Epitaxial Lateral Overgrowth of

Indium Phosphide and Its Application in Heteroepitaxy

Doctoral Thesis by

Yan-Ting Sun

Laboratory of Materials and Semiconductor Physics Department of Microelectronics and Information Technology

Royal Institute of Technology (KTH) Electrum 229, SE-164 40 Kista, Sweden

Stockholm 2003

Epitaxial Lateral Overgrowth of Indium Phosphide and Its Application in Heteroepitaxy A dissertation submitted to the Royal Institute of Technology, Stockholm, Sweden, in partial fulfillment of the requirements for the degree of Doctor of Philosophy. TRITA-HMA REPORT 2003:4 ISSN 1404-0379 ISRN KTH/HMA/FR-03/4-SE Yan-Ting Sun, November 2003 Printed by Universitetsservice US AB, Stockholm 2003

i

Yan-Ting Sun Epitaxial lateral overgrowth of indium phosphide and its application in heteroepitaxy Royal Institute of Technology (KTH), Department of Microelectronics and Information Technology, Kista, Sweden TRITA-HMA Report 2003:4; ISSN 1404-0379; ISRN KTH/HMA/FR-03/4-SE Abstract

Monolithic integration of optoelectronics on silicon is a dream. This thesis deals with the studies on the heteroepitaxy of indium phosphide on silicon substrate towards making that dream come true. Materials growth issues, characterization and defect identification are addressed.

Epitaxial lateral overgrowth (ELOG) technique is used to grow high quality epitaxial indium phosphide on a silicon substrate provided with a low quality indium phosphide seed layer. Hydride vapor phase epitaxy is used for ELOG. The growth parameters were optimized first by carrying out ELOG experiments on an InP substrate. The lateral growth rate is strongly dependent on the orientation of the openings, the highest growth rate being for the openings oriented at 30o and 60o off [110] directions. But the vertical growth rate is relatively unaffected by the opening orientation. The observation of an inhomogeneous and orientation dependent dopant distribution within the same layer has been explained by invoking the bonding configurations exposed to the incorporating dopant atoms in the different emerging planes.

When ELOG of InP is conducted on InP/Si, unlike that on InP substrates, the lateral growth is not symmetric on both sides due to the propagation of defects from the seed layer. For example, a higher concentration of threading dislocations intersecting the surface of the {111}A emerging planes would cause a higher growth rate of these planes. The growth rate of {111}A planes with respect to the others can also be caused by the vapor phase supersaturation as predicated by Burton-Cabrera-Frank model. The determined dislocation density in the ELOG InP on InP/Si is ∼4×107cm-2, which is nearly two magnitude lower than in the seed layer (∼4×109 cm-2). If the seed layer is of a better quality, the ELOG layer will also be. Combination of high resolution x-ray diffraction reciprocal lattice mapping and low temperature photoluminescence indicates that the ELOG InP layer with high aspect ratio is nearly strain-free.

When ELOG of sulfur doped InP is conducted on ring shaped openings on InP/Si substrate instead of stripe openings, octahedral shaped ELOG InP templates with smooth surface are formed. Strain compensated InGaAsP 6 periods multi-quantum wells (MQW) at 1.5 µm wavelength (target value) were grown on these templates by metalorganic vapor phase epitaxy. RT-PL is indicative of a good quality ELOG layers. Optimized ELOG on ring openings may become very attractive for heteroepitaxy of III-V compounds on silicon.

As an extension of ELOG of InP on InP/Si, growth of InP is also conducted on planar Focused-Ion-Beam (FIB)-modified (001) GaAs substrate. The impacts of the III/V ratio, crystallographic orientation of implanted lines and implantation dose were explored. The choice of suitable growth conditions makes it possible to obtain continuous InP wires aligned in all possible directions.

ii

iii

Acknowledgments First of all I would like to thank Prof. Gunnar Landgren for employing me as a Ph.D. student at KTH and creating an attractive and friendly atmosphere to work in. Especially, I would like to express my deepest gratitude to my supervisor, Prof. Sebastian Lourdudoss for his guidance, enthusiasm and continuous involvement in the details of this work. I would also like to thank Gunnar Andersson for his valuable technical support on the HVPE system. I have learnt a lot from the members of our HMA group. I would like to thank those who contributed to the realization of the work presented here, namely: Dr. Denis Jahan, for guiding me through the puzzle of HVPE system and spending time with me at the initial stage of the heteroepitaxy of III-V/Si project; Dr. Egbert Rodríguez Messmer, for helping me to understand the crystallographic model of InP growth; Dr. Jérôme Napierala, for the constructive discussion on supersaturation; Dr. Srinivasan Anand, Dr. Kestutis Makuys and Olivier Douheret, for their efforts and expertise on SCM characterization and analysis; Dr. Krishnan Baskar, for PL characterization and analysis; Dr. David Söderström, for teaching me the processing in clean room; Dr. Henry Radamson, for helping with XRD measurements. I would also like to express my gratitude to Professor Jouni Ahopelto (VTT Microelectronics, Finland), Professor Juan Jiménez, and Dr. Manuel Avella for their valuable collaboration on materials characterization. Dr. Anand and Baskar had valuable comments on the thesis. Many thanks to Agneta Odéen, Margreth Hellberg and Rose-Marie Lövenstig for their help on administrative issues. I am very grateful to the rest of the members of HMA during my time here for creating a friendly and helpful atmosphere: Dr. Carl Asplund, Dr. Carlos Angulo Barrios, Audrey Berrier, Jesper Berggren, Roberta Campi, Dr. Carl-Fredrik Carlström, Nicolae Chitica, Peter Goldmann, Dr. Mattias Hammar, Julius Hållstedt, Dr. Dietmar Keiter, Cyril Menon, Dr. Sebastian Mogg, Mikael Mulot, Fredrik Olsson, Amit Patel, Glenn Plaine, Dr. Fredrik Salomonsson, Dr. Martin Strassner, Petrus Sundberg, Rickard Marcks von Würtemberg, and Gael Mion. I am grateful to Doss’ family: Alphonsa, Cecilia, Pierre and Ilango, for very nice dinner, unforgettable New Year eves and especially, party for my marriage. Finally, I would like to thank my parents and my wife for their encouragement and constant support. Thank you all. Yan-Ting Sun Stockholm, November 2003

iv

v

List of papers Publications included in the thesis A Y.T. Sun, E. Rodríguez Messmer, D. Söderström, D. Jahan and S. Lourdudoss,

“Temporally resolved selective area growth of InP in the openings off-oriented from [110] direction”, Journal of Crystal Growth, Vol. 225, 9-15, 2001.

B Y.T. Sun, S. Anand and S. Lourdudoss “Crystallographic orientation dependence of

impurity incorporation during epitaxial lateral overgrowth of InP”, Journal of Crystal Growth, Vol. 237-239, 1418-1422, 2002.

C S. Anand, Y.T Sun, S. Lourdudoss, M.W. Xu, and W. Vandervorst, “High resolution

electrical characterization of laterally overgrown epitaxial InP”, 15th International conference on indium phosphide and related materials, IEEE, Santa Barbara, CA, USA, 2003, 563-566.

D Y.T. Sun, J. Napierala, and S. Lourdudoss, “Selective area growth of InP on InP

precoated silicon substrate by hydride vapor phase epitaxy”, 14th International conference on indium phosphide and related materials, IEEE, Stockholm, Sweden, 2002, 339-342.

E Y.T. Sun and S. Lourdudoss, “Effect of growth conditions on epitaxial lateral

overgrowth of InP on InP/Si (001) substrate by hydride vapor phase epitaxy”, Proceedings of SPIE Vol. 4997 Photonics Packaging and Integration III, edited by Randy A. Heyler, David J. Bobbins, Ghassan E. Jabbour, (SPIE, Bellingham, WA, 2003) 221-231.

F Y.T. Sun, K. Baskar, and S. Lourdudoss, “Thermal strain in indium phosphide on

silicon obtained by epitaxial lateral overgrowth”, Journal of Applied Physics, Vol. 94, 2746-2748, 2003.

G Y.T Sun, S. Lourdudoss, M. Avella, and J. Jiménez, “Sulfur doped indium phosphide

on silicon substrate grown by epitaxial lateral overgrowth”, manuscript. H Y.T Sun, K. Baskar, J. Berggren and S. Lourdudoss, “InGaAsP multi-quantum wells at

1.5 µm wavelength grown on indium phosphide templates on silicon”, 15th International conference on indium phosphide and related materials, IEEE, Santa Barbara, CA, USA, 2003, 277-280.

I Y.T. Sun, E. Rodríguez Messmer, S. Lourdudoss, J. Ahopelto, S. Rennon, J. P.

Reithmaier, and A. Forchel, “Selective growth of InP on focused-ion-beam-modified GaAs surface by hydride vapor phase epitaxy”, Applied Physics Letters, Vol. 79, 1885-1887, 2001.

vi

Other relevant publications not included in the thesis 1. F. Olsson, G. Mion, Y.T. Sun, P. Sundgren, K. Baskar, G. Salviati, M. Hammar and S.

Lourdudoss, “Selective Area Growth of GaInNAs/GaAs by MOVPE”, Presented on 5th International Workshop on Epitaxial Semiconductors on Patterned Substrates and Novel Index Surfaces (ESPS-NIS) 2003

2. S. Lourdudoss, D. Söderström, C. Angulo Barrios, Y.T. Sun and E. Rodríguez

Messmer, “Semi-Insulating Epitaxial Layers for Optoelectronic Devices” (Invited paper), IEEE International conference on semi-insulating and conducting compounds, Canberra, Australia, 2000

3. S. Lourdudoss, O. Kjebon, Y.T Sun, R. Schatz, D. Söderström, C. Angulo Barrios, and

E. Rodríguez Messmer, “III-V Materials Growth by Hydride VPE for High Frequency Optoelectronic Devices” (Invited paper), 2002 IEEE/LEOS Annual Meeting Conference Proceedings, The 15th Annual Meeting of the Lasers & Electro-Optics Society Glasgow, Scotland

4. Y.T. Sun, E. Rodríguez Messmer, D. Söderström, D. Jahan and S. Lourdudoss, “Temporally Resolved Growth of InP in the Openings Off-Oriented From [110] Direction”, 2000 International Conference on Indium Phosphide and Related Materials, IEEE, Williamsburg, USA, 2000.

vii

Acronyms AFM Atomic Force Microscopy AP-HVPE Atmosphere Pressure Hydride Vapor Phase Epitaxy BCF model Burton-Cabrera-Frank model CL Cathodoluminescence CW Continuous Wave DHS Double Heterostructure DLD Dark Line Defect DSD Dark Spot Defect ELOG Epitaxial Lateral Overgrowth EPD Etch Pit Density FIB Focused Ion Beam FWHM Full Width at Half Maximum HBT Heterojunction Bipolar Transistor HRXRD High Resolution X-Ray Diffraction LAIMCE Low Angle Incidence Microchannel Epitaxy LD Laser Diode LED Light Emitting Diode LPE Liquid Phase Epitaxy LP-HVPE Low Pressure Hydride Vapor Phase Epitaxy MBE Molecular Beam Epitaxy MOEMS Micro-Opto-Electro-Mechanical Systems MIS Metal Insulator semiconductor MOVPE Metal Organic Vapor Phase Epitaxy MQW Multi Quantum Well OEIC Optoelectronic Integrated Circuit PECVD Plasma Enhanced Chemical Vapor Deposition PIC Photonic Integrated Circuit PL Photoluminescence RC Rocking Curve RLM Reciprocal Lattice Mapping SAG Selective Area Growth SCM Scanning Capacitance Microscopy SEM Scanning Electron Microscopy SF Stacking Fault SSRM Scanning Spreading Resistance Microscopy TAD Triple-Axis Diffractometry TD Threading Dislocation

viii

ix

Table of Contents Abstract ................................................................................................................................ i Acknowledgments.............................................................................................................. iii List of papers ...................................................................................................................... v Acronyms .......................................................................................................................... vii 1 Introduction.................................................................................................................. 1 2 Background .................................................................................................................. 6

2.1 Growth techniques ............................................................................................................................... 6 2.1.1 Hydride vapor phase epitaxy (HVPE)......................................................................................... 6 2.1.2 Other growth techniques.............................................................................................................. 7

2.2 Heteroepitaxy by ELOG...................................................................................................................... 8 2.3 Heteroepitaxy on FIB modified surface .............................................................................................. 9 2.4 Crystallographic structures of zinc blende .......................................................................................... 9

2.4.1 {001} planes ..............................................................................................................................10 2.4.2 {110} planes ..............................................................................................................................10 2.4.3 {111} planes ..............................................................................................................................11 2.4.4 Boundary plane in ELOG InP ...................................................................................................11

2.5 Defects in ELOG InP on InP/Si.........................................................................................................12 2.5.1 Impurity......................................................................................................................................12 2.5.2 Dislocations................................................................................................................................12 2.5.3 Planar defects .............................................................................................................................13 2.5.4 Thermal strain ............................................................................................................................13

2.6 Characterization techniques...............................................................................................................14 2.6.1 Photoluminescence ....................................................................................................................14 2.6.2 High resolution x-ray diffraction...............................................................................................15

2.6.2.1 Thermal strain ........................................................................................................................16 2.6.2.2 Dislocation density ................................................................................................................17

2.6.3 Cathodoluminescence................................................................................................................18 2.6.4 Techniques to characterize dopant incorporation .....................................................................19

3 Summary of results and discussions......................................................................... 21

3.1 Epitaxial lateral overgrowth (ELOG) of InP..................................................................................... 21 3.2 Heteroepitaxy of InP/Si by ELOG ....................................................................................................26

3.2.1 Asymmetric growth morphology ..............................................................................................26 3.2.2 Defect reduction.........................................................................................................................28

3.3 Thermal strain in InP ELOG on Si ....................................................................................................29 3.4 Effect of sulfur doping.......................................................................................................................31 3.5 InGaAsP MQW at 1.55 µm ...............................................................................................................33 3.6 Heteroepitaxy of InP on GaAs...........................................................................................................34

4 Summary of appended papers .................................................................................. 38 5 Summary, conclusions and future work.................................................................. 42 Appendix: Properties of InP............................................................................................ 44 References ......................................................................................................................... 45

x

1

1 Introduction This thesis deals with the efforts made towards combining two dissimilar semiconductor materials to exploit their additive advantages. The ultimate aim is to achieve one or more devices on each of these semiconductors on a single platform and thereby monolithic integration. The considered combinations are indium phosphide on silicon and gallium arsenide on indium phosphide. The focus in this thesis is to obtain a good quality material. Hence growth aspects and materials characterization are given due importance. Semiconductors are the raw materials in the microelectronics industry. These are a family of materials that conduct electricity more than the insulators but less than the metals. The dawn of transistor made of semiconductors in the 1940’s revolutionized the world by changing our life-style drastically. Further developments led to the exploration of the use of semiconductors to exploit the interaction between light and electricity although the relation between them had been known for more than a century. This exploration gave birth to a new field called optoelectronics. While silicon, a dominant semiconductor in the microelectronics industry, was soon identified to be inadequate to efficiently exploit light-electricity (photon-electron) interaction, several other semiconductors, called compound semiconductors, e.g., indium phosphide, gallium arsenide etc. were found to be very suitable. This is because the latter ones are direct bandgap materials and silicon is an indirect bandgap material. Without going into the details of physics, one can state that in a direct bandgap material the photon-electron interaction is straightforward and quick and in an indirect bandgap material it is both slow and not straightforward. It is also worthwhile to point out that not all the compound semiconductors are direct bandgap materials. Optoelectronics combines the properties of light with the capabilities of microelectronics. Development of compound semiconductor materials technology, invention and demonstration of semiconductor lasers together with low-loss optical fibers have spurred the progress of optoelectronics industry. Since then, optoelectronics has ceased to make an impact nearly in every walk of life. Tele-communication and data-communication networks, mobile telephones, laptop computers, compact disc players, photography (including digital cameras and digital video display (DVD)), imaging, sensing, detection, distance measurement, bar code readers, high-definition television, health care, analysis and monitoring of exhaust gases, flat-panel displays and super-luminescent traffic lights are only a few examples. Any optoelectronic device (or any semiconductor device) consists of two or more of the suitable semiconductor layers grown on a substrate. A grown layer is called an epitaxial layer if it continues to follow exactly the same crystallographic lattice dimension of the substrate. Said differently, an epitaxial layer is lattice matched to the substrate. In practice, as long as the deviation of their lattice parameters is <0.05%, they are considered to be lattice-matched. On the other hand, a large mismatch induces stress, which in turn causes strain especially in the grown layer. This leads to the generation of several defects (basically bond-breaking) at the interface that can be mobile enough to reach the surface. In such case, the material becomes useless for device fabrication. Many of the compound semiconductors offer a great deal of flexibility in their properties for optoelectronics. Three options are known: (i) by mixing two or more compound semiconductors, one can generate a series of new alloys with different bandgaps and

2

lattice constants, see Figure 1; (ii) by adjacently placing two different compound semiconductors or their alloys, one can create, the so called heterojunctions and (iii) one can alter the band structure of these semiconductors by means of quantum structures and/or by introducing some strain (not large enough to generate defects!). These three options enable a good control on the charge carrier confinement, charge carrier transport/dynamics and photon-electron interaction. Hence both discrete and integrated electronic and optical devices can be fabricated from them. Several of the III-V compound semiconductors are superior to silicon in their electronic and optical properties.

