technique measures the sheet resistance and Hall voltage, which can then be used to extract the aforementioned properties.

Four ohmic contacts are placed along the edge of the sample and the resistances are determined for all adjacent contact combinations (assume contacts 1, 2, 3, and 4 in clockwise formation): a current 퐼12 is applied between contacts 1 and 2, and a voltage 푉34 is measured between contacts 3

푉34 and 4, giving a resistance, 푅12,34 = . 푅23,41 is then determined by applying current 퐼23 퐼12

39 between contacts 2 and 3, and measuring voltage 푉41 between contacts 4 and 1. The two reciprocal measurements, giving 푅34,12 and 푅41,23, as well as the four reversed polarity measurements, giving 푅21,43, 푅32,14, 푅43,21, and 푅14,32, are also done to reduce the measurement error. The sheet resistance 푅푠 is related to these resistances by the Van der Pauw equation:

푒−휋푅푣⁄푅푠 + 푒−휋푅ℎ⁄푅푠 = 1, (2.1) where

푅 + 푅 + 푅 + 푅 푅 + 푅 + 푅 + 푅 푅 = 12,34 34,12 21,43 43,21 and 푅 = 23,41 41,23 32,14 14,32. 푣 4 ℎ 4

To determine the Hall voltage, a magnetic field is applied perpendicular to the sample. A current

퐼 is then applied between the cross contacts (i.e. between contacts 1 and 3 or 2 and 4) while the voltage is measured along the remaining contacts. These measurements are done for all cross contacts combinations at the same current and then repeated with the magnetic field reversed.

The overall Hall voltage 푉퐻 is given by:

푉 − 푉 + 푉 − 푉 + 푉 − 푉 + 푉 − 푉 푉 = 13,+푀 13,−푀 24,+푀 24,−푀 31,+푀 31,−푀 42,+푀 42,−푀 , (2.2) 퐻 8 where +푀 and – 푀 refer to the measurements done at positive and negative magnetic field, respectively. The sheet density 푛푠 can be calculated based on the overall Hall voltage 푉퐻, the applied current 퐼, and the magnetic field strength 퐵:

퐼퐵 푛푠 = . (2.3) 푞|푉퐻|

Dividing the sheet density by the film thickness gives the carrier concentration. The Hall mobility, µ, is determined from the sheet density and sheet resistance discussed above:

40

1 µ = . (2.4) 푞푛푠푅푠

In this work, samples are cleaved into roughly 1cm × 1cm squares; indium contacts are placed at the four corners. Van der Pauw measurements are done as discussed above but at various applied currents to obtain eight I-V curves. A suitable sample will have two sets of four curves, with the curves in each set overlapping or nearly overlapping each other. In addition, the I-V curves should exhibit linear behavior. Only samples satisfying these conditions can be further measured for carrier concentration and mobility.

2.5.2. Seebeck Measurements

Seebeck measurements were conducted using a home-built setup consisting of two thermoelectric modules connected in series such that applying a current will result in heating from the top face of one module and cooling from the top face of the second module. Samples are placed on top of the modules in a “bridge” formation. During operation a current is applied through the modues, and the temperature on each end of the sample as well as the induced voltage are measured. The Seebeck coefficient is determined by dividing the induced voltage by the temperature difference.

41

Chapter 3: Epitaxy of the GaAs/Ge System

42

3.1. Introduction

In this chapter, we investigate the growth and morphology of GaAs/Ge superlattices and alloys grown by MOCVD under a variety of conditions. We focus on this particular III-V/IV system as it has very little lattice mismatch. Investigation of this system will give insight to growth involving III-V/IV heterovalency without the additional complexity arising from lattice mismatched materials.

We would like to note here that the growth of GaAs/Ge superlattice structures has never been attempted using MOCVD previously. For this reason, we have surveyed several structures under various combinations of growth temperature, reactor pressure, substrate orientation, and surface/interface preparation, with the aim of acquiring the conditions that establish sufficient morphological control for the investigations detailed in chapter 4. We emphasize that this is not, and not intended to be, an exhaustive study on the effects of each growth parameter on morphology. As a consequence, the structures observed will often not exhibit high quality superlattice morphologies like those typically seen in GaAs/AlAs superlattices.56 Nevertheless, we will make use of “superlattice” and its associated terminology throughout this chapter for ease of reading and, more importantly, to distinguish between structures with different “periods” or “layer thicknesses” that would be expected from the growth process steps and times.

MOCVD growth of (GaAs)1-x(Ge2)x alloy structures has been attempted previously, as noted in section 1.3.3. Our initial attempts at growing these alloy structures, however, have resulted in morphologies that are vastly different from those observed in the previous study.9 For this reason, we briefly survey several different morphologies of (GaAs)1-x(Ge2)x alloys obtained

43 using various combinations of growth conditions. This also is not intended to be an extensive study on each individual growth parameter’s influence on morphology.

3.2. Growth of GaAs/Ge superlattices

GaAs/Ge superlattices were grown using various combinations of growth temperature, reactor pressure, and substrate orientation. An additional parameter investigated involved modifying the

Ga-to-As ratio on the surface following each deposited GaAs layer. As discussed in section

1.3.2., high quality Ge on GaAs was obtained only when the GaAs surface was Ga-rich. Bai et al. investigated two methods for preparing a Ga-rich surface prior to Ge deposition. The first method involved a short TMGa “pulse” immediately after deposition of GaAs. This pulse of only TMGa will react with surface As and result in a surface construction with a higher Ga-to-As ratio. The second method involved an anneal in N2 ambient; at elevated temperatures without an

AsH3 overpressure As atoms will desorb from the surface, resulting in a higher surface Ga-to-As ratio.30 In this work, TMGa pulses were conducted as follows: upon completion of the GaAs layer, AsH3 flow was shut off for 10 seconds to remove precursors in the ambient (i.e. only N2 is flowing). After 10 seconds, TMGa was flown into the reactor chamber for several seconds.

After TMGa flow was stopped an additional 10 seconds were allotted to purge the ambient, followed by Ge deposition. Samples grown without the TMGa pulse were simply annealed for

20 seconds in N2 ambient prior to Ge deposition. A summary of the conditions used for each sample, as well as the observed superlattice period and growth rates, are shown in Table 3.1. For the deposition of GaAs, layers were grown at a V/III ratio above 100. In-situ monitoring of the

44 reflectivity of the growing surface was done using EpiTT at a 950 nm wavelength. The information obtained can be used to observe growth transitions and incubation times.

Sample Observed Substrate Growth Reactor TMGa pulse Growth Rate Superlattice Orientation Temperature Pressure (Å/s) Period (°C) (Torr) 1 55nm exact (001) 650 100 8 sec, 13 μmol/min GaAs: 1.0-2.3 GaAs/40nm Ge Ge: 0-4.1 2 55nm (001) 6° off 650 100 8 sec, 13 μmol/min GaAs: 0-2.2 GaAs/40nm Ge toward <111>A Ge: 0-4.5

3 50nm exact (001) 500 100 15 sec, 34 GaAs: 0-3.6 GaAs/10nm Ge μmol/min 4 50nm (001) 6° off 500 100 15 sec, 34 GaAs: 0-3.1 GaAs/10nm Ge toward <111>A μmol/min

5 60nm exact (001) 650 250 No pulse, 20 sec N2 GaAs: 1.0-1.4 GaAs/60nm Ge anneal Ge: 10.8-14.4 6 60nm (001) 6° off 650 250 No pulse, 20 sec N2 GaAs: 1.1-1.4 GaAs/60nm Ge toward <111>A anneal Ge: 10.4-14.0

7 30nm exact (001) 650 250 5 sec, 13 μmol/min GaAs: 0.5-2.4 GaAs/70nm Ge Ge: 1.4-18.0 8 40nm (001) 6° off 650 250 5 sec, 13 μmol/min GaAs: 0.4-2.0 GaAs/60nm Ge toward <111>A Ge: 2.0-14.0

9 30nm exact (001) 650 250 No pulse, 20 sec N2 GaAs: 1.0-1.6 GaAs/30nm Ge anneal Ge: 10.0-14.0 10 30nm (001) 6° off 650 250 No pulse, 20 sec N2 GaAs: 0-2.6 GaAs/30nm Ge toward <111>A anneal Ge: 10.0-12.8

11 20nm exact (001) 650 250 5 sec, 13 μmol/min GaAs: 0-1.9 GaAs/25nm Ge Ge: 0-12.4 12 20nm (001) 6° off 650 250 5 sec, 13 μmol/min GaAs: 0-1.7 GaAs/25nm Ge toward <111>A Ge: 0-12.8

13 55nm exact (001) 500 250 5 sec, 34 μmol/min GaAs: 1.0-3.7 GaAs/20nm Ge 14 55nm (001) 6° off 500 250 5 sec, 34 μmol/min GaAs: 1.6-3.2 GaAs/20nm Ge toward <111>A Table 3.1. Summary of grown structures and epitaxy conditions. Thicknesses of the GaAs and Ge layers were determined using TEM. For samples with non-uniform layers, approximate average thicknesses are given. For samples 3, 4, 13, and 14, thicknesses given do not reflect the thickness of the first Ge layer deposited; Ge growth rates are not determined due to difficulty in separating out incubation time.

45

3.2.1. 650°C and 100 Torr growth (samples 1 and 2)

650°C substrate temperatures and a reactor pressure of 100 Torr are the conditions in which the epitaxy of many semiconductors, including GaAs, falls within the mass-transport-limited regime.

In this regime, growth is limited by the rate of precursors diffusing and reaching the substrate surface. MOCVD growth is done in this regime in most cases to obtain high quality films. We attempted to grow GaAs/Ge superlattices at these conditions on both an exact (001) GaAs substrate (sample 1) and a (001) GaAs substrate offcut 6° towards the <111>A direction (sample

2). An 8 second TMGa pulse was chosen for preparation of a high surface Ga-to-As ratio.

Cross-section TEM (XTEM) images of the samples are shown in Fig. 3.1. The presence of antiphase boundaries (APBs) in both samples was confirmed from observing contrast reversal between (002) darkfield and (00-2) darkfield images.

Figure 3.1. (002) Darkfield XTEM image of GaAs/Ge superlattices grown at 650°C and 100 Torr with substrate oriented in (a) the exact (001) direction (sample 1) and (b) (001) 6° offcut towards the <111>A direction (sample 2). Antiphase boundaries (APBs) in the GaAs layers were confirmed through observing contrast reversal in adjacent domains in (002) darkfield and (00-2) darkfield images (not shown).

46

The presence of APBs in the sample using the offcut substrate despite the expectation that their formation be inhibited is likely explained by the loss of planarity of the Ge layers. As discussed in section 1.3, APB-free GaAs films are produced as a result of a double atomic step structure that maintains registry of the domains on either side of the steps. This double step structure is obtained by annealing the surface of offcut (001) Ge; however, the fact that the anneal process does not work for exact (001) Ge suggests that a regular array of steps is necessary for the double step structure to be thermodynamically favorable throughout the surface. As the surfaces of the

Ge layers seen in Fig. 3.1b are non-planar, it is reasonable to assume that there are local areas in which double step structure formation was not favorable.

3.2.2. 500°C and 100 Torr growth (samples 3 and 4)

As discussed in section 1.3, layer-by-layer growth is not thermodynamically favorable for either the deposition of GaAs on Ge or the deposition of Ge on GaAs. Since the samples in the previous section were grown in the mass-transport-limited regime, adatoms on the surface should diffuse to their thermodynamically stable locations (i.e. island formation). Thus, the morphologies observed in Fig. 3.1 are not unexpected.

Samples 3 and 4 were grown at the same reactor pressure (100 Torr), but growth temperatures were reduced from 650°C to 500°C. Images of the samples are shown in Fig. 3.2. A comparison between these samples to the previous samples grown at 650°C reveals the large effect of temperature on the length scale of the Ge interface morphology; this is most clearly observed in the first Ge layer. For the samples grown at 500°C the top interface of the first Ge layer is

47 reasonably planar as expected since the reduced Ge surface diffusion suppresses the ability for adatoms to find and attach to Ge clusters during the deposition process.

