COMPOSITE CONTACT METALLIZATION ON SIC FOR HIGH TEMPERATURE APPLICATIONS IN AIR Except where reference is made to the work of others, the work described in this dissertation is my own or was done in collaboration with my advisory committee. This dissertation does not include proprietary or classified information. Adetayo V. Adedeji Certificate of Approval: Chin-Che Tin Professor Department of Physics John R. Williams, Chair Professor Department of Physics Jianjun Dong Associate Professor Department of Physics Minseo Park Assistant Professor Department of Physics Stephen L. McFarland Acting Dean, Graduate School COMPOSITE CONTACT METALLIZATION ON SIC FOR HIGH TEMPERATURE APPLICATIONS IN AIR Adetayo V. Adedeji A Dissertation Submitted to the Graduate Faculty of Auburn University in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy Auburn, Alabama August 8, 2005 COMPOSITE CONTACT METALLIZATION ON SIC FOR HIGH TEMPERATURE APPLICATIONS IN AIR Adetayo V. Adedeji Permission is granted to Auburn University to make copies of this thesis at its discretion, upon the request of individuals or institutions and at their expense. The author reserves all publication rights. Signature of Author Date of Graduation iii DISSERTATION ABSTRACT COMPOSITE CONTACT METALLIZATION ON SIC FOR HIGH TEMPERATURE APPLICATIONS IN AIR Adetayo V. Adedeji Doctor of Philosophy, August 8, 2005 (B.S., University of Ilorin, Nigeria, 1989) (DICTP, International Center for Theoretical Physics, Italy, 1994) (M.S., Obafemi Awolowo University, Nigeria, 1997) 204 Typed Pages Directed by John R. Williams Composite ohmic and Schottky contacts fabricated on 4H-SiC for applications in air at 350a0a2a1 are described and evaluated in this study. The ohmic contacts were fabricated on pa3 implanted and na3 epitaxial materials, while the Schottky contacts were on na4 epilayers. These contacts were protected from the harsh environment with the following metallization layers: Ta-Si-N served as diffusion/oxidation barrier, Pt-N sputter-deposited at 250a0 a1 promoted adhesion between the diffusion barrier layer and the Au cap layer and it also acts as a barrier to Au diffusion. A gold cap layer was deposited for the purpose of wirebonding to the device or die attach to a carrier. The composite contact?s electrical and physical characteristics were monitored as a function of annealing time in air at 350a0 a1 . For the ohmic contact, characteristics such as the specific contact resistance and SiC sheet resistance determined from TLM measurements were monitored with time of anneal. Contact parameters of nickel ohmic contact on heavily implanted pa3 4H-SiC materials were very stable up to 4000 hours at 350a0 a1 in air. 70-30 wt% Al-Ti ohmic contacts iv on implanted pa3 , 80-20 wt% Ni-Cr and Ni ohmic contacts on epitaxial na3 4H-SiC were also studied. Current density - voltage and capacitance - voltage characteristics of nickel silicide and Ta- Si-N schottky contacts were studied as a function of annealing time in air. Parameters such as the barrier height and ideality factor of the devices were monitored and found to be stable as the devices were annealed in air. Adhesion characteristics of the metallization stack (barrier layer, Pt layer and Au layer) on thermal oxides, Schottky contacts and ohmic contacts were investigated in this study. All the contacts have good adhesion to the metallization as fabricated except for the nickel ohmic contacts that needed Ar ion cleaning to improve its adhesion to the metallization stack. In all cases, inter-diffusion within the metallization and oxidation of the contacts were monitored with RBS (Rutherford backscattering spectrometry) and AES (Auger electron spec- troscopy). v ACKNOWLEDGMENTS I give thanks to God for the gift of life, good health and the opportunity to be able to take part in this study. The guidance and support of my advisor, professor John Williams, throughout the period of my graduate study at Auburn University are acknowledged and highly appreciated. Without his wisdom, patience and persistence, it would be impossible to achieve the goals of this study. I am indebted to Dr. Claude Ahyi, a post-doctoral fellow in our laboratory, for his contribu- tions and useful discussion during the course of this work. The assistance of Dr. Shurui Wang, also a post-doctoral fellow in our laboratory, are appreciated. Valuable contributions of Mrs. Tamara Isaac-Smith for the initiation and successful completion of this work are acknowledged with gratiude. The contributions of Mr. Max Cichon in building and maintaining essential equip- ment needed for the success of this work - including the heating system used during platinum sputter-deposition - are highly appreciated. This acknowledgement will fall short without acknowledging the support, encouragement, and assistance of my beloved sweetheart, Dolapo Abimbola Adedeji, during these years of my study. I truly appreciate our boys, Jethro and Enoch, with thanks to the Lord for their endurance during the period of this study. vi Style manual or journal used Transactions of the American Mathematical Society (together with the style known as ?auphd?). Bibliograpy follows the style used by the American Physical Society. Computer software used The document preparation package TEX (specifically LATEX) together with the departmental style-file auphd.sty. vii TABLE OF CONTENTS LIST OF FIGURES x LIST OF TABLES xvi 1 GENERAL INTRODUCTION AND SILICON CARBIDE 1 1.1 General Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1.1 Why This Work? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.1.3 To Accomplish the Objectives . . . . . . . . . . . . . . . . . . . . . . 3 1.2 Silicon Carbide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.2.2 Crystal Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.2.3 Semiconducting Properties . . . . . . . . . . . . . . . . . . . . . . . . 9 1.2.4 Devices and Applications . . . . . . . . . . . . . . . . . . . . . . . . 11 2 METAL-SEMICONDUCTOR CONTACTS AND DIFFUSION BARRIERS 19 2.1 Metal-Semiconductor Contacts . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.1.1 General Information . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.1.2 Metal-SiC Contacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 2.2 Diffusion and Oxidation Barriers . . . . . . . . . . . . . . . . . . . . . . . . . 41 2.2.1 Diffusion in Solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 2.2.2 Thermodynamics of Diffusion . . . . . . . . . . . . . . . . . . . . . . 46 2.2.3 Diffusion Barriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 3 ANALYTICAL TECHNIQUES 61 3.1 Electrical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 3.1.1 Sheet Resistance Measurement . . . . . . . . . . . . . . . . . . . . . . 61 3.1.2 Contact Resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 3.1.3 Linear Transmission Line Model Measurement . . . . . . . . . . . . . 66 3.1.4 I-V Charateristics of Schottky Diodes . . . . . . . . . . . . . . . . . . 70 3.1.5 C-V Charateristics of Schottky Diodes . . . . . . . . . . . . . . . . . . 74 3.2 Physical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 3.2.1 Rutherford Backscattering Spectrometery (RBS) . . . . . . . . . . . . 77 3.2.2 Auger Electron Spectroscopy (AES) . . . . . . . . . . . . . . . . . . . 92 3.2.3 X-ray Photoelectron Spectroscopy (XPS) [166] . . . . . . . . . . . . . 97 3.2.4 X-ray Diffraction (XRD) . . . . . . . . . . . . . . . . . . . . . . . . . 99 3.3 Mechanical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 3.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 viii 4 EXPERIMENTAL PROCEDURES 106 4.1 Standard Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 4.1.1 Sample Cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 4.1.2 Photolithography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 4.1.3 Sputter-deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 4.1.4 Annealing System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 4.2 Device Fabrication and Measurements . . . . . . . . . . . . . . . . . . . . . . 114 4.2.1 Ohmic Contact Fabrication . . . . . . . . . . . . . . . . . . . . . . . . 114 4.2.2 Schottky Contact Fabrication . . . . . . . . . . . . . . . . . . . . . . . 116 4.2.3 Protective Stack Fabrication and Device Mesurements . . . . . . . . . 118 4.2.4 Wirebonding and Chip Brazing . . . . . . . . . . . . . . . . . . . . . 122 4.2.5 Thermal Aging of Samples . . . . . . . . . . . . . . . . . . . . . . . . 124 4.2.6 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 5 RESULTS AND DISCUSSION 128 5.1 The Choice of Barrier Layer Material . . . . . . . . . . . . . . . . . . . . . . 128 5.1.1 Silicide Barrier Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 5.1.2 Structure of Ta-Si-N . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 5.2 Results on Ohmic Contacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 5.2.1 Contacts Without Protective Metallization . . . . . . . . . . . . . . . . 133 5.2.2 Ohmic Contact with Protective Metallization . . . . . . . . . . . . . . 138 5.2.3 Metallization with Hot Platinum . . . . . . . . . . . . . . . . . . . . . 144 5.3 Results for Schottky Contacts . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 5.3.1 I-V and C-V Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 5.4 Wirebonding and Brazing Results . . . . . . . . . . . . . . . . . . . . . . . . 164 5.4.1 Chip Shear Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 5.4.2 Wirebond Pull and Shear Testing . . . . . . . . . . . . . . . . . . . . . 168 6 CONCLUSIONS 174 BIBLIOGRAPHY 178 ix LIST OF FIGURES 1.1 Typical composite metallization scheme ( [2] ) . . . . . . . . . . . . . . . . . . 4 1.2 (a) basic structural unit of SiC, (b) 3C-SiC stacking sequence, (c) 4H-SiC stack- ing sequence and (d) 6H-SiC stacking sequence. The Jagodzinski (kkk, khkh, kkhkkh) and ABCA?B?C? notations are also shown [24]. . . . . . . . . . . . . 10 1.3 Basic FET structure [29] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.1 I-V Characteristics of ohmic and schottky contact . . . . . . . . . . . . . . . . 20 2.2 Energy level diagram of metal and semiconductor . . . . . . . . . . . . . . . . 21 2.3 Energy level diagram for metal-semiconductor (n-type) contact. For a5a7a6a9a8a10a5a12a11 : (a) just in contact, (b) in equilibrium with built-in potential a13a15a14 , (e) forward biased with a13a7a16 , (f) reverse biased with a13a18a17 . For a5a7a6a20a19a21a5a12a11 : (c) just in contact and (d) at equilibrium [43, 44] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.4 Energy level diagram for metal-semiconductor (p-type) contact. For a5a7a6a9a8a10a5a12a11 : (a) just in contact and (b) in equilibrium. For a5a15a6a22a19a10a5a12a11 : (c) just in contact, (d) at equilibrium with built-in potential a13a7a14 , (e) forward biased with a13a15a16 , (f) reverse biased with a13a18a17 . [43, 44] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.5 Non ideal energy level diagram under forward bias a13a7a16 [45] . . . . . . . . . . . 26 2.6 Image force. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 2.7 Effect of image force lowering on the barrier height. . . . . . . . . . . . . . . . 29 2.8 Effect of doping concentration on the barrier width. (a) low (b) moderate and (c) high doping [44]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 2.9 Atomic energy barrier as a function of atomic position [118]. . . . . . . . . . . 48 2.10 Diffusion barrier in multi-layer structure [140] . . . . . . . . . . . . . . . . . . 53 2.11 Stuffed diffusion barrier [140] . . . . . . . . . . . . . . . . . . . . . . . . . . 56 3.1 (a) Collinear four-point probe measurements (b) van der Pauw four-point probe measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 x 3.2 van der pauw correction factor plot [151] . . . . . . . . . . . . . . . . . . . . 64 3.3 Plot of total contact to contact resistance as a function of inter-pad spacing [159] 67 3.4 TLM resistive network [160]. The contact width W is into the page. . . . . . . 68 3.5 Effect of barrier lowering on barrier height . . . . . . . . . . . . . . . . . . . . 72 3.6 Capacitance measurements under applied bias [151] . . . . . . . . . . . . . . . 75 3.7 Close impact collision and backscattering [166] . . . . . . . . . . . . . . . . . 78 3.8 a23a25a24a21a26a28a27a30a29a32a31 = detector solid angle. S = detector area. d = detector-target distance. t = target thickness. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 3.9 Energy loss during inward and outward trajectory of incident particle [166] . . 83 3.10 Pelletron tandem accelerator at Auburn University . . . . . . . . . . . . . . . . 87 3.11 (a) Sources section of the accelerator. Illustrations of (b) SNICS source (c) He ion source [168] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 3.12 Pelletron charging system [171] . . . . . . . . . . . . . . . . . . . . . . . . . 90 3.13 Illustration of auger electron de-excitation [166] . . . . . . . . . . . . . . . . . 93 3.14 X-ray diffraction geometry [177] . . . . . . . . . . . . . . . . . . . . . . . . . 100 3.15 (a) Bragg-Brentano diffraction geometry for thin films (b) Seemann-Bohlin diffraction geometry [177] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 3.16 Wire bond pull and shear testing . . . . . . . . . . . . . . . . . . . . . . . . . 104 3.17 Chip brazed shear testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 4.1 (a) TLM masks (b) Diode/MOS Mask . . . . . . . . . . . . . . . . . . . . . . 109 4.2 Sputter-deposition system with disk sample holders having heating capability . 112 4.3 Annealing system with enlarged carbon strips showing sample . . . . . . . . . 113 4.4 Sequence for ohmic contact fabrication . . . . . . . . . . . . . . . . . . . . . 115 4.5 Sequence for Schottky contact fabication . . . . . . . . . . . . . . . . . . . . . 117 xi 4.6 Sequence for protective metallization on (a) ohmic contact (b) Schottky contact 119 4.7 Device cross-section of (a) ohmic (b) initial Schottky (c) new Schottky contacts120 4.8 Modified sequence for protective metallization (a) ohmic contact with Mo lift- off (b) Schotky contact with Pt etch-back . . . . . . . . . . . . . . . . . . . . 121 4.9 (a) Large area wire bonded sample for pull and bond shaer testing (b) Chip brazed sample for die shear testing . . . . . . . . . . . . . . . . . . . . . . . . 124 4.10 (a) TLM measurement set-up (b) measurement configuration . . . . . . . . . 126 4.11 (a) Current-Voltage measurement set-up (b) Capacitance-Voltage measurement set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 5.1 RBS spectra of (a) SiC/Mo-Si/Pt (b) SiC/W-Si/Pt and (c) SiC/Ta-Si/Pt . . . . 129 5.2 (a) RBS spectra of SiC/Ta-Si/Ta-Si-N with 0, 2, 5, 10% nitrogen content by flow rate (b) van der Pauw sheet resistivity of Ta-Si-N from TaSi a31 and Taa33 Sia34 targets (c) Nitrogen aomic percentage against nitrogen flow . . . . . . . . . . . . . . . 131 5.3 (a) Sheet resistance of Ta-Si-N on semi-insulating SiC against anneal time in air at 350a0a35a1 (b) RBS spectra of Ta-Si-N on semi-insulating SiC for 0 and 5000hrs (b) 0% and (c) 2% nitrogen flow . . . . . . . . . . . . . . . . . . . . . . . . . 132 5.4 XRD scan of Ta-Si-N with 0, 2, 5, and 10% nitrogen flow (a) as-sputtered sam- ples (b) samples annealed in air at 350a0 a1 for 750hrs (c) samples annealed in evacuated quartz tube at 350a0a35a1 for 750hrs . . . . . . . . . . . . . . . . . . . . 134 5.5 Total contact-to-contact resistance against inter-contact spacing for (a) nickel ohmic contact on pa3 implanted material (b) nickel-chromium (80-20 wt%) ohmic contact on na3 epitaxial material . . . . . . . . . . . . . . . . . . . . . . 135 5.6 Experimental and RUMP simulated RBS spectra of (a) as-deposited nickel con- tact on SiC (b) nickel ohmic annealed on SiC . . . . . . . . . . . . . . . . . . 137 5.7 Specific contact resistance and SiC sheet resistance against annealing time for (a,c) SiC/Ni(ohmic)/Ta-Si-N/Pt-N/Au/Sn/Au (b,d) SiC/Al-Ti(ohmic)/Ta-Si- N/Pt-N/Au/Sn/Au . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 5.8 RBS spectra of p-ohmic contact after 24, 1500 and 4000hrs annealing (a) nickel ohmic contact with 2% nitrogen in the stack (b) Al-Ti ohmic contact with 2% nitrogen in the stack. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 xii 5.9 AES spectra for (a) as-deposited nickel contact with no nitrogen in the metal- lization stack (b) as-deposited nickel with 2% nitrogen in the stack (c) nickel ohmic contact with no nitrogen in the stack (d) nickel ohmic with 2% nitrogen in the stack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 5.10 AES spectra of nickel ohmic contacts annealed in air at 350a0 a1 for 1500hrs (a) nickel ohmic contacts with no nitrogen in the stack (b) nickel ohmic contacts with 2% nitrogen in the stack . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 5.11 (a) Plots of specific contact resistance against annealing time for Ni-Cr (80-20 wt%) ohmic contacts with 2% nitrogen in the stack (b) Plots of SiC sheet re- sistance against annealing time for Ni-Cr (80-20 wt%) ohmic contacts with 2% nitrogen in the stack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 5.12 (a) RBS spectra of Ni-Cr (80-20 wt%) ohmic contacts after 24 and 4000hrs an- nealing; AES spectra of Ni-Cr (80-20 wt%) ohmic contacts with 2% nitrogen in the stack (b) before annealing in air (c) afer annealing in air for 1500hrs . . . . 