Figure 1. Lattice constants and band gaps of semiconductors

III-VTechnology

h Optical capability

h High speedh High

frequencyh High voltage

Si technology

h Large wafer sizesh Low costh Volume

manufacturingh Mature

integration capability

h Memoriesh Microprocessors

Integrate the superior electrical and optical performance of III-V semiconductors with the mature Silicon-based technology to create a new industry of integrated semiconductor circuits

IIIIII--V/Si platform V/Si platform technologytechnology

Figure 2. III-V/Si platform technology offers a broad range of applications,

including microelectronics and optoelectronic products.

However III-V technology is less mature than silicon. Silicon is a single component system. III-V is a multi-component system. A multi-component system will offer several degrees of freedom but not without several constraints along with. In contrast, a single component system is easy to control. Besides, silicon is superior in terms of thermal and mechanical properties. The inherent silicon dioxide, an insulator to isolate integrated components is an asset to silicon. Hence it is believed that several of these factors will always maintain the silicon technology very attractive and contribute to even more maturity. Thus if both III-V and silicon technologies are combined on one single platform

3

one can reap the benefits of both which is the motivation of this thesis. Some of the benefits are indicated in Figure 2. Properties within the III-V semiconductors also vary. If for example InP is grown on GaAs, one can exploit the strengths of both of these and their related materials. Although InP and their related materials (meaning those that are lattice matched to InP) are more suitable for high speed electronics, GaAs and their related materials are normally used because of the higher maturity of the latter. InP and their related materials can produce long wavelength lasers and detectors; GaAs counterparts can produce short wavelength lasers. Heteroepitaxy denotes the epitaxial growth of a material that is different from the substrate on which it is grown. E.g., epitaxy of Ga0.47In0.53As on InP is considered as heteroepitaxy since both are lattice matched but different. But today heteroepitaxy is normally employed to denote the growth of a material whose lattice parameter is not the same as the substrate, e.g. InP on Si, GaN on sapphire, InP on GaAs etc. Throughout this thesis we attribute the latter meaning whenever the word heteroepitaxy is used. In heteroepitaxy one would expect materials full of defects. However, several defect engineering approaches have been taken to suppress these defects: 1) Growth of very thick layers1 – In this case the grown layer is so thick that the defects generated at the interface do not propagate up to the surface and hence a part of the layer at the top is free of defects. 2) Epitaxial lateral overgrowth (ELOG) approach2 – In this case, a bad quality layer is first grown on a substrate after which only a part of the surface is exposed for growth, the remaining surface being masked for direct growth. Under this circumstance, the growth starts on the open surface but when the layer reaches a particular thickness, the layer starts to grow laterally from the opening on the masked region. This growth retains the defects in and near opening region. 3) Compliant substrate approach3 – Compliant substrates offer a new approach to strain management in lattice-mismatched structures. The role of the compliant substrate is to reduce the strain in a mismatched over layer by sharing the strain via deformation of the substrate, or by nucleating and confining defects in the substrate. This is done either by direct bonding or twist bonding of an appropriate layer on a substrate. 4) Focused ion beam (FIB) approach4 – In this case a beam of ions is focused on to a substrate. The beam width is only a few tens of nanometers broad. With appropriate beam energy, the substrate is damaged, which then is used to grow our desired layer. When doing so, that part of the substrate “written” by FIB acts as the active sites for growth (nucleation sites). The layer grows normally only on these parts. When the thickness is sufficient, lateral growth takes place as in ELOG and the same defect suppressing mechanism as in ELOG is operative. This method is not flexible enough to be applicable for any desired combination. This thesis adopts the ELOG and FIB approaches. ELOG was used to grow indium phosphide on silicon and FIB was used to grow indium phosphide on gallium arsenide. The lattice mismatching in these systems are ∼8 % and ∼4%, respectively. Suppression of defects is the most difficult challenge in heteroepitaxy. Let us consider the case of heteroepitaxy of InP on Si. Large mismatches of lattice constant and thermal expansion coefficients between InP and Si could introduce high density of defects in the epitaxial InP layer on Si and cause the degradation of optoelectronic devices fabricated on these InP layers. The reliability of optoelectronic devices built up on heteroepitaxial InP layer on silicon substrate is critically affected by the defect density and residual strain in

4

epi-layer. The degradation mechanism of laser diodes (LDs) and light emitting diodes (LEDs) has been investigated on III-V substrate extensively.5, 6, 7 The performance of optoelectronic device can be affected by three degradation modes: rapid degradation, gradual degradation and catastrophic degradation. They arise mainly from two sources: gliding of misfit dislocation through the device structure and strain in epi-layer will accelerate the degradation process. The degradation mechanism of InP based laser diodes heteroepitaxially grown on Si substrates has been studied by Sasaki, et. al..8 Although no dark line defects (DLDs) develop during the aging tests, which are related to the rapid degradation caused by the climbing motion of dislocation, the density of the dark spot defects (DSDs) related to the threading dislocation in active layer gradually increase leading to the device degradation. The effect of thermal strain on LDs and LEDs degradation has also been investigated. The rapid degradation of GaAs lasers and LEDs grown on silicon substrates observed as the rapid increase in threshold and decrease in external differential quantum efficiency with time has been shown to be associated with the growth of non-radiative regions in the active layer.9 The growth of dislocations is enhanced by the non-radiative carrier recombination and is aided by the presence of a large tensile strain that results from the mismatch in the thermal expansion coefficients. Even a relatively small amount of strain has been shown to aid significantly the growth of non-radiative dislocations.10 The reduction of dislocation density and thermal strain intrinsically present in InP on silicon substrate is essential to obtain reliable operation of InP based optoelectronic devices on Si, comparable to that on InP substrate. Several approaches have been applied to improve the quality of InP on Si. For instance, by inserting low temperature GaAs11, ZnSe12 buffer layer, or super lattice13, 14, 15, performing thermal annealing16, 17, or growing InP on patterned Si substrates18, 19, rather low strain and defect density in epitaxial InP layers on Si can be obtained. However, the threading dislocation density in these InP/Si layers still exceeds 105/cm2. Further decrease of TD density in grown InP/Si layer is imperative for fabrication of optoelectronic device. Epitaxial lateral overgrowth (ELOG) could be the answer for the preparation of device quality InP/Si layers. ELOG has been used to fabricate novel electronic devices extensively because such structure can decease the parasitic capacitance and resistance in Si base CMOS devices.20,

21 It is also used in the fabrication of GaAs waveguide.22 The ability of ELOG to decrease the threading dislocation in heteroepitaxy structure of III-V on Si was first reported in 1980’s.23 It has been used to grow high quality GaN layers on sapphire to fabricate blue LDs with life time of 1000 hours under 30 mW continuous-wave operation at 60°C.24 Recently, InP/InGaAs heterojunction bipolar transistor (HBT) with regrown-base prepared by ELOG has been reported with improved high frequency performance.25 The aim of this thesis is to: (1) Investigate the optimum growth conditions for ELOG InP in LP-HVPE system. (2) Study the growth behavior of ELOG InP on InP/Si substrate and the formation of

defects (planar defects and threading dislocations) during the ELOG InP growth. (3) Explore the possibility to reduce the residual strain in epi-layer and to improve the

crystallographic quality of the ELOG InP on InP/Si. (4) Grow layers of quantum wells on ELOG InP on InP/Si templates and assess its

material quality.

5

(5) Make the ELOG growth mechanism applicable to the growth of InP on focused-ion-beam modified GaAs surface.

The structure of the thesis is as follows. In chapter 2, some background information on the crystallographic structure of III-V compound semiconductors, comparison of several growth techniques for ELOG, defect structures encountered in ELOG InP on InP/Si growth, and characterization techniques for heterostructures will be described. In chapter 3, the major results will be summarized and discussed. In chapter 4, summary of all the papers will be presented. Finally in chapter 5, summary, conclusion and suggestions for future work will be given.

6

2 Background In this section we will give a brief introduction to the growth techniques, ELOG and FIB modified surface for heteroepitaxy, certain details on zinc blende crystallographic structure relevant to ELOG, defects in ELOG InP on InP/Si and the characterization techniques used in the study (photoluminescence, high resolution X-ray diffraction, scanning capacitance microscopy, scanning spreading resistance microscopy).

2.1 Growth techniques Liquid phase epitaxy (LPE), vapor phase epitaxy (VPE), metalorganic vapor phase epitaxy (MOVPE) and molecular beam epitaxy (MBE) are basically the four techniques employed for the growth III-V semiconductors. Among them the first two are near equilibrium techniques and the latter two are far-equilibrium techniques. 2.1.1 Hydride vapor phase epitaxy (HVPE) In this work we have used HVPE to conduct ELOG and growth on FIB modified surface. Low pressure HVPE (LP-HVPE) was used in the former case and atmosphere pressure HVPE (AP-HVPE) in the latter case. The design of HVPE reactor has been presented before.26 The group-III sources are chlorides formed by flowing HCl through the molten metal, say indium to form InCl (Equation 1).

Equation 1 221 HInClHClIn +⇔+

The group-V sources are hydrides, i.e. PH3 and AsH3. Unlike in AP-HVPE, the general growth reaction of InP in LP-HVPE system involves PH3 instead of P2 and P4 (Equation 2) because the number of the collision between PH3 molecules is relatively small at low pressure and a substantial fraction of PH3 is not decomposed completely.27

Equation 2 23 HHClInPInClPH ++⇔+

The relative gas phase supersaturation of the reaction, γ can be defined as the state of advancement of the growth reaction with respect to equilibrium and expressed as:

Equation 3 [ ] [ ][ ] [ ] 1

2

3 −⋅⋅

⋅= TK

HHClPHInCl

γ

where [i] is the partial pressure of each gaseous species in reactor and KT is the reaction constant of the Equation 2. It is related to the partial pressures of gaseous species at equilibrium, [i]eq:

Equation 4 [ ] [ ][ ] [ ] eqeq

eqeq

T PHInCl

HHClK

3

2

⋅

⋅=

7

The deposition of InP can occur by either Burton-Cabrera-Frank (BCF) spiral growth28 or the two-dimensional (2D) nucleation.29 The latter mechanism applies to the growth of crystals from the vapor phase when the existing substrate surface is perfect, i.e. dislocation free. In this case the spontaneous formation of the first nuclei on the surface (that provides the necessary steps for the attachment of the freshly arriving adatoms) requires a critical supersaturation above which the 2D growth rate, R2D increases exponentially with the supersaturation γ according to:29

Equation 5

−∝

22'exp

TCR Dγ

where C’ is a constant. For the substrate containing high density dislocations, the surface steps, where the adatoms attach themselves, are provided by the emergence points of the dislocations having Burgers vector with a component normal to the surface. The relationship between the growth rate and the supersaturation is then given by BCF model:28

Equation 6

=

γγ

γγ 1

1

2

tanhCR BCF

where C and γ1 are constants for a given temperature. RBCF approximates to C(γ2/γ1) for low supersaturation (γ<γ1) and to Cγ for high supersaturation (γ>γ1). Therefore BCF model predicts a quadratic R(γ) curve for low supersaturation and gradual transition to a linear law as the supersaturation increases above the critical value given by γ1. For more general situation with group of dislocations, Equation 6 can be modified by a dislocation density factor ε:28

Equation 7

=

εγγ

γγ

ε 1

1

2

tanh' CR BCF

In InP/Si seed layer, on which ELOG will be conducted in this work, the density of threading dislocation terminating on (001) surface plane is close to 109cm-2. As predicated by Equation 7, such a high density of dislocations will contribute to a high growth rate on the planes that they intersect. 2.1.2 Other growth techniques Liquid phase epitaxy (LPE) was used for the early research on III-V semiconductors and is still in use because of its simplicity to grow simple device structures. ELOG III-V growth was exploited in LPE on both III-V30, 31, 32 and Si2, 33 substrates. Metalorganic vapor phase epitaxy (MOVPE) is a powerful growth technique in compound semiconductor industry especially due to its ability to grow heterostructures with abrupt interface and well controlled thickness, which is necessary for the formation of optical and electronic device structures. Some efforts have been made to investigate the feasibility to use MOVPE to selectively grow heterostructures in masked areas on GaAs or InP

8

substrate for the purpose of novel photonic devices. However, MOVPE generally operates with gas phase supersaturation far from equilibrium. This leads to the ready nucleation of materials on both the masked and unmasked regions, diminishing the selectivity of this process when using the commonly used growth precursors. As regards the selective epitaxy, constraints on the morphology and uniformity of the epitaxial layer usually limit the accessible range of growth parameters in MOVPE. There are also reports on the autodoping.34 The choice of growth precursors has the greatest influence on the attainment of selectivity under conditions yielding a high material quality. Both the co-introduction of chloride species (HCl, CCl4, CHCl3 etc.) and conventional metalorganic sources and the use of new metalorganic sources can give rise to selective growth.35 Molecular beam epitaxy (MBE) is another growth technique used extensively for the formation of compound semiconductor heterostructures with abrupt interface and well controlled thickness. The selective growth by conventional MBE is difficult due to the high sticking coefficient of group III materials on the mask. Nevertheless, selectivity in MBE has been achieved by periodic supply molecular beam epitaxy (PSE/MBE)36 and the lateral overgrowth has also been realized by using the low angle incidence microchannel epitaxy (LAIMCE) in MBE system.37 The growth rates are very low in MBE and MOVPE.