The growth of GaAs on the Ge layers at 500°C is highly non-uniform, however; there appears to be GaAs islands in various degrees of coalescence. Experiments by Larsen et al. suggest that at low temperatures the reaction to form GaAs occurs only with TMGa and AsH3 molecules that have been adsorbed onto the surface.57 It is possible that at 500°C the reduced surface diffusion lowered the probability of TMGa and AsH3 meeting, resulting in slower nucleation of GaAs. On the other hand, adatom attachment to any GaAs nuclei that do form is significantly more favorable thermodynamically; further growth of the nuclei is less likely to be limited by the need to have both the TMGa and AsH3 molecule present at the same time. The end result is that island growth still occurs, and thicker GaAs layers would be necessary to achieve full coalescence and film planarity. Attempts to grow thin GaAs layers would result in the incomplete states of coalescence as observed in Fig. 3.2.

48

Figure 3.2. XTEM images of GaAs/Ge superlattices grown at 500°C and 100 Torr. (a) (002) darkfield and (c) (00-2) darkfield images of sample 3, grown using a substrate oriented in the exact (001) direction. (b) (002) darkfield and (d) (00-2) darkfield images of sample 4, grown using a substrate oriented (001) 6° offcut towards the <111>A direction. APBs were determined from observing contrast reversal in adjacent domains in the (002) darkfield and (00-2) darkfield images. Stacking faults (SF) are observed in sample 3, and are also present in sample 4 (not shown).

EpiTT in-situ reflectivity monitoring indicated an approximately 100 second incubation time during the deposition of the first Ge layer in the 500°C growths. This was not observed in the samples grown at 650°C. Substrate orientation also affected the growth rate of the first Ge layer; the growth rate of Ge after incubation was 0.5 nm/s when using the exact (001) substrate but only 0.1 nm/s when using the (001) 6° offcut substrate. The GeH4 flow rate for both samples was the same at 629 μmol/min. Furthermore, subsequent deposited Ge layers had even lower

49 growth rates and/or longer incubation times. The incubation times for those layers were difficult to determine from the reflectivity results due to the slow growth rate and non-uniformity effects from the growing surface. These results suggest that in addition to a reduction in the Ge surface diffusion, the lower growth temperature also significantly reduced the rate of adatom adsorption.

This reduction in adatom adsorption is affected by the surface structure; the more irregular the surface the lower the adsorption rate.

3.2.3. 650°C and 250 Torr growth (samples 5 – 12)

In the mass-transport-limited regime, the growth rate of GaAs can be estimated from the boundary layer model by the equation:58,59

푣 1⁄2 푟 = (푐표푛푠푡푎푛푡)푝 ( ) , (3.1) 𝑔 푇푀퐺푎 푃

where 푝푇푀퐺푎 is the partial pressure of TMGa, 푣 is the gas velocity, and 푃 is the overall system pressure. It is assumed in this case that the TMGa is completely depleted at the growing surface.

Lower growth rates are observed in instances where increased deposition occurs at the reactor walls since the effective TMGa near the surface is lower than the input partial pressure.58

The discrepancy between the input partial pressure and the effective pressure at the surface likely contributes to the significant failure of equation 3.1 in predicting the growth rate of Ge; we observed an increase in the Ge growth rate when the total system pressure was increased from

100 Torr to 250 Torr, similar to observations by Ayers and Ghandhi.60 A detailed study of mechanisms involved is beyond the scope of this thesis; the empirical observation of increased

Ge growth rate with system pressure is of primary interest.

50

Figure 3.3. Dependence of Ge growth rate on total system pressure. Growth was done at 750°C on GaAs substrates with orientations (001) 3° offcut towards <110>, (110), and As terminated (111).60

Images of the samples grown at 650°C and 250 Torr using the N2 anneal surface preparation method (samples 5 and 6) are shown in Fig. 3.4. Deposition of the first Ge layer on the exact

(001) GaAs substrate resulted in a planar Ge film (Fig. 3.4a). We suspect that the observed increase in Ge growth rate reduced the time surface Ge adatoms have to diffuse before becoming incorporated into the bulk, thereby increasing the likelihood of nucleation rather than growth of clusters. Subsequent deposited layers are observed to be less planar, particularly near the APBs formed in the GaAs layers. This suggests that APBs must be suppressed in order to achieve highly planar GaAs/Ge superlattice layers.

The morphology shown in the sample grown using the offcut substrate (Fig. 3.4b) indicates that the method of using annealed offcut substrates alone is not sufficient to suppress APBs. There is

51 a noticeable wave-like morphology throughout the film starting from the GaAs substrate itself.

Step-bunching is one possible explanation for the wave-like surface of the substrate; many groups have observed step-bunching of grown GaAs on offcut substrates under various conditions, although there is not a clear consensus on some details of the underlying mechanisms.61–63 Another possible explanation would be the deposition of Ge inducing a wave- like surface through an exchange or intermixing mechanism. This is less likely however, as such a mechanism would reasonably be expected to occur in the sample shown in Fig. 3.1b as well, with more pronounced undulations due to the slower Ge growth rate. Irrespective of the origin of this wave-like characteristic, it likely affects the thermodynamic favorability of the double step surface formation much like in the sample grown at low pressure (sample 2, Fig. 3.1b).

Figure 3.4. (002) Darkfield XTEM images of GaAs/Ge superlattices grown at 650°C and 250 Torr using the N2 anneal surface preparation method. (a) Sample 5, grown on a substrate oriented in the exact (001) direction. (b) Sample 6, grown on a substrate oriented (001) 6° off towards the <111>A direction. APBs in the GaAs layers were confirmed through observing contrast reversal in adjacent domains in (002) darkfield and (00-2) darkfield images (not shown).

Images of samples grown at 650°C and 250 Torr using the TMGa pulse surface preparation method (samples 7 and 8) are shown in Fig. 3.5. Surprisingly, the use of the TMGa pulse results in Ge layers that are significantly less planar; Ge growth appears to be suppressed or significantly reduced in localized regions. Experiments by Pristovsek et al. on As desorption

52 from GaAs surfaces indicate that the Ga-rich β2(2×4) surface reconstruction is obtained by

64 annealing GaAs at 650°C for 27 seconds in either H2 or N2 ambient. The morphologies observed in Fig. 3.4a suggest that 20 seconds of annealing is sufficient to drive off enough As from the surface to enable planar Ge film growth. The samples grown using a TMGa pulse had a

10 second anneal in N2 ambient prior to and after the pulse. Since a 20 second anneal was sufficient to obtain a uniform Ge layer, the coverage of Ga atoms provided from the TMGa pulse was likely excessive. This result contrasts with those found by Bai et al. and is likely related to the difference in growth temperature. They obtained smooth Ge on GaAs using a TMGa pulse when growing at temperatures below 500°C, where TMGa partially decomposes to form methylgallium complexes 30. At 650°C, however, TMGa is expected to completely decompose and deposit Ga onto the surface. Unlike methylgallium complexes, which form self-limiting monolayers on the GaAs surface, excess Ga atoms are not self-limiting and can cluster. It is likely that in the samples shown in Fig. 3.5, this excess Ga locally inhibited or slowed Ge growth.

Figure 3.5. (002) Darkfield XTEM images of GaAs/Ge superlattices grown at 650°C and 250 Torr using the TMGa pulse surface preparation method. (a) Sample 7, grown on a substrate oriented in the exact (001) direction. (b) Sample 8, grown on a substrate oriented (001) 6° off towards the <111>A direction. APBs in the GaAs layers were confirmed through observing contrast reversal in adjacent domains in (002) darkfield and (00-2) darkfield images (not shown).

53

3.2.3.1. Reduction in period thicknesses (samples 9 – 12)

Superlattice samples with smaller period thicknesses were also grown at 650°C and 250 Torr; images of the samples are shown in Figure 3.6. The use of a TMGa pulse results in morphologies different to that seen in the thicker period samples (Fig. 3.5). Localized reduction in growth rate is observed for both Ge and GaAs, resulting in “well” formations, extending through multiple layers. The sample grown using the N2 anneal on the exact (001) substrate (Fig.

3.6a) in contrast exhibits morphology more consistent with what would be expected from strictly a growth step and time adjustment. The sample grown using the N2 anneal on the offcut substrate (Fig. 3.6c) is also relatively more comparable to its thicker period counterpart than the samples using the TMGa pulse; however, there is significantly more breakdown of the alternating GaAs-Ge layer structure. The deposition of GaAs, in particular, appears to be exhibiting similar issues as observed in the low temperature growths (Fig. 3.2). This suggests that, at these growth conditions, the intended layer thicknesses are near the critical thickness needed for islands to fully coalesce into a film. Indeed, we will see in the next chapter that further reductions in the period thicknesses will result in morphologies with cell-like structures for both substrate orientations.

54

Figure 3.6. (002) Darkfield XTEM images of GaAs/Ge superlattices grown at 650°C and 250 Torr. Sample 9 (a) and sample 11 (b) were grown on substrates oriented in the exact (001) direction, while sample 10 (c) and sample 12 (d) were grown on substrates oriented (001) 6° offcut towards the <111>A direction. N2 anneal was used in (a) and (c) while TMGa pulsing was used in (b) and (d). Precursor flows and flow times during deposition of each GaAs and Ge layer were the same for all samples. APBs in the GaAs layers were confirmed through observing contrast reversal in adjacent domains in (002) darkfield and (00-2) darkfield images (not shown).

3.2.4. 500°C and 250 Torr growth (samples 13 and 14)

Morphologies of GaAs/Ge superlattices grown at 500°C and 250 Torr are shown in Figure 3.7.

These samples exhibited many similarities with the samples grown at the same temperature but lower pressure (see section 3.2.2) along with an increase in Ge growth rate as a result of the higher pressure. The incubation time for the deposition of the first Ge layer was approximately

50 seconds. Growth rate was again affected by substrate orientation; the growth rate of Ge after

55 incubation was 1 nm/s on the exact (001) substrate and 0.4 nm/s on the (001) 6° offcut substrate.

As in the low pressure samples, the incubation times and exact growth rates of subsequent deposited Ge layers were difficult to determine from the reflectivity results.

A significant density of stacking faults can be observed in the samples. Stacking faults were also observed in the low pressure samples but at a lower density. We speculate that the combination of low temperature and higher pressure is causing a larger reduction in surface diffusion lengths than desired, increasing the likelihood of forming these defects.

Figure 3.7. (002) Darkfield XTEM images of GaAs/Ge superlattices grown at 500°C and 250 Torr. (a) substrate oriented in the exact (001) direction (sample 13). (b) substrate oriented (001) 6° offcut towards the <111>A direction (sample 14). Stacking faults are observed in both samples.

3.3. Growth of (GaAs)1-x(Ge2)x alloys

(GaAs)1-x(Ge2)x alloys were grown by codeposition of GaAs and Ge using several different growth temperatures, reactor pressures, and substrate orientations; details are shown in Table 3.2.

An additional sample grown consisted of a nominal 100nm thick (GaAs)1-x(Ge2)x alloy film followed by 200nm of GaAs (sample 22 in Table 3.2). In all samples, a thin (~40nm) GaAs

56 homoepitaxial layer was first grown on the substrates followed by a 20 second N2 ambient anneal prior to deposition of the alloy film. Compositions were determined using EDX.

Sample Composition Substrate Growth Temperature Reactor Growth rate Orientation (°C) Pressure (Torr) (Å/s) 15 (GaAs)0.73(Ge2)0.27 exact (001) 650 100 1.6

16 (GaAs)0.73(Ge2)0.27 (001) 6° off toward 650 100 1.6 <111>A

17 (GaAs)0.72(Ge2)0.28 exact (001) 650 250 7.8

18 (GaAs)0.72(Ge2)0.28 (001) 6° off toward 650 250 7.8 <111>A

19 (GaAs)0.77(Ge2)0.23 (001) 6° off toward 650 250 7.8 <111>B

20 (GaAs)0.83(Ge2)0.17 exact (001) 575 250 6.7

21 (GaAs)0.71(Ge2)0.29 exact (001) 700 250 8.2

22* (GaAs)0.77(Ge2)0.23 exact (001) 650 250 7.8 Table 3.2. Summary of epitaxy conditions of the (GaAs)1-x(Ge2)x alloys. Compositions were determined using EDX. Sample 22 – “double heterostructure” with 200nm GaAs and GaAs substrate cladding the alloy film.