146 5.13 Total contact-to-contact resistance of (a) cold and hot Pt in the metallization stack with no nitrogen content (b) cold and hot Pt in the metallization stack with 2% nitrogen content (c) nickel ohmic contact on na3 epilayer SiC with hot Pt and 2% nitrogen in the metallization stack after different anneal times in air . 148 5.14 RBS spectrum of a nickel ohmic contact on na3 epitaxial 4H-SiC. . . . . . . . . 149 5.15 Specific contact resistance and SiC sheet resistance for nickel ohmic contacts on na3 epitaxial 4H-SiC with hot Pt and 2% nitrogen content in the metallization stack.150 5.16 Specific contact resistance and SiC sheet resistance of nickel ohmic contacts on pa3 implanted 4H-SiC with hot Pt in the metallization stack. . . . . . . . . . . . 151 5.17 Plots of (a) total contact-to-contact resistance against inter-contact spacing (b) Specific contact resistance against temperature and (c) SiC sheet resistance against temperature for nickel ohmic contact on pa3 implanted 4H-SiC with 2% nitrogen in the metallization stack. . . . . . . . . . . . . . . . . . . . . . . . . 152 5.18 Nickel silicide Schottky devices (a) forward J-V dependence on annealing time (b) reverse J-V dependence on annealing time (c) barrier height and ideality factor against annealing time. . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 5.19 Bar chart of current density at -10V reverse bias for nickel silicide and Ta-Si-N Schottky devices (a) 0hr nickel silicide contacts (b) 0hr Ta-Si-N contacts (c) 1500hr nickel silicide contacts (d) 1500hr Ta-Si-N contacts . . . . . . . . . . 156 xiii 5.20 Nickel silicide Schottky devices (a) forward characteristics with 0% nitrogen (b) reverse characteristics with 0% nitrogen (c) forward characteristics with 2% nitrogen (d) reverse characteristics with 2% nitrogen. . . . . . . . . . . . . . . 157 5.21 (e) Forward characteristics with 5% nitrogen (f) Reverse characteristics with 5% nitrogen (g) Forward characteristics with 10% nitrogen (h) Reverse char- acteristics with 10% nitrogen. . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 5.22 Ta-Si-N Schottky devices (a) forward characteristics with 0% nitrogen (b) re- verse characteristics with 0% nitrogen (c) forward characteristics with 2% ni- trogen (d) reverse characteristics with 2% nitrogen. . . . . . . . . . . . . . . . 159 5.23 (e) Forward characteristics with 5% nitrogen (f) reverse characteristics with 5% nitrogen (g) forward characteristics with 10% nitrogen (h) reverse characteris- tics with 10% nitrogen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 5.24 Nickel silicide Schottky devices (a) barrier height against annealing time (b) ideality factor against annealing time. Ta-Si-N Schottky devices (c) barrier height against annealing time (d) ideality factor against annealing time. . . . . 161 5.25 Plots of (A/C)a31 against applied bias for nickel silicide diodes with (a) 0% ni- trogen (b) 2% nitrogen (c) 5% nitrogen (d) 10% nitrogen content in the metallization stack. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 5.26 Plots of (A/C)a31 against applied bias for Ta-Si-N diodes with (a) 0% nitrogen (b) 2% nitrogen (c) 5% nitrogen (d) 10% nitrogen content in the metallization stack.163 5.27 Nickel silicide Schottky contacts with 2% nitrogen in the metallization stack (a) RBS spectra (b) AES spectra after 1500hr anneal in air. . . . . . . . . . . . . 165 5.28 Chip shear strength with cold Pt in the metallization stack against annealing time. 166 5.29 Chip shear strength with hot Pt and 2% nitrogen in the metallization stack against annealing time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 5.30 Wirebond pull and shear strengths against annealing time for (a) metallization with 2% nitrogen and hot Pt on a thermal oxide (b) metallization stack with 2% nitrogen and hot Pt on a nickel silicide Schottky contact (c) Au wire failure mode and sheared wedged bond. . . . . . . . . . . . . . . . . . . . . . . . . . 169 5.31 RBS and AES results for nickel ohmic contacts (a) RBS of ohmic annealed con- tacts (b) AES of ohmic annealed contacts (c) RBS of RIE cleaned ohmic contacts (d) AES of RIE cleaned ohmic contacts . . . . . . . . . . . . . . . . 170 xiv 5.32 RBS and AES results for nickel ohmic contacts (a) RBS of Ar ion cleaned ohmic contacts (b) AES of Ar ion cleaned ohmic contacts. . . . . . . . . . . . . . . 171 5.33 (a) wirebond pull and shear strength of Ar ion cleaned nickel ohmic contacts (b) picture of failed wirebond and (c) picture of good pull and shear bonding. . 173 xv LIST OF TABLES 2.1 Ideal MS contanct dependence on work functions . . . . . . . . . . . . . . . . 21 2.2 Some ohmic contacts on n-type a36 -SiC. The layers in multi-layered contacts are seperated with slashes, layers at the surface to the interface with SiC proceed from right to left . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 2.3 Some ohmic contacts on p-type a36 -SiC. The layers in multi-layered contacts are seperated with slashes, layers at the surface to the interface with SiC proceed from right to left . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.4 Some ohmic contacts on n-type a37 -SiC. The layers in multi-layered contacts are seperated with slashes, layers at the surface to the interface with SiC proceed from right to left . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 2.5 Some ohmic contacts on p-type a37 -SiC. The layers in multi-layered contacts are seperated with slashes, layers at the surface to the interface with SiC proceed from right to left . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 2.6 Elemental diffusion barrier layer for semiconducting devices . . . . . . . . . . 58 2.7 Binary compound or alloy diffusion barrier layer for semiconducting devices . 59 2.8 Tenary compound or alloy diffusion barrier layer for semiconducting devices . 59 4.1 Profile of 4H-SiC purchased from Cree research Inc. . . . . . . . . . . . . . . 106 5.1 Various ohmic contacts fabricated in this study. The metal in contact with the SiC is on the left for the case of two or more layers. The metal sheet resistance (a38 a39a41a40 a11a43a42 ) column has values before/after contact annealing. t is the contact metal thickness. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 xvi CHAPTER 1 GENERAL INTRODUCTION AND SILICON CARBIDE 1.1 General Introduction 1.1.1 Why This Work? The need for cost effective high power/high temperature semiconductor devices is essential to support advancements in power electronic and microelectronic technology that are important and much needed as we enter the 21st century. Application areas are numerous, including [1] : a44 High power switching for electric utilities, industrial motor control and electric vehicles. a44 Sensors, controls and power management for the aerospace and automotive industries. a44 Control electronics for the nuclear power industry and down-hole well logging in the petroleum drilling industry. The technological implications of advanced electronics are huge, and so also are the eco- nomic implications. Millions of dollars can be saved for aircraft and spacecraft as a result of reduced space and size, higher efficiency for power/fuel usage, simpler system design, and im- proved long-term reliability [2]. In the automotive industry, high temperature, on-engine sensing will lead to higher fuel efficiency and less pollution of the environment. The US Department of Defence requires wide band gap devices and component solutions for more electric ships, aircraft and vehicles, space platforms, and directed energy missions. A near-term goal for the Air Force is the development of advanced motor drives for flight control 1 and space vehicle applications, and converters/inverters for power conditioning applications in more electric aircraft and directed energy systems [2]. It is well known that Si-based devices are not capable of high temperature or harsh envi- ronment operation. Therefore, major improvements in system performance for extreme environ- ments will result from the successful development of wide band gap semiconductor materials like SiC, the group III-nitrides (GaN, AlN and BN) and diamond. Diamond technolgy is still in its early stages of development. The direct band gap nitrides are being developed for blue-UV optoelectronic applications. SiC is the most developed of the wide band gap semiconductors, and the 4H polytype is particularly suited for high power/high temperature applications because of its intrinsic properties such as larger band gap energy, higher breakdown field strength, higher saturated carrier velocity, higher thermal conductivity and more isotropic bulk carrier mobility [3]. For high power applications, 4H-SiC material properties promise higher power density (reduced size and weight), less cooling, improved operating effi- ciency and greater reliability. The advantages of SiC for high power/high temperature applications can be translated into practical devices only if advanced metallization and packaging technologies are developed that allow devices to operate under extreme conditions. The development of harsh environment met- allization schemes that are compatible with high temperature packaging technology is the focus of this work. 1.1.2 Objectives The goals of this work are the following: a44 Develop an advanced SiC metallization process for ohmic and Schottky contacts that is suitable for long term operation (a45 10 khr) in air at 350a0 a1 . 2 a44 Develop a protection scheme against oxidation and metal-interdiffusion, thereby preserv- ing the characteristics of the metal-semiconductor interface and the electrical performance of the contacts. a44 Make the metallization process compatible with high temperature packaging techniques. 1.1.3 To Accomplish the Objectives Typical composite contact layout For high temperature/high power devices, composite (multilayer) contacts that are com- patible with the techniques and procedures used for packaging (wire bonding and die attach) have been proposed [2]. A typical composite contact is shown in figure 1.1. The simplest composite contact will have three metallization layers: (1) ohmic or Schottky contact, (2) diffu- sion/oxidation barrier layer and (3) cap layer. An ohmic contact must have symmetric current - voltage (I-V) characteristics with the lowest possible specific contact resistance. A Schottky contact is a metal-semiconductor contact with asymmetric I-V characteristics. It should be able to turn-on at low forward voltages, and the reverse leakage current should be as low as possible. The diffusion/oxidation barrier layer must have metal-like conduction while preventing, or at least slowing, the intermixing of stack layers as well as reducing the oxidation of the contact layer which can lead to increased contact resistance. Because of the non-ideality of the barrier layer, it may be necessary to introduce other layers (e.g., a suitable adhesion layer) in the metallization stack. Finally, the cap layer must be suitable for packaging through wirebonding or die attach. 3 Figure 1.1: Typical composite metallization scheme ( [2] ) Ohmic contact Ohmic contacts can be fabricated to either n-type or p-type SiC. n-type ohmic contact - Nickel-based contacts on n-SiC have been more widely reported com- pared to other metals [4, 5, 6]. Specific contact resistance between a46a48a47a49a46a2a50a32a4a7a51 and a46a48a47a52a46a35a50a32a4a53a31a54a23a55a39 a56a58a57 a31 can be found in literature, depending on the processing method and doping concentration. Nickel-based contacts are known to be thermally stable, and so are appropriate for this work. Nickel reacts with SiC at temperatures a59a9a60a61a50a62a50 a0 a1 , forming silicides such as Ni a31 Si which is an excellent ohmic contact [5, 6]. p-type ohmic contact - Aluminum based metal alloys (AlTi, NiAl, etc) are widely reported to form ohmic contacts to p-SiC [7, 8]. Al-Ti forms an ohmic contact to moderately doped p-SiC when annealed at temperatures above a63a62a50a61a50 a0a35a1 [7, 8]. A great disadvantage for the Al-based con- tact is the possibile oxidation of Al during long-term operation. However, ohmic contacts with excellent uniformity and reproducibility have been reported with Al-Ti alloys [9, 10, 11]. The 4 70-30 weight percent Al-Ti alloy is good for p-SiC ohmic contacts because of reduced spiking of Al into the SiC. The contact is more reproducible, and the contact layer/SiC interface is much smother following the contact anneal. 70-30 Wt% Al-Ti will be used in this work for ohmic con- tact on p-SiC, and we will investigate the thermal stability of this alloy at a64a66a65a30a50 a0a2a1 in air. Ni-based alloys will also be investigated as possible candidates for ohmic contacts to heavily doped p-SiC (a45a10a46a35a50a62a31a68a67 a56a54a57 a4 a34 ). Schottky contact Nickel silicide will be sputter-deposited for Schottky contacts on na4 4H-SiC. Pure Ni has been considered, as well. However, Ni reacts with SiC at about a69a61a50a62a50 a0 a1 to form nickel silicide. This reaction could affect the Schottky characteristics during the long-term thermal stability investigation we intend to carry out. Sputtered nickel silicide is expected to more stable. Diffusion/oxidation barrier layer An effective diffusion barrier is necessary, especially where it is impossible to eliminate thermodynamic driving forces for reactions. An extensive body of literature exists for barrier layers used in silicon VLSI/ULSI applications. Transition metal silicides (e.g. TaSi a31 , MoSi a31 , WSi a31 ) have been used successfully. These silicides will be studied as possible diffusion barriers in SiC composite contacts. The rate of mass transport through a polycrystalline material is noticeably enhanced by migration along grain boundaries. The silicides we will consider for diffusion barriers are some- times polycrystalline or amorphous depending on the deposition process. Diffusion through a polycrystalline material can be reduced significantly by ?stuffing? the grain boundaries with 5 small elements such as nitrogen or carbon [139]. Grain boundaries are not a problem if the mate- rials are amorphous; however, crystallization during long-term operation at elevated temperature is a potential problem. We will sputter the silicide barrier layers in a70a72a71a61a27a2a73 a31 gas mixtures in order to produce amor- phous barriers or to improve the properties of polycrystalline diffusion/oxidation barrier layers. Cap layer and packaging issues Layers above the barrier layer have to be oxidation/corrosion resistant and must be com- patible with the diffusion barrier (adhesion and stability). Cap layers need to be suitable for the envisioned packaging sheme: e.g. wirebonding, die attach or brazing. All composite metallization schemes will be developed for a64a66a65a30a50 a0 a1 operation in air. Com- posite contact stacks will be evaluated electrically (LTLM, I-V, C-V) and physically (RBS, AES/XPS, XRD). Mechanical techniques (shear and pull testing) will be used to evaluate the durability of the wirebonds and die attachments. Devices will be subjected to temperature cy- cling at a64a61a65a62a50 a0a2a1 in air. 1.2 Silicon Carbide 1.2.1 Introduction Silicon carbide, originally called mineral moissanite, has been known to exist naturally since the late 1800s. It gained attention because of its chemical, thermal and mechanical stability [12]. Based on these properties, SiC has been used largely in industry for grinding and cutting. Acheson [13] was the first to devise a method of producing SiC on a large scale, and the largest market originally was for use as an abrasive. However, other applications were soon exploited 6 - including cheap heating elements with long lifetimes, coating of uranium grains for the fuel elements of nuclear reactors, reinforced materials, etc. Polycrystalline SiC with binders was first used in commercial electrical devices as lightning arrestors to protect equipment from power surges [12, 13]. Major scientific interest started when crystallographers observed that hexagonal SiC is ca- pable of crystallizing into many different forms called polytypes. ZnS is another semiconductor that occurs with similar structural variations. SiC is however unique because the semiconductor bandgap is polytype dependent, typically between 2.2 eV for cubic SiC to 3.3 eV for the wurtzite structure [12]. SiC is sometimes reffered to as both the oldest and youngest semiconductor [12]. Lossew [14] in 1907 discovered the electroluminescence of the semicondutor while experimenting with SiC crystals. The first systematic analysis of the semiconductor properties was conducted in 1945 by Busch [15] using material from the Acheson SiC crystal collections. Because of the dif- ficulty of growing good quality single crystals at the time, progress was slow towards exploiting the semiconductor characteristics of this material. Cree, Inc. introduced a 25 mm single crystal wafer of 6H-SiC in 1990 [16]. However, known material problems such as the existence of micropipe defects (small holes penetrating the wafers parallel to the growth direction) persisted. Micropipe diameters varied from less than one micron to tens of microns, and the micropipe density was about a46a35a50a30a74 a56a54a57 a4a53a31 . By the year 2000, noticeable improvements in wafer quality had been achieved. Seventy-five millimeter wafers with a65a62a50a75a39a76a46a35a50a62a50a30a77a79a78a7a80a81a78a83a82a85a84a86a27 a56a58a57 a31 were available. Wafers with as few as a64a32a87a88a65a85a77a79a78a7a80a81a78a83a82a85a84a86a27 a56a54a57 a31 have been reported recently [16]. In 2004, Nakamura, et al.