2.2 Heteroepitaxy by ELOG Here we describe the principle of ELOG through a schematic description given in Figure 3.

Si substrate

III-V seed

Mask Mask

ELOG

Threading dislocation

Opening

ELOG

Figure 3. Schematic representation of ELOG

First a thin seed III-V layer (~1 µm) is deposited on Si substrate by MOVPE or MBE. Then a dielectric mask (SiO2 or Si3N4) is deposited by the well established techniques like plasma enhanced vapor phase deposition (PECVD). Using conventional photolithography and wet chemical etching, openings are formed in the mask. During the initial stages of growth in HVPE, selective area growth (SAG) is achieved, i.e., growth takes place selectively only in the openings and not on the mask. This is very important to achieve ELOG and in HVPE it is very easy to achieve because of its near-equilibrium nature. When the growth exceeds the top of mask, it will extend laterally. As mentioned in the Introduction, this technique may lead to a filtering of defects: above the windows, the microstructure of the seed is reproduced whereas the laterally grown material (over the mask) is less of defects. The dielectric mask blocks the propagation of threading dislocations (TD), which arise from the seed. Recently, growth mechanism, physical properties and defect reduction in heterostructures of III-V semiconductors on Si prepared by liquid phase epitaxy (LPE) has been investigated. However, the work in vapor phase epitaxy (VPE) is still limited.

9

One more characteristic that is often encountered in ELOG is growth anisotropy. It corresponds to the occurrence of different growth rates on different crystallographic planes. This has been observed in HVPE since at least two decades ago.38 These features are strongly supported by the analysis of the orientation dependence of the growth rate associated with surface kinetics. The crystallographic properties of the boundary planes in ELOG will be addressed in section 2.4.

2.3 Heteroepitaxy on FIB modified surface In this method, the substrate is first subjected to focused ion beam implantation, e.g. that of Ga on GaAs. Afterwards, growth of e.g., InP is done in HVPE, see Figure 4.

Figure 4. Schematic description of heteroepitaxy on FIB modified surface

It has been found that a selective growth takes place only on the lines “written” by FIB in HVPE but not in other techniques.4 The dose, energy and the size of the beam are the variables that control an uninterrupted growth on the “written” lines. The growth features can also depend upon the growth parameters and the crystallographic orientation of the FIB lines. In this thesis, we have attempted to grow on the FIB line “written” along all the crystallographic directions.

2.4 Crystallographic structures of zinc blende The crystallographic structure of semiconductor is essential for understanding the process of epitaxy and the generation of defects during growth. InP crystallizes in a zinc blende (sphalerite) lattice. The zinc blende structure consists of two interpenetrating, face-centered cubic (fcc) sublattices consisting of only indium and phosphorus atoms in each sublattice. Each sublattice is shifted by a/4 [111] relative to the other fcc sublattice, where a is the length of the fcc lattice constant. The zinc blende structure of InP is shown in Figure 5, where open circles are indium atoms and solid circles are the phosphorus atoms. The most common surface used for the growth of III-V’s on InP and GaAs is (001). The {001}, {110}, and {111} basic planes in III-V compounds are of particular interest, which are important to understand the III-V semiconductor growth. Other surfaces with higher

10

Miller indices can be resolved on an atomic scale into stepped structures built up from combinations of these three basic sets of planes.39 During the selective growth, boundary planes surrounding the ELOG InP are formed because of the anisotropic growth rates. For the ELOG InP on (001) substrate, the normally encountered boundary planes are {111}, {110} planes, and other high index planes that can be decomposed to basic planes. Knowledge on their properties will facilitate to understand ELOG, nucleation of defects and strain relaxation after growth.

In

P

Figure 5. Lattice structure of InP, ○= In; •=P.

2.4.1 {001} planes

In

P

Figure 6. Atomic structure of {001} planes in InP crystal.

An ideal (001) plane is shown in Figure 6. It consists of alternating planes of indium or phosphorus atoms. The In and P planes are geometrically equivalent and are equally spaced. Each In (P) atom in indium (phosphorus) plane is bonded downward into the crystal to two P (In) atoms. Under typical growth conditions in LP-HVPE system, where V/III ratio is larger than one, the growth can be treated as indium incorporation limited, since under the condition of excess PH3, total adsorption of phosphorus atoms takes place readily. The growth on {001) plane can be seen as layer by layer growth. Incorporation of In atoms on {001} planes will cause net change of dangling bonds equal to zero 40, if surface reconstruction is neglected There is no substantial nucleation barrier for crystal growth on {001} planes.

2.4.2 {110} planes Atomic structure of {110} planes is shown in Figure 7. {110} planes contain equal number of In and P atoms. Atoms in {110} planes form planar zigzag chains, each atom being bonded to two adjacent atoms in the chain, with its other two bonds extending out on each side to the remaining tetragonal position. Growth on planar {110} planes can encounter nucleation barrier to initiate the growth of a new layer because the net change of dangling bonds is 2 after incorporation of new atoms. Once the first atom is in place,

11

atoms of the opposite type can deposit adjacent to it without change in the number of dangling bonds.

PIn Figure 7. Atomic structure of {110} planes in InP crystal

2.4.3 {111} planes The structure of {111} planes in InP crystal is shown in Figure 8. This family of planes is significant for understanding the growth profile and dislocation reduction in ELOG InP. The {111} double layer contains two closely spaced planes of In and P atoms and denoted as {111}A and B planes, respectively. Each atom in one plane has three bonds to the other plane. The fourth bond extends normal to the plane of the double layer to connect to an atom in the next such double layer. Since the two sides of the {111} double layer are not chemically equivalent, crystals bounded by {111} facets would be expected to show polarization effects during the selective growth. Growth on {111}B planes is lethargic in a phosphorus rich vapor ambient, because further attachment of indium atoms at {111}B planes will cause the net change of dangling bonds to be +2.39 It means that indium atoms have to overcome high energy barrier to deposit on {111}B planes.

PIn

Figure 8. Atomic structure of {111} planes in InP crystal

In zinc blende crystal structure, {111} planes are gliding planes for dislocation that lies along <110> directions because {111} planes are close-packed, which gives rise to the largest inter-planar spacing, d, and lowest Peierls stress. {111} planes also have a low density of bonds perpendicular to the plane. Both factors facilitate gliding.41 Planar defects, such as stacking faults and microtwin can only occur along <111> directions during heteroepitaxy of InP on Si.42 2.4.4 Boundary plane in ELOG InP During ELOG, boundary planes (surrounding the grown InP layer) are formed. For growth in stripe openings along [110] and [110] directions, the boundary planes are normally low index planes such as {111} or {110}. For those ELOG layers grown in openings along misoriented directions, which often give rise to high lateral growth rate, the

12

boundary planes could be difficult to identify instantly. Nevertheless, it is important to identify these and their atomic structures to understand the anisotropic growth mechanism of lateral overgrowth. The indices of boundary planes can be derived from the interplanar angle between the boundary plane and the substrate surface (001) plane by:43

Equation 8 )''')((

'''cos222222 lkhlkh

llkkhh++++

++=φ

where φ is the interplanar angle and (hkl) and (h’k’l’) are reference surface and boundary planes, respectively.

2.5 Defects in ELOG InP on InP/Si Certain semiconductors properties, e.g., free carrier mobility, and carrier lifetime change dramatically with the introduction of defects and impurities at the parts per million or parts per billion level. Scientists and engineers have made great efforts to take control on the imperfections in semiconductor substrates and epitaxial layers. Heteroepitaxy of dissimilar materials system, such as III-V on Si, raises more challenges in the aspects of impurity level control, suppression of dislocation density and releasing thermal strain. 2.5.1 Impurity Control of the impurity properties in epitaxial layer is essential for the performance of optoelectronic devices. Impurities and intentional dopant can alter the electrical conductivity and carrier lifetime.44 The main impurity encountered in HVPE is Si that is introduced into the grown crystal via chlorosilane generated by the reaction of HCl, H2, and the quartz (SiO2) reactor tube.45 High purity InP can be grown by LP-HVPE26 with background doping concentration as low as 1014cm-3. The source for n-type doping in HVPE is H2S. Some studies have been conducted on the mechanism of silicon46, 47 and sulfur incorporation in III-V semiconductors during VPE growth.48, 49, 50, 51 As n-type dopants, S atoms will replace V atoms (P sites) and Si will replace III atoms (In sites). Because of the presence of polar axes (<111>) in III-V semiconductor crystal, for the same dopant, non-uniform incorporation coefficients on different crystallographic planes were noticed during the epitaxial growth. Crystallographic orientation dependence of impurity incorporation has been studied for Si and S in III-V semiconductors with surface planes lying between {100} and {111}A/B.52, 53, 54, 55 The hierarchy of impurity incorporation efficiency as (111)B>(100)>(111)A has been observed. 2.5.2 Dislocations The lattice constant of InP is larger than that of Si. This lattice mismatch causes strain in the epitaxial layer, which increases with the layer thickness. The mismatch can be accommodated by elastic deformation of the lattice, i.e. InP layer is compressed in the plane of growth and expanded along the direction of growth. When the thickness of the epi-layer is larger than the critical thickness, the associated lattice mismatch strain energy is larger than the dislocation energy. Then the dislocations are generated to relieve the misfit strain energy. These dislocations are called misfit dislocations.

13

Two types of misfit dislocations are often observed in heteroepitaxial InP layer on Si: perfect edge dislocation and inclined dislocations (60° in the case of InP on Si). The dislocation can be specified in terms of Burgers vector. For the growth axis of [001], the

perfect edge dislocation at the interface between InP and Si has Burgers vector2a <110>.

Both the dislocation line and the Burgers vector lie in the (001) plane. They are very efficient in accommodating mismatch strain because their Burgers vectors are parallel to the growth plane. The inclined dislocations have their Burgers vector at an angle of 45° to the plane of the substrate and 60° to the dislocation line. Since their Burgers vectors are at an angle to the surface plane, more of these dislocations are required to accommodate a given mismatch than edge dislocations. The dislocations can never just end within a crystal, although it can form a closed loop within a crystal or a half loop with both their ends touching the surface. The part of a misfit dislocation propagating upto the free surface is called threading dislocation (TD). The threading dislocation can extend to the device structures grown on ELOG and degrade the performance of devices. The threading dislocation in ELOG InP on InP/Si will glide on {111} planes along <110> direction. The gliding of TDs could be blocked by the edge of dielectric mask on patterned substrate.56 Dislocations in semiconductor will generate shallow levels due to elastic strain field around the core and deep levels associated with dangling bonds.57 They also interact with impurities. The dislocation density in substrate can be decreased by introducing certain dopants.58, 59, 60 The effect of impurity on the generation of dislocations and their properties have been studied in III-V semiconductors.61, 62, 63 2.5.3 Planar defects The most important planar defects observed in heteroepitaxy of III-V on Si is stacking faults (SF) and micro-twins. Initially the formation of stacking faults was attributed to lattice mismatch and thermal expansion coefficient mismatch.64, 65 Recently, it has been argued that these mismatches play a minor role in the nucleation of stacking faults and twins.42, 66 The new model on the formation of stacking faults and twins are based on the assumption that the early stage of nucleation is faceted by low-energy planes such as {111}. Errors in stacking on the {111} facets during deposition give rise to stacking faults and twins. Twins and stacking faults can propagate along two out of four (111) planes, namely, the (111) and (111) planes. The propagation along these planes usually leads to the mutual annihilation of the twins and stacking faults and thus they are confined to a short distance from the interface, but some still propagate into the epi-layer and intersect the surface.67 2.5.4 Thermal strain Another problem we have to solve in heteroepitaxy of InP on Si is the residual strain in the epitaxial film. There are two major sources of strain: one is the lattice mismatch and the other is the difference in the thermal expansion coefficient between the epitaxial layer and substrate. For III-V semiconductors on Si, such as GaAs and InP, the lattice mismatch strain should be compressive because the epitaxial layers have larger lattice constant than the substrate, while the thermally induced strain should be tensile because the layers have larger thermal expansion coefficients than the substrate. It has been reported that for the heteroepitaxial layers with a lattice mismatch exceeding about 4%, almost the entire lattice mismatch is accommodated by the generation of misfit dislocations.68 However, in III-V

14

semiconductors grown on Si, tensile strain has been constantly observed, even for GaP that is nearly lattice matched Si.69 Therefore, it is reasonable to consider the residual strains in InP layers on Si to be dominated by the thermally induced strains after growth during the sample cool down from growth temperature to room temperature. The thermal strain in InP on Si epi-layer can be estimated by Equation 9:

Equation 9 dTSiInP )(// ααε −= ∫

where αInP and αSi are thermal expansion coefficients for InP and Si respectively, and ε// is the strain in the growth plane.

2.6 Characterization techniques Characterization of epitaxial layer is important for analyzing and understanding the mechanism of imperfection generation and distribution in crystal. Especially for heteroepitaxy of InP/Si, characterization techniques should be able to provide the information about the level of strain, defect density and distribution, and impurity distribution in grown layer. In addition, the ELOG growth from the immediate vicinity of openings extends only tens of micrometers over dielectric mask. Therefore the characterization techniques should be able to operate in small dimensions with good spatial resolution. 2.6.1 Photoluminescence Photoluminescence (PL) has been used to examine the strain in heteroepitaxy structures, such as InGaP/GaAs70, InGaAsP/InP71, InGaAs/GaAs72, 73, and GaAs/Si74, InP/Si75, GaP/Si.76 When the epitaxial layer has a larger lattice constant than the substrate, as is the case of GaAs or InP on Si, the epi-layer is under biaxial compressive strain. This strain can be decomposed into a hydrostatic component, and a shear component. The hydrostatic component changes the volume of crystal lattice of epi-layer and causes a shift in the PL peak wavelength. The shear component of the strain is proportional to the asymmetry in the strain parallel and perpendicular to the stress plane. It splits apart the degenerate valence bands at k=0 (Γ point), Figure 9. Each valence band participates in carrier excitation and recombination. Hence shear strain component splits singlet PL peaks into doublets.77 These effects make PL such an effective probe of stress. Under strain, the gaps ELH and EHH associated respectively with the conduction band to light hole transition, and the conduction band to heavy hole transition, become:

Equation 10 //11

1211

11

1211 )2()(2 ε

++

−−−=

CCC

bC

CCaEE GLH

Equation 11 //11

1211

11

1211 )2()(2 ε

+−

−−−=

CCC

bC

CCaEE GHH

where EG is the unstrained band gap, the Cij are elastic stiffness coefficients of InP at room temperature,78 a is the hydrostatic deformation potential, b is a shear deformation potential, and ε is the strain. Since the elastic constant and deformation potentials in bulk III-V semiconductors are known, the energies measured from the PL data yield strain values.