The morphologies of the alloy samples grown at 650°C and 100 Torr (samples 15 and 16) are shown in Fig. 3.8. These structures clearly have different morphologies from the structures observed by Norman et al. (see Fig. 1.8), which showed pronounced phase separation but not in a columnar formation. For their structures, they suggested a growth mechanism where excess Ge segregated to the growing surface. The accumulation of Ge at the surface causes surface faceting, and upon reaching a critical concentration the excess Ge precipitates out, conformal to the faceted surface. This process is repeated throughout the growth, resulting in the phase separated morphology observed.9 The columnar morphology we observe in Fig. 3.8 is closer in appearance to those observed by Petroff et al. in ultra-thin period Ga1-xAlxAs/Ge superlattices grown by MBE. The growth mechanism proposed for those structures involved nucleation of

57

Ga1-xAlxAs islands on Ge which, due to the thinness of the attempted layers, remain exposed following the deposition of Ge. Consequently, the subsequent Ga1-xAlxAs film preferentially grew on those exposed islands. This process is repeated through the rest of the growth, forming

65 columns of Ga1-xAlxAs and Ge. We suspect a similar mechanism occurs with our alloy samples, with the difference being that the columns are Ge-rich and GaAs-rich alloy compositions rather than GaAs and Ge columns. This would also suggest that the growth of our alloys is similar in nature to the growth of many ternary and quaternary III-V alloys below their miscibility gaps.66–70

Based on the growth mechanism suggested by Norman et al. for their structures, it would appear that certain growth conditions can completely inhibit nucleation of Ge on GaAs. This nucleation is necessary in order to generate the columnar structure. Due to the numerous differences with the exact conditions used during MOCVD epitaxy, and likely many more that are not discussed, it is difficult to pinpoint which growth conditions may be responsible for the different morphologies. One possibility is the difference in surface preparation prior to depositing the alloy. As we discussed in section 1.3.2., Bai et al. found As-rich surfaces to be unfavorable for

Ge adatom incorporation.30 Since the alloy structures codeposit Ga, Ge, and As, an As-rich substrate surface in the samples investigated by Norman et al. may explain the segregation of Ge toward the surface as Ga and As incorporate into the film. If the growth operated in excess AsH3 as is typically done in growth of GaAs, this segregation could continue, with the end result being the growth mechanism described above. Our structures, in contrast, used the 20 second N2 ambient anneal process discussed earlier in the chapter for the superlattice structures; this would give the Ga-rich surface that enables Ge nucleation.

58

Figure 3.8. (002) Darkfield XTEM images of (GaAs)0.73(Ge2)0.27 alloys grown at 650°C and 100 Torr. a) substrate oriented in the exact (001) direction (sample 15). b) substrate oriented (001) 6° off towards the <111>A direction (sample 16).

In chapter 1, we discussed that phase separation of (GaAs)1-x(Ge2)x is a thermodynamically favorable process but largely inhibited in the bulk due to the low bulk diffusion of GaAs and Ge.

Growth of a pure metastable (GaAs)1-x(Ge2)x alloy therefore requires inhibiting the Ge to Ge and

GaAs to GaAs segregation process that would occur at the surface of the growing film. The contrast seen in Fig. 3.8 indicates that the conditions used in those samples were insufficient.

Fig. 3.9 show the morphologies of the alloy samples grown at 650°C and 250 Torr (samples 17 and 18). These structures were grown at a much faster rate, reducing the surface diffusion length of adsorbed atoms and thus reducing the length scale of the phase separation.

59

Figure 3.9. (002) Darkfield XTEM images of (GaAs)0.72(Ge2)0.28 alloys grown at 650°C and 250 Torr. a) substrate oriented in the exact (001) direction (sample 17). b) substrate oriented (001) 6° off towards the <111>A direction (sample 18).

The morphology of the alloy sample grown at 650°C and 250 Torr on a GaAs substrate orientated (001) 6° off towards the <111>B direction (sample 19) is shown in Fig. 3.10. This is vastly different from the sample in Fig. 3.9b, which was grown using all the same growth conditions except for the misorientation being along the <111>A direction (the different thicknesses grown are unlikely to be the cause for the morphology differences). While there are some localized streaks near the substrate-film interface in Fig. 3.10a that appear to bear some similarity to the streaks in Fig. 3.9b, the majority of the film has a much finer scale of inhomogeneity as seen at higher magnification in Fig. 3.10b. This large difference in morphologies shown for the two substrate orientations suggests that the surface reconstruction of the GaAs substrate orientated (001) 6° off towards <111>B significantly reduces surface adatom diffusion.

60

Figure 3.10. a) (002) Darkfield XTEM image of a (GaAs)0.77(Ge2)0.23 alloy grown at 650°C and 250 Torr. with substrate oriented (001) 6° off towards the <111>B direction (sample 19). b) magnified image of box in a)

Figure 3.11. a) (002) Darkfield XTEM image of a (GaAs)0.83(Ge2)0.17 alloy grown at 575°C and 250 Torr with substrate oriented in the exact (001) direction (sample 20). b) (002) Darkfield XTEM image of a (GaAs)0.71(Ge2)0.29 alloy grown at 700°C and 250 Torr with substrate oriented in the exact (001) direction (sample 21).

61

The morphologies of the alloy samples grown at 575°C and 700°C (samples 20 and 21) are shown in Figs. 3.11a and 3.11b, respectively. These morphologies are consistent with the results observed thus far. A lower growth temperature will result in reduced surface mobility; thus we observe a structure with a finer scale of inhomogeneity. In contrast, by increasing the growth temperature, the surface atom mobility increases, resulting in a structure exhibiting coarser inhomogeneity.

An attempt at growing a double heterostructure consisting of GaAs clads sandwiching a layer of

(GaAs)0.77(Ge2)0.23 is shown in Fig. 3.12. The morphology of the alloy is reasonably consistent with the morphology shown in Fig. 3.9a, which had the same growth condition. No APBs are observed in the upper GaAs layer; since the substrate is not offcut, this would suggest that the alloy film maintains predominantly a zincblende crystal lattice at the surface. The determination of the zincblende-diamond cubic order-disorder transition of the alloy was a key topic of interest in several theoretical studies that followed the original experimental study on the (GaAs)1-x(Ge2)x alloys; although a consensus does not appear to have been reached on many of the details, the average composition of the alloy in Fig 3.12 would be comfortably within the zincblende regime in all the studies.3,4,6,8,71,72 The structures we observed are not the single metastable phase; however, since no APBs are observed it is likely that there are very few, if any, regions near the alloy surface with higher than 30% Ge composition, the composition at which the zincblende- diamond cubic transition occurs according to a few of those studies.6,8 If there are regions at the surface that are diamond cubic in nature they may be sufficiently small so that rather than nucleate GaAs islands, they are overgrown by the coalescence of GaAs from the adjacent zincblende regions, preserving a single domain.

62

Figure 3.12. (002) Darkfield XTEM image of a GaAs – (GaAs)0.77(Ge2)0.23 – GaAs structure grown at 650°C and 250 Torr with substrate oriented in the exact (001) direction (sample 22).

3.4. Summary

In this chapter, we have investigated the growth of GaAs/Ge superlattices and alloys grown by

MOCVD at a variety of different growth conditions. A key focus of the investigation was to understand how the morphologies changed as various conditions were chosen in an attempt to limit growth kinetics, since “ideal” (i.e. flat, abrupt interfaces) superlattices and single phase alloys are thermodynamically unfavorable growth morphologies. In the superlattice structures, we observe morphologies closest to the “ideal” structure when the growth rate of Ge was increased and individual layers were sufficiently thick. Decreasing the thickness of the layers, however, resulted in a deviation of the morphology to a more randomized nanostructure. This result is not unexpected, since controlling the growth kinetics merely creates a pseudo-layer-by- layer growth mechanism where a high density of islands nucleate and coalesce to form a smooth film. Loss of the alternating band morphology indicates that the thicknesses being deposited must be below the critical thickness required for full coalescence. In applications requiring a

63 precise, regular microstructure, this randomized nanostructure may not be useful; however, the nanostructure may not necessarily be detrimental for thermoelectrics. In the next chapter, we will study the thermoelectric properties of these superlattice nanostructures.

In the alloy structures, we observed a decrease in the length scale of phase separation when using conditions that limit growth kinetics. The morphologies of our alloy structures are in significant contrast to those investigated by Norman et al., and may possibly be the result of different surface reconstructions prior to depositing the alloy film. The different morphologies may explain why we observe room temperature photoluminescence from our samples. We will explore this topic further in chapter 5.

64

Chapter 4: Thermoelectric Properties of III-V/IV Nanostructures

65

4.1. Introduction

In this chapter, we investigate the impact of GaAs/Ge superlattice morphologies and interface densities on their thermoelectric properties. GaAs/Ge is a model system for studying the impact of heterovalent interfaces; as we noted in chapter 1, such interfaces can in principle produce more disruption to phonon transport in comparison to isovalent interfaces. We also study the properties of a (GaAs)0.77(Ge2)0.23 alloy structure as well as an (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy structure. As we will discuss later in the chapter, several phonon scattering mechanisms are very weak in the GaAs/Ge system; thus, it is not likely to be the ideal III-V/IV system for thermoelectrics. (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16, however, encompasses those phonon scattering mechanisms and thus should give a better idea to what the overall III-V/IV material space can offer.

4.1.1. Structural quality of materials

Studies on thermoelectric properties of superlattices have often focused on materials systems such as GaAs/AlAs73,74 and Si/Ge75,76; the fabrication of high quality structures from these materials is well established, allowing for investigations of high level physical processes. Such high level investigations are not feasible on material systems such as III-V/IV superlattices, for which high quality structures have not been demonstrated. Thermoelectric studies on a more practical scope, however, do not have this limitation. In fact, it is not necessarily obvious whether or not particular crystal imperfections in a given structure would be detrimental overall to the thermoelectric figure of merit. Past studies on GaAs/AlAs77 and Si/Ge78 superlattices have found that the cross-plane thermal conductivities are lower than the in-plane thermal

66 conductivities. In addition, Luckyanova et al. observed that experimental results for thermal conductivity were lower than the results from simulations using density functional perturbation theory (DFPT), and suggested that this was a result of the simulations only accounting for atomistic mixing and not local layer thickness fluctuations.77 These observations suggest that a traditional high quality material structure may exhibit the lowest thermal conductivities, one of the aims for maximizing the figure of merit. If the reduction in thermal conductivity attributed to a crystal imperfection outweighs its negative effect (if any) on the electrical properties, then such an imperfection would actually be desirable.

This work does not investigate how the structural quality of GaAs/Ge superlattices affects the thermoelectric properties. We simply wish to note that there is merit to investigating “poor” quality superlattices for thermoelectrics; this is relatively unusual in the general field, where low quality materials are nearly always inferior to high quality materials.

4.2. Design of experiment

4.2.1. GaAs/Ge superlattices

The superlattice samples investigated used the growth methods discussed in section 3.2.3 (650°C growth temperature, 250 Torr system pressure, N2 anneal following GaAs deposition). Several different “period thicknesses” were chosen; deposition times for the samples were adjusted to obtain a nominal 500nm total film thickness. The samples have a 50:50 GaAs:Ge overall composition. Images of the samples grown on semi-insulating (001) exact GaAs substrates (on- axis samples) are shown in Fig. 4.1, while images of the samples grown on semi-insulating (001)

GaAs substrates offcut 6° toward the <111>A direction (offcut samples) are shown in Fig. 4.2.

67

As mentioned in the section 3.2.3, attempts at reducing the superlattice “period thickness” results in loss of the regular alternating formation of GaAs and Ge due to incomplete coalescence of

GaAs islands (see morphological difference between Figs. 4.1a and 4.1d as well as between Figs.

4.2a and 4.2d). This results in cell-like structures where GaAs “cells” are isolated by ribbon-like bands of Ge. Further attempts at reducing the period thickness also appears to result in loss of

Ge film coalescence in the on-axis samples (Figs. 4.1e-f). In the offcut samples, an unusual but pseudo-regular morphological formation can be observed (Figs. 4.2e-f). Due to the various morphologies present, a comparison between the structures and their effect on the thermoelectric properties must be done in terms of the interface densities. (As in the previous chapter, we will still continue to refer to the samples as superlattices for convenience.)