[17] reported a modified growth method that has led to ultra-high quality SiC single crystals that are virtually dislocation free. 7 1.2.2 Crystal Structure The fundamental unit of SiC has a tetrahydral structure with four nearest neighbor C atoms bonded to Si (a26a89a80 a1 a74 ) or four nearest neighbor Si atoms bonded to C (a1 a26a89a80 a74 ). Bonding between Si and C is primarily a90a91a63a61a63a66a92a94a93 covalent, the rest of the bonding is ionic. [20]. Polytypism is the most remarkable feature of SiC. Polytypism may be defined as the ability of a substance to crystallize into a number of different modifications in all of which two dimensions of the unit cell are the same and the third dimension is a variable, integral multiple of a common length [12]. Two base vectors of the hexagonal unit cell are a95a96a15a97 a24a98a95a96 a31 a24a99a64a32a87a100a50a66a101a30a63 ? a70 while the third base vector a95 a56 is an integral multiple of a102a32a87a88a65a79a46a35a63 ?a70 . The difference between the polytypes is in the stacking order of the basic building blocks which are close packed Si-C tetrahedra. In all polytypes, any Si or C atom has the same first and second coordination. Every atom has four first neighbors of the other atom and twelve second neighbors of the same atom. There are over 200 known polytypes, and some polytypes do not have a repeating pattern until after hundreds of layers [21]. SiC is a bi-layer compound, above the center of a triangle formed by three Si atoms is a carbon atom. The three Si atoms form in an hexagonal network forming the first Si layer of the structure. The fouth Si atom of the tetrahydral belongs to the second layer of the Si network. The Si atom in the second layer bonds directly to the first carbon layer atoms directly above it. The atoms of the third Si layer are connected also in unilateral position to the atoms of the second carbon layer. They may occupy a position having its projection on the atom of the first Si layer or at the center of the triangle not covered by the projections of the atoms of the first carbon layer. Identical successive layers of tetrahydra are oriented either parallel or anti-parallel. Different structures arise as a result of the characteristic succession or alteration of tetrahydral layers, forming a repeating unit. If every 8 layer is parallel to a preceeding layer, a cubic structure is formed, but if anti-parallel, a hexagonal structure is formed [20]. The Ramsdell notation [22] for polytypes is the most common notation, while Zhadanov [23] is another notation that is sometimes used. Ramsdell notaion has a bravis lattice type of structure (H - hexagonal, C - cubic, R - rombohydral). Preceeding this letter is a number (eg. 3C, 4H, 6H), which specifies how many bi-layers form the repeating structure along the stacking direction. The most common of SiC polytypes being developed for electronic applications are 3C (a37 -SiC), 4H and 6H (a36 -SiC). 3C-, 4H- and 6H-SiC stacking sequence are shown in figure 1.2 [24]. There is a close correlation between crystallograhpic and semiconducting properties. Even within a given polytype, some important electrical properties are nonisotropic, that is, they are strong functions of crystallographic direction (e.g. electron mobility for 6H-SiC) [25]. It is known experimentally that growth temperature [26] and pressure [27] affect polytype distribu- tion, and impurities may also have an affect [13, 26]. 1.2.3 Semiconducting Properties Silicon carbide is a wide band gap semiconductor (a103a105a104a75a24 2.2-3.3 eV) like GaN (a103a105a104a75a24 3.4 eV) and AlN (a103 a104 a24 6.2 eV) [16], and SiC is the only known compound semiconductor that can be oxidized to form a high quality native oxide (SiO a31 ). Silicon carbide can be doped in- situ during growth by introducing nitrogen or phosphorous for n-type conductivity and either aluminum or boron for p-type conductivity. Doping concentration ranges from a46a35a50 a97 a74 a56a58a57 a4 a34 to a59a99a46a35a50 a97a43a106 a56a58a57 a4 a34 [28]. Selective doping can be achieved by high energy implantation. The usual doping method for Si of thermal diffusion is not practicable for SiC because of the very low 9 Figure 1.2: (a) basic structural unit of SiC, (b) 3C-SiC stacking sequence, (c) 4H-SiC stack- ing sequence and (d) 6H-SiC stacking sequence. The Jagodzinski (kkk, khkh, kkhkkh) and ABCA?B?C? notations are also shown [24]. 10 diffusion coefficients below 2000a0 a1 for all dopant species. For similar implant and post-implant anneal conditions, activation percentages are lower for Al and B compared to N and P, so that selective p-type doping is more challenging. Large band gap energy and correspondingly low intrinsic carrier concentration allow SiC to retain its semiconducting characteristics at much higher temperatures compared to Si, since the intrinsic carrier concentration is a pre-factor in the equation governing undesired junction reverse-bias leakage current. As the temperature increases, the intrinsic carrier concentration increases exponentially. The leakage current increases as well, and eventually at high enough temperature, semiconductor device operation is overcome by uncontrolled conductivity as the intrinsic carrier concentration exceeds the intentional device doping [25]. The breakdown field and high thermal conductivity of SiC, coupled with high operational junction temperature, theoretically permit extremely high power density and operating efficiency. 1.2.4 Devices and Applications FET Structure All field-effect transistors (FETs), whether MOSFETs, MESFETs or JFETs have similar structure [29]. The gate contact metal and conducting channel form a parallel plate capacitor. The sheet charge density a107a83a11 of the mobile carriers at any point, x, along the channel is controlled by the potential difference between the gate and the channel. a107 a11 a24 a1a109a108a111a110 a13a53a112a114a113a75a39a115a13a116a90a118a117a12a93a89a39a115a13a18a119a114a120 (1.1) a13a18a119a121a24 threshold voltage, a13a53a112a114a113a55a24 voltage difference between the gate and the source, a13a122a90a123a117a114a93a124a24 channel voltage with respect to the source, and a1 a108 a24 gate capacitance per unit area a90a123a125 a108 a27a85a29 a108 a93 . The 11 Figure 1.3: Basic FET structure [29] drain current a126a54a127 depends on a107a53a11 and the drain-to-source voltage a13a15a128a129a113 because a13a116a90a118a117a12a93 varies with a117 and tends to reduce a107a83a11 . The threshold voltage is an important parameter in FETs because it separates the above- threshold regime (or on-state where the charge induced in the device channel is proportional to the gate voltage swing), and the below-threshold regime (or off-state). The potential barrier separating the source and the drain determines the transistors off-current and hence the on-to-off current ratio [29]. MOS techology The vast majority of semiconductor integrated circuits rely on silicon metal-oxide-semiconductor FETs (MOSFETs). Success with Si-MOSFETs has naturally encouraged the development of 12 high-performance inversion channel SiC-MOSFETs. The difference in the quality of the ther- mal oxide grown on Si and SiC has prevented SiC-MOSFETs from realizing their full potential [25]. 4H-SiC MOSFET deficiencies can be attributed to differences in Si and SiC thermal ox- ide quality and an interface structure that causes SiO a31 /SiC to exhibit high interface state den- sities (a8a130a46a2a50 a97 a34 a82a85a13a131a4 a97 a56a54a57 a4a53a31 near the conduction band edge), high fixed oxide charge densities a90a132a45a22a46a35a50 a97a133a97 a39a134a46a35a50 a97 a31 a56a58a57 a4a53a31 a93 , charge trapping, carrier oxide tunneling, and roughness-related scatter- ing of inversion channel carriers. The SiO a31 /SiC interface can be much rougher than the SiO a31 /Si interface because of the off-angle polishing needed for homoepitaxial growth as well as step- bunching that can occur during annealing to activate implanted dopant species [25]. High-temperature devices Silicon becomes a conductor at temperatures above about a102a30a50a61a50 a0 a1 because of a narrow band gap that allows the intrinsic carrier concentration to increase rapidly with temperature. For higher temperature applications, and in high power systems where juction self-heating leads to much higher temperature, Si devices will not be suitable. Hence, transition to SiC is essential for high power/high temperature applications. Long-term reliability of SiC devices operaing at temperatures between a64a62a50a61a50a135a39a136a69a61a50a62a50 a0 a1 is a challenge. High temperature gate-insulator reliability is needed for the successful realization of MOSFET-based integrated circuits. Gate-to-channel Schottky diode leakage limits the peak operating temperature of SiC MESFETs to around a137a66a50a61a50 a0 a1 [25]. Bipolar SiC transistors exhibit poor gain, but improvements in SiC crystal growth and surface passivation could greatly improve SiC BJT performance to make SiC bipolar devices a more viable technology [30]. A common 13 obstacle to all technologies is reliable long-term operation of contacts, interconnects, passivation and packaging at a138a21a8a41a64a62a50a61a50 a0 a1 . Increasingly capable and economical SiC integrated circuits will continue to evolve as SiC crystal growth and device fabrication technology contnue to improve. Currently however, non- ideality in SiC epilayer doping and thickness, surface morphological defects, and slow charge trapping/detrapping phenomena cause unwanted device I-V drift, and limit the yield, size, and manufacturability of SiC high-temperature ICs [31]. High power devices High power switching devices operate with high electric fields and high current densities, placing great electrical stress on the device. [25]. Current power switching devices are reaching a fundamental limit imposed by silicon?s low breakdown field. Silicon carbide has about 10 times the breakdown strength of Si though SiO a31 /SiC interface quality is not yet as good as that of SiO a31 /Si. Nevertheless, power MOSFETs, IGBTs and various types of MOS-controlled thyristors (MCTs) can all be fabricated on SiC. Because of the higher breakdown field, SiC power devices can have specific-on resistances up to 400 times lower than similar Si devices [32]. a40a140a139a142a141a131a24 a137a32a13a135a31 a143 a77a12a141a18a144a111a11a111a103 a34 a145 (1.2) a13 a143 is the breakdown voltage, a103 a145 is the semiconductor critical field, a77a12a141 is the carrier mobility, and a144a111a11 is the semiconductor permittivity. For SiC, Ea145 is about seven times larger compared to Si. Early on, commercially available SiC epilayers were limited to about a46a35a50a85a77 a57 in thickness with a maximum blocking voltage of about 1600 V. However, with the introduction of thick 14 layer growth techniques, the epilayer thickness barrier has been removed, and the maximum blocking voltage has increased to a8 10 kV for a single device [33, 34]. Micopipes have been a serious problem for SiC device quality and yield, but the desity of micropipes has dropped drastically from hundreds per square centimeter in the early 1990s to a45 1 per square centimeter in the best material currently available. Apart from micropipes, the density of screw dislocation defects in SiC wafers and epilayers has been measured to be of the order of several thousand per square centimeter. While these defects are not nearly as detrimental to device performance as micropipes, experiments have shown that they can degrade material properties such as breakdown field strength and minority carrier lifetime [25]. For SiC power devices to function successfully, peripheral breakdown owing to edge related electric field crowding must be avoided through careful device design and proper choice of a passivating dielectric material. The peak voltage of most prototype high-voltage SiC devices has been limited by edge related breakdown, especially in SiC devices capable of blocking voltages above a few kilovolts [25]. The high power diode rectifier is a building block of power conditioning circuits. SiC power rectifiers are similar to Si diode device in many ways, except that the current densities, blocking voltages, power densities and switching speeds are typically much higher in SiC. High breakdown field and the wide energy band gap allows SiC Schottky diodes to operate with these improved characteristics [25]. Hybrid Schottky/p-n rectifier structures combine p-n juction reverse blocking capability with low Schottky forward turn-on, and these devices are extremely useful for application- optimized SiC rectifiers [35, 36]. 15 Three terminal power switches (MISFETs, IGBTs and thyristors) that use small drive sig- nals to control large voltages and currents are also critical building blocks of high-power con- version circuits. Silicon carbide solid state switches are similar in structure to Si switches. They are designed to maximize power density via vertical current flow using the substrate as one of the device terminals. Operating current is presently limited by deficiencies in material qual- ity. Nontraditional power switching topologies have also been proposed to reduce the effects of oxide and other material deficiencies while maintaining normally-off insulated gate operation. For example, lateral and vertical doped-channel power MOSFETs and JFETs can be completely depleted by built-in potential at zero gate bias [32]. RF and microwave electronics Based on the electronic and thermal properties, several SiC polytypes should perform better than the semiconductors currently used for high-frequency and microwave electronic devices. There is interest in SiC devices for applications such as microwave power amplifiers that can be used in phased-array radars, base station transmitters for mobile communications, high- frequency and broad band radar transmitters, and other applications [37]. RF and microwave devices that can be fabricated from SiC include metal-semiconductor field-effect transistors (MESFETs), static induction transistors (SITs), bipolar junction transistors (BJTs), heterojuc- tion bipolar transistors (HBTs) and impact avalanche transit-time (IMPATT) diodes. Silicon carbide RF devices are used for high-frequency solid state high-power application at frquencies from around 600 MHz (UHF band) to about 10 GHz (X-band). High breakdown voltage and high thermal conductivity coupled with high carrier saturation velocity allow SiC RF transistors to handle much higher power densities than their Si or GaAs counterparts. RF operation at higher drain bias is made possible by the high breakdown field of SiC, and this leads 16 to high SiC MESFETs output power densities. High thermal conductivity minimizes channel self-heating so that phonon scattering does not seriously degrade channel carrier velocity and current [39]. Silicon carbide is an ideal semiconductor for the fabrication of high power microwave de- vices operating in the 1 - 10 GHz range [38, 39]. Short channel MESFETs have been operated with an a146a35a147 of 22 GHz and an a146 a6a149a148a151a150 of 50 GHz, and static induction transistor (SITs) have reached power levels of 470 W (1.36 W/mm) at 600 MHz and 38 W (1.2 W/mm) at 3 GHz [39]. Optoelectronics and sensors Wide band gap semiconductor like SiC can be used to realize short-wavelength blue and ultraviolet optoelectronics. 6H-SiC blue p-n juction light emitting diodes (LEDs) were the first commercially available SiC devices, and they were the first mass produced LEDs to cover the blue wavelength range (250 - 280 nm). They were not efficient diodes (efficiency a19a152a46a153a92 ) since SiC is an indirect band gap material (i.e., the positions of the conduction band minimum and valence band maximum do not coincide in crystal momentum space), in which case, transistion from the valence band maximum to conduction band minimum is phonon assisted. Despite their inefficiency, they were commercially successful from 1989 to 1995 [25]. SiC blue LEDs have now been replace by the brighter and more efficient direct band gap GaN LEDs. SiC has proved to be much more efficient at absorbing short-wavelength light, which allows the fabrication of UV-sensitive photodiodes that serve as excellent flame sensors for turbine- engine combustion monitoring and control. 6H-SiC has been used to fabricate diodes with low dark current, as well as sensors that are not sensitive to the near infrared wavelengths that are produced by heat and solar radiation [25]. 17 The high temperature capabilities of SiC allow the fabrication of catalytic metal-SiC and metal-insulator-SiC (MIS) prototype gas sensors that show great promise as combustion engine emission monitors [40, 41]. These structures enable rapid detection of changes in hydrogen and hydrocarbon content to sensitivities of parts per million. The sensors are small enough to be place anywhere in an engine. When fully developed, these sensors could assist in active combustion control to reduce harmful pollution emissions from automobile and aircraft engines. [42]. 