15

Figure 9. An in plane biaxial strain effect on the electronic energy-bandgap

structure near the Γ point in zinc-blende-type semiconductors. 79

In this work, 514.5 nm line Ar- ion laser was used in the PL setup. The PL emission was detected by a liquid-nitrogen cooled Ge detector. The samples were mounted on a cold finger and cooled by a closed cycle helium refrigerator. The PL measurements were carried out at different temperatures from 10 K to 295 K. 2.6.2 High resolution x-ray diffraction High resolution x-ray diffraction (HRXRD) has been developed into a powerful tool for the nondestructive investigation of heterostructures. The principle of this technique is based on the elastic scattering of x-ray radiation by the crystal lattice structure and the collection of scattered signal by an open detector or a crystal analyzer. The information of crystal structure and the distance between two scattering crystal planes can be obtained by Bragg Law: Equation 12 λθ nd =sin2

where λ is the x-ray wavelength, θ is the angle between the incident x-ray beam and the surface of a set of scattering planes and d is the spacing between the planes. The configuration of x-ray diffractometer used in this work is shown in Figure 10. By placing a 4-crystal Ge (220) mono-chromator after the Cu x-ray source with an angular divergence of 12 arcsec, Cu Kα1 (λ=1.5406 Å) radiation is separated from other radiation components of the x-ray source. The sample stage can be adjusted independently. Thereby, the angle of incidence ω, the diffraction angle 2θ, the angle Φ of rotation around the surface normal, and the angle Ψ of rotation around an in plane surface direction can all be varied. The detector can be set in either receiving-slit mode, which is used to record rocking curve (RC), or in triple-axis mode, which is ideal for reciprocal lattice mapping (RLM) measurement. A 2-crystal analyzer is placed after the specimen and before the detector, in order to restrict its angular acceptance to 12 arcsec and leads to the triple axis mode. This has the effect of separating the effects of strains and tilts on the measurement.

16

Figure 10. Schematic view of triple-axis x-ray diffractometry used in this work. (ω is x-ray incident angle, θ is Bragg angle, Φ is rotation angle, Ψ is tilt angle)

2.6.2.1 Thermal strain In this work, the strain in ELOG InP on InP/Si was also estimated from high resolution x-ray diffraction reciprocal lattice mapping (HRXRDRLM)80,81,82 obtained from triple-axis x-ray diffractometry (TAD) by performing a series of ω-2θ scans with a gradual off-set of ω value. The peaks from Si substrate, seed InP layer and ELOG InP layer with different lattice constants will be separated along ω-2θ direction in reciprocal lattice mapping, as shown in Figure 11.

ELOGInP

Seed InP

Si

Qy×10000(rlu)

Qx×10000(rlu)

Figure 11. High resolution x-ray diffraction reciprocal lattice mapping (HRXRDRLM) at 004 reflection of ELOG InP on InP/Si grown structure

The lattice constant of ELOG InP parallel to the growth direction, a⊥ELOG can be calculated from the Bragg angles of Si substrate, θSi and ELOG InP, θELOG:

17

Equation 13 ELOG

SiSiELOG aa

θθ

sinsin

×= ⊥⊥

The strains in ELOG InP layer normal (ε⊥) and parallel (ε//) to the substrate surface can be calculated by considering a tetragonal distorted film using the elasticity theory:83

Equation 14 1−=⊥

⊥InP

ELOG

aa

ε

Equation 15

×−= ⊥

12

11// 2C

Cεε

where aInP is the lattice constant of bulk InP, 5.8686Å, and C11 and C12 are the elastic constants of InP at room temperature.

2.6.2.2 Dislocation density The HRXRD can also be used to measure the dislocation density in heteroepitaxial layers. Dislocations are commonly present at the interface as misfit dislocations to accommodate the strain and in the epitaxial layer as threading dislocations generated by local plastic deformation due to thermal or mechanical strain. The misfit dislocations at the interface of heterostructure will give a specified relaxation of strain between the substrate and the epi-layer. This causes quantifiable shifts in the peak position and adds diffuse scattering.84, 85 On the other hand, threading dislocations inside the epi-layer will not shift the rocking curves but would rather broaden it. Gay. et. al.86 have described the theory of x-ray diffraction in metal crystal with dislocation. The threading dislocation density in semiconductor epi-layer has been estimated from the diffused scattering x-ray data by means of reciprocal space mapping87 and from the full width at half maximum (FWHM) values of rocking curves.88, 89 In this work, threading dislocation density in ELOG InP on InP/Si substrate is estimated from the FWHMs of (004), (115) and (117) Bragg reflections in the ω scan direction. If βm (hkl) is the measured rocking curve FWHM of the (hkl) reflection, then90

Equation 16 )()()(

)()()()(222

2220

2

hklhklhkl

hklhklhklhkl

rL

dm

βββ

ββββ

ε

α

+++

++=

where β0(hkl) is the intrinsic rocking curve width for the crystal being examined, βd(hkl) is the intrinsic rocking curve width from the monochromator, βα(hkl) is the rocking curve broadening caused by the angular rotation at the dislocations, βε(hkl) is the rocking curve broadening caused by the strain surrounding the dislocations, βL(hkl) is the rocking curve broadening due to the crystal size, and βr(hkl) is the rocking curve broadening due to the curvature of the specimen. β0(hkl) and βd(hkl) are usually less than 10 arcsec and can often be neglected. Crystal size broadening, βL(hkl) can be neglected for layers thicker than 1.0 µm. Usually, curvature broadening can also be neglected since the epi-layer is much thinner than the substrate. Thus the rocking curve broadening mostly arises from the angular rotation at the dislocations and the strain surrounding the dislocations, i.e. βα(hkl)

18

and βε(hkl), respectively. The square of strain broadening βε2(hkl) is proportional to tan2θ, 91 where θ is the Bragg angle. Thus Equation 16 can be simplified as:

Equation 17 θββ εα222 tan)()( Khklhklm +=

where Kε is a proportionality constant. The FWHMs of three measured rocking curves of reflections (004), (115) and (117) and the corresponding Bragg angles, θ (for each reflection) can be inserted into Equation 17 and the function βm

2 versus tan2θ is plotted as a straight line with intercept βα2 and slope Kε.

Figure 12. FWHM2 versus tan2θ for ELOG InP on InP/Si layer grown in

openings aligned at 60° off [110] direction.

As an example, the FWHM2 of rocking curves measured on ELOG InP on InP/Si layer grown in openings aligned at 60° off [110] direction is plotted versus tan2θ in Figure 12. Then dislocation density, D can be determined by:89

Equation 18 2

2

36.4 bD αβ=

where b is the Burgers vector of threading dislocation. 2.6.3 Cathodoluminescence Cathodoluminescence (CL) is a technique to measure the luminescence arising from the excitation caused by an electron beam. It is an important imaging mode in modern scanning electron microscopy. Even though the mechanism of light emission from CL is similar to photoluminescence, the electron beam excitation leads to the emission by all mechanisms present in the semiconductor, while PL emission, on the other hand, may strongly depend on the excitation energy. Image contrast in CL may arise from a number of different mechanisms, e.g., enhanced non-radiative recombination, emission at undetectable wavelengths, variations in radiative efficiency, optical scattering, and optical absorption. Careful interpretation of CL contrast can reveal the presence of

19

crystallographic defects, impurity inhomogeneties, gettering, stress variations, space-charge regions, and interfaces. CL has been used to study the defects and stress in heterostructures since it has the following advantages: a) high spatial resolution and b) availability of more depth-resolved information by varying electron beam energy. 2.6.4 Techniques to characterize dopant incorporation Characterizing the impurity distribution in epitaxial layer is important because its influence on the electronic and photonic properties of the material. The continuous down scaling of the semiconductor device size and the use of complex 2-D, 1-D and 0-D structures have led to the development of advanced carrier profiling techniques with high lateral resolution. Among theses, scanning capacitance microscopy (SCM)92 and scanning spreading resistance microscopy (SSRM)93, 94, 95 are capable of measuring doping with nm-spatial resolution and complement each other. Recent works have demonstrated their ability to study the orientation dependent impurity incorporation in selective area growth and related issues in buried heterostructure lasers.92, 94

The SCM and SSRM set up are shown in Figure 13. Both techniques are based on atomic force microscope (AFM) equipped with a conductive tip that is biased relative to the sample.

ContactMode AFM

CapacitanceData

CapacitanceMeasurementElectronics

X-Y StageX-Y Stage

Sample

Bias Voltage

Conductive Probe

ContactMode AFM

Current &Resistance Data

LogarithmicCurrentAmplifier

X-Y StageX-Y Stage

Sample

Conductive Probe

Bias Voltage

(a) (b)

Figure 13. Schematic diagram of (a) SCM and (b) SSRM set up

In SCM, the tip and the sample form a metal-insulator-semiconductor (MIS) structure, the insulator (in this work) being the native oxide on cleaved surface formed by the exposed air. As the tip scans the sample, the capacitance information is obtained by a resonant capacitance sensor that operates at 915MHz. The tip-sample capacitance is modulated by low frequency ac (5-100kHz) bias and lock-in detection is used to measure the differential capacitance (dC/dV). In accordance with the high frequency MIS capacitance voltage relationship for an n-type sample, in the dC/dV image, low and high doped regions in the sample will appear bright and dark, respectively, and the signal magnitude is related to the doping level (Figure 14(a)). In SSRM, the spreading resistance value is derived from the measured electrical current between the biased tip and the sample, which is a function of the local carrier concentration at the surface region surrounding the probe’s tip. The spreading resistance is primarily determined by the resistivity of the material at the surface region, which is closely related to the local carrier concentration (and mobility) in this surface region. The spatial resolution of SSRM mainly depends on the tip radius and pressure. Under the measurement condition/configuration, the “doping contrast” is such that low (high) doped

20

regions appear dark (bright). Irrespective of this, the resistance is low (high) for high (low) doped region (Figure 14(b)).

Undoped InP

S-doped InP (a)

5µm

S-doped InPUndoped InP

(b)

5µm

Figure 14. Two-dimensional map of electrical properties of ELOG InP in 45° off [110] openings taken by (a) SCM and (b) SSRM. The SCM image contains

only a part of the structure shown in the SSRM image.

21

3 Summary of results and discussions The published original work will be discussed in this part. The behavior of InP epitaxial lateral overgrowth was investigated on (001) InP substrate prior to that on the InP precoated silicon substrate (Paper A, B, and C). The crystallographic orientation dependence of lateral overgrowth rate and impurity incorporation during the ELOG growth was studied. Then the optimized growth conditions were used in ELOG InP on InP/Si growth. Papers D, E, F, G and H contain the major results toward fulfilling the objectives of this thesis. The defect reduction and thermal strain reduction were accomplished by ELOG growth at optimized growth conditions. The effect of impurity incorporation on dislocations was investigated in sulfur doped ELOG InP on InP/Si growth. Then InGaAsP multi-quantum wells (MQWs) at 1.55 µm were grown on ELOG InP on InP/Si templates which gave rise to promising luminescence spectra. In paper I, the application of ELOG on the fabrication of nanostructure was explored by selective growth of InP on focused-ion-beam modified GaAs surface.

3.1 Epitaxial lateral overgrowth (ELOG) of InP Crystallographic orientation dependent growth rate of III-V semiconductor on planar substrate has been studied in a vapor phase epitaxy system in the kinetic limiting region.38 During the ELOG process several facet planes are formed simultaneously surrounding the grown layer. This will give rise to different growth rates on different planes under the same growth conditions leading to anisotropic lateral overgrowth rate on patterned substrate. The aspect ratio between the lateral and vertical growth rate is an important parameter to characterize the ELOG growth. A large aspect ratio is desired from the point of view of defect reduction and device fabrication. To this end, temporally resolved selective growth of InP consisting of alternating sulfur doped and undoped layers was initiated from the stripe openings aligned at various directions on (001) InP substrate with 15° interval. The [110] view cross-section scanning electron microscopy images of stain etched ELOG InP grown in the openings aligned at 15°, 30°, 45°, 60°, and 75° off [110] directions are shown in Figure 15. The lateral and vertical growth rates are shown in Figure 16. The lateral growth rates both at the bottom and at the top are depicted in the same figure. The maximum lateral overgrowth rate both at the top and the bottom occur for the openings aligned at 30° off [110] direction, as shown in Figure 16. But, there is no obvious dependence of vertical growth rate on the opening orientation. By carefully examining the emerging boundary planes of ELOG InP layer in cross-section SEM image, we postulate that the high lateral overgrowth rate is due to the formation of high index boundary planes. Such planes have high density of surface steps and kinks. The Miller indices of the boundary planes are calculated from interplanar plane angle, φ between the boundary planes and (001) surface according to Equation 8 and listed in Table 1, as long as φ is accessible from Equation 19 (paper A):

Equation 19(a)

=

θφ

φcos

'tanarctan (0°<θ<45°)

22

(a)

ELOGOpening

(b)

ELOG Opening

(c) ELOG Opening

(d) ELOG

Opening

(e) ELOG

Opening

Figure 15. Cross-section SEM micrograph of ELOG InP grown in openings oriented at: θ = (a) 15°, (b) 30°, (c) 45°, (d) 60°, and (e) 75° off [110] directions, [110] view.

Equation 19(b)

−

=)90cos(

'tanarctanθ

φφ (45°≤θ<90°)

where φ’ is the angle in (110) cleaved plane measured from the SEM photos and θ is the misorientation angle of the stripe openings. These high index boundary planes in the misoriented openings are vicinal planes, which have high density of steps at the growth front over the Si3N4 mask. These crystallographic steps on boundary plane can supply the seed for continuous lateral growth over the mask, although there is no nucleation site on the mask.

23

Figure 16. Relationship between stripe openings orientation and lateral

overgrowth rate on (001) InP substrate

θ φ' φ Index of

(degree) (degree) (degree) boundary plane 15 48 49 {2 2 1} 30 42 46,5 {3 3 1} 45 36 45 {1 1 0} 45 148 135 {1 1 0} 60 142 138 {3 3 1} 75 152 151 {3 1 1}

Table 1. Boundary planes identified through Equation 8 and Equation 19

The dependence of lateral and vertical growth rates on gas phase composition was studied by varying the input flow rate of InCl and PH3. The cross-sectional view of ELOG InP is trapezoidal at all our growth conditions. Two groups of experiments were conducted on (001) InP substrate patterned with the openings aligned at 30° off [110] direction and separated by 200 µm. On one set of samples ELOG growth was carried out under constant PH3 flow of 120 sccm and InCl flow rates of 8, 12, 18, and 30 sccm. On the other set of samples InCl flow was held constant at 18 sccm and PH3 flow rate was varied as 90, 120 and 240 sccm. In both cases, the growth was conducted at 620°C, at a total pressure of 20 mbar with a total flow rate of 900 sccm. The carrier gas was N2. Lateral growth rate that is measured at the top of ELOG, vertical growth rate and aspect ratio as functions of InCl and PH3 flow rates for the two sets of experiments are plotted in Figure 17 and Figure 18. The lateral growth rate passes through a maximum as PH3 or InCl is increased (as shown in Figure 17 (a) and Figure 18 (a)). The vertical growth rate increases monotonically when InCl flow rate is higher than 12 sccm (Figure 17 (a)), but it is almost constant as a function of PH3 flow rate (Figure 18 (a)). The dependence of aspect ratio (lateral growth rate/vertical growth rate) on PH3 and InCl flow rates is shown in Figure 17 (b) and Figure 18 (b). The inverse dependence of the lateral growth rate on the indium chloride flow rate after the maximum may indicate a competitive adsorption process of InCl and HCl for the available P adsorption sites on boundary planes.96

24

Figure 17. (a) Lateral and vertical growth rate and (b) aspect ratio of ELOG InP

in 30° off [110] openings versus InCl flow rate; growth temperature=620°C; total pressure=20mbar, and total flow=900sccm.