Figure 4.1. (002) Darkfield XTEM image of GaAs/Ge superlattice on-axis samples with interface densities of (a) 0.02 nm-1, (b) 0.037 nm-1, (c) 0.03 nm-1, (d) 0.109 nm-1, (e) 0.12 nm-1, and (f) 0.106 nm-1. Bright bands/regions correspond to GaAs while dark bands/regions correspond to Ge. Antiphase boundaries (APBs) can be seen as dark vertical lines in the GaAs regions. See section 4.2.1.1 for discussion on the interface density.

68

Figure 4.2. (002) Darkfield XTEM image of GaAs/Ge superlattice offcut samples with interface densities of (a) 0.022 nm-1, (b) 0.039 nm-1, (c) 0.038 nm-1, (d) 0.101 nm-1, (e) 0.105 nm-1, and (f) 0.068 nm-1. Bright bands/regions correspond to GaAs while dark bands/regions correspond to Ge. Antiphase boundaries (APBs) can be seen as dark vertical lines in the GaAs regions. See section 4.2.1.1 for discussion on the interface density.

4.2.1.1. Determination of interface density

The interface density was estimated by measuring a total interface length within an image and dividing by the encompassing area. The method is as follows: multiple cross-section darkfield

TEM images at reasonably high magnifications (60000x-80000x) are taken at various sections of the material using the (002) two-beam condition to obtain contrast between GaAs and Ge (GaAs sections bright, Ge sections dark). A threshold is then obtained using an image processor for each image; an example is shown in Fig. 4.3. For each threshold image, the perimeter of all black regions is calculated. This is summed with the perimeter of the black regions in the reversed threshold image, resulting in a perimeter equivalent to 2x the interface length and the

69 original image border. The interface length is then extracted by subtracting the image border and halving the result. Many images at various locations are used to obtain an average interface density in order to reduce the measurement error. The total interface length includes all interfaces regardless of interface orientation, and it was assumed to remain consistent through the sample perpendicular to the images.

Note that the interface density of the offcut sample shown in Fig. 4.2f is significantly lower than that of the sample shown in Fig. 4.2d. This is despite the fact that the sample in Fig. 4.2f was grown with a significantly lower intended “period thickness.” A similar result was also seen in the on-axis samples. Incomplete coalescence of both GaAs and Ge likely contributes to a lower interface density since there would be a higher amount of GaAs on GaAs and Ge on Ge deposition, which would not create more interface length. Additionally, since the layers were intended to be extremely thin (~2nm), bulk diffusion at the near surface may reduce the formation of heterovalent bonds and thus lower interface densities.

Figure 4.3. Threshold (right) of an (002) darkfield TEM image (left). An image processor can be used to calculate the perimeter of the dark regions. This perimeter is then summed with the perimeter calculated for the reversed image (result: 2x interface length + original image border). Total interface length is then extracted by subtracting the original image border and dividing the result by two.

70

4.2.2. III-V/IV alloy structures

Growth conditions of the (GaAs)0.77(Ge2)0.23 alloy structure investigated in this chapter are given in section 3.3 (see Fig. 3.10 for the offcut structure; the on-axis structure has the same morphology as the sample shown in Fig. 3.9a, but grown to a thickness of ~1μm). An

(In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy structure was grown by codeposition of the constituent materials to a nominal thickness of 1μm. The substrates were semi-insulating (001) exact GaAs

(on-axis sample) and (001) GaAs 6° offcut toward the <111>B direction (offcut sample). The sample was grown at 575°C and 250 Torr system pressure. A thin (~40nm) GaAs homoepitaxial layer was first grown on the substrates followed by a 20 second N2 ambient anneal prior to deposition of the alloy structure. No apparent defects or phase separation were observed in the sample; an image of the sample grown on the on-axis substrate is shown in Fig. 4.4.

Figure 4.4. (002) Darkfield XTEM image of (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy grown on (001) exact GaAs. Defects and phase separation were not observed in the structure.

71

4.3. GaAs/Ge structures

4.3.1. Thermal conductivity

A representative trace of the measured reflectance signal (part of the method of measuring film thermal conductivity) along with the corresponding model fits is shown in Fig. 4.5, where the solid blue curve represents the best model fit from the heat diffusion theory and the dashed curves represent the model fits by adjusting the best fit thermal conductivity by ±10% to show measurement sensitivity. Generally, we achieved excellent fitting between the heat diffusion model and the experimentally measured reflectance signal by using the sample thermal conductivity and the metal-sample interface conductance as free parameters. We also note that the measurement is very sensitive to the thermal conductivity we would like to probe, as verified by the dashed curves shown in Fig. 4.5.

Figure 4.5. Representative measured phase signal along with the best model fit based on the thermal diffusion model. The dashed curves represent the model fits by adjusting the best fit thermal conductivity by ±10% to show the sensitivity of the measurement to the underlying thermal conductivity. 72

Figure 4.6. Thermal conductivities and the corresponding interface density of the on-axis samples. The alloy sample is shown on the right. Thermal conductivities of bulk GaAs (triangle) and Ge (*) are shown at 0 interface density for comparison.

Figure 4.6 shows the correlation between the thermal conductivity and interface density of the on-axis samples. The thermal conductivity of pure GaAs and Ge are shown as well. The general trend of decreasing thermal conductivities as the interface density increases, as well as a significant decrease in comparison to that of the constituent materials, matches that of cross- plane thermal conductivities for GaAs/AlAs superlattices74 and Si/Ge superlattices.75 However, in contrast to those two systems, the lowest observed thermal conductivity corresponded to the alloy sample. Similar results were seen in the offcut samples, shown in Fig. 4.7. These results

73 are consistent with the idea that the alloy samples are phase separated with fine/nanoscale inhomogeneity (see section 3.3). The structures essentially can be considered to have a very high density of interfaces for phonons to scatter at. This is in addition to actual alloy scattering within the nano-regions themselves, which is absent in the non-alloyed GaAs and Ge layers of the superlattices.

Figure 4.7. Thermal conductivities and the corresponding interface density of the offcut samples. The alloy sample is shown on the right. Thermal conductivities of bulk GaAs (triangle) and Ge (*) are shown at 0 interface density for comparison.

While a reduction in thermal conductivity is observed in these GaAs/Ge structures, the reduction is much less relative to that seen in Si/Ge superlattices, which can attain themal conductivities on the order of 3-5 W/m-K. This can be attributed in part to the very thin periods achievable in

74

Si/Ge superlattices (~3-7 nm)75 which would correspond to an effective interface density of 0.3-

0.67 nm-1, significantly higher than the interface densities observed in the GaAs/Ge superlattices.

Additionally, Si and Ge have a large difference in density and elastic constants, both of which play a significant role in scattering phonons at interfaces.75,79 GaAs and Ge, in contrast, have very similar densities and elastic constants. Nevertheless, the substantial reduction observed in the GaAs/Ge structures from the bulk constituent values is evidence that III-V/IV heterovalent interfaces can play a significant role in obtaining low thermal conductivities.

4.3.2. Electrical mobility

Van der Pauw measurements of the on-axis superlattice samples showed non-linear behavior in the I-V results; the resistance decreased as the applied current was increased. Bai et al. determined that by preparing a Ga-rich surface prior to Ge deposition, a substantial portion of the

Ge film becomes highly p-doped. This occurs due to the tendency of Ga to exchange with Ge and segregate to the growth surface, resulting in a gradual incorporation of Ga as an unintentional dopant into the Ge film. A Ge film with thickness of 120nm was shown to be completely p-type, with a carrier concentration in the 1018-1019 cm-3 range.80 Given the different conditions used for the growth of our superlattice samples, the Ge regions of the superlattices may not necessarily have similar results; however, their observations indicate a high unlikelihood that the Ge regions in the superlattice samples would be moderately or highly doped n-type (i.e. regions are p-type or possibly doped n-type with ≪1017 cm-3 carrier concentration).

Any unintentional of GaAs by Ge, on the other hand, is expected to be n-type with carrier concentrations greater than 1017 cm-3. As a result, a depleted or low carrier concentration region

75 is formed in the superlattice films. At low currents, this region limits electrical transport to a thinner region nearer to the film surface. At higher currents, transport breaks through this weak barrier to conduct through the entire thickness of the film.

This non-linear I-V behavior was not observed in most of the offcut superlattice samples nor the alloy structures; the characteristic was seen only in the offcut sample with the lowest interface density (see Fig. 4.2a). In the case of the alloy samples one would not expect any formation of a low carrier region parallel to the plane of the film. In the case of the superlattices it is likely that these regions are formed at an angle, given the morphologies, and thus pass through the entire cross section of the film. Electrical transport through the layer would require passing through these regions at any applied current. If these barriers to transport are very weak relative to the magnitudes of the currents used, the I-V results would be effectively linear. In the offcut sample with the lowest interface density, we suspect that the layers were sufficiently thick so that the low carrier region formed parallel to the plane of the film.

GaAs/Ge Interface Density κ (W/m-K) n (cm-3) µ (cm2/V-s) S (µV/K) Sample (nm-1) superlattice .039 26.7 1.6×1017 480 -182 (offcut)

superlattice .068 21.8 5.2×1017 1130 -1.3 (offcut)

superlattice .101 16.9 3.0×1017 1490 -197 (offcut)

superlattice .105 17.2 3.0×1017 1550 -195 (offcut)

alloy -- 15.4 1.4×1017 1140 -83 (offcut)

alloy -- 15.9 1.5×1017 1090 -31 (on-axis) Table 4.1. Thermoelectric transport properties of several offcut superlattice samples and the GaAs/Ge alloy structures.

76

The in plane electrical properties of the offcut superlattice samples and the GaAs/Ge alloy structures are shown in Table 4.1. Interestingly, the electrical transport results for the superlattice samples with higher interface densities are comparable to those of bulk GaAs81 and bulk Ge82 (see Figs. 4.8 and 4.9). In addition, although the measurement was conducted in the plane of the film, electrons traveling through the material must still cross a large number of interfaces (see Fig. 4.2d-f). The results suggest that the transport behavior of electrons is largely unaffected by GaAs/Ge interfaces. A possible explanation for this is the close alignment of the conduction bands of GaAs and Ge at the interfaces.83,84 Electrons traveling through the material do not “perceive” such an interface and thus experience negligible interface-related scattering as a result. The electrical properties of the sample with the lower interface density (.039 nm-1) are much lower in comparison. We suspect that the presence of APBs (see Fig. 4.2b) is responsible for the reduced properties. The samples with higher interface densities do not appear to have a significant density, if any, of APBs, which is consistent with the electrical results being much higher and comparable to that of bulk GaAs and Ge. Due to the vertical nature of the APBs, it is possible that they have a minimal impact on the cross plane electrical properties. Since GaAs/Ge interfaces do not appear to significantly degrade electrical transport, the cross plane electrical properties of the sample with low interface density may be comparable to those of the samples with higher interface densities.

77

Figure 4.8. Electron mobility of bulk GaAs.81 The samples in Table 4.1 are plotted for comparison.

Figure 4.9. Electron mobility of bulk Ge. 1. Low temperature measurement (77K) 2. Room temperature measurement (300K).82 The samples in Table 4.1 are plotted for comparison.

78

4.3.3. Seebeck coefficient

When one end of a material is heated, charge carriers will diffuse toward the cold end. If only one type of carrier is present, this will to a built-in voltage. The Seebeck coefficient is a measurement of this induced voltage from the temperature differential. Maximizing the Seebeck coefficient thus requires that the structure contains predominantly one carrier type, among other considerations. If there are significant amounts of both carriers, they would both diffuse toward the cold end and cancel out each other’s contribution to the induced voltage. The Seebeck coefficients of the samples discussed in the previous sections are listed in Table 4.1. The results do not show any obvious relation between morphology and the Seebeck coefficient. However,

Fig. 4.8 suggests the possibility of a high background concentration of acceptors in the materials.

Should this be the case, the samples investigated all have lower Seebeck coefficients then could potentially be obtained given the same carrier concentrations. In addition, based on Fig. 4.8 the alloy structures may have an even higher background concentration of acceptors than most of the superlattices, which would be consistent with the lower observed Seebeck coefficients in the alloy structures. This is far from conclusive, however, since one of the superlattices exhibits an extremely low Seebeck coefficient.