18 CHAPTER 2 METAL-SEMICONDUCTOR CONTACTS AND DIFFUSION BARRIERS 2.1 Metal-Semiconductor Contacts 2.1.1 General Information The metal-semiconductor (MS) contact is a crucial part of all solid-state devices. An ideal MS contact can either be rectifying (Schottky) or non-rectifying (ohmic). High quality ohmic contacts are necessary to connect semiconductor devices to external circuit, and stable Schottky contacts are essential for switching or provide rectification, either as a stand alone device or as part of more complex circuits [43]. An ideal MS contact should, on an atomic scale, be an intimate contact between metal and semiconductor with no layer of any type between them. There should be no interdiffusion be- tween them and no adsorbed impurities or surface charges at the MS interface. However, in practical contacts, ideal conditions are seldom achieved and MS devices operate below theo- retical predictions. Non-idealities in modern-day structures primarily affect the barrier height characterizing the contact [44]. Controlling the surface parameters that affect the Schottky bar- rier height guides the development of ohmic and rectifying contacts in semiconducting systems [45]. Ideally, it should be possible to predict the behavior of a metal on a semiconductor if the work function of both are known. But contact performance, as related to surface properties in SiC for example, is poorly understood, and most of the metal-SiC behavior cannot be predicted [45]. 19 Figure 2.1: I-V Characteristics of ohmic and schottky contact Schottky contacts are characterized by asymmetric current-voltage (I-V) curve as shown in figure 2.1, similar to a p-n junction. An ohmic contact may be considered a limiting case of Schottky contact in which Schottky characteristics have been converted to ohmic characteristics after certain processing steps. However, some MS contacts may be ohmic as prepared if the barrier height at the MS interface is low enough. In these cases, asymmetric I-V characteristics does not occur [46]. Schottky-Mott and Bardeen limit for barrier height The Schottky model for ideal, intimate MS contacts is based on energy level diagrams [45] as illustrated in figure 2.2. a5a83a154 is the metal work function. The semiconductor work function is a5a12a113a155a24a157a156a159a158a76a90a160a103 a145 a39a55a103a161a16a129a93a153a162a16 a143 where FB means ?flat band? conditions - i.e. no band bending at the MS interface. The electron affinity of the semiconductor at the semiconductor surface is a156a155a24a157a90a91a103a161a139a163a39a49a103 a145 a93a153a162a11a91a164a153a17a166a165a2a148 a145a132a167 . The energy difference between the conduction energy ( a103 a145 ) 20 Figure 2.2: Energy level diagram of metal and semiconductor n-type p-type semiconductor semiconductor a5a15a6a76a8a41a168a105a11 Rectifying Ohmic a5a15a6a76a19a41a168a105a11 Ohmic Rectifying Table 2.1: Ideal MS contanct dependence on work functions and the Fermi energy level at flat band is the other term added to electron affinity to obtain the semiconductor work function expression. The semiconductor can be n-type or p-type depending on the dopant. The energy level diagrams of metal-semiconductor combination (n-type and p- type) are shown in figure 2.3 with different metal and semiconductor work functions. The barrier to reverse current flow which is not affected by the applied potential is defined as the barrier height for the metal-semiconductor pair [43]. It may also be defined as the energy difference between the Fermi level in the metal and the bottom of the conduction band in the semiconductor at the interface [45]. 21 Figure 2.3: Energy level diagram for metal-semiconductor (n-type) contact. For a5a7a6a169a8a170a5a12a11 : (a) just in contact, (b) in equilibrium with built-in potential a13a15a14 , (e) forward biased with a13 a16 , (f) reverse biased with a13a18a17 . For a5a15a6a76a19a136a5a12a11 : (c) just in contact and (d) at equilibrium [43, 44] 22 Figure 2.4: Energy level diagram for metal-semiconductor (p-type) contact. For a5a7a6a9a8a10a5a12a11 : (a) just in contact and (b) in equilibrium. For a5a7a6a20a19a76a5a83a11 : (c) just in contact, (d) at equilibrium with built-in potential a13a7a14 , (e) forward biased with a13 a16 , (f) reverse biased with a13a79a17 . [43, 44] 23 For the ideal MS contact illustrated in figures 2.3 and 2.4, we have chosen the flat band condition and assumed no electric field exists within the semiconductor. This implies that the semiconductor terminates at the surface without distortion of the electron energy levels and that no surface states exist. For the case of an n-type semiconductor in which the metal work function is greater than semiconductor work function, (a5a15a6a76a8a136a5a12a11 ), and taking vacuum level to be the same in the metal and semiconductor, electrons flow from the semiconductor to the metal, leaving positive donor ions, and accumulating at the surface of the metal. The resulting dipole electric field opposes further electron flow, and in equilibrium, the Fremi levels in the metal and semiconductor are equal, a103a161a16 a154 a24a171a103a72a16a114a113 . If the semiconductor is uniformly doped, the charge density is uniform in the depletion layer and the electric field a172 is linear with distance. The barrier energy of the Schottky barrier contact is a5 a143 a24a173a5a7a6a20a39a134a156a174a11 . This descrption is referred to as the Schottky limit, and a5 a143 is directly proportional to a5a7a6 . a5 a143 is the barrier energy that electrons going from metal to semiconductor encounter, while electrons going from semiconductor to metal encounter built-in potential energy a175a61a13a7a14 determined by the band bending in the semiconductor at equilibrium. The fact that a13a7a14 is bias dependent and a5 a143 is not is the reason for the asymmetric I-V characteristics of Schottky diodes. The Schottky model predicts that it is possible to obtain a rectifying contact or an ohmic contact by simply choosing a metal with the appropriate work function. This is not always true however, especially for diodes fabricated on III-V semiconductors. Recent evidence for SiC suggests that SiC approximates Schottky behavior [45]. MS contacts are never ideal, and Bardeen [47] proposed that if surface states exist at the MS interface in sufficient number, a5 a143 would be independent on a5 a6 . Surface states are electronic states (e.g., dangling bonds) localized at the semiconductor surface due to the termination of 24 bulk periodicity. Consider a neutral semiconductor surface, with say a neutral energy level a5a53a139 measured relative to the valence band. If the states are filled to an energy greater than a5a53a139 , the surface possesses a net negative charge, and the states are acceptor-like in behavior. If the states are filled to a level below a5a7a139 , the surface has a net positive charge, and the states behave in a donor-like manner [45]. Assume a very thin (a45 1nm) insulating layer separates the metal and semiconductor, such that electrons can flow though it with little or no restriction. This layer still support a potential difference, and if the number of surface state is large, the Fermi level at the surface of the semi- conductor will be a5 a139 . The energy difference a5 a6 a39a122a5 a139 for the Schottky model will appear entirely across the thin interfacial layer, since the charges in surface states will fully accomodate the necessary potential difference. Thus, these states contribute some of the electrons and positive space charge needed to bring the structure to equilibrium. The height and width of the barrier shrink substantially [48]. Hence, a5 a143 is independent of a5a7a6 , and a103a161a16 a154 a24a157a103a161a16a114a113 at the surface of the semiconductor. The barrier height for Bardeen contact is a5 a143 a24a152a103 a104 a39a176a5a53a139 . The Fermi level is ?pinned? by the surface states at a5a53a139 above the valence band. This is known as Bardeen limit. The Fermi level is found experimentally to be ?pinned? at a5 a139a124a177 a97 a34 a103a105a104 which impies the barrier height a5 a143 should be typically close to a31 a34 a103 a104 . Most common semiconductors (Si, Ge, GaAs, GaP) have enough surface states to pin the Fermi level at about a97 a34 a103 a104 . Surface states are often seen as a pronounced peak in the density of states plot at an energy a103a105a178a105a158 a97 a34 a103a179a104 [48]. In practice, the value of the barrier height a5 a143 will be somewhere between the Schottky and Bardeen limits [45]. A general relationship that combines both the surface states and workfun- tion was first given by Cowley and Sze [49] as a5 a143a181a180 a24a41a182a174a90a160a5a15a6a41a39a183a156a89a11a151a93a28a158a184a90a166a46a149a39a185a182a28a93a54a90a160a103 a104 a39a115a5a7a139a153a93 (2.1) 25 Figure 2.5: Non ideal energy level diagram under forward bias a13 a16 [45] For a case where no electric field inside the semiconductor (i.e., the flat band condition) a182a186a24 a125 a108 a125 a108 a158a187a175a62a188a131a11a166a189 (2.2) where a125 a108 is the permittivity of the interfacial layer, a189 is the thickness. a188a131a11 is the surface state density per unit energy per unit area. a175a62a188 a11 a189a191a190a192a125 a108 a193 a182 a177 a46 Schottky limit a175a62a188a131a11a166a189a191a194a192a125 a108 a193 a182a186a190a195a46 Bardeen limit A typical diagram for a non-ideal MS interface under forward bias a13a7a16 is shown in figure 2.5. An insulating layer of thickness a189 exist between the metal and semiconductor and the interface states are filled to a5a53a139 . 26 Figure 2.6: Image force. Image Force Lowering The practical barrier energy in a Schottky diode is slightly smaller than expected. One likely reason is attributed to image force lowering which occurs when an electron outside the metal induces virtual positive image charge in the metal because of the requirement that the electric field be perpendicular to the interface. The coulomb attractive force on the electron towards the metal surface due to its virtual positive charge is a196 a108 a24a10a39a72a175a85a172a159a24a76a39 a175 a31 a137a62a197a198a144a111a11a35a90a132a102a85a117a12a93 a31 a24a76a39 a175 a31 a46a35a69a30a197a181a144a151a11a142a117 a31 (2.3) The electrostatic energy associated with this image force is determined by the work the image force does as the electron moves from point x to a199 a103 a108 a24 a200 a196 a108 a90a118a117a12a93a79a29a66a117a201a24a76a39 a175 a31 a46a35a69a85a197a198a144a111a11a133a117 (2.4) 27 a144a111a11 is the high frequency permittivity of the semiconductor rather than static permittivity because the thermal velocity of the electron as it approaches the surface is quite high (a45a202a46a2a50 a33 a57 a84a30a4 a97 ) [50], and the semiconductor might not be fully polarized in the short time the electric field is produced. The image potential energy has to be added to the potential energy due to the Schottky barrier as shown in figure 2.7. a103a76a24a21a103 a108 a158a55a103a140a11a163a24a76a39a140a175a62a13 a108 a90a123a117a114a93a198a39a52a175a61a13a15a14a35a90a123a117a114a93 a103a122a90a123a117a114a93a203a24a10a39 a175a30a31 a46a2a69a30a197a198a144a111a11a68a117 a39a52a175a61a13a15a14a2a90a118a117a12a93 (2.5) Since the image potential is only important near the surface, and the Schottky barrier field is constant near the surface, the maximum potential energy occurs at a position a117 a6 where the resultant electric field vanishes - that is, where the image force field is equal and opposite to the depletion region field. a172a204a6a149a148a151a150a48a24 a175 a46a2a69a30a197a198a144a111a11a68a117 a31 a6 a193 a117a7a6a184a24a157a205 a175 a46a35a69a85a197a198a144a111a11a133a172a204a6a149a148a151a150 (2.6) The amount of barrier lowering can be determined by finding the extremum, a127a151a206a89a207a150a35a208 a127 a150 a24a184a50 a96a204a209 a117a201a24a134a117a15a6 a210 a5a83a14 a108 a24a134a117a15a6a140a172a66a6a149a148a151a150a72a158a22a211 a175 a46a35a69a85a197a198a144a111a11a133a172a204a6a149a148a151a150a83a212 a97a142a213 a31 a24a21a102a85a117a15a6a72a172a204a6a149a148a151a150 a210 a5a83a14 a108 a24a121a102a86a172a204a6a149a148a151a150a94a211 a175 a46a35a69a30a197a181a144a151a11a68a172a204a6a149a148a151a150a83a212 a97a43a213 a31 a24a21a102a174a211 a175a85a172a66a6a149a148a151a150 a46a35a69a30a197a181a144a151a11a7a212 a97a43a213 a31 (2.7) The effect of the image force is to lower the barrier that the electron has to overcome in passing from the metal into the semiconductor by the amount a210 a5a53a14 a108 . For this contribution to the barrier energy to be present, there must be electrons in the conduction band near the top 28 Figure 2.7: Effect of image force lowering on the barrier height. of the barrier. This requirement does not affect other contributions to a5 a143 from work function difference, surface state charge, etc. [48, 50]. Holes can be attracted to the metal by image force as well. Hole energy is measured down- ward from the top of the valence band, so that the effect of the image force is to bend the valence band upward near the metal surface. 2.1.2 Metal-SiC Contacts Ohmic contacts to SiC An ohmic contact to a semiconductor is usually produced by heavily doping the surface region of the semiconductor immediately under the contact. It is known that, to first order, the equilibrium barrier height is not affected by an increase in semiconductor doping, but that the depletion width decreases with increased doping. Reducing the depletion width increases the 29 Figure 2.8: Effect of doping concentration on the barrier width. (a) low (b) moderate and (c) high doping [44]. carrier tunneling probability, and contact resistance is generally observed to drop with increased doping as depicted in figure 2.8. When the semiconductor doping exceeds a45a22a46a35a50 a97a132a214 a56a58a57 a4 a34 (moderate doping regime), signifi- cant tunneling can take place through the thin upper portion of the barrier. For doping exceeding a45a202a46a35a50 a97a43a106 a56a58a57 a4 a34 (high doping regime), the entire barrier becomes so narrow that low energy major- ity carriers can tunnel the barrier; that is, the barrier becomes effectively transparent to carrier flow [44]. The specific contact resistance ra145 is the parameter that characterizes ohmic contact. This parameter depends on the doping concentration of the semiconductor, the characteristics of the semiconductor surface (e.g., epilayer or implant/activation) and the high temperature anneal conditions for forming the ohmic contact. 30 n-type a36 -SiC Ohmic contacts on n-type a36 -SiC (4H and 6H) have been studied over the last two decades. The quality of SiC, both bulk and epitaxial, has improved steadily over the same period. The availability of heavily doped n-type material has improved also because of better understand- ing of doping processes - both implant/activation and doping during epitaxial growth. Specific contact resistance of the order of a46a2a50 a4a7a51 a23 a56a58a57 a31 are common for doping concentration of about a46a35a50 a97a43a106 a56a58a57 a4 a34 . Much of the contact work was previously done for 6H-SiC, but recently the focus has shifted to 4H-SiC because of its superior mobility characteristics [46]. Metal silicide formation has been the most dominant mechanism for ohmic contact forma- tion, on either 4H- or 6H-SiC polytypes. Results have shown that the specific contact resistance increases with increasing barrier height, but decreases as the doping concentration increases. For heavily doped materials, the specific contact resistance was predicted to increase dramatically. Nickel silicide ohmic contact formed during the high temperature (a45a184a60a62a50a61a50 a0a35a1 ) annealing of nickel deposited on SiC is the most widely used contact on n-type SiC [51, 52, 4, 53, 54]. Reported specific contact resistance vary from between a46a124a47a49a46a2a50a32a4a7a51 and a46a135a47a49a46a35a50a204a4a53a31a2a23 a56a54a57 a31 . With proper pro- cessing, nickel makes a relatively good ohmic contact on moderately doped (a45a152a46a35a50 a97a43a214 a56a54a57 a4 a34 ) SiC [46]. The first ohmic contact fabricated on 6H-SiC was deposited nickel which was subsequently annealed at high temperatures by Palmour, et al [55]. Nickel ohmic contacts on 6H-SiC have been characterized electrically and physically by Crofton et al [5]. Specific contact resistances a19a41a65a114a47a161a46a35a50 a4a7a51 a23 a56a58a57 a31 were measured following 2 minute anneals of the samples in vacuum at a60a66a65a30a50 a0a2a1 . Nickel layers were deposited on epilayers with doping concentrations between a101a7a39a48a60a89a47a215a46a2a50 a97a43a214 a56a58a57 a4 a34 . Rutherford backscattering spectroscopy and Auger electron spectroscopy have shown that the 31 high temperature anneal leads to the reaction of Ni with SiC forming nickel silicide. Nickel silicide formation has also been observed by Liu, et al [6]. During the formation of nickel silicide, carbon is set free at the interface and migrates towards the surface of silicide layer. Apart from the formation of nickel silicide, other mechanisms such as vacancy formation might also contribute to forming the ohmic contact. As-sputtered nickel silicide on SiC epilayers does not form an ohmic contact. Ohmic contacts result only from the reaction of Ni and SiC at high temperature. Liu, et al [6] included chromium in the contact layer, for example Ni/W/Cr, in order to improve contact stability while retaining the low specific contact resistance of pure nickel. It was suggested that chromium reacts with the carbon liberated by the reaction of nickel with SiC. Stable a1 a71a153a34 a1 a31 and a73a186a80 a31 a26a89a80 compounds are the product of this coupled reaction. The choice of metal for the formation of a silicide contact to n-type SiC is not limited to Ni. Metals such as Co, Hf, and Ta can also form silicides on SiC with physical and electrical properties similar to nickel silicide contact [46]. Binary alloys and multilayer contacts have also been investigated as possible ohmic contact to n-type SiC. Table 2.2 shows a list of ohmic contacts for n-type a36 -SiC and some typical values of the specific contact resistance. TiN reported by Glass, et al [56, 57] and TiW reported by Crofton, et al [54] are among the special cases of ohmic contact where x-ray photoelectron spectroscopy (XPS) has shown that a thin insulating layer (0.5 - 1.5nm) of silicon nitride at the MS interface formed a metal-insulator- semiconductor (MIS) structure and promoted ohmic behavior. TiN has a low work function which is favorable for an ohmic contact on n-type material. Ohmic behavior was not observed without the formation of a thin Si-N layer [58]. 