Figure 18. (a) Lateral and vertical growth rate and (b) aspect ratio of ELOG InP

in 30° off [110] openings versus PH3 flow rate; growth temperature=620°C; total pressure=20mbar, and total flow=900sccm

As noticed in the cross section SEM micrographs of ELOG InP shown in Figure 15, the impurity distribution in the same grown layer is inhomogeneous and higher concentration of impurities are incorporated in the lateral overgrowth on top of the mask. The impurity incorporation in ELOG InP is related to the crystallographic property of boundary planes. The impurity concentration is characterized by SCM and SSRM with high spatial resolution. The cross-sectional SCM image reflecting the impurity distribution in ELOG InP layers grown in the openings oriented at 30° off [110] direction is shown in Figure 19. In the SCM cross-sectional image, even numbered InP layers (L2, L4….) are intentionally doped with sulfur (n≈1×1018 cm-3) and odd numbered layers (L1, L3….) are unintentionally doped InP (n≈2×1015 cm-3). Letters v and l indicate vertical and lateral sections of growth. Regarding the ratio between lateral and vertical growth rates, the undoped ELOG InP layer (L3) has a value much larger than that of the S doped ELOG InP layer (L2). Incorporation of S decreases the lateral overgrowth rate. This can be due to the sulfur doping mechanism in HVPE growth of InP, where sulfur and phosphorus atoms compete for the incorporation sites on growth front of the lateral growth. The lateral overgrowth is a kinetically limited process, which means the growth rate of InP is affected by the crystallographic properties of the growth front plane. So does the incorporation of dopant. The boundary plane of ELOG InP in 30° off [110] direction has been identified as {331}B plane, which is P rich plane. The {331}B plane is composed of (110) terrace, on which

25

b1

b2

a1

a2 (a)

substrate

L2-l

L3-l L3-v L4-l

L2-v

L4-v L5-v

L1-l L1-v

L5-l

(b)

Figure 19. (a) Cross-section SCM (dc/dv) images of ELOG InP grown in openings 30°

off [110] directions−[110] view. (b) Variation of the SCM signal along the lines indicated on the image, low signal⇒ high doping; high signal⇒ low doping.

there are two stable P sites and each of the sites is connected with three dangling bonds from underlying (111)In layer, as shown in Figure 20. This means that, from the energy point of view, P and S atoms will incorporate on this plane more easily. And high S concentration along <331> direction will decrease the lateral growth rate.

26

Figure 20. The surface bonding structure of {331}B plane. The relevant

incorporation sites for dopant atoms are indicated.

3.2 Heteroepitaxy of InP/Si by ELOG From the study of ELOG InP growth on (001) InP substrate, we understand the relationship between the opening orientation and aspect ratio of ELOG. The ELOG InP growth was attempted on InP precoated (001) Si substrate with 4° miscut toward <111> direction. The work is discussed in Paper D and E. With the same growth conditions and sample pattern geometry as in ELOG of InP on InP, a different morphology was observed in ELOG growth on Si substrate, e.g., the cross sections of ELOG InP on InP/Si layers grown in openings aligned at 30° and 60° off [110] directions shown in Figure 21. These are different from Figure 15 (b) and (d) in their growth rates (or aspect ratios) and asymmetric growth profiles. This difference can be due to the existence of a high density of threading dislocations in the precoated seed InP layer grown by MOVPE. As estimated by HRXRD, the dislocation density can be as high as 109 cm-2 in the seed InP layer. Such a high density of threading dislocation will propagate during growth and exit on the boundary planes of ELOG to enhance the growth rate as predicted by BCF model (Equation 7). The asymmetric growth is caused by the nature and abundance of the dislocations on {111}A or {111}B planes since these are the boundary planes.41 3.2.1 Asymmetric growth morphology A highly asymmetric growth morphology was observed for the ELOG InP on InP/Si grown in [110] oriented openings with well developed lateral overgrowth on one side bounded by {111}A and {111}B planes but less developed lateral growth on the other side

27

(a)

OpeningELOG ELOG

(b) ELOGELOG

Opening

Figure 21. SEM micrograph of ELOG InP on InP/Si grown in openings aligned

at (a) 30° and (b) 60° off [110] directions. [110] view.

bounded exclusively by {111}B planes. (Paper D). This asymmetric growth morphology indicates an inhomogeneous threading dislocation distribution between the {111}A and {111}B planes in the seed layer. The dislocations lying on {111}A planes are referred to as α dislocations, while those on {111}B planes, β dislocations.62, 97 The nature of α and β dislocations are different because they lie on different {111} double layer in III-V semiconductors. Because of this, the generation barrier is also different for α and β dislocations. The result is that their propagation velocities are different under applied stress, which leads to different density for α and β dislocations. Based on the observed growth morphology, where the (111)A plane intersecting the (111)B plane had a high growth rate, we postulate that the subdensity of β dislocations is much higher than that of α type dislocations in the seed layer. The asymmetric growth morphology can also be influenced by the relative gas phase supersaturation, γ, over the surface of ELOG InP layer grown in the openings as predicted by BCF model, Equation 7. γ can be controlled by changing the growth temperature, introducing extra HCl in growth ambient, changing reaction species or changing the pattern geometry, such as the distance between two openings. In Paper D, the effect of adding HCl was studied. After introducing extra HCl into growth ambient, growth morphology of ELOG InP on InP/Si appears symmetric due to the decreased growth rate on {111}A plane intersecting with high density TDs, since the growth mechanism transits from spiral growth to two-dimensional growth under decreased gas phase supersaturation. The similar phenomenon was reported in homoepitaxial growth of InP in HVPE.98 In Paper E, the effect of gas phase supersaturation on growth morphology and defect density was studied systematically at a set of III-V ratios by changing the flow rates of PH3 and InCl as indicated in Equation 3. The dependence of supersaturation on the distance

28

between the openings was also studied. The detailed experimental parameters are shown in Table 2.

Group Sample Pattern PH3 flow (sccm)

InCl flow (sccm)

V/III ratio

A Direction: [110], Opening width: = 6 µm,

Opening separation = 400 µm.

120 12 10

B -do- 240 12 20 C -do- 240 8 30

I

D -do- 120 24 5 E Direction: 30 o off [110],

Opening width: = 6 µm, Opening separation = 100 µm.

120 12 10 II

F Direction: 30 o off [110], opening width: = 6 µm,

Opening separation = 400 µm.

120 12 10

Table 2. Summary of the wafer pattern and growth parameters. Total flow=900 sccm; Tg=620°C; Total Pressure=20 mbar.

The hierarchy of gas phase supersaturation is: γB(V/III=20) = γD(V/III=5) > γC(V/III=30) > γA(V/III=10). Whether the growth morphology is symmetric or asymmetric depends upon the growth rate on {111}A planes with respect to the {111}B planes. The growth rate on {111}A planes (with a constant threading dislocation) can be varied by changing the supersaturation. We found the growth rate on the {111}A is approximately related to the vapor phase supersaturation by a parabolic function, as predicated by BCF model. The growth rate of {111}A planes was calculated by measuring the normal to this plane from the edge of the openings. But the relative growth rate of {111}A with respect to {111}B can be also assessed by the emerging length of the former with respect to the latter from the SEM cross-section. At high supersaturation, V/III ratios affect the growth morphology profoundly; at low V/III ratio, symmetric growth morphology is formed due to the emerging {110} facet planes at both sides of ELOG InP on InP/Si. But this vanishes for high V/III ratio due to the high growth rate on these planes. Similar dependence of growth rate hierarchy of boundary planes on V/III ratio has been observed in selective area growth of GaAs, which is related to kinetically controlled growth mechanism100. The effect of the pattern geometry on the gas phase supersaturation was studied by changing the separation distance between the openings aligned at 60° off [110] directions. At large separation distance, the local concentration of the adsorbed species is high in relation to the consumption rate due to reaction. This situation enhances the supersaturation at the interface of ELOG InP layer. On the contrary, when the opening distance decreases, more gaseous species are consumed and the supersaturation is dropped. For the ELOG InP on InP/Si growth in misaligned openings, vertical growth rate is more susceptible to the fluctuation of gas phase supersaturation than lateral growth rate. Lower supersaturation will give lower vertical growth rate and higher aspect ratio. 3.2.2 Defect reduction The defects encountered in heteroepitaxy of III-V on Si are threading dislocations and planar defects. The defects in ELOG InP on InP/Si layer were revealed by means of a 20s

29

etching in Hubert’s solution (2H3PO4:1HBr)101 at room temperature. By wet chemical etching, two kinds of etch pits have been identified. One is isolated etch pit from the threading dislocations. Another one is the groove shaped pit from planar defects, such as micro-twins and stacking faults (Paper E). Their densities are calculated from the optical microscope inspection. For comparison, the TD density was also estimated from the FWHM of HRXRD rocking curves.86, 89 The relationship between gas phase supersaturation and defect density in ELOG InP on InP/Si was established. The threading dislocation density derived from HRXRD is higher than EPD. This could be due to that EPD (estimated from chemical attack) represents a bundle of threading dislocations.89 No etch pits from stacking faults have been noticed in the seed InP layer on Si substrate after wet etching, while the EPD due to threading dislocation is rather high. Density of etch pits arising from threading dislocation is in general lower in ELOG InP on InP/Si growth from the opening along [110] direction than in the seed layer. The stacking faults in ELOG layer could only be generated during the growth. It has been reported that the stacking faults in InP have rather low generation energy compared to those in other III-V compound semiconductors102. The growth rate on the defected {111}A is enhanced due to the spiral growth, which causes these planes susceptible to stacking faults.66 The interference from atomic steps caused by emerging threading dislocations on {111}A boundary planes during the growth can nucleate stacking faults easily. These stacking faults will continue growing until they intersect each other. Depending on the nature of the stacking faults, different types of stair-rod dislocations form at their intersection and the growth of the stacking faults are terminated.41, 42 The density of stacking faults is affected by both gas phase supersaturation and V/III ratio. Micro-twins are observed in ELOG InP on InP/Si layers grown in the openings aligned at 60° off [110] direction. The etch pit from these planar defects are oriented along [110] direction and their densities strongly depend on the gas phase supersaturation. The formation of micro-twins outside the seed openings and across the whole grown layer could indicate that the lateral overgrowth involves the formation and coalition of {111} facet planes along the edge of ELOG InP layer over Si3N4 mask. As pointed out before, the interaction of {111} planes and threading dislocation increases the possibility for stacking faults. Since the probability of formation of stacking faults is proportional to the area of {111} facets available before the coalescence of nuclei, we speculate a lower density of {111} facets in ELOG InP on InP/Si layer at low supersaturation.

3.3 Thermal strain in InP ELOG on Si The thermal strain in ELOG InP on InP/Si substrate has been studied in Paper F by means of HRXRD reciprocal lattice mapping and low temperature (10 K) PL. The thermal strain in ELOG InP on InP/Si layer is believed to be caused by the mismatch of thermal expansion coefficients of InP and Si3N4 mask. As mentioned in section 2.6.1, strain will change the volume of crystal lattice and cause red shift of InP peak wavelength in PL spectrum. In HRXRDRLM, the strain can be estimated from the separation of peaks from ELOG InP and seed InP. Both techniques show similar dependence of thermal strain on the aspect ratio of ELOG InP layer. A higher aspect ratio gives rise to a lower thermal strain at the measured temperature, as shown in Table 3. The effect of thermal strain on the optical property of ELOG InP on InP/Si layer with different aspect ratios grown in the openings aligned at 30° and 60° off [110] direction was studied by PL at several temperatures, as shown in Figure 22. The aspect ratio for ELOG InP on InP/Si in 30° off

30

FWHM⊥XRD ε// XRD λPL ∆Eg ε// PL Sample Aspect ratio (arcsec) (ppm) (nm) (meV) (ppm) [110] 0.5 39 626 882 9.7 1070 30° off [110] 2.3 38 476 879 4.6 510 60° off [110] 2.8 36 348 878 3.3 360

Table 3. Thermal strain ε// , in ELOG InP on InP/Si layers estimated from HRXRDRLM data (295 K) and LT-PL (10 K) shift ∆Eg for different aspect ratios and openings orientation on a Si3N4 mask. ∆Eg is the bandgap difference between InP on plain InP substrate and ELOG InP on InP/Si.

840 860 880 900 920 940

x10

(a)

295K150K

80K

40K

10K

Inte

nsity

(a.u

.)

Wavelength (nm)

850 860 870 880 890 900 910 920 930 940 950 960 970 980

(b)

295K150K

40K

80K

10K

Inte

nsity

(a.u

.)

Wavelength (nm)

Figure 22. Temperature dependent PL measurement of ELOG InP on InP/Si grown in the openings aligned at (a) 30° off and (b) 60° off [110] directions

[110] openings is 2.3 and that for 60° off [110] openings is 2.8. For both samples, the peak wavelength shifts toward low energy at the same time as the peak gets broader as the

31

temperature increases. At all temperatures, the spectra of the 60° off openings sample have narrower linewidth than those of 30° off samples. We notice doublet peaks in PL spectra at 40K and 80K for the growth in 30° off samples with comparable magnitude, which is less obvious for the growth in 60° off openings. The splitting of valence band due to tensile thermal strain has been reported in GaAs grown on Si by MBE as inferred from its LT-PL spectrum at 77 K.103 The splitting of valence band in ELOG InP on InP/Si could also be due to the thermal strain. The absence of doublet peak in the 60° off [110] ELOG InP on InP/Si sample indicates a lower thermal strain in the grown layer than that in 30° off [110] growth.

3.4 Effect of sulfur doping In our case, the dislocation density in seed InP/Si layer is already high. The dopant effect on dislocation behavior and density reduction during the ELOG growth is studied in Paper G. Octahedral shaped undoped and sulfur doped ELOG InP on InP/Si templates were grown in ring shaped openings. The templates are bounded by {110}, {111}A and {111}B facet planes. The growth morphology was characterized by AFM, as shown in Figure 23. Step like features oriented along [110] direction are observed on the surface of undoped ELOG InP on InP/Si templates, which are obscure on S-doped templates. These [110] surface steps could arise from the dislocations gliding through the epitaxial layer, which are often observed in lattice mismatched heterostructures consisting of III-V materials.

(a)

[110]

[110]

(b)

[110]

[110]

Figure 23. AFM surface morphology images of (a) undoped and (b) sulfur

doped ELOG InP templates on Si substrate.