4.4. Estimation of thermoelectric potential for the GaAs/Ge system

As we noted at the beginning of the chapter, the GaAs/Ge system is a model III-V/IV system that is unlikely to have the best thermoelectric properties; however, it is still instructive to see where this system would stand in the field. Based on the results detailed in Table 4.1, it is clear that the

79 samples we investigated are far from being good thermoelectric materials; however, it is also reasonable to believe that the properties are not at their optimized levels. In particular, the vast majority of thermoelectric materials have typically been shown to maximize zT at degenerate or near degenerate carrier concentrations (1019 – 1020 cm-3).33,85 State-of-the-art room temperature thermoelectric materials consist of alloys of Bi2Te3 and Sb2Te3, with maximium zT of about

1.33,85,86 In the GaAs/Ge system, we observed in section 4.3.2 that the electron carrier concentration – mobility relation appears to be similar to that of GasAs and Ge. Assuming this

19 -3 relation holds for a (GaAs)0.77(Ge2)0.23 alloy with a doping level of 10 cm the mobility would be approximately 300 cm2/V-s. This would result in an electrical conduction of 480 Ω-1 cm-1. At this doping level, it is also likely that the thermal conductivity is still dominated by phonon transport and thus not significantly changed from ~16 W/m-K. To obtain a room temperature zT of 1, the Seebeck coefficient of this alloy would need to be ~ 1050 µV/K. Such a Seebeck coefficient is typically seen only in insulators;85 it is highly unlikely that this can be obtained in a

20 -3 highly doped (GaAs)0.77(Ge2)0.23 alloy. If the alloy instead has a doping level of 10 cm the mobility would be approximately 100 cm2/V-s, giving an electrical conduction of 1600 Ω-1 cm-1.

If we assume a best case scenario where the thermal conductivity is unchanged from 16 W/m-K, then the Seebeck coefficient must be ~580 µV/K to obtain a room temperature zT of 1. This is still highly unlikely to be attainable. A room temperature zT of 0.3, however, would require a

Seebeck coefficient of 320 µV/K, which is within the range of values seen in doped semiconductors and thus may possibly be attainable. A (GaAs)0.77(Ge2)0.23 alloy with a zT of 0.3 clearly would not replace Bi2Te3 alloys in most applications; however, there may be particular situations where the sacrifice in efficiency is worth accessing the unique advantages of the

80

GaAs/Ge system (e.g. higher compatibility with conventional device fabrication techniques may allow for form factors/lower costs in devices that cannot be realized with Bi2Te3 alloys).

In addition to the carrier concentration, the operating temperature is also a parameter that must be considered in determining a material’s effectiveness as a thermoelectric. As we noted in chapter 1, the peak zT of Bi2Te3 alloys occurs around room temperature, but for other materials like SiGe alloys the peak zT occurs at temperatures of 800°C – 900°C (see Fig. 1.10). Generally, the relation between peak zT and operating temperature is dictated by the bandgap of a material. zT typically increases with temperature up until the onset of intrinsic conduction.85,86 Intrinsic conduction negatively affects the zT by reducing the Seebeck coefficient (high concentration of both carriers) as well as increasing the electronic contribution to thermal conductivity. Since intrinsic conduction in a material with a large bandgap occurs at higher temperatures, the zT would peak at a higher temperature. The GaAs/Ge system can be expected to have higher bandgaps than PbTe or CoSb3 alloys; thus, based on Fig. 1.10, it is likely that the system would have peak zT at temperatures higher than 600°C. (The bandgap of the (GaAs)1-x(Ge2)x alloys in this work is the subject of further investigation in chapter 5; if we take the bandgaps given by

Newman et al. to be accurate, the zT of the (GaAs)0.77(Ge2)0.23 alloys would likely peak at temperatures between 700°C – 800°C.) Measurements of the thermoelectric properties at those elevated temperatures would be needed to further gauge the potential of the GaAs/Ge system.

Composition is another parameter that needs to be considered. Since the electrical conductivity of the GaAs/Ge system appears to be similar to bulk GaAs and bulk Ge, this would suggest that the electrical conductivity is relatively insensitive to composition. Thus, the ideal composition would depend on its effect on the thermal conductivity and Seebeck coefficient.

81

4.5. Thermoelectric properties of (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16

The thermoelectric properties of the (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy structures are shown in

Table 4.2 (The properties of the (GaAs)0.77(Ge2)0.23 alloys discussed above are also shown for ease of comparison). The thermal conductivity of these samples are as expected much lower than that of the (GaAs)0.77(Ge2)0.23 alloys, with the presence of indium and silicon creating a larger mismatch in density and elastic constants and thus enhancing phonon scattering. In addition, the (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy structures have significantly lower thermal conductivities than the constituent materials – based on Figs. 4.10 and 4.11, the thermal conductivity of In0.1Ga0.9As is ~12 W/m-K while the thermal conductivity of Si0.1Ge0.9 is ~25

W/m-K.

Sample n µ S σ κ zT (cm-3) (cm2/V-s) (µV/K) (Ω-1 cm-1) (W/m-K) (300 K)

17 -4 (GaAs)0.77(Ge2)0.23 1.4×10 1140 -83 25.5 15.4 3.4×10 (offcut)

17 -5 (GaAs)0.77(Ge2)0.23 1.5×10 1090 -31 26.2 15.9 4.8×10 (on-axis)

17 -4 (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 7.3×10 390 -42 45.6 4.6 5.2×10 (offcut)

17 -3 (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 7.8×10 380 -73 47.4 4.3 1.8×10 (on-axis)

Table 4.2. Thermoelectric properties of (GaAs)0.77(Ge2)0.23 and (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy structures.

82

Figure 4.10. Dependence of thermal resistivity of In1-xGaxAs on composition. Experimental values shown as a solid line. Dotted line denotes a theoretical fit.87

88 Figure 4.11. Dependence of thermal conductivity of Si1-xGex on composition.

83

4.5.1. Optimization of (In1-xGaxAs)1-z(Si1-yGey)z

As with the case of the (GaAs)0.77(Ge2)0.23 alloy structures, it is unlikely that the

(In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy structures investigated have the best thermoelectric properties for the material system. Optimizing the (In1-xGaxAs)1-z(Si1-yGey)z material system, however, is significantly more complex since there are three adjustable composition parameters, in addition to considering the carrier concentration and operating temperature. Based on the thermal conductivities dependences shown in Figs. 4.10 and 4.11, one can reasonably assume that the lowest thermal conductivity of an (In1-xGaxAs)1-z(Si1-yGey)z alloy structure would be observed in a sample consisting of an In0.5Ga0.5As constituent, corresponding to a thermal conductivity of ~ 4.5 W/m-K, and a Si1-yGey constituent with y = 40-80% composition, corresponding to a thermal conductivity of ~10 W/m-K. We observed significant reduction in the (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy, relative to the constituents, when composition z was just

16%. Thus if composition z is maintained at 16% for the new composition combination, the thermal conductivity of the overall alloy should be lower than the 4.5 W/m-K of In0.5Ga0.5As, which would make it lower than that of the investigated (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy. (A more appropriate composition z may result in further reduction; the exact composition, however, would need to be investigated. Minimum thermal conductivities of many III-V ternary alloys occur near 50% compositions, but there are a few with minimums slightly away from 50% –

GaAsP being one example, having the minimum at 60% P.87,89).

Unlike the GaAs/Ge system, where the electron mobility appears insensitive to composition, the electron mobility of (In1-xGaxAs)1-z(Si1-yGey)z is strongly affected by not only composition x and

84 composition y (see Figs. 4.12 and 4.13, respectively) but also strongly affected by composition z, given the significant lower mobility of the (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy (see Table 4.2) compared to In0.1Ga0.9As and Si0.1Ge0.9. Thus, if we focus entirely on reducing the thermal conductivity and use the compositions discussed above, the electron mobility resulting from that composition is highly unlikely to be optimal. Composition z, in particular, may actually need to be relatively small even if a lower thermal conductivity could be obtained at higher z. We can expect a bow-shaped relation for the electron mobility as a function of composition z for

(In0.1Ga0.9As)1-z(Si0.1Ge0.9)z, similar to that seen for InGaAs (Fig. 4.12) and SiGe (Fig. 4.13).

This would mean very low electron mobilities if the composition z approaches 50%. Similarly, we can see from Fig. 4.12 that the electron mobility of In1-xGaxAs is much greater at low Ga compositions, which suggests that optimizing (In1-xGaxAs)1-z(Si1-yGey)z may actually require a low composition x.

Given the complexity of determining the optimal compositions (note that we have not even considered the Seebeck coefficient), it is difficult to comment on the possibility of obtaining a zT of 1 for (In1-xGaxAs)1-z(Si1-yGey)z alloys like we did for (GaAs)1-x(Ge2)x. We note, however, that if we can obtain the same electron mobilities shown in Table 4.2 at carrier concentrations of

7.8×1019 cm-3 and also maintain the same thermal conductivity, Seebeck coefficient need only be

~ 172 µV/K to achieve a zT of 1.

85

90 15 -3 Figure 4.12. (Solid curves) Dependence of electron mobility of In1-xGaxAs on composition. 1 – n=3×10 cm ; 2 – n=4×1016 cm-3; 3 – n=2.3×1017 cm-3

91 Figure 4.13. Dependence of electron mobility of Si1-xGex on composition.

86

4.6. Summary

In this chapter, we observed a significant decrease in the thermal conductivities of GaAs/Ge superlattices relative to bulk GaAs and bulk Ge. This decrease can be attributed primarily to the heterovalent interfaces; most other phonon scattering mechanisms are weak due to the similar masses, densities, and elastic constants of GaAs and Ge. The thermal conductivity decreased with increasing interface density; however, the lowest thermal conductivities were observed in the (GaAs)0.77(Ge2)0.23 alloys, in contrast to previous experimental studies done for the

GaAs/AlAs system and Si/Ge system. This result suggests that the alloys are indeed phase separated. The length scale of the phase separation is small; thus the alloys can be considered as structures with an extremely high density of interfaces. The electron mobilities of the structures were very similar to highly compensated bulk GaAs and did not appear to be strongly affected by

GaAs/Ge interfaces. This may be due to the close alignment of the GaAs and Ge conduction bands, which results in minimal electronic interface scattering.

Based on the transport property results obtained, it is highly unlikely that the GaAs/Ge material system will be suitable as thermoelectric materials, even after accounting for several potential optimizations of the structures we investigated – the relatively high thermal conductivities is likely the limiting factor. This high thermal conductivity, however, is the consequence of significantly decreased effectiveness of various phonon scattering mechanisms as a result of using the highly similar GaAs and Ge. More complex alloys, such as (In1-yGayAs)1-x(Si1-zGez)x, would retain these scattering mechanisms; indeed, we observe much lower thermal conductivities in (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloys. These thermal conductivities were also lower than the expected thermal conductivities of the constituting III-V and IV alloys, indicating that the heterovalent mixing still plays a part in the reduction. Further investigations are needed

87 to determine if this material system, or potentially other III-V/IV systems, can be suitable for thermoelectric applications.

88

Chapter 5: Optical Properties of III-V/IV Alloy Structures

89

5.1. Introduction

In this chapter, we investigate the optical properties of III-V/IV alloy structures grown by

MOCVD. Our focus is primarily on trying to understand the origin of room temperature photoluminescence (PL) signals observed in our alloys, a surprising but very interesting result that may potentially lead to a variety of optical applications for these material systems. We will attempt to explain the results as it relates to the morphology, since it is clear upon comparison with the previous studies (luminescence signals only observed at extreme cryogenic temperatures for very different sample morphologies – see Fig. 1.8 and section 3.3 for comparison)9,20 that morphology must play a significant role.

Table 5.1 summarizes the alloy structures we will discuss in this chapter. All samples were grown at 250 Torr system pressure on semi-insulating (001) exact GaAs substrates. The general morphology is noted; TEM images can be found in the figures listed.