32 SiC carrier Annealing Method of a71 a145 Metallization conc.(a56a58a57 a4 a34 ) condition a71 a145 a90a160a23 a56a54a57 a31a2a93 measurement Ref. Ni a137a18a87a88a65a75a47a52a46a2a50 a97a43a216 1000a0 C, 20s a46a85a87a88a101a75a47a183a46a35a50 a4a15a74 TLM [53] Ni a137a18a87a88a101a75a47a52a46a2a50 a97a132a214 950a0 C, 5mins mid a46a35a50a32a4a53a31 4-pt. probe [54] Ni a101a204a87a100a60a131a47a52a46a2a50 a97a132a214 950a0 C, 2mins a19a136a65a131a47a52a46a2a50 a4a7a51 TLM [5] Ni a60a32a87a100a63a131a47a52a46a2a50 a97a43a216 1050a0 C, 5mins a46a2a50a32a4 a34 a39a136a46a2a50a32a4a15a74 TLM [58] Ni a137a18a87a88a65a75a47a52a46a2a50 a31a142a67 1000a0 C, 5mins a46a215a47a183a46a35a50 a4a7a51 Cont. area [4] Ni a64a32a87a88a102a75a47a52a46a2a50 a97a43a216 1000-1200a0 C a46a85a87a100a64a131a47a183a46a35a50 a4 a33 a39 TLM [87] and a46a30a87a217a137a116a47a183a46a35a50 a97a132a214 1min a64a32a87a100a69a131a47a183a46a35a50 a4a7a51 Ni a102a75a47a183a46a35a50 a97a132a214 a39 950-1000 a0 C a137a116a47a183a46a35a50a32a4a15a74a85a39 TLM [88] a102a75a47a183a46a35a50 a97a132a106 a45a10a46a35a50 a4a7a51 Mo a102a75a47a183a46a35a50 a97a132a214 a39 950-1000 a0 C a137a116a47a183a46a35a50a32a4a15a74a85a39 TLM [88] a102a75a47a183a46a35a50 a97a132a106 a45a10a46a35a50a204a4 a33 Ni-Cr a137a18a87a88a101a75a47a52a46a2a50 a97a132a214 950a0 C, 5min a46a85a87a100a63a131a47a183a46a35a50 a4 a34 a39 circular [54] (60-40 Wt%) TLM Ti a102a75a47a183a46a35a50 a97a132a214 a39 as-deposited a46a215a47a183a46a35a50a32a4a53a31a153a39 circular [89] a46a215a47a183a46a35a50 a31a142a67 a19a136a102a131a47a52a46a2a50 a4 a33 TLM W a64a131a47a183a46a35a50 a97a132a214 a39 1200-1600 a0 C a65a75a47a183a46a35a50a32a4 a34 a39 4-pt probe [90] a46a215a47a183a46a35a50 a97a132a106 a46a215a47a183a46a35a50a32a4a15a74 TiW a137a18a87a88a101a75a47a52a46a2a50 a97a132a214 600a0 C, 5min a101a204a87a100a63a131a47a183a46a35a50 a4a15a74 circular [54] TLM Mo a8a218a46a215a47a52a46a2a50 a97a132a106 as-deposited a45a10a46a191a47a52a46a2a50a32a4a15a74 4-pt probe [92] TLM Ta a8a218a46a215a47a52a46a2a50 a97a132a106 as-deposited a45a10a46a191a47a52a46a2a50a32a4a15a74 4-pt probe [92] TLM Ni, Ni/W a46a2a50 a97a43a216 a39a25a46a35a50 a97a132a214 1000-1050a0a2a1 a46a2a50 a4 a34 a39a136a46a2a50 a4a7a51 TLM [93] Ni/Ti/W 5-10min Cr/W a46a2a50 a97a43a216 a39a25a46a35a50 a97a132a214 1000-1050a0 a1 a46a2a50a32a4a53a31a163a39a136a46a2a50a32a4a15a74 TLM [93] Cr/Mo/W 5-10min TiC a137a116a47a183a46a35a50 a97a132a106 etched at a46a2a64a61a50a62a50 a0a2a1 a46a85a87a100a64a131a47a183a46a35a50 a4 a33 TLM [94] for 15mins in a219 a31 Table 2.2: Some ohmic contacts on n-type a36 -SiC. The layers in multi-layered contacts are seper- ated with slashes, layers at the surface to the interface with SiC proceed from right to left 33 p-type a36 -SiC It is more difficult to form ohmic contacts to p-type SiC by simply reducing the Schottky barrier height because of the large bandgap and large work function of SiC. Aluminum (pure and alloy) is convetionally used to create ohmic contacts on p-type material [58]. The low melt- ing point and rapid oxidation characteristics of aluminum makes processing the contact difficult. The melting point can be increased by using aluminum alloys, and the most widely used alloys are Al-Ti of different compositions. The thermodynamic driving force for aluminum oxidation is very high, and this places restrictions on using aluminum-based contacts [7, 60]. Another difficulty is the fact that aluminum is very volatile at moderate annealing temperature. Work re- ported by Crofton, et al [9] have shown that 90/10 wt% Al/Ti alloy layer annealed at a46a2a50a61a50a61a50 a0 a1 can loose aluminum to the annealing environment, thus increasing the metal sheet resistance signifi- cantly. Using aluminum-based alloys can place strict requirements to processing and passivating contacts. As shown in table 2.3, the earliest reported ohmic contacts to p-type SiC contain either aluminum or boron [59] and were annealed at very high temperatures a45a20a46a153a101a30a50a61a50 a0 a1 . An enhanced p-type concentration at the SiC surface, probably due to recrystallization from solution rather than aluminum diffusion into the SiC, was observed. Ohmic contact on p-type SiC are now achieved primarily by reducing the depletion width via high doping concentration rather than by reduction of the Schottky barrier height. Tunneling current dominates the electron transport for highly doped SiC, and the resulting specific contact resistance varies according to the relation [61, 62] a71 a145a163a220a41a221a2a222a32a223 a211 a5 a143 a224 a73 a212 (2.8) 34 SiC carrier Annealing Method of a71 a145 Metallization conc.(a56a58a57 a4 a34 ) condition a71 a145 a90a91a23 a56a58a57 a31 a93 measurement Ref. Al a46a85a87a100a63a131a47a52a46a2a50 a97a132a214 700a0 C, 10min a46a30a87a100a101a131a47a185a46a35a50 a4 a34 TLM [54] Al a63a94a47a185a46a35a50 a97a132a214 800a0 C, 10min a46a35a50 a4a53a31 a39a136a46a2a50 a4 a34 TLM [58] Al-Si NR 1700a0 C NR - [59] Al-Si NR 900-1000a0 C NR - [95] Al-Ti a65a131a47a185a46a35a50 a97 a33 - 1000 a0 C, 5min a102a32a87a225a60a94a47a185a46a35a50a32a4a53a31 - Circular [96] a102a131a47a185a46a35a50 a97a132a106 a46a30a87a100a65a131a47a185a46a35a50a32a4 a33 TLM Si-B NR 1700-2000a0 C NR - [59] Ta a8a218a46a191a47a183a46a2a50 a97a132a106 as-deposited a101a131a47a183a46a2a50 a4a15a74 TLM [92] 4-pt probe Ti a8a218a46a191a47a183a46a2a50 a97a132a106 as-deposited a64a94a47a183a46a2a50 a4a15a74 TLM [92] 4-pt probe Mo a8a218a46a191a47a183a46a2a50 a97a132a106 as-deposited a102a131a47a183a46a2a50a32a4a15a74 TLM [92] 4-pt probe Al/Ti Al implant 500a0 C, 20min a65a32a87a225a69a94a47a185a46a35a50 a4a15a74 4-pt probe [97] dose: a46a191a47a183a46a35a50 a97 a33 (1650 a0 C, 30min) Al/Ti/Ai a60a94a47a185a46a35a50 a31a142a67 600-800a0 C a46a179a39a115a102a75a47a52a46a2a50 a4a15a74 TLM [98] W NR 1900a0 C NR - [59] Al/W/Au- NR 1900a0 C, 120s a102a226a39a115a65a75a47a52a46a2a50a32a4a15a74 4-pt probe [90] W/W Table 2.3: Some ohmic contacts on p-type a36 -SiC. The layers in multi-layered contacts are seper- ated with slashes, layers at the surface to the interface with SiC proceed from right to left 35 a71 a145 is the specific contact resistance, a5 a143 is the schottky barrier height, N is the carrier concentra- tion. It may be possible to reduce or elliminate the need for annealing contacts to either n-type or p-type SiC if the surface can be doped heavily enough. As deposited Mo, Ta, and Ti contacts on p+ SiC epitaxial layers grown by chemical vapor deposition (CVD) have been reported to yield ohmic contacts [63]. Specific contact resistances measured by Kuphal?s 4-point method were of the order of a46a35a50 a4a15a74 a23 a56a58a57 a31 , though the method does not account for nonuniform current densities. To reduce the Schottky barrier height on a p-type material, metals with high work function must be used, though empirical evidence shows that the Schottky barrier height has a weak dependence on metal work function because of partial pinning of the Fermi level. Platinum (high work function (5.65eV), high melting point and high resistance to oxidation) has displayed ohmic behavior both as deposited and after contact anneals at a63a61a65a62a50 a0a2a1 [58]. Work on Al-Ti contacts [9] including surface studies of etched Al-Ti contact layers have shown many pits of significant size and density at the contact surface, suggesting that aluminum diffusion may not actually be doping SiC surface, but may instead have resulted in enhanced field emission because of the creation of many hemispherical intrusions into the SiC surface similar to those observed by Braslau [10] for Au-Ge contacts on GaAs. Specific contact resistance as low as a137a155a47a41a46a35a50a204a4a7a51a2a23 a56a58a57 a31 have been reported for a p-doping concentration of a102a122a47a49a46a35a50 a97a43a106 a56a58a57 a4 a34 in 6H-SiC [64] for non aluminum-base contact. Double layers (160nm of silicon on 50nm of cobalt) were annealed at a65a30a50a61a50 a0a2a1 for 5 hours followed by a60a62a50a61a50 a0a2a1 for 2 hours. Cobalt silicide was formed, and the absence of graphite was observed at the interface by RBS. The authors used silicon to prevent the formation of a carbon rich phase, since residual carbon has been reported as the cause of high contact resistance. 36 n-type a37 -SiC Nickel contacts annealed between a60a61a50a62a50a129a39a155a46a35a102a61a65a62a50 a0a54a1 are commonly used for a37 -SiC [65, 66, 67, 68]. Au-Ta, Cr, TaSi a31 and Al are some of the metals that have been used for ohmic contact to n-type a37 -SiC [66]. Steckl and Su [67] have used as-deposited and annealed nickel for rectifying and ohmic contact, respectively, on the same device. Daimon, et al [65] reported that annealed nickel and as-deposited aluminum formed ohmic contacts on n-type a37 -SiC (100) with low carrier concen- trations (a65a227a47a25a46a35a50 a97 a51a228a39a10a46a94a47a25a46a35a50 a97a142a216 a56a54a57 a4 a34 ). High specific contact resistance for both contacts were attributed to the low doping concentrations. Table 2.4 shows a number of ohmic contact metallization schemes. Most of the contacts were annealed at temperatures above 800a0a35a1 . There is no simple formular for creating ohmic contact on a37 -SiC. Ohmic contact formation may be due to silicide formation, as in the case of Ni, or both silicide and carbide formation (Cr, Ta, W, Ti, Mo), or neither silicide nor carbide as for Au and Ag [69]. Many of the studies reported the specific contact resistance as shown in table 2.4. Multilevel metallization schemes were investigated by Shor, et al [70] based on Ti and W for high temperature (650 - 750a0 a1 ) applications. Electronic and optoelectronic extended opera- tion at high temperature need such a multilevel metallization scheme. Multilevel contact metal- lization addresses concerns like reactivity, oxidation and diffusivity within the metallization and with the SiC itself. The most promising metallization scheme reported was Au/Pt/TiN/Ti, which remained ohmic for 31 hours at 650a0a2a1 and then became rectifying. TiN layer was believed to act as a diffusion barrier. 37 SiC carrier Annealing Method of a71 a145 Metallization conc.(a56a54a57 a4 a34 ) condition a71 a145 a90a91a23 a56a58a57 a31 a93 measurement Ref. Al a65a131a47a183a46a35a50 a97 a51 as-dep a46a30a87a225a69a94a47a183a46a2a50a32a4 a97 3-cont. [66] Ni a65a131a47a183a46a35a50 a97 a51 a46a153a102a62a65a62a50 a0a2a1 , 5 min a46a30a87a229a137a122a47a183a46a2a50 a4 a97 3-cont. [66] Cr a65a131a47a183a46a35a50 a97 a51 a46a153a102a62a65a62a50 a0a2a1 , 5 min a101a32a87a225a50a94a47a183a46a2a50 a4a53a31 3-cont. [66] Ti a46a35a50 a97a43a216 a39a136a46a35a50 a97a43a214 a64a61a50a62a50 a0a2a1 , 30-90 min a101a32a87a225a69a228a39a49a60a32a87a88a102a75a47a52a46a2a50 a4 a34 4-point [99] W a46a35a50 a97a43a216 a39a136a46a35a50 a97a43a214 as-dep - a46a30a87a100a65a131a47a183a46a2a50a32a4a53a31 4-point [99] a69a61a50a62a50 a0a2a1 , 10min Ta a65a131a47a183a46a35a50 a97a43a106 as-dep - a101a131a47a183a46a35a50a204a4 a216 a39 circular TLM [100] a46a35a50a62a50a61a50 a0a2a1 , 1hr a137a15a87a225a64a94a47a183a46a2a50 a4a7a51 Re a65a131a47a183a46a35a50 a97a43a106 as-dep - a46a191a47a183a46a35a50 a4a15a74 a39 circular TLM [100] a60a61a50a62a50 a0a2a1 , 30min a46a191a47a183a46a35a50 a4 a33 Pt a65a131a47a183a46a35a50 a97a43a106 as-dep - a69a94a47a183a46a35a50a204a4a7a51a85a39 circular TLM [100] a65a62a50a62a50 a0a2a1 , 30min a46a191a47a183a46a35a50 a4 a33 Au/Pt/Ti a46a35a50 a97 a51a149a39a136a46a35a50 a97a142a216 650a0 C, 1hr a46a30a87a230a46a191a47a183a46a2a50a32a4a15a74 4-pt probe [70] Au/Pt/W a46a35a50 a97 a51a149a39a136a46a35a50 a97a142a216 650a0 C, 8hr a102a131a47a183a46a35a50a204a4a15a74 4-pt probe [70] Au/Pt/- a46a35a50 a97 a51 a39a136a46a35a50 a97a142a216 650a0 C, 31hr a46a30a87a229a137a122a47a183a46a2a50 a4a15a74 4-pt probe [70] TiN/Ti Pt/TiW/Ti a46a35a50 a97 a51a149a39a136a46a35a50 a97a142a216 650a0 C, 3hr a102a32a87a225a69a94a47a183a46a2a50a32a4a15a74 4-pt probe [70] a138 a96 a26a89a80 a31 a65a131a47a183a46a35a50 a97 a51 a63a66a65a30a50 a0a2a1 , 5 min a102a131a47a183a46a35a50 a4a53a31 3-cont. [66] a138a161a80a142a26a174a80 a31 a46a35a50 a97a43a216 a39a136a46a35a50 a97a43a214 a46a35a50a62a50a61a50 a0 a1 , 10s + a46a30a87a230a46a191a47a183a46a2a50a32a4a15a74 4-point [99] a137a32a65a30a50 a0 a1 , 390min a231 a26a89a80 a31 a46a35a50 a97a43a216 a39a136a46a35a50 a97a43a214 a46a35a50a62a50a61a50 a0a2a1 , 10s + a64a79a87a225a60a94a47a183a46a2a50 a4a15a74 4-point [99] a137a32a65a30a50 a0a2a1 , 390min Table 2.4: Some ohmic contacts on n-type a37 -SiC. The layers in multi-layered contacts are seper- ated with slashes, layers at the surface to the interface with SiC proceed from right to left 38 SiC carrier Annealing Method of a71 a145 Metallization conc.(a56a58a57 a4 a34 ) condition a71 a145 a90a91a23 a56a58a57 a31a35a93 measurement Ref. Al a46a215a47a183a46a35a50 a97 a51 a63a61a63a62a50 a0 a1 , 3min a64a79a87a230a46a191a47a183a46a2a50a32a4a53a31 3-cont. [66] Ni a46a2a50 a97a43a216 a39a136a46a35a50 a97a132a214 as-dep - a137a15a87a230a46a191a47a183a46a2a50a32a4a53a31 - 3-cont. [68] a101a62a50a62a50 a0a2a1 , 15min a102a32a87a225a63a94a47a183a46a2a50 a4a53a31 Au-Ta-Al a46a215a47a183a46a35a50 a97 a51 a46a153a102a30a50a61a50 a0 a1 , 30min a137a15a87a100a101a131a47a183a46a2a50a32a4 a97 3-cont. [66] (91:2:7 at%) TaSi a31 /Al a46a215a47a183a46a35a50 a97 a51 a46a153a102a30a50a61a50 a0a54a1 , 30min a102a32a87a225a50a94a47a183a46a2a50 a4 a97 3-cont. [66] Table 2.5: Some ohmic contacts on p-type a37 -SiC. The layers in multi-layered contacts are seper- ated with slashes, layers at the surface to the interface with SiC proceed from right to left p-type a37 -SiC There are few reported studies on ohmic contacts to p-type a37 -SiC. Some of the contacts in literature are listed in table 2.5. Specific contact resistances reported for aluminum contact or aluminum alloy contact are high [58]. The high annealing temperature does not seem to help the contact characteristics significantly, probably because of the high driving force for aluminum oxidation, causing aluminum to diffuse away from the interface and react with oxygen. Variable and sometimes conflicting results for contacts on a37 -SiC may be attributed to [58] (1) differences in crystal quality of a37 -SiC layer due to defect densities that are much higher than in a36 -SiC polytypes and (2) differences in the state of the surface prior to metal deposition. Schottkty contacts to SiC Schottky contact on both n- and p-type SiC are reported in the literature for different contact metals. The first Schottky contact on 6H-SiC were made by fusing small pellets of Si-Al and Si-B alloys to single crystal SiC at 1700a0a2a1 and a8 2000a0a2a1 , respectively [59]. These diodes showed sharp breakdown at voltages a45a232a65a30a50a66a13 and high reverse current prior to breakdown. A 39 forward current of 0.5A at 4.5V is typical of these diodes [43]. After this pioneering work, a lot of 6H-schottky barrier diodes have been fabricated. Mead and Spitzer [71] and Mead [72] had early reports of barrier height studies on 6H-SiC Schottky contacts. Al and Au were used as metal contacts on n-type SiC. They found that Schottky barrier height, determined by capacitance-voltage technique, was almost independent of the work function of the metals used for the contact. Hagen [73] drew similar conclusion for Au, Ag and Al contact on both p- and n-type samples of cleaved and etched 6H- and 15R- polytypes. Schottky diodes on a37 -SiC were first reported by Yoshida et al [74] on epitaxially grown n- type layers with Au metal contact. Barrier heights in the range of 1.11 - 1.15eV were determined by C-V and photoelectron techniques. Waldrop and Grant [75] investigated the effect of the choice of metal on 3C-SiC Schottky barrier characteristics. Metals such as Pd, Au Co, Ti, Ag, Tb, and Al were considered, but only Pd, Au and Co formed schottky contacts. The interface chemistry and the Schottky barrier height were studied using x-ray photoelectron spectroscopy (XPS), I-V and C-V techniques. Shenoy, et al [76] have also reported Pt-Schottky barrier diodes on a233 a4 a27a85a233a28a3 3C-SiC grown on a233 a3 -Si substrates. The diodes had low specific-on resistance (a69a32a87a234a46a226a47a155a46a35a50a32a4a15a74a2a23 a56a58a57 a31 ) and a break- down voltage of about 85V for a a50a79a87a229a137a61a77 a57 thick drift layer. Schottky barrier heights of about 0.85V with an ideality factor of 1.25 were obtained. High-voltage Schottky barrier diodes on 4H-SiC were first reported by Itoh et al [77, 78]. The contact metals used include Au, Ni and Ti. I-V and C-V techniques were used for barrier height determination. They reported barrier height dependence on work function because surface ?pinning? of the Fermi level did not occur. The room temperature breakdown voltage was around 800V, and a current density of a46a35a50a61a50a30a70 a56a58a57 a4a53a31 was obtained for forward bias of 1.67V. 40 The breakdown voltage of Ti/4H-SiC diodes increased to 1100V when a highly resistive layer is formed at the periphery of the contact, serving as edge termination [79]. The edge termination was formed by boron ion implantation followed by high temperature heat treatment to remove damage to the lattice sustained during implantation. Raghunathan et al [80] have reported high breakdown voltage (1000V) for Ti/4H-SiC. Weitzel et al [81] reported 1400V breakdown on 4H-SiC with an on-resistance of a46a30a87a88a65 a57 a23 a56a58a57 a31 . Saxena and Steckl [82, 83] have also reported Ni and Pt Schottky contacts on 4H-SiC. Barrier heights extracted from common techniques (C-V, I-V, XPS) have generally indicated Fermi level ?pinning?. All nickel contacts (ohmic and schottky) on 6H-SiC have been reported by several authors [84, 85, 86, 83], and breakdown voltages exceeding 1000V at 25 and 300a0 a1 have been observed. Most of the diodes used an oxide for surface passivation and device isolation. 