The improved morphology and reduced dislocation density in ELOG could be due to sulfur inhibiting the movement of threading dislocations (called impurity hardening effect) 58, 59, 60 extending from the seed InP layer. Dislocation pinning in S-doped ELOG InP on InP/Si templates is observed in low temperature cathodoluminescence (LT-CL) at 77 K, as shown in Figure 24. The lateral overgrown part of InP growth on the top of Si3N4 mask is almost free of dislocations. In the CL spectrum of ELOG InP:S, a broader peak with emission energy as high as 1.5 eV (825 nm) is observed, in addition to the band edge (BE) emission of InP at 1.417 eV (875 nm), whilst only BE transition is observed in the undoped ELOG InP. The intensity distribution of BE emission in S-doped sample coincides with the dislocation distribution in panchromatic image, as shown in Figure 25(a). The highest intensity of luminescence occurs in the epi-layer overgrown in the central part indicating a good crystalline quality

32

with respect to the other regions. In contrast, the intensity distribution of emission at 825 nm is rather uniform across the surface of S-doped template, as shown in Figure 25(b).

Figure 24. Panchromatic CL image of S-doped ELOG InP on InP/Si templates. Dark spots on the surface of

templates are threading dislocations pinned by sulfur atoms due to impurity hardening effect.

(a)

(b)

Figure 25. Monochromatic image of CL with center wavelength at (a) 875 nm

and (b) 825 nm taken by CCD detector.

The band structure corresponding to the high energy transition is not clear, but comparably high energy emission (1.563 eV) has been reported by Williams et. al. in heavily tin doped InP grown by LPE.104 However, the high energy emission around 1.5 eV requires the dopant concentration as high as 8.5×1019/cm3 in tin doped InP. In samples studied in this work, the nominal sulfur concentration is 1.5×1018/cm3 in S-doped InP grown by LP-HVPE and the highest reported emission energy for S-doped InP (n ≈ 5×1018/cm3) is 1.404 eV at 77 K. The 1.515 eV emission requires much higher sulfur concentration. Combining all these observations and results, we postulate that the sulfur concentration is enhanced by diffusion through the core of threading dislocations from seed InP layer during the growth. The exact value of sulfur concentration around dislocation needs further investigation.

33

3.5 InGaAsP MQW at 1.55 µm In Paper H, InGaAsP MQWs at 1.55 µm were grown on undoped and sulfur doped ELOG InP on InP/Si templates respectively. The surface morphology after the growth of quantum wells is shown in Figure 26. For quantum wells grown on undoped templates, line shaped surface defects are observed along [110] especially close to the edges. These defects could be from the surface steps on templates. The surface is rather clean after growth of MQWs on S-doped templates. Some deposition of quantum wells is seen on {111}A and {110} facets, but absent on the {111}B facets. Certain amount of surface undulation can be seen close to one of the edges of templates.

[110]

[110][001]

{111}A{110}

{111

}B

(a)

[110]

[110][001]

{111}A

{111

}B

{110}

MQW onInP seed

Undulation

(b)

Figure 26. SEM micro image of Quantum wells grown on (a) undoped and (b)

sulfur doped ELOG InP on InP/Si templates.

(a) [110]

[110]

[110]

[110](b)

Figure 27. 80K CL Panchromatic micro-image of InGaAsP quantum wells

grown on (a) undoped and (b) S-doped ELOG InP on InP/Si templates.

Defect distribution in the quantum wells was characterized by CL at 80K. The panchromatic micro-images of quantum wells grown on undoped and sulfur doped templates are shown in Figure 27. [110] Dark line defects (DLDs) are observed in the samples with quantum wells on the undoped templates. It seems that the distribution of DLDs coincides with the distribution of surface defects shown in SEM micrographs. That means the surface steps on templates will generate line defects in quantum wells. For the

34

quantum wells grown on S-doped templates, few DLDs and high density dark spot defects (DSDs) are observed. These DSDs could be the threading dislocations in templates that penetrate the quantum wells layer. A long curved dark line is noticed in (b) close to the edge of templates, which corresponds to the surface undulation observed in SEM photos. Figure 28 shows the RT-PL curves of MQWs grown on ELOG InP on InP/Si templates. As shown in the figure, the PL peak wavelength of reference MQW is 1546nm and the peak wavelengths of MQW on undoped and S-doped ELOG templates are around 1570 nm. Although we have reduced dislocation in our ELOG by two orders of magnitude it has still enough dislocations to cause DSDs in spite of that the PL is not quenched indicating a good quality materials.

Figure 28. Room temperature PL spectrum from 6 InGaAsP MQW at 1.55µm grown

on ELOG InP:S/Si, ELOG InP on InP/Si and planar (001)InP:S substrate.

A red-shift of approximately 24 nm with respect to the reference MQW is observed at room temperature. Large red shift of QW peak wavelength has been observed in selective epitaxy of InGaAsP structures, which is caused by thicker wells and richer indium content. The latter is due to the difference in diffusion coefficients of In and Ga under MOVPE growth conditions.105 This occurs even though we removed the Si3N4 mask before MQW growth. As shown in Figure 26, the InP template is bounded by {110}, {111}A and B facet planes. The interfacet migration of adsorbed reactant atoms between these boundary planes and (001) surface plane will change the growth rate and composition of InGaAsP on (001) plane, where the adsorbed species will migrate from low growth rate boundary plane to faster growing plane106. As seen in Figure 26, there is deposition on {111}A planes, but not on {111}B. Species atoms adsorbed on {111}B could diffuse to (001) surface causing thicker layer. The difference of diffusion coefficient between In and Ga will cause In richer layers. Both these effects will shift the MQW to lower energy. In addition to this surface migration effect, the remaining threading dislocations in ELOG InP templates will also enhance the vertical growth rate and cause thicker layer.

3.6 Heteroepitaxy of InP on GaAs Selectivity is an imperative factor for ELOG. However, the dimension of openings is limited by the wavelength of light source in photolithography. This limits the application of ELOG technique to grow nano-sized structures in dissimilar materials system. Focused particle based pattern generation technology, such as ion and electron beam, are capable of achieving very high resolution. In Paper I, we investigated maskless SAG of InP on FIB

35

modified GaAs surface in AP-HVPE system. Continuous growth along [110] direction has been reported while the growth along [110] comprised of small detached islands107, 108. This is due to the formation of low growth rate (111)B facet planes under growth conditions of V/III >1. By manipulating the V/III ratio, the growth rate of the facet planes can be controlled to achieve continuous growth in FIB patterns along [110] direction. Spokes of wheel shaped patterns were formed by 100 keV Ga+ at a beam current of 10 pA FIB implantation on GaAs (001) substrate at different line dose, 0.2, 0.4, 0.8, and 2×1010 cm-1. The angular interval of the spokes is 15°. The estimated diameter of the beam is 20 nm. Then InP growth was carried out in an atmospheric HVPE reactor. The dependence of SAG morphology on III/V ratio was also studied by setting III/V ratios at 10, 12, 20 and 0.2 with constant InCl flow.

(a)

[110]

[110]

(b)

[110]

[110]

Figure 29. Selective growth of InP on FIB modified GaAs planar substrate under III/V ratio of (a) 0.2 and (b) 12. The line dose is 2×1010 cm-1 for both

samples.

The morphology of selective InP growth under III/V ratio of 0.2 is different from that grown under III/V ratio larger than 1 (III/V=12), as shown in Figure 29. For the growth at III/V ratio of 0.2 (Figure 29 (a)), growth on the FIB lines oriented along all the directions except along [110] merge together to form a single layer. Besides, the morphology is rather rough. For the growth at III/V ratio of 12 (Figure 29 (b)), the continuous selective grown InP wires in spoke patterns are well separated with islands scattered between the wires close to the edge of patterned area. The islands are formed uniformly outside the ion implanted area. The selectivity is caused by the higher probability of nucleation on the implanted areas due to an increased number of steps and kinks, and once the nuclei are formed, they act as an effective sink for the migrating surface species leading to the selectivity. The line dose has strong effect on the selective growth of InP in FIB implanted patterns. The minimum dose that is required to initiate continuous InP selective growth is different for different orientations of the FIB lines. The exact [110] direction needs the smallest threshold dose (0.2×1010 cm-1) to create sufficient surface steps and kinks to enhance selective growth in implanted lines. For the lines aligned along 30° and 45° off [110] direction, the threshold dose to obtain continuous growth is 0.8×1010 cm-1. The InP growth rate on FIB lines is assumed to be proportional to that of the islands. The growth rate of islands can be measured in cleaved cross-section as shown in Figure 30.

36

(a)

InP IslandGaAs

(b)

(c)

(d)

(e)

(f)

(g)

(h)

Figure 30. InP islands grown on GaAs substrate at different III/V ratios: 10, (a) [110] view, (b) [110] view; 12, (c) [110] view, (d) [110] view; 20 (e) [110] view, (f) [110] view; 0.2 (g) [110] view, (h) [110] view.

37

[110] 15° 30°45°

-15°-30°

-45°(a)

[110] 15° 30°45°

-15°-30°

-45°(b)

Figure 31. AFM cross section of InP wires grown in FIB lines on GaAs

substrate. (a) [110] view and (b) [110] view. III/V=12.

The vertical growth rate decreases with PH3 flow. At III/V ratio of 0.2, the islands elongate along [110] direction with lengths in the range of 10 µm. As III/V ratio increases, the size of islands decreases in both directions of [110] and [110] but their density is much higher than that at III/V ratio of 0.2. The lateral and vertical growth rates of InP wires grown on FIB lines at III/V ratio of 12 are characterized by atomic force microscopy (AFM), as shown in Figure 31. The lateral growth rate increases, as the FIB lines are misaligned by an angle from the low index directions. This is similar to that observed in ELOG InP growth in openings on Si3N4 mask. The vertical growth rate is high at the low index directions, i.e. [110], [110] and [010] directions on (001) plane. However, in ELOG the vertical growth rate is relative constant.

38

4 Summary of appended papers Paper A Temporally resolved selective area growth of InP in the openings off-oriented from [110] direction Temporally resolved selective area growth of InP on patterned substrates with openings off-oriented from [110] direction was studied by low pressure hydride vapor phase epitaxy system. Lateral overgrowth and vertical growth were analyzed. The lateral growth rate was observed to be strongly dependent on the orientation of the openings. The maximum lateral growth rate was achieved when the openings oriented at 30o and 60o off [110] direction. The vertical growth rate was relatively constant, independent of the opening orientation. The growth behavior of InP in openings aligned at low index directions was explained by dangling bond theory. A new phenomenon of inhomogeneous and orientation dependent dopant distribution within an overgrown layer was observed in stained cross-sections by SEM. Contribution: Experiments design, ELOG growth, analysis and writing Paper B Crystallographic orientation dependence of impurity incorporation during epitaxial lateral overgrowth of InP Temporally resolved epitaxial lateral overgrowth (ELO) of 12 alternating layers of unintentionally doped and S-doped InP layers has been conducted in a low pressure hydride vapor phase epitaxy reactor. The growth was conducted in the openings on (001) n-InP substrate and they were oriented along 30o off [110] direction. Based on the analysis of the cleaved cross-sections by scanning electron microscopy and scanning capacitance microscopy, an inhomogeneous dopant distribution has been observed within the same ELO layer. This is explained by invoking different bonding configurations exposed to the incorporating dopant atoms in the different emerging planes. Contribution: ELOG InP growth, SEM analysis, interpretation and writing Paper C High resolution electrical characterization of laterally overgrown epitaxial InP Dopant incorporation in epitaxial lateral over-growth of InP is investigated by scanning capacitance microscopy (SCM) and scanning spreading resistance microscopy (SSRM). Dramatic variations in doping contrast in the laterally overgrown regions are observed both in SCM and SSRM measurements implying highly inhomogeneous dopant incorporation. The observed doping variations are explained by surface bonding configurations in the different emerging planes during growth. The results show that methods such as SCM and SSRM not only provide detailed electrical information on a nanometer scale, but also give insights into the growth mechanisms.

39

Contribution: ELOG InP growth and discussion. Paper D Selective area growth of InP on InP precoated silicon substrate by hydride vapor phase epitaxy Selective area growth of InP is carried out on InP precoated (001) 2° off Si substrate in low pressure hydride vapor phase epitaxy system under different gas phase supersaturation. Epitaxial lateral overgrowth can be formed on Si3N4 mask under certain growth condition, which has improved crystallographic quality than the InP seed layer. A crystallographic model is proposed to demonstrate the observed asymmetric growth profile of selective area growth. Contribution: ELOG InP/Si growth, analysis, interpretation and writing Paper E Effect of growth conditions on epitaxial lateral overgrowth of InP on InP/Si (001) substrate by hydride vapor phase epitaxy Epitaxial lateral overgrowth of (ELO) InP on (001) InP/Si substrate is explored in a low pressure hydride vapor phase epitaxy system under various growth conditions. The effect of gas phase supersaturation on boundary plane formation of ELO and the behavior of dislocations in the grown layers are investigated. We found that the growth rate on (111)A boundary plane is determined by Burton-Cabrera-Frank model, which predicts a parabolic relationship between gas phase supersaturation and growth rate. Formation of (111)A plane will cause stacking faults in the grown InP layer. They will interact and annihilate each other and introduce fresh dislocations during the growth. Gas phase supersaturation can also be changed by varying opening separation distance. Low gas phase supersaturation is obtained by decreasing the distance between two openings. It gives rise to a lower staking fault density due to the suppression of nucleation of {111} facet plane at the edge of ELO. Etch pit density (EPD) and X-ray diffraction (XRD) techniques are used to estimate the dislocation density. Full width at half maximum of rocking curve at (004), (115) and (117) reflections were used to calculate the dislocation density in ELO InP/Si. Experimentally measured etch pit density is smaller than the dislocation density derived from XRD data. In general the dislocation density is dependent on gas phase supersaturation. In this work, we demonstrate that it is possible to grow high quality InP layer on silicon substrate by epitaxial lateral overgrowth technique under optimized growth conditions. Contribution: ELOG InP/Si growth, analysis, interpretation and writing Paper F Thermal strain in indium phosphide on silicon obtained by epitaxial lateral overgrowth High-resolution X-ray diffraction reciprocal lattice mapping and low-temperature photoluminescence (PL) were used to study the thermal strain in InP layers grown on Si (001) substrate by hydride vapor-phase epitaxial lateral overgrowth (ELO) technique.