Alloy Composition General Morphology Growth Temperature Image

1 (GaAs)0.81(Ge2)0.19 speckled 650°C Fig. A1

2 (GaAs)0.83(Ge2)0.17 speckled 575°C Fig. 3.11a

3 (GaAs)0.88(Ge2)0.12 speckled 575°C Fig. A2

4 (GaAs)0.94(Ge2)0.06 speckled 575°C Fig. A3

5 (GaAs)0.77(Ge2)0.23 speckled 650°C Fig. A4

6 (GaAs)0.71(Ge2)0.29 striations/columns 700°C Fig. 3.11b

7 (GaAs)0.8(Si0.1Ge0.9)0.2 speckled 650°C Fig. A5

8 (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 No apparent features 575°C Fig. 4.4

9 (In0.15Ga0.85As)0.8(Si0.1Ge0.9)0.2 striations/columns 650°C Fig. A6 Table 5.1. Alloy structures investigated. Compositions were determined using EDX.

90

5.2. Optical measurements of the (GaAs)1-x(Ge2)x alloys

Room temperature photoluminescence spectra of samples 1 – 4 are shown in Fig. 5.1. The spectra beyond 1650nm could not be obtained with the InGaAs detector used for this particular measurement; however, we estimate the full width at half maximum (FWHM) to be greater than

250nm. Peak wavelengths for the samples shown were all approximately 1550nm (0.8 eV) despite the change in composition. Luminescence intensity, however, decreased as the composition of Ge decreased in the alloys.

Transmission and reflection spectra were measured for samples 2 – 4. The absorption coefficient,

훼, was estimated by the following equation:92

1 1 − 푟 훼 = 푙푛 ( ) (5.1) 푑 푡 where 푑 is the thickness of the film, r is the reflectance, and t is the transmittance. Fig. 5.2 shows the absorption coefficients obtained as a function of excitation energy, indicating an absorption edge at around 0.8 eV. Additionally, the absorption is weaker with decreased Ge compositions.

Fig. 5.3 shows the photoluminescence spectrum of the (GaAs)0.81(Ge2)0.19 alloy measured at 77K.

The peak wavelength was observed at approximately 1495nm (0.83 eV). No peaks were observed in the wavelength ranges that would correspond to the indirect bandgap transition of Ge, which can be observed at low temperatures,80 casting doubt into the possibility that the 0.8 eV transition we observe at room temperature corresponds to the direct gap transition of Ge.

91

Figure 5.1. Room temperature photoluminescence spectra of (GaAs)1-x(Ge2)x alloy samples with Ge compositions less than 20%. The Luminescence peaks of all samples at approximately 1550nm. Luminescence intensity decreases with decreasing Ge composition.

Figure 5.2 Absorption coefficient dependence on excitation energy for the (GaAs)0.83(Ge2)0.17, (GaAs)0.88(Ge2)0.12, and (GaAs)0.94(Ge2)0.06 alloys. The absorption edge for each sample appears to be around 0.8 eV. (Note: A measurement artifact is responsible for the apparent absorption between 0.6 – 0.8 eV.)

92

Figure 5.3. 77K photoluminescence spectrum of (GaAs)0.81(Ge2)0.19. Peak wavelength at approximately 1495nm.

Room temperature photoluminescence spectra of the “speckled” (GaAs)0.81(Ge2)0.19 and

(GaAs)0.77(Ge2)0.23 alloys are shown in Fig. 5.4. The peak wavelength of (GaAs)0.77(Ge2)0.23 lies at approximately 1720nm (0.72 eV), in contrast to the samples with lower Ge compositions. The photoluminescence spectrum of the “striated” (GaAs)0.71(Ge2)0.29 alloy is also shown; the peak occurs at approximately 1760nm (0.7 eV), and there appears to be a shoulder peak at approximately 1900nm (0.65 eV). These transitions are all lower than the direct gap transition of

Ge, nor should they correspond to the indirect transition of Ge since the measurements were conducted at room temperature. Fig. 5.5 shows the photoluminescence spectrum at 77K of the

“striated” (GaAs)0.71(Ge2)0.29 alloy. Multiple peaks/shoulders are resolved and is likely due to the noticeably more heterogeneous morphology. Samples with speckled morphologies did not appear to exhibit multiple peaks, as in the case of (GaAs)0.81(Ge2)0.19 in Fig. 5.3.

93

Figure 5.4. Room temperature photoluminescence spectra of the (GaAs)0.81(Ge2)0.19, (GaAs)0.77(Ge2)0.23, and (GaAs)0.71(Ge2)0.29 alloy samples. Luminescence peak for (GaAs)0.81(Ge2)0.19 at approximately 1550nm (0.8 eV) is consistent with the results obtained in Fig. 5.1. Luminescence peaks for (GaAs)0.77(Ge2)0.23 and (GaAs)0.71(Ge2)0.29 at approximately 1720nm (0.72 eV) and 1760nm (0.7 eV), respectively.

Figure 5.5. 77K photoluminescence spectrum of (GaAs)0.71(Ge2)0.29, with multiple luminescence peaks resolved. These peaks likely correspond to different striated regions in the structure. 94

5.3. Optical measurements of (GaAs)1-y(Si1-xGex)y and (In1-xGaxAs)1-z(Si1-yGey)z

The room temperature photoluminescence spectrum of the (GaAs)0.8(Si0.1Ge0.9)0.2 alloy is shown in Fig. 5.6. The spectrum of the (GaAs)0.81(Ge2)0.19 alloy is also shown for comparison. The peak wavelength for (GaAs)0.8(Si0.1Ge0.9)0.2 is approximately 1530nm (0.81 eV), which is a slight blueshift compared to (GaAs)0.81(Ge2)0.19. This shift is more noticeable at low temperature, as seen in Fig. 5.7.

Photoluminescence measurements were conducted on two different (In1-xGaxAs)1-z(Si1-yGey)z samples (samples 8 and 9). The room temperature and low temperature photoluminescence spectra of the “featureless” (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy is shown in Figs. 5.8 and 5.9, respectively. At room temperature, there appears to be two dominant energy transitions, one at

1550nm (0.8 eV) and the other beyond 1650nm (0.75 eV). Only one dominant transition is observed at low temperature, however. Room temperature and low temperature photoluminescence spectra of the “striated” (In0.15Ga0.85As)0.8(Si0.1Ge0.9)0.2 alloy is shown in Figs.

5.10 and 5.11, respectively. Multiple luminescence peaks were observed; the low temperature measurement in particular revealed four distinct peaks. These results can be attributed to the significant degree of phase separation observed in the morphology.

95

Figure 5.6. Comparison of the photoluminescence spectra at room temperature for (GaAs)0.8(Si0.1Ge0.9)0.2 and (GaAs)0.81(Ge2)0.19. The addition of Si causes a blueshift in the peak luminescence.

Figure 5.7. Comparison of the photoluminescence spectra at 77K for (GaAs)0.8(Si0.1Ge0.9)0.2 and (GaAs)0.81(Ge2)0.19. The blueshift in the peak luminescence is significantly larger at low temperature.

96

Figure 5.8. Photoluminescence spectra at room temperature for (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 and (GaAs)0.81(Ge2)0.19.

Figure 5.9. Photoluminescence spectra at 77K for (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 (actual signal × 10) and (GaAs)0.81(Ge2)0.19.

97

Figure 5.10. 77K photoluminescence spectrum of the (In0.15Ga0.85As)0.8(Si0.1Ge0.9)0.2 alloy.

Figure 5.11. 77K photoluminescence spectrum of the (In0.15Ga0.85As)0.8(Si0.1Ge0.9)0.2 alloy. Peaks are observed at 1495nm (0.83 eV), 1710nm (0.73), 1860nm (0.67), and 1880nm (0.66). The significant degree of phase separation observed in the morphology likely causes the numerous luminescence peaks.

98

5.4. Possible explanations for the origin of photoluminescence

The past investigations discussed in chapter 1 indicate that (GaAs)1-x(Ge2)x metastable alloys with 19%, 17%, 12%, and 6% Ge compositions should have approximate bandgaps of 0.85 eV,

0.9 eV, 1.04 eV, and 1.22 eV, respectively.5,6,8 Clearly, this is not observed in our alloys with the corresponding compositions, all of which appear to have room temperature transitions at 0.8 eV. While a 0.8 eV transition happens to be the direct gap transition for Ge, the alloys with 23% and 29% Ge show smaller energy transitions at room temperature. It is quite likely that the cause of the observed luminescence in these higher Ge alloys is the same for the luminescence in the lower Ge alloys; thus it is very unlikely that the observed 0.8 eV transitions in the lower Ge alloys is from pure Ge quantum dots/nanocrystals/etc. Here, we speculate on some mechanisms that may be involved in causing the photoluminescence observed.

5.4.1. Bandgap narrowing due to composition fluctuation

Various groups studying III-V ternary and quaternary alloys exhibiting non-random atomic arrangements have observed reductions in the energy gap relative to the homogeneous random

66,93–95 alloy structure. Composition fluctuations in the (GaAs)1-x(Ge2)x alloys can be expected to cause a similar phenomenon. Fig. 5.12 is a possible band diagram of the alloys. Regions with higher Ge compositions than the alloy average would have smaller bandgaps. Holes generated by the excitation source would relax to these regions before recombining, resulting in luminescence peaks at longer wavelengths than expected for the homogeneous alloy. Moreover, composition fluctuation can also explain the relative broadness of the peaks, since localized regions can form such that they have higher Ge compositions in comparison to the immediate

99 surroundings. Holes falling into these regions would then recombine even if the region does not have the smallest bandgap. This results in significant luminescence signal across a relatively broad range of wavelengths.

5.4.2. Radiative and non-radiative defects

TEM imaging did not reveal any observable defects in the (GaAs)1-x(Ge2)x alloys; however, it is difficult to ignore the possibility of point defects or other defects not observable by TEM, given the necessity of forming III-IV and IV-V bonds. If present, there is the possibility that the defects are optically active and cause the luminescence observed in our samples. Several groups

96– have observed near-infrared photoluminescence in Ge nanocrystals embedded in SiO2 matrix,

99 and one group has proposed that the origin of the luminescence is from defects rather than the nanocrystals.100

Non-radiative defects may also play a role in the luminescence spectra observed. The luminescence spectra of the (GaAs)1-x(Ge2)x samples with Ge compositions less than 20% suggest that composition fluctuations resulted in regions containing 21% Ge, the composition corresponding to a 0.8 eV bandgap. 5,6,8 The density of these high Ge regions naturally would be lower in the samples with lower average Ge composition (see Fig. 5.12), which could explain the decreasing intensity of the peak with decreasing Ge composition. A lower density of these regions, however, should not result in a reduction to the overall luminescence yield; this reduction may be due to non-radiative trap states that only fall within the energy gaps for the low

Ge regions, thus quenching any luminescence at shorter wavelengths.

100

5.4.3. Recombination at interfaces due to spatial separation of photoexcited carriers

In type II band alignment, the lowest energy states for electrons and holes are in different regions. this can result in recombination between spatially separated electrons and holes; the energy transition would be lower than the local recombination transition. Based on the band alignments of Ge, GaAs, and InAs84 and the likelihood that the alloy is heterogeneous as in the other alloys, it is likely that this type of recombination is the cause for the lower energy transition (beyond

1650nm) observed for the (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy in Fig. 5.8, while the higher energy transition (1550nm) is due to local recombination. At low temperatures carriers have a longer lifetime, enabling them to fall to the band edges before recombining. Thus, we observe only one peak in Fig. 5.9.

Recombination between spatially separated electrons and holes may possibly occur in the

(GaAs)1-x(Ge2)x alloys as well; we have suggested in chapter 4 that the conduction bands of all compositions of (GaAs)1-x(Ge2)x alloys are approximately aligned given the minimal degradation of the electron mobility. (We have assumed an approximately aligned conduction band in Fig.

5.12 for this reason). In this case, since the transition energy would be equivalent to that of the local recombination, the luminescence spectra cannot be used to distinguish between the recombination mechanisms. It is important to note the possibility of both mechanisms, however.

Depending on the exact loss mechanisms in the material, the spectra and intensities we observed may be dominated by one of the mechanisms. We can consider a hypothetical case in which the luminescence is exclusively due to space separated electron-hole recombination at interfaces and luminescence elsewhere is quenched by non-radiative defects. This would suggest that luminescence intensity can be increased by increasing interface density.