2.2 Diffusion and Oxidation Barriers 2.2.1 Diffusion in Solids Fick?s laws are the ealiest mathematical basis for macroscopic diffusion. For an inhomoge- nous single phase binary alloy with diffusion coefficient (a188 a97 ) of component 1, the concentration gradient in steady state is defined by Fick?s first law as a235 a97 a24a76a39a140a188 a97 a211a129a236 a1 a97 a236 a117a49a212 a147 a158 a1 a97a15a237 a24a76a39 a210 a147 a1 a97 a158 a1 a97a166a237 (2.9) where a235 a97 is the flux of atoms of component 1 at a given time, a1 a97 is the concentration and v is the velocity of mass moving because of the application of forces such as electromigration, 41 thermal or chemical potential gradient [118]. An inhomogenous specimen becomes homogenous if annealed long enough, and the net flow of matter will cease. The diffusion coefficient is called the self-diffusion coefficient if no external forces and chemical gradient are involved. For diffusion under a chemical gradient, where the diffusion is affected by the motion of all atomic species, the diffusion process is called inter-diffusion with a chemical diffusion coefficient. Under non-steady state conditions where the concentration is changing with time, Fick?s second law states that the rate of change of concentration is equal to the gradient of the flux a236 a1 a236 a209 a24a218a39 a210 a95 a235 a24a184a188 a210 a31 a1 (2.10) It is generally assumed that D is not a function of position. Fick?s second law is the basis for most of the diffusion measurements and calculations in solids. It can be applied for different kinds of sample geometry [118]. Solving a diffusion equation with all the possible driving forces included can be a chal- lenging mathematical problem. The simplest mathematical treatment assumes the absence of chemical driving forces, which is strictly valid only if thermodynamic equlibrium is established between two samples with different diffusion coefficients. Chemical driving forces may be ab- sent if the components of the diffusion couple are identical and hence have identical diffusion properties, or if the two components of the diffusion couple are chemically different but co-exist in thermodynamic equilibrium. For example, a metal can be in equilibrium with its oxide [119]. The solution to Fick?s second law depends largely also on sample geometry. Several geometry specific solutions are described below. 42 Thin film on bulk sample Consider an infinitesimally thin layer of areal density M with thickness less than the diffu- sion distance a90a160a188 a209 a93 a97a142a213 a31 such that it can be represented by a Dirac delta function a189 . Assuming no surface flux or flux through imperfections, the boundary conditions in one-dimension are a1 a90a123a117a198a238a111a50a66a93a129a24a134a38a239a189a18a90a118a117a12a93 and a240a153a241 a207 a67a54a242 a147 a208 a240 a150 a24a184a50 The solution to Fick?s second law is a1 a90a123a117a198a238 a209 a93a203a24 a38 a102 a224 a197a198a188 a209 a221a54a222a79a223 a211 a39 a117a12a31 a137a204a188 a209 a212 (2.11) M is the total amount of diffusant per unit area, t is the diffusion time and C is the concentration at position x and time t. This geometry is widely used in radioactive tracer work compatible with a delta function type of diffusant distribution and infinitesimally small mass thickness. Most thin film work on bulk layers is conducted with an initial thickness of the source of diffusant (h) that is greater than the diffusion distance a90a160a188 a209 a93 a97a142a213 a31 when the following boundary conditions are satisfied a1 a90a123a117a198a238a111a50a66a93a129a24 a1 a139 if a243a227a59a176a117a159a59a136a50 , a1 a90a118a117a181a238a151a50a61a93a129a24a184a50 for a117a244a8a41a243 and for a209 a8a136a50 , a240a153a241 a207a67a54a242a147 a208 a240 a150 a24a184a50 , the solution is given by a1 a90a118a117a181a238 a209 a93a203a24 a1 a139 a102a98a245 a82a35a71a66a146a246a211 a117a75a158a25a243 a102 a224 a188 a209 a212 a39a49a82a35a71a61a146a246a211 a117a122a39a115a243 a102 a224 a188 a209 a212a161a247 (2.12) Diffusion couple with constant surface composition For a diffusion couple with two samples having uniform initial concentrations of a1 a139 and a1 a97 , the initial boundary conditions at a209 a24a184a50 can be witten as 43 a1 a90a123a117a198a238a111a50a66a93a122a24 a1 a97 for a117a10a19a99a50 and a1 a90a118a117a181a238a151a50a61a93a122a24 a1 a139 for a117a202a8a99a50 . The solution to the diffusion equation is then a1 a90a123a117a198a238 a209 a93a198a39 a1 a139 a1 a97 a39 a1 a139 a24 a46 a102 a245 a82a35a71a61a146 a56 a211 a117 a102 a224 a188 a209 a212a72a247 a87 (2.13) In the diffusion couple solution, if the difference between a1 a97 and a1 a139 is large such that a117a227a194 a102 a224 a188 a209 the diffusion coefficient will vary with composition along x. To determine the diffu- sion coefficient at known composition, Boltzmann-Matano analysis is used [118]. Diffusion in structurally inhomogenous sample Thin film diffusion samples are almost never structurally homogenous. That is, defects like dislocations and grain boundaries are always present. Diffusion occurs in the lattice through equilibrium point defects such as vacancies, interstitial atoms, divacancies, etc, but engineering materials are mostly polycrystalline in nature, and they contain non-equilibrium defects such as dislocations, stacking faults and grain boundaries. Diffusion along these defects is greater than in monocrystalline samples. The diffusive process in thin films may be largely controlled by grain boundary diffusion, especially at low temperature (a45a202a46a2a50a61a50 a0a2a1 ) [118]. In polycrystalline materials, there is simultaneous diffusion within the grain and along the grain boundaries, and these two diffusion mechanism are coupled. The ease with which lattice atoms go into the grain boundary or vice-versa will determine the diffusion kinetic regime that prevails in the whole sample. According to Harrison [120], three types of kinetic regime labelled A, B, and C are possible. A-Kinetics: Extensive lattice diffusion that causes the diffusion fields from adjoining grains to overlap. 44 B-Kinetics: Each boundary is assumed to be isolated and the flux at large distance from a grain boundary approaches zero. C-Kinetics: Lattice diffusion is considered negligible, and significant atomic transport occurs only within the boundaries. Grain boundary diffusion measurements in bulk materials invariably involve B-kinetics. Little evidence has been obtained from profiling experiments on bulk materials that clearly sup- port the existence of A- or C-kinetics. In this respect, thin films are both unique and challenging as candidates for diffusion studies because the very high density of structural defects makes it possible to observe any of the three kinetic regimes given the appropriate annealing condition [121, 122]. Diffusion in amorphous materials Diffusion in amorphous solids is a common phenomenon. According to the Stokes-Einstein equation [123], the frictional force impeding the motion of particle i of radius a71 a108 is given by the viscosity of the liquid in which it moves. Einstein?s diffusion coefficient is, a188a248a24a21a249 a143 a138a105a27a62a250 where a250 is the friction coefficient. The Einstein equation becomes the Stokes-Einstein coefficient given by a188 a108 a24a184a249 a143 a138a105a27a54a70a140a197a114a71 a108 . The general feeling is that this equation has broad validity despite the fact that it was derived for a macroscopic sphere moving steadily through a viscous fluid. However, there is nothing in the equation to adjust for the existence of measurable diffusion in glassy solids or metallic systems, where the viscosity may depend on time in some way. Diffusion in solids can be studied via experimental, statistical and atomic mechanism [124]. One of the reasons for the poor understanding of mass transport in amorphous solids is the lack of a consistent approach to the problem. First, proper knowledge of the experimental characteristics 45 of diffusion is needed. The next thing is to propose an atomic mechanism and then build its proper statistical mechanisms in order to compare the experimental data to the macroscopic consequences of the atomic behavior. Experimental diffusion coefficients are frequently expressed in terms of a pre-exponent a188a124a139 and the activation energy a103a105a251 , a188a9a24a184a188 a139 a221a2a222a32a223 a211 a39 a103 a251 a252 a138 a212 a87 (2.14) The temperature range over which this equation is valid is quite narrow, two to three orders of magnitude for the diffusion coefficient at most [125, 126, 127]. Determining a188a75a139 precisely can also be difficult. Inconsistencies between various authors for the same alloy point to measure- ment error and cast doubt on the stated values of a188a124a139 and a103a105a251 . 2.2.2 Thermodynamics of Diffusion Metallizations in microelectronics are composed of many layers of dis-similar materials, and the knowledge of the stability of the interfaces between different materials is important. The total Gibbs free energy of the system (the two materials in contact and the interface be- tween them) can be decreased by different processes including enrichment of the components at the interface with respect to one or both of the components or via chemical reaction of the components followed by the formation of additional phases [129]. The change in Gibbs free energy of the system can be used as a criteria for possible reaction products that form at the interface. The Gibbs free energy is defined from the combination of the first and second laws of thermodynamics. In a closed system the Gibss free energy is given by the relation a253a121a254a76a255 a158a159a78a12a13a21a39a183a138a226a26a52a24a184a219 a39a52a138a140a26 (2.15) 46 where U, V, H, and S are the internal energy, volume, enthalpy and entropy of the system, respectively. In an open system however, the Gibbs free energy also depends on the number of moles of each of the components, and the differential form of the free energy equation becomes a29 a253 a90a118a138a105a238 a1 a238a166a233 a108 a93a129a24a10a39a226a26a174a29a109a138a115a158a176a13a191a29a28a78a131a158 a2 a108 a182 a108 a29a66a233 a108 (2.16) where a182 a108 is the chemical potential of component i. The equilibrium state of the system can be investigated with the Gibbs free energy function. There are three stable equilibrium states: (i) complete, (ii) partial and (iii) local thermody- namic equilibrium [130]. A system in complete equilibrium has its Gibbs free energy function minimum, a29 a253 a24 a50 . Systems in equilibrium with respect to only certain components are said to be in partial equilibrium. For local equilibrium, equilibrium exists only at the interfaces be- tween different phases present in the system. In thin films systems, complete equilibrium is seldom achieved, and the concept of local equilibrium is therefore important. Local equilibrium is treated using chemical potential. In the treatment of multicomponent open systems, the most common process considered in defining the thermodynamic functions for a solution is called the mixing process. The mixing process is the change in state experienced by the system when appropriate amounts of compo- nents in their reference states are mixed together forming a homogenous solution brought to the same temperature and pressure as the initial state [131]. The molar Gibbs free energy of mixing or formation can be expressed as a210 a253 a6a184a24 a210 a219a124a6a136a39a183a138 a210 a26a53a6 (2.17) 47 Figure 2.9: Atomic energy barrier as a function of atomic position [118]. The mixing process is strongly influenced by forces between atoms and molecules (a210 a219a124a6 ). The fundametal cause of mixing, however, is entropy (a210 a26a53a6 ). Diffusion in solids involve atomic exchange with some kind of lattice imperfection, and it is usually thermally activated. We will consider some thermodynamic factors involved in determining the diffusion coefficients in chemically inhomogenous samples. The energy barrier seen by atoms as a function of atomic position is depicted in figure 2.9. Each atom sees the same energy barrier in the absence of a driving force, the probability of a forward or backward jump is the same, and there is no net velocity in any direction. Atoms make a jump if given enough free energy a253 a6 to go over the barrier and occupy an equivalent equilibrium position by an exchange with defect state. The probability (W) that an atom will acquire an energy a253 a6 is given by [118] a231 a24a4a3a85a139 a221a54a222a79a223 a211a114a39 a253 a6 a252 a138a52a212 (2.18) 48 a3a85a139 is the atomic vibration frequency. However, a successful jump is subject to the availability of a defect state in the adjoining position. The average probability of finding a defect is a73a191a127a226a24a6a5 a221a54a222a79a223 a211a114a39 a253 a165 a252 a138a94a212 (2.19) where a253 a165 is the free energy needed to form the defect and Z is the coordination factor. The frequency of successful jumps can be written as a7 a24 a231 a73a191a127a226a24a8a5a9a3a85a139 a221a54a222a79a223 a245 a39 a90 a253 a6a187a158 a253 a165 a93 a252 a138 a247 (2.20) In 3-D, the difussion coefficient can be expressed as a188a9a24 a96 a31 a69 a146 a7 (2.21) where a is the nearest neighbor atomic distance and f is a correction factor (f a10 1) for non- random jump. Hence, a231 a24 a46 a69 a96 a31 a146a11a5a12a3a85a139 a221a54a222a79a223 a245 a90a132a26 a165 a158a176a26a53a6a215a93 a252 a247 a221a54a222a79a223 a39 a245 a90a160a219 a165 a158a55a219a124a6a215a93 a252 a138 a247 (2.22) a210 a219 a165 and a210 a219a135a6 are the enthalpies of formation and motion of defects, respectively, and a26 a165 and a26a53a6 are the corresponding entropies. The temperature dependence of the diffusion coefficient is normally written as a188a248a24a21a188a124a139 a221a54a222a79a223 a211a114a39a14a13a252 a138 a212 (2.23) 49 with a188a75a139a161a24 a97 a51 a96 a31a2a146a11a5a12a3a85a139 a221a2a222a32a223a16a15 a207 a113a18a17 a3 a113a20a19 a208 a21 a22 and a13 a24a184a219 a165 a158a176a219a135a6 . The diffusion of atoms in a solid can occur via the movement of interstitial atoms in the lattice. Lattice atoms can be displaced into the interstitial sites (self-interstitial), and the vacancy left behind is annihilated at an interface, surface or by another interstitial atom. Self-interstitial diffusion is of high probability at high temperature because higher initiation energy is available. When an interstitial atom is formed and vacancy is left behind, the defect is called a Frenkel defect. Frenkel defects dominate in radiation damaged solids. Frenkel defects are very mobile and contribute to mass transport. Interstitial atoms may also be foreign or impurity atoms which are generally of small size, for example, H, C, O and N. Because of their sizes, they move into the interstitial sites easily, and their formation emergies (a26 a165 and a219 a165 ) are negligible compared with the motion energy. Hence, the diffusion of foreign or impurity interstitials may be very fast. Diffusion kinetics change dramatically in the presence of non-equilibrium vacancies and in- terstitial atoms. Deviation from stoichiometry in intermetallic componds can yield non-equilibrium vacancies. In non-stoichiometric alloys, the size of the species that is larger in concentration may dictate whether the diffusion will be enhanced. Finite driving force on atoms For non-zero driving force on an individual atom (linear chemical diffusion regime), a net flux of atoms in a particular direction is observed. A finite mass velocity term is included in Fick?s first law. Free energies between adjoining sites are not the same, and the change in free energy between adjacent sites can be written as a210 a253 a6a134a24 a96 a196 . The number of forward and backward jumps are not the same, and the net number of jumps can be written as 50 a7 a3 a39 a7 a4 a24 a7 a90a91a82 a4a24a23 a39a49a82 a3 a23 a93 (2.24) where a125a215a24 a97 a31a26a25 a112 a19 a21 a119 a7 a3a159a39 a7 a4 a102 a7 a24a28a27a30a29a32a31a34a33 a211 a96 a196 a252 a138a116a212 (2.25) Possible driving forces in thin film metallizations include [118]: Electromigration a35 a5a140a82a35a172 , where a5a226a82 is the effective charge, and a172 is the electric field Thermomigration a35 a36a38a37a40a39 a119a16a41 a240 a119 a240 a150 where a13a43a42 is the heat of transport. Chemical inhomogeniety a36 a240a45a44 a240 a150 a41 a35 a39 a252 a138 a36 a240a38a46a48a47a34a49 a240 a150 a41 where a182 is the activity coefficient. Stress field a35 a50 a240a18a51 a240 a150a53a52 where U is the interaction energy. Chemical composition variations give rise to chemical potential gradients in non-ideal solid reactions. The atomic force over N lattice planes due to a chemical potential grdient can be written as a196 a24 a21 a119 a148a55a54 a90a57a56a32a31 a1 a31 a182 a31 a39a58a56a32a31 a1 a97 a182 a97 a93 for a binary solution with components 1 and 2. a182 a97 and a182 a31 are the activity coefficients for non-ideal solution, a1 a97 and a1 a31 are the concentration terms. Generally if a46a35a50a124a19 a241a24a59a60a49a61a59 a241a63a62a64a49a65a62 a19a218a46a2a50a61a50a61a50 over 10 atomic planes, a50 a148 a16 a21 a119 a52 a24a67a66a69a68 a4 a66a71a70 a66 a190a169a46 This is refered to as the linear regime and a1 a97a133a237 a24 a1 a97 a7 a96 a31 a196 a27 a252 a138 a72a85a71 a237 a24 a128 a62 a16 a21 a119 a24a4a73 a97 a196 a238 the Nernst-Einstein relation where a73 a97 is the mobility. Fick?s first law without a concentration gradient becomes a235 a97 a24a4a73 a97 a196 a1 a97 . 