40

Good agreement is found between the PL and X-ray measurements. We show that strain in the grown ELO InP/Si layers is affected by the aspect ratio (width to height ratio) of ELO InP layer. Almost strain free InP layer with high crystallographic quality is obtained on Si substrate, which is similar to that of a homoepitaxial InP layer. Contribution: ELOG InP/Si growth, HRXRD analysis, interpretation and writing Paper G Sulfur doped indium phosphide on silicon substrate grown by epitaxial lateral overgrowth The epitaxial lateral overgrowth (ELOG) of sulfur doped InP from ring shaped openings on SiNx masked InP/Si substrate in low pressure hydride vapor phase epitaxy system was investigated. Octahedral shaped ELOG InP templates with smooth surface were formed and studied by cathodoluminescence (CL). High energy transition at 825 nm (1.52 eV) due to the band filling effect caused by high concentration sulfur atoms trapped in threading dislocations was observed in spectra at 80 K. The band edge transition at 875 nm (1.42 eV) in CL spectra has no red shift caused by thermal strain. As observed in panchromatic image, defect free area was surrounded by high density threading dislocations that were pinned by sulfur atoms due to impurity hardening. The quality of the ELOG InP templates is promising for the integration of photonic active layer on Si substrate. Contribution: ELOG InP/Si growth, participate CL, AFM measurement, results analysis and writing Paper H InGaAsP multi-quantum wells at 1.5 µm wavelength grown on indium phosphide templates on silicon Sulfur doped InP templates were deposited in ring patterns on Si substrate by epitaxial lateral overgrowth (ELOG) in low pressure hydride vapor phase epitaxy (LP-HVPE) system. The morphology of ELOG InP was examined by atomic force microscopy (AFM) and the surface is fairly smooth. The thermal strain in these ELOG InP templates was characterized by room temperature photoluminescence (PL) and no red shift of peak wavelength induced by thermal tensile strain in InP templates was detected. Strain compensated InGaAsP 6 periods’ multi-quantum wells (MQW) at 1.5 µm wavelength were grown on these templates by metalorganic vapor phase epitaxy (MOVPE). Morphology of these MQW was characterized by AFM and a rough surface is noticed. The optical property of the MQW on ELOG InP/Si template was characterized by room temperature PL and a large red-shift of 24 nm was detected. This could be due to the thicker and indium richer InGaAsP layers grown on non-planar substrates, where interfacet migration of reactant atoms adsorbed on boundary planes of ELOG InP play an important role. Contribution: ELOG InP/Si template growth, part of PL analysis, results interpretation and writing

41

Paper I Selective growth of InP on focused-ion-beam-modified GaAs surface by hydride vapor phase epitaxy The growth of InP islands on planar Focused-Ion-Beam (FIB)-modified (001) GaAs substrate was investigated in a hydride vapor phase epitaxy system. InP grew selectively on the FIB-implanted lines forming continuous stripes, whereas isolated islands were observed outside the implanted area. The impacts of the III/V ratio, crystallographic orientation of implanted lines and implantation dose were explored. The choice of suitable growth conditions makes it possible to obtain continuous InP wires aligned in all possible directions. The results of this work could be used for the fabrication of future optoelectronic integrated circuits, which would include nanoscale structure, e.g. quantum-wire optical device with GaAs electronic circuits. Contribution: SEM measurements, result analysis, and writing

42

5 Summary, conclusions and future work The thesis work can be summarized as follows. (i) Several InP ELOG experiments were carried out first on InP substrate with the

intention of gaining knowledge to apply to the same on InP/Si. We used an Aixtron LP-HVPE system for this purpose. From the temporally resolved InP ELOG (with n-doped and unintentionally doped InP layers) on InP substrates with openings off-oriented from [110] direction, the lateral growth rate was observed to be strongly dependent on the orientation of the openings, the highest lateral growth rate being for the openings oriented at 30o and 60o off [110] directions. But the vertical growth rate was relatively unaffected by the opening orientation.

(ii) The above experiments also indicated an inhomogeneous and orientation dependent dopant distribution within the same layer. This was qualitatively observed from the stained cross-sections and later confirmed by SCM and SSRM analyses. This has been explained by invoking the bonding configurations exposed to the incorporating dopant atoms in the different emerging planes.

(iii) When InP ELOG was conducted on InP/Si, unlike that on pure InP substrates, the lateral growth was not symmetric on both sides because of larger growth rates along certain directions, e.g. along <111>A. These were explained to be due to the influence of the defects (in the seed layer) that propagate even during ELOG. For example, a higher concentration of threading dislocations intersecting the surface of the {111}A emerging planes would cause a higher growth rate of these planes. Another example is that if a (111)A plane and a (111)B plane emerge simultaneously in such a way that they intersect each other, the defects on these planes will also intersect each other. This means, in this case, that plane containing a larger number of defects will enhance the growth of the other plane. From the fact that {111}A planes have higher growth rates, we believe that in the seed layer, the defect density on {111}B is larger than that on {111}A. The growth rate on the emerging plane can also be varied by changing the supersaturation. We found the growth rate on the {111}A is approximately related to the vapor phase supersaturation by a parabolic function, as predicated by BCF model. This also explains the asymmetric growth. In our study we varied the supersaturation both by means of excess HCl and by varying the distance between the openings.

(iv) Dislocation density in the InP ELOG on InP/Si was estimated from the full width at half maximum of XRD rocking curves of (004), (115) and (117) reflections and compared with EPD obtained by chemical revelation. These values were ~ 4×107 cm-2. Thus threading dislocation density has been reduced from ∼ 4×109 cm-2 in the seed layer. If the seed layer is of a better quality, the ELOG layer will also be.

(v) Combination of HRXRDRLM and LT-PL measurements indicate that in InP ELOG layers on InP/Si the strain is almost non-existent (as in a homoepitaxial layer) for the layers with high aspect ratio.

(vi) When ELOG of sulfur doped InP was conducted on ring shaped openings on InP/Si substrate instead of straight openings, octahedral shaped ELOG InP templates with smooth surface were formed as indicated by AFM. CL indicates a good quality ELOG in the middle of the template. Strain compensated InGaAsP 6 periods’ multi-quantum wells (MQW) at 1.5 µm wavelength (target value) were grown on these templates by metalorganic vapor phase epitaxy (MOVPE). Although the surface was slightly rough it exhibited RT-PL with a large red-shift of approximately 24 nm. RT-PL indicates the good quality of the ELOG layers, although CL measurement

43

indicates some DSDs. The red-shift was expected due to the thicker and indium richer InGaAsP layers grown on non-planar substrates.

(vii) As an extension of ELOG of InP on InP/Si, we conducted the growth of InP islands on planar Focused-Ion-Beam (FIB)-modified (001) GaAs substrate in an atmospheric HVPE. The impact of the III/V ratio, crystallographic orientation of implanted lines and implantation dose was explored. The choice of suitable growth conditions makes it possible to obtain continuous InP wires aligned in all possible directions.

The major conclusions that can be drawn and some suggestions for future work are: (i) The symmetric growth achieved on both sides of the straight openings in InP ELOG

on InP substrates is apparently difficult on InP/Si substrates due to a high density of dislocations in the seed layer. The surface morphology was also not uniform. However the ring openings yielded better symmetrical growth with a better surface morphology. From this one can conclude that a more effective defect hindering mechanism is operative in the ELOG from ring openings. We suggest that further investigation on this with various dimensions of the rings, distance between the rings etc. One can also investigate the possibility of a complete merger. The latter can also be used as a template in future Micro-Opto-Electro-Mechanical systems (MOEMS) and photonic crystal applications.

(ii) GaAs on Si may be attractive for fabricating e.g. short wavelength VCSELs for optical interconnects on silicon. ELOG can be attempted on the ring openings since it appears to be very attractive.

(iii) EPD in the ELOG layer has been reduced to ∼4×107 cm-2 from ∼4×109 cm-2 in the seed layer. Thus ELOG definitely reduces the defect density. Since we had only one source of the seed layer, we suggest to obtain other sources with 1×105 cm-2 which can hopefully lead to an ELOG layer of EPD with ∼1×103 cm-2.

44

Appendix: Properties of InP

Quantity Unit InP

Crystal Structure _ Zinc blende

Lattice Constant Å 5.8686

Band Gap Energy

300K eV 1.350

2K eV 1.425

Thermal Expansion 10-6/K 4.56

Elastic Constant

C11 dyn/cm2 1.022×1012

C12 dyn/cm2 0.576×1012

Deformation Potential

a eV -6.694

b eV -1.55

45

References 1 H. Mori, M. Sugo, and Y. Itoh, “Heteroepitaxy of InP on Si substrates”, Adv. Mater., 5 208-209 (1993) 2 S. Naritsuka, T. Nishinaga, M. Tachikawa, and H. Mori, “InP layer grown on (001) silicon substrate by epitaxial lateral overgrowth”, Jpn. J. Appl. Phys., 34 L1432-L1435 (1995) 3 A. Bourret, “Compliant substrates: a review on the concept, techniques and mechanisms”, Appl. Surf. Sci., 164 3-14 (2000) 4 J. Ahopelto, M. Sopanen, H. Lipsanen, S. Lourdudoss, E. Rodriguez Messmer, E. Hofling, J.P. Reithmaier, A. Forchel, A. Petersson, and L. Samuelson, “Maskless selective growth of InGaAs/InP quantum wires on (100) GaAs”, Appl. Phys. Lett., 70 2828-2830 (1997) 5 M. Fukuda, Reliability and degradation of semiconductor lasers and LEDs, Artech House, Norwook, (1991) 6 G.H. Olser, C.J. Nuese, and M. Ettenberg, “Reliability of vapor-grown InGaAs and InGaAsP heterojunction laser structures”, IEEE J. Quantum Electron., QE-15 688-693 (1979) 7 O. Ueda, “Materials-related reliability aspects of III-V optical devices”, Mater. Sci. and Eng., B20 9-20 (1993) 8 T. Sasaki, H. Mori, M. Tachikawa, and T. Yamada, “Aging testes of InP-based laser diodes heteroepitaxially grown on Si substrates”, J. Appl. Phys., 84 6725-6728 (1998) 9 J.P. Van Derziel and R.D. Dupuis, “Degradation of GaAs lasers and light-emitting diodes on silicon substrates”, Mater. Sci. and Eng., B1 37-45 (1988) 10 P. Peroff and R.L. Hartman, “Defect structure introduced operation of heterojunctions GaAs laser”, Appl. Phys. Lett., 23 469-471 (1973) 11 M. Sugo, M. Yamaguchi, and M.M. Al-Jassim, “Heteroepitaxial growth of InP on Si substrates”, J. Crystal Growth, 99 365-370 (1990) 12 M.K. Lee, D.S. Wuu, H.H. Tung, J.H. Chang, and Y.F. Lin, “Growth and characterization of InP epi-layers on ZnSe-coated Si substrates by low-pressure metalorganic vapor phase epitaxy”, App. Phys. Lett., 53 107-109 (1988) 13 A. Seki, F. Konushi, J. Kudo, and M. Kuba, “Epitaxial growth of InP on Si by OMVPE-defect reduction in epitaxial InP using InAs/sub x/P/sub 1-x//InP superlattices”, J. Crystal Growth, 93 527-531 (1988)

46

14 T.E. Crumbaker, H.Y. Lee, M.J. Hafich, and G.Y. Robinson, “Growth of InP on Si substrates by molecular beam epitaxy”, App. Phys. Lett., 54 140-142 (1989) 15 K. Samonji, H. Yonezu, Y. Takagi, K. Iwaki, N. Ohshima, J.K. Shin, K. Pak, “Reduction of threading dislocation density in InP-on-Si heteroepitaxy with strained short-period superlattices”, App. Phys. Lett., 69 100-102 (1996) 16 Y. Kohama, Y. Kadota, and Y. Ohmachi, “Heteroepitaxy of InP on Si substrates by MOCVD”, J. Elechem. Soc., 136 3853-3856 (1989) 17 H. Mori, M. Sugo, and Y. Itoh, “Heteroepitaxy of InP on Si substrates”, Adv. Mate., 5 208-209 (1993) 18 R.F. Schnabel, L. Grundmann, A. Krost, J. Christen, F. Heinrichsdorff, D. Bimberg, and H. Cerva, “Defect reductions and strain relaxation mechanisms in InP grown on patterned Si(001)”, 6th International Conference on Indium Phosphide and Related Materials, 1994, 640-643 19 A. Bakin, D. Piester, I. Behrens, H.H. Wehmann, E. Peiner, D. Fehly, and A. Schlachetzki, “Growth of InP layers on nanometer-scale patterned Si substrates”, Crystal Growth & Des., 3 89-93 (2003) 20 G.W. Neudeck, P.J. Schubert, J.L. Glenn,J.A. Friedrich, W.A. Klaasen,R.P. Zingg, and J.P. Denton, “Three dimensional devices fabricated by silicon epitaxial lateral overgrowth”, J. Electro. Mater., 19 1111-1117 (1990) 21 L. Jastrzebski, “SOI by CVD: epitaxial lateral overgrowth (ELO) process-review”, J. Crystal Growth, 63 493-526 (1983) 22 N. Vodjdani, M. Erman and J.B. Theeten, “Structural analysis and optical characterization of low loss GaAs waveguides fabricated by selective epitaxy”, J. Crystal Growth, 71 141-148 (1985) 23 B-Y. Tsaur, R.W. McClelland, J.C.C. Fan, R.P. Gale, J.P. Saöerno, B.A. Vojak, and C.O. Bozler, “Low-dislocation-density GaAs epi-layers grown on Ge-coated Si substrates by means of lateral epitaxial overgrowth”, App. Phy. Lett., 41 347-349 (1982) 24 T. Miyajima, T. Tojyo, T. Asano, K. Yanashima, S. Kijima, T. Hino, M. Takeya, S. Uchida, S. Somiya, K. Funato, T. Asatsuma, T. Kobayashi, M. Ikeda, “GaN-based blue laser diodes”, J. Phy.: Condensed Matter, 13 7099-7114 (2001) 25 Yingda Dong, Yae L. Okuno and Umesh K. Mishra, “MOCVD selective growth of InP through narrow openings and its application to InP HBT extrinsic base regrowth”, J. Crystal Growth, in press 26 D. Jahan, D. Söderström, and S. Lourdudoss, “LP-HVPE growth of S, Fe and undoped InP”, 11th International conference on indium phosphide and related materials, IEEE, Davos, Switzerland, 1999, 123-126

47

27 K. Grüter, M. Deschler, H. Jürgensen, R. Beccard, and P. Balk, “Deposition of high quality GaAs films at fast rates in the LP-CVD system”, J. Crystal Growth, 94 607-612 (1989) 28 W.K. Burton, N. Cabrera and F.C. Frank, “The growth of crystals and the equilibrium structure of their surfaces”, Philos. Trans. R. Soc. London, Ser. A, 243 299-358 (1951) 29 B. Mutaftschiev, “Crystal growth and dislocations” in F.N.R. Nabarro (ed.), Dislocation in Solids, Vol. 5, North-Holland, Amsterdam, (1980) 30 S. Zhang and T. Nishinaga, “LPE lateral overgrowth of GaP”, Jpn. J. Appl. Phys., 29 545-550 (1990) 31 S. Naritsuka and T. Nishinaga, “Epitaxial lateral overgrowth of InP by liquid phase epitaxy”, J. Crystal Growth, 146 314-318 (1995) 32 S. Zhang and T. Nishinaga, “Epitaxial lateral overgrowths of GaAs on (001) GaAs substrates by LPE: Growth behavior and mechanism”, J. Crystal Growth, 99 292-296 (1990) 33 W.Y. Uen and T. Nishinaga, “Liquid-phase epitaxial lateral overgrowth of GaAs on 0.3°-misoriented epitaxial Si substrates”, Mater. Chem. & Phys., 42 231-236 (1995) 34 S. Nozaki, J.J. Murray, and A.T. Wu, “New mechanism for Si incorporation in GaAs-on-Si heteroepitaxial layers grown by metalorganic chemical vapour deposition”, App. Phys. Lett., 55 1674-1676 (1989) 35 Nils Nordell, “Metalorganic vapour phase epitaxy for advanced III-V devices”, Doctoral Dissertation, Stockholm, 1993 36 G. Bacchin, T. Nishinaga, “Fabrication of submicrometer structures by PSE/MBE”, J. Crystal Growth, 211 389-394 (2000) 37 G. Bacchin, A. Umeno and T. Nishinaga, “Lateral growth in molecular beam epitaxy by low angle incidence microchannel epitaxy”, Appl. Surf. Sci., 159-160 270-276 (2000) 38 D.W. Shaw, “Kinetic aspects in the vapour phase epitaxy of III-V compounds”, J. Crystal Growth, 31 130-141 (1975) 39 R.C. Sangster, “Model studies of crystal growth phenomena in the III-V semiconducting compounds”, in R.K. Willardson, H.L. Goering(Eds.), compound Semiconductors, Volume 1, Chapter 28, Reinhold Publishing Corporation, NY, 241-253 (1962) 40 S. Lourdudoss, E. Rodríguez Messmer, O, Kjebon, and G. Landgren, “Temporally resolved regrowth of InP”, J. Crystal Growth, 152 105-114 (1995) 41 J.P. Hirth and J. Lothe, Theory of dislocations, John Wiley & Sons (1982)