101

Figure 5.12. Possible band diagram schematic of the (GaAs)1-x(Ge2)x alloys. Based on our results in section 4.3.2, we estimate that the conduction band of (GaAs)1-x(Ge2)x at all compositions is approximately aligned. Due to the composition fluctuations in the samples, regions with higher Ge compositions than the sample average, and thus smaller bandgaps, may exist. This causes a narrowing of the bandgap. We suspect that by reducing the Ge composition (bottom), these high Ge composition regions are still present, but simply reduced in density.

5.5. Summary

In this chapter, we have investigated the room temperature photoluminescence observed in our

III-V/IV alloys grown by MOCVD. The luminescence peaks of (GaAs)1-x(Ge2)x were not consistent with the bandgap-composition correlation in past studies; specifically, the highest energy transition observed was 0.8 eV even though it is expected to be larger as the composition of Ge in the alloys is reduced from 20%. This 0.8 eV transition is unlikely to be caused by Ge nanocrystals or similar structures; we observed smaller transitions at 23% and 29% Ge compositions, which would not be consistent with the direct gap transition in Ge. We suspect that the heterogeneous nature of the alloys plays a role in this observed bandgap narrowing.

Composition fluctuations result in localized regions with higher Ge composition, and thus smaller energy gap, than the alloy average. Carriers, specifically holes, will relax to these high

102

Ge regions and then recombine, resulting in a lower energy transition than expected. In addition, due to the close alignment of the conduction bands of (GaAs)1-x(Ge2)x alloys it is possible that recombination occurs at the interfaces, between spatially separated electrons and holes, rather than by local recombination.

By further alloying (GaAs)1-x(Ge2)x with elements such as Si and In, we observe shifts in the energy transition. In particular, alloying with a small amount of Si results in a larger energy transition than 0.8 eV, which was not successfully achieved by lowering the Ge composition.

This suggests the possibility of engineering a range of low bandgaps using III-V/IV alloys instead of attempting to grow the homogeneous metastable (GaAs)1-x(Ge2)x alloy using MOCVD, which has thus far been unsuccessful.

103

Chapter 6: Conclusions and Future Work

104

6.1. Summary of results

In this work, we studied the thermoelectric and optical properties of III-V/IV heterovalent structures grown using MOCVD. For thermoelectrics, we speculated that III-V/IV nanostructures have the potential to be better materials than isovalent nanostructures due to increased phonon scattering from heterovalent interfaces. Observation of room temperature photoluminescence in our III-V/IV alloy structures can have significant implications for low bandgap optical devices at the GaAs lattice constant. The following sections summarizes the results of our investigations.

6.1.1. MOCVD of GaAs/Ge superlattices and (GaAs)1-x(Ge2)x alloys

We investigated the growth of GaAs/Ge superlattices and alloys using MOCVD, surveying the effects of a variety of different growth conditions, to gain insight in epitaxy involving heterovalent materials. Initial attempts at growing GaAs/Ge superlattices using standard conditions for typical GaAs growth and Ge growth (650°C and 100 Torr system pressure) resulted in morphologies that are highly irregular and defective. A qualitative improvement in superlattice morphology was seen when the growth rate of Ge was increased as a result of increasing the system pressure to 250 Torr and GaAs surfaces were annealed to achieve a high surface Ga-to-As ratio prior to depositing Ge. This improvement can be attributed to establishing pseudo-layer-by-layer growth, in which a high density of islands is nucleated and coalesce into a film layer relatively quickly. As the thickness of the layers was decreased below the critical thickness needed for complete coalescence of islands, the morphology deviated towards a more randomized nanostructure.

105

Composition fluctuations were observed in (GaAs)1-x(Ge2)x alloys. The intensity of the fluctuations qualitatively was higher in the alloys grown at a slow rate or high temperature; the morphologies of these alloys exhibited vertical striations. In contrast, alloys grown at a fast rate and lower temperatures resulted in speckled morphologies. These results are typical of a material that thermodynamically favors phase separation. By reducing the ability of surface atoms to migrate to their thermodynamically favorable locations, a structure with a lower degree of phase separation can be achieved.

6.1.2. Thermoelectric properties of III-V/IV structures

We observed a significant decrease in the thermal conductivities of GaAs/Ge nanostructures and alloys relative to bulk GaAs and bulk Ge. The magnitude of the decrease is not as large as the decrease observed in GaAs/AlAs and Si/Ge superlattices with respect to their bulk constituents; however, it is known that the decrease observed in those materials are primarily due to phonon scattering from atomic mass differences. GaAs and Ge have very similar masses, limiting the effect of mass difference phonon scattering. These results have two implications – first, a further reduction in thermal conductivity should be seen in structures containing additional elements, and second, the observed reduction in the GaAs/Ge structures from their bulk thermal conductivities can be attributed to the presence of the heterovalent interfaces. We have confirmed the first implication through investigating an (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy, which had a thermal conductivity of 4.3 W/m-K, significantly lower than that of the GaAs/Ge structures. The second implication suggests that our original hypothesis is correct, in which a

III-V/IV structure should have lower thermal conductivities than a homovalent structure. This

106 appears to be the case for the (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy, which is unlikely to be at the optimal composition yet already has a thermal conductivity comparable to some state-of-the-art thermoelectric materials.

The electron mobilities of the GaAs/Ge structures were very similar to bulk GaAs and bulk Ge, indicating that electron scattering due to GaAs/Ge interfaces was minimal. This result would be consistent with a close band alignment of the GaAs and Ge conduction bands. More importantly, this suggests that the mere presence of heterovalent interfaces does not degrade the electrical conductivity but rather the differences with the material constituents themselves do. Thus heterovalent interfaces are desirable for thermoelectrics as they contribute to a reduction in thermal conductivity but not electrical conductivity.

6.1.3. Optical properties of III-V/IV alloys

The room temperature photoluminescence of various compositions of (GaAs)1-x(Ge2)x alloys grown by MOCVD was investigated. 0.72 eV and 0.7 eV energy transitions were seen in alloys with 23% Ge and 29% Ge, respectively. However, we observed a maximum energy transition of

0.8 eV, even as the Ge composition was decreased from 19% to 6%. This result is not consistent with the (GaAs)1-x(Ge2)x bandgap-composition correlation in literature, which indicate a consistently increasing bandgap as Ge composition is decreased from 30% to 0%. The composition fluctuations observed in the alloy morphologies is likely the cause of this bandgap narrowing. By alloying with Si, we observed a larger energy transition than 0.8 eV. We speculate that this occurs due to the higher bandgap of SiGe relative to Ge.

107

6.2. Future directions

Several areas for further work are described below.

a. Doping of III-V/IV structures – As we have noted in chapter 4, peak zT for thermoelectric

materials generally occurs at carrier concentrations in the 1019 – 1020 cm-3 range. Doping

of III-V/IV structures is complicated by the fact that the typical dopants for III-V

materials are not the same as the dopants for group IV materials. Determining how to

dope the structure and control the concentration is a necessary step in order to truly study

the viability of the materials for thermoelectrics.

b. Uniformity of carrier type within III-V/IV structures – The Seebeck coefficient is

maximized when only one carrier type is present within the structure. Given the

unavoidable issue of unintentional doping in III-V/IV structures, it is possible that both p-

type and n-type regions exist within the structure, dependent on the method of growth.

This issue must be considered in investigations to optimize the Seebeck coefficient.

c. Optimization of III-V/IV compositions for thermoelectric properties – The thermoelectric

properties of the (In0.1Ga0.9As)0.84(Si0.1Ge0.9)0.16 alloy we investigated is unlikely to be at

the limit of what the III-V/IV system can occur. Further studies can be done in the large

design space available to optimize not only the various compositions (constituents and

overall) but also potentially alloying with other elements.

108 d. Identification of loss mechanisms in the photoluminescence of III-V/IV alloys – A better

understanding of the loss mechanisms and how to eliminate or minimize them is needed

in order to evaluate the possibility of III-V/IV alloys as viable optical materials.

109

Appendix A

(GaAs)1-x(Ge2)x Alloy TEM images:

Figure A1. (002) Darkfield XTEM image of the (GaAs)0.81(Ge2)0.19 alloy. The sample was grown by codeposition of GaAs and Ge at 650°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX.

110

Figure A2. (002) Darkfield XTEM image of the (GaAs)0.88(Ge2)0.12 alloy. The sample was grown by codeposition of GaAs and Ge at 575°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX.

111

Figure A3. (002) Darkfield XTEM image of the (GaAs)0.94(Ge2)0.06 alloy. The sample was grown by codeposition of GaAs and Ge at 575°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX.

112

Figure A4. (002) Darkfield XTEM image of the (GaAs)0.77(Ge2)0.23 alloy. The sample was grown by codeposition of GaAs and Ge at 650°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX.

113

(GaAs)0.8(Si0.1Ge0.9)0.2 Alloy TEM image:

Figure A5. (002) Darkfield XTEM image of the (GaAs)0.8(Si0.1Ge0.9)0.2 alloy. The sample was grown by codeposition of GaAs and SiGe at 650°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX.

114

(In0.15Ga0.85As)0.8(Si0.1Ge0.9)0.2 Alloy TEM image:

Figure A6. (002) Darkfield XTEM image of the (In0.15Ga0.85As)0.8(Si0.1Ge0.9)0.2 alloy. The sample was grown by codeposition of InGaAs and SiGe at 650°C and 250 Torr on a semi-insulating (001) GaAs substrate. A ~40nm GaAs homoepitaxial layer was deposited on the substrate to ensure a high quality surface. Prior to deposition of the alloy, a 20 second N2 anneal was conducted to reduce the amount of As on the surface. Compositions were determined by EDX.

115

References

1 J.R. Lang, J. Faucher, S. Tomasulo, K. Nay Yaung, and M. Larry Lee, Appl. Phys. Lett. 103, 092102 (2013).

2 O. Romanyuk, T. Hannappel, and F. Grosse, Phys. Rev. B 88, 115312 (2013).

3 K.E. Newman, J.D. Dow, B.A. Bunker, L.L. Abels, P.M. Raccah, S. Ugur, D.Z. Xue, and A. Kobayashi, Phys. Rev. B 39, 657 (1989).

4 J.E. Greene, J. Vac. Sci. Technol. B 1, 229 (1983).

5 K.E. Newman, A. Lastras-Martinez, B. Kramer, S.A. Barnett, M.A. Ray, J.D. Dow, J.E. Greene, and P.M. Raccah, Phys. Rev. Lett. 50, 1466 (1983).

6 K.E. Newman and J.D. Dow, Phys. Rev. B 27, 7495 (1983).

7 I. Banerjee, D.W. Chung, and H. Kroemer, Appl. Phys. Lett. 46, 494 (1985).

8 B. Koiller, M.A. Davidovich, and R. Osório, Solid State Commun. 55, 861 (1985).

9 A.G. Norman, J.M. Olson, J.F. Geisz, H.R. Moutinho, A. Mason, M.M. Al-Jassim, and S.M. Vernon, Appl. Phys. Lett. 74, 1382 (1999).

10 J. Fan, L. Ouyang, X. Liu, J.K. Furdyna, D.J. Smith, and Y.-H. Zhang, J. Cryst. Growth 371, 122 (2013).

11 J. Fan, X. Liu, J.K. Furdyna, and Y.-H. Zhang, Appl. Phys. Lett. 101, 121909 (2012).

12 H.-X. Deng, B. Huang, and S.-H. Wei, Comput. Mater. Sci. 98, 340 (2015).

13 M.A. Haase, J. Qiu, J.M. DePuydt, and H. Cheng, Appl. Phys. Lett. 59, 1272 (1991).

14 P. Sharma, T. Milakovich, M.T. Bulsara, and E.A. Fitzgerald, ECS Trans. 50, 333 (2013).

15 T. Milakovich, R. Shah, S. Hadi, M. Bulsara, A. Nayfeh, and E. Fitzgerald, in 2015 IEEE 42nd Photovolt. Spec. Conf. PVSC (2015), pp. 1–4.

16 M. Diaz, L. Wang, D. Li, X. Zhao, B. Conrad, A. Soeriyadi, A. Gerger, A. Lochtefeld, C. Ebert, R. Opila, I. Perez-Wurfl, and A. Barnett, Sol. Energy Mater. Sol. Cells 143, 113 (2015).

17 S.A. Barnett, M.A. Ray, A. Lastras, B. Kramer, J.E. Greene, P.M. Raccah, and L.L. Abels, Electron. Lett. 18, 891 (1982).