51 2.2.3 Diffusion Barriers Introduction It is known that reliable contact and metallization schemes for microelectronics require thin film diffusion barriers. Though atomic diffusion in the solid phase is a relatively slow process, the distances (film thicknesses) are small, and diffusion can be significant, even at room temperatures in some cases. The idea of a diffusion barrier came to prominence because Al- based metallizations that are commonly used in VLSI technology for contacts and interconnects are highly reactive with Si. Aluminum must be kept away from Si where such reactions could be detrimental to device performance. Electromigration problems with the Al-based interconnects forced researchers look for alternative metals and copper with lower electrical resistivity and higher resistance to electromigration, became a candidate [132, 133, 134, 135]. However, Cu is very mobile in metals and semiconductors even at quite modest temperatures. Copper creates deep traps in Si, resulting in serious degradation of the device performance and reliability [136, 137, 138]. Therefore the need for a very effective diffusion barrier layer became paramount. The initiative of diffusion barrier effort for VLSI technology has been extended to contact metallizations for harsh environment and high power devices fabricated with wide band gap materials (SiC, GaN, etc.). Effective barriers will prevent rapid deterioration of the contact properties and device performance as the result of oxidation and inter-mixing driven by high temperature operation. Reliable long-term operation in harsh environment is the primary goal of this work. 52 Figure 2.10: Diffusion barrier in multi-layer structure [140] Definitions A diffuion barrier material (a74 ) physically separates material (A) from material (B) as shown in figure 2.10. Material B may be an ohmic or Schottky contact or the semiconductor, while material A is suitable for connection with an external circuit. An ideal diffusion barrier should meet the following conditions [139] a44 The transport of A across a74 and of B across a74 should be small a44 The loss rate of a74 into A and into B should be small a44 a74 should be stable thermodynamically against A and B a44 There should be strong adhesion of a74 with A and with B a44 The specific contact resistance of A on a74 and B on a74 should be small a44 a74 should be laterally uniform in thickness and structure 53 a44 a74 should be resistant to mechanical and thermal stresses a44 a74 should be highly conductive (thermally and electrically) Most of the materials used as diffusion barrier may not be able to satisfy all the conditions listed, and compromises usually have to be made. Diffusion barriers can be loosely classified in three groups: (a) sacrificial barriers (b) stuffed barriers and (c) amorphous barriers. When two materials, A and B are separated by diffusion barrier a74 , layer a74 only slows down the eventual mixing of A and B, and the ultimate state of equilibrium is not eliminated by a74 . The question then is, how long can a74 serve its purpose of separating A and B? [139] (a) Sacrificial barrier The length of time that a sacrificial barrier will be useful can be estimated. This means that the point of possible failure is foreseeable. A typical sacrificial barrier is illustrated in figure 2.10. It is essential to fully characterize the compound formed betwween a74 - A and a74 - B, and to know the reaction rates and the activation energies. Predicting the time it will take for a74 to be totally consumed at a given temperature allows the pre-determination of how much of a74 is needed for a given temperature-time cycle to prevent its total consumption. If a74 is totally consumed, it is assumed that the metallization will fail catastrophically. The advantage of the sacrificial barrier is its adaptability. Many binary metal combinations form compounds, so that the choice of material is wide, even taking into consideration other constraints which are important for the proper functioning of the barrier. However, compounds a74 -A and a74 -B formed in the process must be compatible with the constraints - e.g., for electrical contacts, a74 -A and a74 -B must be good conductors, maintaining good thermal and mechanical properties and resistant to corrosion and/or oxidation. The diffusion of A and B in a74 must be 54 negligible compared with the growth of a74 -A and a74 -B. (b) Stuffed barriers Grain boundaries and other structural defects in the diffusion barrier material could serve as rapid diffusion paths when there is no thermodynamic driving force for a74 to react with A and B as in the case of sacrificial barriers. It is essential to be able to stop the diffusion of A and B along the grain boundaries and structural defects. One way to do this is to eliminate the diffusion paths by plugging the paths with suitable impurity atoms or molecules (e.g., carbon or a73 a31 ), and this process is called ?stuffing? the barrier. There are evidence to suggest that stuffing the barrier can indeed be effective. The interpretation of this evidence in terms of atomistic mechanisms is largely conjecture, but the resulting effect can be striking [139] Reactive sputtering is one of the ways of stuffing barriers. Sputtering Mo for example, in an Ar/a73 a31 gas mixture can produce a nearly stoichiometric a38a75a72 a31 a73 layer which works well as stuffed barrier [139, 140]. Electrically conductive nitrides (a138a105a80a142a73a244a238a111a38a75a72 a31 a73 ), borides (a138a105a80a76a73 a31 ) and carbides (TiC, NbC) used as diffusion barriers have had varying success. Stuffing the barriers generally makes it less conducting. (c) Amorphous barriers Apart from plugging the diffusion paths with light atoms and molecules, the diffusion path could be removed by eliminating the grain boundaries altogether by making the diffusion bar- rier either single crystalline or amorphous. It is difficult to practically make single crystalline barriers, and many researchers work with amorphous barriers. However, amorphous barrier are metastable, i.e., they are transformed to polycrystalline films at some elevated temperature and the consequences are the problems associated with polycrystalline barriers. It is also possible that the amorphous layer a74 can react with A and/or B, in which case amorphous barrier becomes sacrificial barrier. 55 Figure 2.11: Stuffed diffusion barrier [140] Pure elemental metals do not form an amorphous phase at room temperature, so amorphous metallic alloys are often used as diffusion barriers. There are a number of empirical rules on how to select the elements to obtain amorphous metallic alloys. Generally, the larger the differences between the constituents in terms of atomic size, crystalline structure and electronegativity, the easier it is for the constituents to form metallic alloys. The two parameters used to characterize metallic alloys are the crytallization temperature (a138 a145 ) and the reaction temperature (a138a7a17 ) of the constituents. Amorphous ternary interstitial alloys of the form MSiN (M = V, Nb, Ta, Cr, Mo, W) have been investigate as diffusion barriers for Cu [141]. Amorphous binary transition metal nitrides of some transition metals have been reported [142, 143, 144, 145]. For tungsten, titanium and tantalum nitrides, researchers have observed that adding silicon stabilizes their amorphous struc- ture. Generally, adding metalloids like (B, C, Si, P, N) to binary amorphous structures has the potential of stabilizing them. The number of ternary amorphous conducting alloys that can be 56 conceived in the way is very large. Beside the combination of a26a89a80a132a150a62a73a78a77 with transition metals listed above, a1 a150a62a73a78a77 and a73a140a150a62a73a79a77 have also received attention as possible candidates. Diffusion barrier literature In the literature, many materials have been tried as diffusion barriers - from single metal to binary and ternary materials, silicides, nitrides, and carbides barriers. The binary compounds are known for their excellent stability while the tenary systems are primarily amorphous and better at slowing the diffusion process [129]. Many of the composite layers studied involve at least the Si/barrier/Cu structure, with additional layers are sometimes included for adhesion purposes. Elemental metallic barriers - Pure metals with high melting points have been used as diffusion barriers on electrical contacts because of the conductivity requirement. But pure metals are polycrystalline, and grain boundaries and other defects serve as paths for rapid diffusion, even when the solubility of the materials in the contact is extremely low. Table 2.6 shows examples of single element diffusion barriers. Polycrystalline films can be more effective barrier if they are stuffed; however, a compromise between the amount of stuffing and other barrier properties such as conductivity and adhesion has to be reached. Binary compound and amorphous materials - In seitching from elemental to compound films, the number of possible materials for the barrier layer increases enormously. However, relatively few binary compound and amorphous barriers have been studied experimentally. The stability of Ti/Pt/Au metallization structure was improved by replacing titanium with tungsten because tungsten has very low solubility in gold and does not react with it. However, to promote adhesion and improve corrosion resistance, titanium was added to the tungsten at levels above the solid solubility requirement [146], and a more effective diffusion barrier was thereby created. Table 57 sample structure annealing ambient barrier effect. (hr/a0 C) failure mode Ref. SiC/Ni/Au Vacuum 990/300 Au reaction with Ni [107] GaN/Ti/Al/Mo/Au NR 360/500 NR [101] Al a31 Oa34 /Ti/Mo/Au NR /a19a136a63a62a50a61a50 Au diff. throu Mo gb [102] Si/Ti/Cu vacuum 0.5/973 reaction at a65a62a50a62a50 a0a2a1 [103] Si/Ta/Cu vacuum 0.5/600 a138 a96 a26a174a80 a31 and a1a81a80 a34a153a26a89a80 [104] Si/Ta/Cu N a31 :H a31 = 9:1 0.5/300 reaction at a137a204a50a62a50 a0 a1 [104] Si/W/Cu vacuum NR Cu diff throu W gb [105] Si/Cr/Cu vacuum NR Cu diff to Si [105] Si/W/Ta/Cu vacuum 0.5/450 NR [105] Table 2.6: Elemental diffusion barrier layer for semiconducting devices 2.7 showed examples of binary compounds and binary alloys used as diffusion barriers. Ternary amorphous barrier - Some attention is currently being given to amorphous ternary alloys as candidates for barrier layer in microelectronics. Binary amorphous alloys of refractory metal silicides films sputtered in argon and nitrogen gas mixtures to form a M-Si-N ternary system, have shown good properties and examples of such systems in litrature are listed in table 2.8. Conducting oxide diffusion barriers - Oxidation of contacts operating at elevated temper- atures is one problem that leads to rapid contact degradation. Even common diffusion barrier materials like TiN and W are sensitive to oxidation [147]. Using thermodynamically stable con- ducting oxides as diffusion barrier at elevated temperatures in air is one option for generating a good diffusion barrier. Ruthenium dioxide (RuO a31 ) is the most thermodynamically stable oxide of ruthenium when formed at temperatures greater than a102a62a50a62a50 a0a2a1 . Ruthenium dioxide is almost as conducting as ruthe- nium metal, and has better conductivity than some silicides that are used as diffusion barriers in integrated circuits [a82a84a83 a164 a24a184a69a79a87a100a101a75a47a183a46a35a50 a4a7a51 a23a187a39 a56a58a57 , a82a71a83 a164 a180 a59 a24a41a137a18a87a100a69a94a47a183a46a2a50 a4 a33 a23a187a39 a56a54a57 ] [147]. 58 sample structure annealing ambient barrier effect. (hr/a0 C) failure mode Ref. SiC/NiCr/Au Vacuum 2500/300 NR [107] SiC/Ti/TaSi a31 /Pt air 200/600 diff. and oxid. [108] Si/TaC/Cu vacuum /775 diff. and react. of Cu [109] Si/Ta a31 N/Cu vacuum /650 react. of Cu [109] Si/TiN/Cu vacuum 0.5/600 react. form Cu-Si [106] Si/HfN/Cu vacuum 0.25/500 struct. (SEM) [115] Si/HfN/Pd vacuum 0.25/500 struct.(SEM) [115] Si/HfN/Au vacuum 0.25/500 struct.(SEM) [115] Si/a-Wa216 a31 Sia97a132a214 /Cu vacuum 1/700 crystall. at a63a62a50a61a50 a0a54a1 [105] Si/ZrN/Al vacuum 0.5/600 diffusion [116] Si/ZrB a31 /Al vacuum 2/625 diffusion [117] Si/a-Ni a51a133a67 Nb a74a133a67 /Cu vacuum 1/600 NR [105] Si/a-Wa216 a51 N a31a68a74 /Cu vacuum 0.5/750 reaction [110] Si/Ta-N/Cu vacuum 0.5/750 reaction [111] Si/Ta-Si/Cu/Ta a70a72a71a228a39a49a73 a31 /550 crystal. and react. [136] Table 2.7: Binary compound or alloy diffusion barrier layer for semiconducting devices sample structure annealing ambient barrier effect. (hr/a0 C) failure mode Ref. Si/SiO a31 /W-Si-N/Cu Vacuum 1/650 crystall. [112] Si/SiO a31 /Mo-Si-N/Cu Vacuum 0.5/850 crystall. [112] Si/Ta-Si-N/Cu Vacuum 0.5/900, a46a35a50 a34 /350 crystall. [112] Si/a-TiPN a31 /Cu vacuum 0.5/600 struct. and elect. [105] Si/TiSi a31 /a-TiPN a31 /Cu vacuum 0.5/700 struct. and elect. [106] Si/a-Ti-Si-N/Cu vacuum 0.5/900 TiN grain growth [113] GaAs/a-Si-Ti-W/Au vacuum 944/300 electrical [114] Table 2.8: Tenary compound or alloy diffusion barrier layer for semiconducting devices 59 There are ternary conducting oxides which may be more difficult to prepare, for example, Bi a31 Ru a31 Oa216 , SrVOa34 , etc. and oxides with an unstable conducting phase such as a102 a1 a71a86a85 a31 a90 a56 a72a85a233a28a29 a80 a56a58a209 a80a132a233a88a87a15a93a16a35 a1 a71 a31 a85a226a34a62a90a118a233a89a72a85a233a187a39 a56 a72a85a233a114a29 a80 a56a58a209 a80a132a233a88a87a18a93a179a158 a97 a31 a85 a31 . Most of the conducting oxides have not been studied with the idea of using them as diffusion barriers, and there is much potential in this area. For example, conducting oxides such as zinc oxide (ZnO) [148], indium tin oxide (ITO) [149], rhodium oxide (a40a191a243a63a85 a31 ) [150] and so on, could turn out to be good diffusion barriers. 