48

42 P. Pirouz, F. Ernst and T.T. Cheng, “Heteroepitaxy on (001) Silicon: growth mechanisms and defect formation”, in H.K. Choi, R. Hull, H. Ishiwara R.J. Nemanich (ed.) Heteroepitaxy on Silicon, Mater. Res. Soc. Proc. 116 Pittsburg, PA (1988) 43 J.D.H. Donnay and G. Donnay, “Crystal geometry”, in J.S. Kasper and K. Lonsdale (Eds.), International Tables for X-ray Crystallography, Volume II, Section 3, The Kynoch Press, England, 101-158 (1959) 44 H.J. Queisser and E.E. Haller, “Defects in semiconductors: some fatal, some vital”, Science, 281 945-950 (1998) 45 M.E. Weiner, “Si contamination in open flow quartz systems for the growth of GaAs and GaP”, J. Electrochem. Soc., 119 496-504 (1972) 46 A. Trassoudaine, O. Parillaud, E. Goumet, D. Castelluci, E. Gil-Lafon, and R. Cadoret, “Kinetic study of Si incorporation in InP by the hydride vapour phase epitaxy”, J. Crystal Growth, 192 402-409 (1998) 47 E. Veuhoff, T.F. Kuech, and B.S. Meyerson, “A study of silicon incorporation in GaAs MOCVD layers”, J. Electrochem. Soc., 132 1958-1961 (1985) 48 D. Grundmann, H. Jürgensen, M. Heyen, J. Korec, P. Balk, “Sulfur incorporation during epitaxial growth of InP in the In-HCl-PH3-H2 system”, J. Ele. Mate., 14 749-767 (1985) 49 J. Korec, D. Grundmann, and M. Heyen, “Modeling of sulfur incorporation during low pressure CVD of GaAs (100) in the Ga-HCl-AsH3-H2-H2S system”, J. Electrochem. Soc., 131 1433-1438 (1984) 50 G.B. Stringfellow, “The role of impurities in III/V semiconductors grown by organmetallic vapor phase epitaxy”, J. Crystal Growth, 75 91-100 (1986) 51 R.A. Logan, T. Tanbun-Ek, and A.M. Sergent, “Doping of InP and GaInAs with S during metalorganic vapor-phase epitaxy”, J. Appl. Phys., 65 3723-3725 (1989) 52 B. Lee, S.S. Bose, M.H. Kim, A.D. Reed, and G.E. Stillman, “Orientation dependent amphoteric behavior of group IV impurities in the molecular beam epitaxial and vapor phase epitaxial growth of GaAs”, J. Crystal Growth, 96 27-39 (1989) 53 A.G. Thompson, “The effects of substrate misorientaion on silicon doping efficiency in MOVPE grown GaAs”, J. Crystal Growth, 124 457-462 (1992) 54 R. Bhat, C. Caneau, C.E. Zae, M.A. Koza, W.A. Bonner, D.M. Hwang, S.A. Schwarz, S.G. Menocal, and F.G. Favire, “Orientation dependence of S, Zn, Si, Te, and Sn doping in OMCVD growth of InP and GaAs: application to DH lasers and lateral p-n junction arrays grown on non-planar substrates”, J. Crystal Growth, 107 772-778 (1991) 55 M. Kondo, C. Anayama, N. Okada, H. Sekiguchi, K. Domen, and T. Tanahashi, “Crystallographic orientation dependence of impurity incorporation into III-V compound

49

semiconductors grown by metalorganic vapor phase epitaxy”, J. Appl. Phys., 76 914-927 (1994) 56 G. P. Watson, D.G. Ast, T.J. Anderson, and B. Pathangey, “The barrier to misfit dislocation glide in continuous, strained, epitaxial layers on patterned substrates”, J. Appl. Phys., 74 3103-3110 (1993) 57 B.G. Yacobi and D.B. Holt, “Cathodoluminescence scanning electron microscopy of semiconductors”, J. Appl. Phys., 59, R1-R26 (1986) 58 Y. Seki, H. Watanabe, and J. Matsui, “Impurity effect on grown in dislocation density of InP and GaAs crystal”, J. Appl. Phys., 49 822-826 (1978) 59 A.S. Jordan, G.T. Brown, B. Cockayne, D. Brasen, and W.A. Bonner, “An analysis of dislocation reduction by impurity hardening in the liquid-encapsulated czochralski growth of <111> InP”, J. Appl. Phys., 58 4383-4389 (1985) 60 G. Jacob, M. Duseaux, J. P. Farges, M. M. B. van den Boom and P. J. Roksnoer, “Dislocation-free GaAs and InP crystals by isoelectronic doping ”, J. Crystal Growth, 61 417-424 (1983) 61 I. Yonenaga and K. Sumino, “Impurity effects on the generation, velocity, and immobilization of dislocations in GaAs”, J. Appl. Phys., 65 85-92 (1988) 62 I. Yonenaga and K. Sumino, “Effects of dopants on dynamic behaviour of dislocations and mechanical strength in InP”, J. Appl. Phys., 74 917-927 (1993) 63 S. Kalisetty, M. Gokhale, K. Bao, J.E. Ayers, and F.C. Jain, “The influence of impurities on the dislocation behavior in heteroepitaixal ZnSe on GaAs”, Appl. Phys. Lett., 68 1693-1695 (1996) 64 Tetsuo Soga, Shuzo Hattori, Shiro Sakai, Masanari Takeyasu, and Masayoshi Umeno, “Characterization of epitaxially grown GaAs on Si substrates with III-V compounds intermediate layers by metalorganic chemical vapor deposition”, J. Appl. Phys., 57 4578-4582 (1985) 65 D. K. Biegelsen, F. A. Ponce, A. J. Smith, and J. C. Tramontana, “Initial stages of epitaxial growth of GaAs on (100) silicon”, J. Appl. Phys., 61 1856-1859(1987) 66 F. Ernst and P. Pirouz, “Formation of planar defects in the epitaxial growth of GaP on Si substrate by metal organic chemical-vapor deposition”, J. Appl. Phys., 64 4526-4530 (1988) 67 H.H. Wehmann, G.P. Tang, A. Koch, M. Seibt, and A. Schlachetzki, “Twin formation during epitaxial growth of InP on Si”, Solid State Phen., 47-48 547-552 (1996) 68 H, Nagai, “Structure of vapour-deposited Ga/sub x/In/sub 1-x/As crystals”, J. Appl. Phys., 45 3789-3794 (1974)

50

69 A. Yamamot and M. Yamaguchi, “Heteroepitaxy of GaAs, GaP and InP on Si by OMVPE and their application to thin film solar cells”, Mat. Res. Soc. Symp. Proc., Vol. 116 285-296 (1988) 70 G.H. Olsen, C.J. Nuese, and R.T. Smith, “The effect of elastic strain on energy band gap and lattice parameter in III-V compounds”, J. Appl. Phys., 49 5523-5529 (1978) 71 C.P. Kuo, S.K.Vong, R.M. Cohen, and G.B. Stringfellow, “Effect of mismatch strain on band gap in III-V semiconductors” J. Appl. Phys., 57 5428-5432 (1985) 72 D. A. Dahl, L. J. Dries, F. A. Junga, W. G. Opyd, and P. Chu, “Photoluminescence in strained InGaAs/GaAs superlattices”, J. Appl. Phys., 61 2079 (1987) 73 D. Gershoni, J. M. Vandenberg, S. N. G. Chu, H. Temkin, T. Tanbun–Ek, and R. A. Logan, “Excitonic transitions in strained-layer InxGa1-xAs/GaAs quantum wells”, Phys. Rev., B 40 10017-10020 (1989) 74 B.A. Wilson, “Novel applications of optical techniques to the study of buried semiconductor interfaces”, IEEE. J. Quantum Electron., QE-25, 1012-1017 (1989) 75 D.S. Wuu, R.H. Horng, and M.K. Lee, “Indium phosphide on silicon heteroepitaxy: Lattice deformation and strain relaxtion”, J. Appl. Phys., 68 3338-3342 (1990) 76 M. Sugo, N. Uchida, A. Yamamoto, T. Nishioka, and M. Yamaguchi, “Residual strains in heteroepitaxial III-V semiconductor films on Si (001) substrates”, J. Appl. Phys., 65 591-595 (1989) 77 L.A. Coldren and S.W. Corzine, Diode Lasers and Photonic Integrated Circuits, John Wiley & Sons, Inc., pp531 (1995) 78 Properties of Indium Phosphide, INSPEC, The Institute of Electrical Engineering, chapter 1, pp.10 (1991) 79 S. Adachi, Physical Properties of III-V Semiconductor compounds, John Wiley & Sons, Inc., pp274 (1992) 80 P.F. Fewster and N.L. Andrew, “Determining the lattice relaxation in semiconductor layer systems by x-ray diffraction”, J. Appl. Phys., 74 3121-3125 (1994) 81 P. van der Sluis, “Determination of strain in epitaxial semiconductor layers by high-resolution x-ray diffraction”, J. Phys. D: Appl. Phys., 26 A188-A191 (1993) 82 H. Heinke, M. O. Möller, D. Hommel and G. Landwehr, “Relaxation and mosaicity profiles in epitaxial layers studied by high resolution X-ray diffraction”, J. Crystal Growth, 135 41-52 (1994) 83 J. Hornstra and W.J. Bartels, “Determination of the lattice constant of epitaxial layers of III-V compounds”, J. Crystal Growth, 44 513-517 (1978)

51

84 M.S. Goorsky, M. Meshkinpour, D.C. Streit, and T.R. Block, “Diffuse x-ray scattering from misfit dislocations at semiconductor hetero-interfaces”, J. Phys. D: Appl. Phys., 28 A92-A96 (1995) 85 V.M. Kaganer, R. Köhler, M. Schmidbauer, and R. Opitz, “X-ray diffraction peaks due to misfit dislocations in hteroepitaxial structrure”, Phys. Rev., B 55 1793-1810 (1997) 86 P.Gay, P.B. Hirsch, and A. Kelly, “The estimation of dislocation densities in metals from x-ray data”, Acta Metal., 1 315-319 (1953) 87 E. Koppensteiner, A. Schuh, G. Bauer, V. Holý, G.P. Watson, and E.A. Fitzgerald, “Determination of threading dislocation density in hetero-epitaxial layers by diffuse x-ray scattering”, J. Phys. D: Appl. Phys., 28 A114-A119 (1995) 88 Syed B. Qadri and J.H. Dinan, “X-ray determination of dislocation density in epitaxial ZnCdTd”, Appl. Phys. Lett., 47 1066-1068 (1985) 89 J.E. Ayers, “The measurement of threading dislocation densities in semiconductor crystals by X-ray diffraction”, J. Crystal Growth, 135 71-77 (1994) 90 M.J. Hordon and B.L. Averbach, “X-ray measurements of dislocation density in deformed copper and aluminum single crystals”, Acta Metal., 9 237-246 (1961) 91 B.E. Warren, theory presented in the Course on X-Ray and Crystal Physics, Massachusetts Institute of Technology (1957) 92 S.Anand, “Another dimension in devices characterization”, IEEE Circuits and Devices 16, 12-18 (2000) 93 P. De Wolf, M. Geva, T. Hantschel, W. Vandervorst, and R. B. Bylsma, “Two-dimensional carrier profiling of InP structures using scanning spreading resistance microscopy”, Appl. Phy. Lett., 73 2155-2157 (1998) 94 P. De Wolf, M. Geva, C. L. Reynolds, T. Hantschel, W. Vandervorst, R. B. Bylsma, “Two-dimensional carrier profiling of InP-based structures using scanning spreading resistance microscopy”, J. Vac. Sci. Technol., A 17 1285-1288 (1999) 95 M.W. Xu, T. Hantschel, and W. Vandervorst, “Reliable 2-D carrier profiling with SSRM on InP-based devices”, Proc. Indium Phosphide and Related Materials, Nara, Japan 541-544 (2001) 96 D.W. Shaw, “Epitaxial GaAs kinetic studies: {001} orientation”, J. Electrochem. Soc., 117 683-687 (1970) 97 S.K. Choi, M. Mihara and T. Ninomiya, “Dislocation velocities in GaAs”, Jpn. J. Appl. Phys., 16 737-745 (1977)

52

98 C. Friger, R. Gleichmann, C. Pelosi and G. Attolini, “Relationship between dislocation generation, vapour phase supersaturation and growth rate in InP layers obtained by vapour phase epitaxy”, Mate. Sci. and Eng., B10 197-207 (1991) 100 E.Gil-Lafon, J. Napierala, D. Castelluci, A. Pimpinelli, R. Cadoret, and B. Gérard, “Selective growth of GaAs by HVPE: keys for accurate control of the growth morphologies”, J. Crystal Growth, 222 482-496 (2001) 101 A. Huber and N.T. Linh, “Metallographic display of crystal defects in InP”, J. Crystal Growth, 29 80-84 (1975) 102 H. Gottscharlk, G. Patzer and H. Alexander, “Stacking fault energy and ionicity of cubic compounds”, Phys. Status Solidi (A), 45 207-217 (1978) 103 H.P. Lee, Y.-H. Huang, X. Liu, H. Lin, J.S. Smith, E.R. Weber, P. Yu, S. Wang and Z.Liliental-Weber, “The photoluminescence and TEM studies of patterned GaAs films on Si substrate grown by molecular beam epitaxy”, Mat. Res. Soc. Symp. Proc., Vol. 116 219-226 (1988) 104 E.W. Williams, W. Elder, M.G. Astles, M. Webb, J.B. Mullin, B. Straughan, and P.J. Tufton, “Indium phosphide”, J. Electrochem. Soc., 120 1741-1760 (1973) 105 C.H. Joyner, S. Chandrasekhar, J.W. Sulhoff and A.G. Dentai, “Extremely large band gap shifts for MQW structures by selective epitaxy on SiO2 masked substrates”, IEEE Photon. Tech. Lett., 4 1006-1009 (1992) 106 A. Madhukar, “Growth of semiconductor heterostructures on patterned substrates: defect reduction and nanostructures”, Thin Solid Films, 231 8-42 (1993) 107 H.J. Lexec, J. Ahopelto, A. Usui, and Y. Ochiai, “Focused-ion-beam surface modification for selective growth of InP wires on GaAs”, Jpn. J. Appl. Phys., Part 1, 32 6251-6257 (1993) 108 J. Ahopelto, H. Lezec, Y. Ochiai, A. Usui, and H. Sakaki, “Maskless InP wire formation on planar GaAs substrates”, Appl. Phys. Lett., 64 499-501 (1994)