116

18 K. Cadien, A. Eltoukhy, and J. Greene, Vacuum 31, 253 (1981).

19 J.E. Greene, S.A. Barnett, K.C. Cadien, and M.A. Ray, J. Cryst. Growth 56, 389 (1982).

20 S.M. Vernon, M.M. Sanfacon, and R.K. Ahrenkiel, J. Electron. Mater. 23, 147 (1994).

21 Woodall Research Group, UC Davis, Bandgap Vs Lattice Constant Diagr. (n.d.).

22 I. Ansara, SpringerMaterials MSI Eureka (1994).

23 Growth Modes - Retrieved from Wikimedia Commons. Public Domain (2012).

24 D.K. Biegelsen, R.D. Bringans, and J.E. Northrup, Mater. Sci. Eng. B 14, 317 (1992).

25 S.M. (Steve M. Ting, Monolithic Integration of III-V Semiconductor Materials and Devices with Silicon, Thesis, Massachusetts Institute of Technology, 1999.

26 D.B. Holt, J. Phys. Chem. Solids 30, 1297 (1969).

27 H. Kroemer, J. Cryst. Growth 81, 193 (1987).

28 Y. Li, G. Salviati, M.M.G. Bongers, L. Lazzarini, L. Nasi, and L.J. Giling, J. Cryst. Growth 163, 195 (1996).

29 D.K. Biegelsen, R.D. Bringans, J.E. Northrup, and L.-E. Swartz, Phys. Rev. B 41, 5701 (1990).

30 Y. Bai, K.E. Lee, C. Cheng, M.L. Lee, and E.A. Fitzgerald, J. Appl. Phys. 104, 084518 (2008).

31 X.-S. Wang, K.W. Self, and W.H. Weinberg, J. Vac. Sci. Technol. A 12, 1920 (1994).

32 I. Banerjee, H. Kroemer, and D.W. Chung, Mater. Lett. 2, 189 (1984).

33 G.J. Snyder and E.S. Toberer, Nat. Mater. 7, 105 (2008).

34 L.D. Hicks and M.S. Dresselhaus, Phys. Rev. B 47, 12727 (1993).

35 L.D. Hicks, T.C. Harman, and M.S. Dresselhaus, Appl. Phys. Lett. 63, 3230 (1993).

36 G. Chen, J. Heat Transf. 119, 220 (1997).

37 G. Chen and M. Neagu, Appl. Phys. Lett. 71, 2761 (1997).

38 B. Poudel, Q. Hao, Y. Ma, Y. Lan, A. Minnich, B. Yu, X. Yan, D. Wang, A. Muto, D. Vashaee, X. Chen, J. Liu, M.S. Dresselhaus, G. Chen, and Z. Ren, Science 320, 634 (2008).

117

39 D. Li, Y. Wu, P. Kim, L. Shi, P. Yang, and A. Majumdar, Appl. Phys. Lett. 83, 2934 (2003).

40 A.I. Hochbaum, R. Chen, R.D. Delgado, W. Liang, E.C. Garnett, M. Najarian, A. Majumdar, and P. Yang, Nature 451, 163 (2008).

41 A.I. Boukai, Y. Bunimovich, J. Tahir-Kheli, J.-K. Yu, W.A. Goddard Iii, and J.R. Heath, Nature 451, 168 (2008).

42 Y. Hu, L. Zeng, A.J. Minnich, M.S. Dresselhaus, and G. Chen, Nat. Nanotechnol. 10, 701 (2015).

43 L. Zeng, K.C. Collins, Y. Hu, M.N. Luckyanova, A.A. Maznev, S. Huberman, V. Chiloyan, J. Zhou, X. Huang, K.A. Nelson, and G. Chen, Sci. Rep. 5, 17131 (2015).

44 L. Zeng, V. Chiloyan, S. Huberman, A.A. Maznev, J.-P.M. Peraud, N.G. Hadjiconstantinou, K.A. Nelson, and G. Chen, Appl. Phys. Lett. 108, 063107 (2016).

45 D.J. Friedman, J.F. Geisz, S.R. Kurtz, and J.M. Olson, J. Cryst. Growth 195, 409 (1998).

46 T.W. Kim, T.J. Garrod, K. Kim, J.J. Lee, S.D. LaLumondiere, Y. Sin, W.T. Lotshaw, S.C. Moss, T.F. Kuech, R. Tatavarti, and L.J. Mawst, Appl. Phys. Lett. 100, 121120 (2012).

47 J.F. Geisz, S. Kurtz, M.W. Wanlass, J.S. Ward, A. Duda, D.J. Friedman, J.M. Olson, W.E. McMahon, T.E. Moriarty, and J.T. Kiehl, Appl. Phys. Lett. 91, 023502 (2007).

48 R.E. Camacho-Aguilera, Y. Cai, N. Patel, J.T. Bessette, M. Romagnoli, L.C. Kimerling, and J. Michel, Opt. Express 20, 11316 (2012).

49 X. Sun, J. Liu, L.C. Kimerling, and J. Michel, IEEE J. Sel. Top. Quantum Electron. 16, 124 (2010).

50 C.J. Pinzone, MOCVD Apparatus - Retrieved from Wikimedia Commons. Public Domain (n.d.).

51 R.V. Petrova, Quantitative High-Angle Annular Dark Field Scanning Transmission Electron Microscopy for Materials Science, University of Central Florida Orlando, Florida, 2006.

52 A.J. Schmidt, X. Chen, and G. Chen, Rev. Sci. Instrum. 79, 114902 (2008).

53 D.G. Cahill, Rev. Sci. Instrum. 75, 5119 (2004).

54 A.J. Schmidt, Optical Characterization of Thermal Transport from the Nanoscale to the Macroscale, Thesis, Massachusetts Institute of Technology, 2008.

55 L. van der Pauw, Philips Res Rep 13, 1 (1958).

118

56 M.N. Luckyanova, J. Garg, K. Esfarjani, A. Jandl, M.T. Bulsara, A.J. Schmidt, A.J. Minnich, S. Chen, M.S. Dresselhaus, Z. Ren, E.A. Fitzgerald, and G. Chen, Science 338, 936 (2012).

57 C.A. Larsen, S.H. Li, N.I. Buchan, G.B. Stringfellow, and D.W. Brown, J. Cryst. Growth 102, 126 (1990).

58 H. Heinecke, E. Veuhoff, N. Pütz, M. Heyen, and P. Balk, J. Electron. Mater. 13, 815 (1984).

59 G.B. Stringfellow, Organometallic Vapor-Phase Epitaxy: Theory and Practice (Academic Press, 1999).

60 J.E. Ayers and S.K. Ghandhi, J. Cryst. Growth 89, 371 (1988).

61 J. Ishizaki, K. Ohkuri, and T. Fukui, Jpn. J. Appl. Phys. 35, 1280 (1996).

62 M. Kasu and N. Kobayashi, J. Appl. Phys. 78, 3026 (1995).

63 T. Fukui and H. Saito, Jpn. J. Appl. Phys. 29, L483 (1990).

64 M. Pristovsek, T. Trepk, M. Klein, J.-T. Zettler, and W. Richter, J. Appl. Phys. 87, 1245 (2000).

65 P.M. Petroff, A.C. Gossard, A. Savage, and W. Wiegmann, J. Cryst. Growth 46, 172 (1979).

66 K. Mukherjee, D.A. Beaton, T. Christian, E.J. Jones, K. Alberi, A. Mascarenhas, M.T. Bulsara, and E.A. Fitzgerald, J. Appl. Phys. 113, 183518 (2013).

67 G.B. Stringfellow, J. Cryst. Growth 65, 454 (1983).

68 S. Mahajan, Mater. Sci. Eng. B 30, 187 (1995).

69 S.N.G. Chu, S. Nakahara, K.E. Strege, and W.D.J. Jr, J. Appl. Phys. 57, 4610 (1985).

70 A. Diéguez, F. Peiró, A. Cornet, J.R. Morante, F. Alsina, and J. Pascual, J. Appl. Phys. 80, 3798 (1996).

71 G. Giorgi, M.V. Schilfgaarde, A. Korkin, and K. Yamashita, Nanoscale Res. Lett. 5, 469 (2010).

72 B.-L. Gu, J. Ni, and J.-L. Zhu, Phys. Rev. B 45, 4071 (1992).

73 T. Yao, Appl. Phys. Lett. 51, 1798 (1987).

74 W.S. Capinski, H.J. Maris, T. Ruf, M. Cardona, K. Ploog, and D.S. Katzer, Phys. Rev. B 59, 8105 (1999).

119

75 S.-M. Lee, D.G. Cahill, and R. Venkatasubramanian, Appl. Phys. Lett. 70, 2957 (1997).

76 T. Borca-Tasciuc, W. Liu, J. Liu, T. Zeng, D.W. Song, C.D. Moore, G. Chen, K.L. Wang, M.S. Goorsky, T. Radetic, R. Gronsky, T. Koga, and M.S. Dresselhaus, Superlattices Microstruct. 28, 199 (2000).

77 M.N. Luckyanova, J.A. Johnson, A.A. Maznev, J. Garg, A. Jandl, M.T. Bulsara, E.A. Fitzgerald, K.A. Nelson, and G. Chen, Nano Lett. 13, 3973 (2013).

78 W.L. Liu, T. Borca-Tasciuc, G. Chen, J.L. Liu, and K.L. Wang, J. Nanosci. Nanotechnol. 1, 39 (2001).

79 E.T. Swartz and R.O. Pohl, Rev. Mod. Phys. 61, 605 (1989).

80 Y. Bai, M.T. Bulsara, and E.A. Fitzgerald, J. Appl. Phys. 111, 013502 (2012).

81 D.L. Rode, Semicond. Semimet. 10, 1 (1975).

82 V.I. Fistul, M.I. Iglitsyn, and E.M. Omelyanovskii, Mobility of Electrons in Strongly Doped with (AMER INST CIRCULATION FULFILLMENT DIV, 500 SUNNYSIDE BLVD, WOODBURY, NY 11797-2999, 1962).

83 G.A. Baraff, J.A. Appelbaum, and D.R. Hamann, J. Vac. Sci. Technol. 14, 999 (1977).

84 C.G. Van de Walle and J. Neugebauer, Nature 423, 626 (2003).

85 C. Wood, Rep. Prog. Phys. 51, 459 (1988).

86 F.D. Rosi, Solid-State Electron. 11, 833 (1968).

87 S. Adachi, J. Appl. Phys. 54, 1844 (1983).

88 H. Stohr and Z. W. Klemm, Anorg. Allgem. Chem. 241, 305 (1954).

89 M.A. Afromowitz, J. Appl. Phys. 44, 1292 (1973).

90 D. Chattopadhyay, S.K. Sutradhar, and B.R. Nag, J. Phys. C Solid State Phys. 14, 891 (1981).

91NSM Arch. - Silicon Germanium SiGe - Mobil. Hall Eff. (n.d.).

92 M. Tabbal, T. Kim, D.N. Woolf, B. Shin, and M.J. Aziz, Appl. Phys. A 98, 589 (2010).

93 A.G. Norman, T.-Y. Seong, I.T. Ferguson, G.R. Booker, and B.A. Joyce, Semicond. Sci. Technol. 8, S9 (1993).

120

94 S.W. Jun, T. Seong, J.H. Lee, and B. Lee, Appl. Phys. Lett. 68, 3443 (1996).

95 K.A. Mäder and A. Zunger, Phys. Rev. B 51, 10462 (1995).

96 S. Takeoka, M. Fujii, S. Hayashi, and K. Yamamoto, Appl. Phys. Lett. 74, 1558 (1999).

97 R. Hayashi, M. Yamamoto, K. Tsunetomo, K. Kohno, Y. Osaka, and H. Nasu, Jpn. J. Appl. Phys. 29, 756 (1990).

98 C. Bostedt, T. van Buuren, T.M. Willey, N. Franco, L.J. Terminello, C. Heske, and T. Möller, Appl. Phys. Lett. 84, 4056 (2004).

99 Y. Maeda, Phys. Rev. B 51, 1658 (1995).

100 Y.M. Niquet, G. Allan, C. Delerue, and M. Lannoo, Appl. Phys. Lett. 77, 1182 (2000).

121