60 CHAPTER 3 ANALYTICAL TECHNIQUES 3.1 Electrical Analysis 3.1.1 Sheet Resistance Measurement The four-point probe technique is the most common technique for thin film sheet resistivity measurements. Parasitic resistances such as the contact resistance (a40 a145 ) and spreading resistance (a40a226a11a91a90 ) makes the result of two-point probe measurements difficult to interpret [151]. These resis- tances are negligible in the four-point probe configurations because the potential differences are measured with zero or very small currents flowing through the potential probes. There are dif- ferent probe configurations that can be adopted for four-point measurements, including collinear four-point probe and van der Pauw probe arrangements. Collinear four-point probe For four probes on a straight line in contact with a semiconductor and with probe separation distance d, the resistivity of an arbitrarily shaped finite sample is given by [151], a82a131a24a21a102a86a197a181a29 a196 a90a132a13a163a27a30a126a32a93 (3.1) F is a correction factor that depends on sample geometry and is a product of several inde- pendent correction factors, a196 a24 a90a92a56a32a31a140a102a61a93 a196 a97 a196 a31 a196 a34 a197 (3.2) 61 Figure 3.1: (a) Collinear four-point probe measurements (b) van der Pauw four-point probe measurements a196 a97 , a196 a31 , a196 a34 are correction factors for sample thickness, lateral sample dimension and probe placement relative to to the sample edges. For very thin samples, a196 a31 a24 a196 a34a72a24a202a46 [151], and a82a131a24 a197 a209 a56a32a31a226a102 a90 a13 a126 a93a203a24a134a137a18a87a88a65a30a64a66a102 a209 a90 a13 a126 a93 (3.3) The sheet resistivity a82a32a11 in a23 /sq is a82 /t or a82a32a11a163a24a134a137a18a87a88a65a30a64a66a102a131a211 a13 a126 a212 (3.4) 62 van der Pauw four-point probe The theoretical basis for sheet resistivity measurements on irregularly shaped sample was developed by van der Pauw using conformal mapping [152, 153, 154]. Flat, arbitrarily shaped sample must meet the following conditions: a44 the contacts are at the circumference of the sample a44 the contacts are sufficiently small a44 the sample is uniformly thick a44 the surface of the sample is singly connected, that is, the sample does not contain any isolated hole. Under these conditions, the sample resistivity is given by a82a75a24 a197 a209 a56a32a31a226a102 a90a160a40 a97 a31a151a242 a34 a74 a158a55a40 a31 a34 a242a74 a97 a93 a102 a196 (3.5) where a40 a97 a31a151a242 a34 a74 a24a21a13a7a34 a74 a27a85a126 a97 a31 , F is a correction factor satisfying the relation a40a72a17a179a39a136a46 a40a72a17a109a158a184a46 a24 a196 a56a32a31a226a102 a56 a72a61a84a86a243 a4 a97 a211 a221a54a222a79a223 a90a92a56a32a31a226a102a62a27 a196 a93 a102 a212 a87 (3.6) For a symmetric and uniform sample, a40a72a17a116a24 a40 a97 a31a151a242 a34 a74 a27a30a40 a31 a34 a242a74 a97 a24a232a46 and a196 a24a232a46 . A plot of the corection factor F against a40a72a17 is shown in figure 3.2. The sheet resistance for a source current of 4.532mA can be written as a82 a11 a24a41a137a18a87a88a65a62a64a61a102 a90a91a40 a97 a31a111a242 a34 a74 a158a55a40 a31 a34 a242a74 a97 a93 a102 a196 a24 a90a160a13a7a34 a74 a158a176a13 a74 a97 a93 a102 a196 (3.7) 63 Figure 3.2: van der pauw correction factor plot [151] where F can be a function of Ra17 , i.e., F = F(Ra17 ). Measurement of sheet resistance a82a79a11 Sourcing current a126 a97 a31 a24a99a137a15a87a100a65a62a64a66a102 a57 a70 between conacts 1 and 2, the potential difference between contacts 3 and 4 (a13a7a34 a74 ) was measured using electrometer. The potential difference measured in millivolts translates directly to resistance in ohms, and the sheet resitance is obtained using equation 3.7. 3.1.2 Contact Resistance Contact resistance is characterized by two parameters: (a) the contact resistance (a40 a145 ) in ohms and (b) the specific contact resistance or specific contact resistivity ra145 in ohm-a56a58a57 a31 . 64 The specific contact resistance includes in its definition not only the metal-semiconductor in- terface but also the regions immediately above and below the interface [151]. For the metal- semiconductor interface only, the specific interfacial resistance a82 a108 is defined as, a82 a108 a24a9a236 a13 a236 a235a186a162a93a95a94 a67 a87 (3.8) Practically, the parameter of interest when contact resistance is measured is the specific contact resistance (a71 a145 ). It is independent of contact area, and it is a convenient parameter when comparing contacts of different sizes. The current density J in a metal-semiconductor contact is generally a function of the applied voltage and the barrier height. The current density under conditions of thermionic emission is [155] a235 a24a184a70 a42a30a42 a138 a31 a221a2a222a32a223 a90 a39a72a175a62a5 a143 a252 a138 a93a161a211 a221a2a222a79a223 a90 a175a61a13 a252 a138 a93a89a39a136a46 a212 (3.9) This expression is used to obtain the following expression for the specific interfacial resistance, a82 a108 a90a118a138a72a103a124a93a129a24a9a236 a13 a236 a235a227a162 a93a95a94 a67 a24a96a82 a97 a221a2a222a32a223 a90 a175a30a5 a143 a252 a138 a93 (3.10) where a82 a97 a24 a21a97 a251 a39a98a39 a119 (TE = Thermionic Emission) For the thermionic-field emission (TFE), the interfacial resistance is [156] a82 a108 a90a118a138 a196 a103a124a93a129a24 a1 a97 a82 a97 a221a2a222a79a223 a90 a175a62a5 a143 a103 a139 a93a58a238 (3.11) 65 and for field emission (FE),the expression for a82 a108 is a82 a108 a90 a196 a103a124a93a203a24 a1 a31 a82 a97 a221a54a222a79a223 a90 a175a62a5 a143 a103a105a139a68a139 a93a151a87 (3.12) a1 a97 and a1 a31 are functions of the donor concentration (a73a135a128 ), the temperature (T) and the barrier height (a5 a143 ). a103 a139a68a139 is a characteristic energy for tunneling process [157] a103 a139a68a139 a24 a175a61a243 a137a62a197 a205 a73 a128 a252 a11 a144 a139 a57 a42 (3.13) a103 a139 is related to a103 a139a68a139 by [158] a103a105a139a72a24 a196 a139a142a139a89a99a61a100a69a101a30a33a48a90 a103a105a139a68a139 a252 a138 a93 (3.14) These are expressions for the specific interfacial resistance. However, it is not possible to obtain accurate theoretical expressions for the specific contact resistance. Experimental measurements yield the specific contact resistance which is contained implicitly in the specific interfacial resis- tance. That is, it is difficult to compare a71 a145 with theory and a82 a108 with experiment. 3.1.3 Linear Transmission Line Model Measurement The transmission line model (TLM) diagramed in figure 3.3 is a simple method of measur- ing the total resistance (a40a72a119 ) between two adjacent contact pads as a function of the inter-pad spacing (L). The ease with which the specific contact resistance and the semiconductor sheet re- sistance are estimated from the plot of the total resistance against inter-pad spacing is one of the advantages of using the TLM method compared to other methods. The total resistance between any two adjacent contacts is [159] a40a72a119a246a24a21a102a62a40 a145 a158a55a40a228a11 (3.15) 66 Figure 3.3: Plot of total contact to contact resistance as a function of inter-pad spacing [159] where a40 a145 is the contact resistance, and a40a226a11 is the resistance contribution from the semiconductor bwtween the pad edges. a40a161a119a155a24a21a102a62a40 a145 a158 a40 a11a43a42a69a102 a231 (3.16) The contact resistance a40 a145 between metal and semiconductor can be described with a resistive network shown in figure 3.4 where [160] a40 a97 a24 a17a104a103 a105 a25 a150 and a40 a31 a24a184a40 a11 a21 a25 a150 a105 . a71 a145 is the specific contact resistance ( a23a49a39 a56a54a57 a31 ), a40 a11 a21 is the modified sheet resistance of semicon- ductor (a23a72a27a62a84a35a175 ) under the contact, and W is the width of the contact. Using Kirchoff?s law, the relation between voltages and currents at a117 and a117a131a158 a210 a117 can be written as, a13a122a90a118a117a75a158 a210 a117a12a93a198a39a115a13a122a90a118a117a12a93a203a24a184a126a53a90a118a117a12a93a133a40 a31 a24a134a126a53a90a118a117a12a93 a83a11a106a92a107 a105 a210 a117 . In the limit a210 a117a108a35a170a50 , a29a79a13 a29a66a117 a24a134a126a53a90a118a117a12a93 a40 a11 a21 a231 (3.17) 67 Figure 3.4: TLM resistive network [160]. The contact width W is into the page. Also, a126a53a90a118a117a75a158 a210 a117a12a93a89a39a183a126a83a90a123a117a114a93a203a24 a93 a207a150a35a208 a83 a62 a24a121a13a122a90a123a117a114a93 a105 a17 a103 a210 a117 , and in limit limit a210 a117a109a35a170a50 , a29a204a126a83a90a123a117a114a93 a29a66a117 a24a21a13a122a90a118a117a12a93 a231 a71 a145 a87 (3.18) Combining the voltage and current equations yeilds a second order differential equation a29 a31 a126a53a90a118a117a12a93 a29a66a117 a31 a24 a231 a71 a145 a29a79a13 a29a66a117 a24a134a126a53a90a118a117a12a93 a40 a11 a21 a231 a231 a71 a145 a24 a40 a11 a21 a71 a145 a126a53a90a118a117a12a93 (3.19) a127 a59a76a110 a207a150a35a208 a127 a150 a59 a24 a110 a207a150a153a208 a111a71a112 a59 where a102 a119a246a24a14a113 a17 a103 a83a11a106a92a107 a102 a119 is referred to as the ?transfer length? or penetration length, and it is the characteristic length 68 over which the current flows under the contact. The general solution to the second order differ- ential equation is a126a83a90a123a117a114a93a203a24a184a126a54a139 a27a114a29a32a31a24a33 a36 a127 a4 a150 a111a71a112 a41 a27a114a29a32a31a24a33 a36 a127 a111 a112 a41 (3.20) and a13a122a90a123a117a114a93a203a24a184a126a54a139 a102 a119a198a40 a11 a21 a231 a99a45a100a86a27a114a33 a36 a127 a4 a150 a111a71a112 a41 a27a30a29a32a31a34a33 a36 a127 a111a71a112 a41 a87 (3.21) The contact resistance is a40 a145 a24 a13a116a90a160a50a61a93 a126a83a90a91a50a66a93 a24 a102 a119a181a40 a11 a21 a231 a99a45a100a69a101a30a33a159a211 a29 a102 a119a105a212 (3.22) For a29a75a194 a102 a102 a119 , a99a61a100a69a101a115a33 a36 a127a111a71a112 a41 a177 a46 and a40 a145 a24 a111a86a112 a83a11a106a92a107 a105 . Hence a40a161a119 becomes a40a72a119a246a24 a102 a102 a119a198a40 a11 a21 a231 a158 a40 a11a43a42a69a102 a231 (3.23) If the semiconductor sheet resistance under the contact is not significantly modified, we have a102 a150a191a177 a102 a102 a119 . For a29a75a194 a102 a119 a193 a40 a11 a21 a177 a40 a11a43a42 , and a40a72a119a155a24 a102 a224 a71 a145 a40 a11a43a42 a231 a158 a40 a11a43a42a69a102 a231 (3.24) This equation implies a linear relationship between a40a161a119 and L. a71 a145 and a40 a11a43a42 are calculated from the slope and intercept of the linear curve. If a102 a150a117a116a24a22a102 a102 a119 , that is a40 a11 a21 a116a24a9a40 a11a43a42 , the correct value of a71 a145 can be found by additional measurement of the contact end resistance (a40 a206 ) [159]. The standard technique for measuring a40 a206 is to pass a constant current between two adjacent contact pads and then measure the poten- tial difference between one of the current pads and a third adjacent pad. It can be shown that a40 a206 69 is expressed as [161] a40 a206 a24 a13 a126 a24 a224 a40 a11 a21 a71 a145 a231 a46 a27a114a29a32a31a24a33 a36 a127 a111a71a112 a41 a87 (3.25) However, a40 a11 a21 a24 a17a114a103a111a71a112 a59 , so that a40 a206 a24 a71 a145 a102 a119 a231 a46 a27a114a29a32a31a24a33 a36 a127 a111a86a112 a41 a40 a145 a40 a206 a24a28a99a45a100a118a27a30a33a244a211 a29 a102 a119 a212 (3.26) 3.1.4 I-V Charateristics of Schottky Diodes The Schottky diode is a ?majority carrier device?, and minority charge carriers play in- significant role in the determination of the I-V characteristic. For an n-type Schottky barrier, electron injection from the semiconductor into the metal dominates the observed current be- cause of the relatively low potential barrier. However, recombination and hole injection currents still exist [44]. Injection of electrons or holes over a potential barrier is reffered to as thermionic emission current. High mobility semiconductors have I-V characteristics given by thermionic emission theory, provided the forward bias is not too large [50]. The thermionic emission current can be derived as follows. An electron in the semiconductor is capable of jumping over the potential barrier if it has a velocity a237 a150 and kinetic energy, a249a159a87a225a103a76a24 a46 a102 a57 a42 a141 a237 a31 a150 a59a25a175a18a90a160a13a7a14 a108 a39a115a13a15a251a174a93 a162 a237 a150 a162a204a59 a237 a6 a108 a141 a254 a245 a102a85a175 a57 a42 a141 a90a132a13a15a14 a108 a39a115a13a18a251a203a93 a247 a97a142a213 a31 (3.27) where Va251 is the applied potential. If there are n electrons per unit volume in the semiconductor with velocity va150 then the current will be, a29a204a126a35a11a98a119a140a6a21a24a76a39a72a175a30a70 a237 a150a85a233a109a90 a237 a150a66a93 . Integrating over the velocity 70 differential, a126a35a11a98a119a226a6a184a24a76a39a72a175a30a70 a200 a4 a178 a19a11a120a122a121 a4a63a123 a237 a150a30a233a129a90 a237 a150a66a93a79a29 a237 a150 a233a129a90 a237 a150a61a93a129a24 a211 a137a62a197 a252 a138 a57 a42 a141 a31 a243 a34 a212 a221a2a222a32a223 a90a91a103a72a16a246a39a49a103 a145 a93 a252 a138 a221a2a222a32a223 a39 a211 a57 a42 a141 a102 a252 a138 a212 a237 a150 a31 a126a35a11a98a119a226a6a184a24a134a70 a1 a42 a138 a31 a221a2a222a32a223 a90a142a39 a5 a143 a252 a138 a93 a221a54a222a79a223 a90 a175a61a13a15a251 a252 a138 a93 (3.28) where a1 a42 a254 a211 a57 a42 a141 a57 a139a53a212 a137a62a197a181a175 a57 a139 a252 a31 a243 a34 a24a202a46a153a102a62a50a30a70 a56a58a57 a4a53a31 a249 a4a53a31 a211 a57 a42 a141 a57 a139a53a212 (3.29) Richardson constant a24a202a46a153a102a30a50a62a70 a56a58a57 a4a53a31a54a249a52a4a53a31 Electrons going from the metal into semiconductor see a constant potential a5 a143 , where a5 a143 is defined in figure 3.5. a126a58a6a124a119a191a11a153a90a132a13a18a251a174a93a129a24a184a126a58a6a124a119a215a11a153a90a132a13a18a251a185a24a184a50a61a93a129a24a76a39a72a126a35a11a98a119a226a6a75a90a160a13a15a251a155a24a21a50a61a93a129a24a76a39a72a70 a1 a42 a138 a31 a221a2a222a32a223 a90a142a39 a5 a143 a252 a138 a93a129a24a134a126a35a11 (3.30) Total current is a126a124a24a184a126a35a11a98a119a140a6a176a158a187a126a58a6a124a119a215a11a149a24a134a126a35a11a98a119a226a6a55a158a187a126a58a6a124a119a191a11a153a90a160a13 a251 a24a184a50a66a93a129a24a134a126a35a11 a211 a221a54a222a79a223 a90 a175a61a13a18a251 a252 a138 a93a89a39a136a46 a212 a87 (3.31) For a13 a251 a8 a few kT/q, a126a75a24a134a126a35a11 a221a54a222a79a223 a97 a93a20a125a21 a119 and for a13a18a251a49a8 a few kT/q in the reverse bias mode, a126a75a24a218a39a72a126a35a11 The expressions above are for ideal diodes. However, real diodes exhibit phenomena such as breakdown voltage and reverse current that are not constant because the potential barrier a5 a143 is not constant. The barrier height a5 a143 associated with image force barrier lowering is expressed 71 Figure 3.5: Effect of barrier lowering on barrier height as a5 a143 a24a184a5 a143a198a180 a39 a210 a5 a143 (3.32) a175a62a5 a143 a24a134a175a61a13a15a139a129a158a187a175a62a13a7a14 a108 a24a134a175a18a90a132a13a18a139a109a158a55a13a7a14 a108 a93 (3.33) The ideal barrier height becomes a good approximation if the diode is operated at a strong for- ward bias. I-V characteristics of a real diode depend on the barrier height that can be a function of the applied bias, under this condition an ideality factor (n) can be introduced into the thermionic current expression by including a linear, voltage dependent term in the modified barrier height expression. The thermionic current becomes [151], a126a75a24a184a126 a11a72a211 a221a2a222a79a223 a90 a175a61a13 a251 a233 a252 a138 a93a198a39a41a46 a212 a24a134a126 a11 a221a2a222a32a223 a90 a175a61a13 a251 a233 a252 a138 a93 a211 a46a179a39 a221a2a222a32a223 a90a142a39 a175a62a13 a251 a233 a252 a138 a93 a212 a87 (3.34) 72 a126a35a11a163a24a134a70 a1 a42 a138a140a31 a221a2a222a32a223 a90a142a39 a97a104a126a18a127 a21 a119 a93 , and the current density a235 a24a134a126a79a27a58a70 so that a235 a24 a1 a42 a138 a31 a221a2a222a79a223 a90a68a39 a175 a252 a138 a90a91a5 a143 a39a115a13a15a251a174a27a85a233a181a93a166a93 a211 a46a179a39 a221a54a222a79a223 a39 a175a61a13a18a251 a252 a138 a212 (3.35) For a13a18a251a185a194 a97a21 a119 , a235 a24 a1 a42 a138 a31 a221a54a222a79a223 a90a142a39 a175 a252 a138 a90a160a5 a143 a39a115a13a18a251a203a27a86a233a198a93a133a93a151a87 (3.36) Taking the natural log of this equation yields a56a32a31 a235 a24 a1a81a128 a39 a175 a252 a138 a5 a143 a158 a175 a233 a252 a138 a13a15a251 (3.37) where a1 a128 a24a129a56a32a31 a1 a42 a138 a31 . A plot of a56a32a31 a235 against a13a15a251 should be linear for a13a15a251 a8 a few kT/q. The ideality factor n is determined from the slope, while the barrier height a5 a143 is determined from the intercept of the linear plot. a233a159a24 a175 a252 a138a49a90a132a84a20a130a131a72a151a78a83a82a85a93 a5 a143 a24a133a132a134a56a32a31a198a90 a1 a42 a138 a31 a93a198a39a58a56a32a31 a235 a108a57a135 a252 a138 a175 (3.38) However, removing the condition a13a15a251a115a194a98a175a61a27 a252 a138 , a linear graph could be obtained for all a13a15a251 if what is ploted against a13a15a251 is a56a32a31a137a136 a138a139 a97 a4a24a140a64a141a55a142 a207 a4a144a143a60a145 a125 a107 a112 a208a147a146a24a148 . The Richadson constant a1 a42 is one major uncertainty in the J-V expression. It is modified to take into account the effective mass of electrons in the semiconductor, quantum-mechanical reflection of those electrons which are able to negotiate the barrier, and phonon scattering of the electrons between the top of the barrier (as determined by image forces) and the surface of the metal [50]. It can also depend on experimental factors such as semiconductor surface cleaning [162], annealing conditions, contact metal thickness and method of metal deposition [163]. 73 Series resistance (a71a86a11 ) is also present during diode operation. It depends on the semiconduc- tor resistivity, contact resistance and sometimes on geometrical factors [151]. Series resistance can be included in the thermionic current expression by replacing (a13a15a251 ) by (a13a25a39a159a126a61a71a85a11 ), where (a13 ) is the measured voltage across the entire diode including substrate and contact resistance. The current expression becomes a126a124a24a134a126a35a11a140a211 a221a2a222a32a223 a90 a175a18a90a132a13a184a39a49a126a61a71a86a11a151a93 a233 a252 a138 a93a198a39a136a46 a212 a87 (3.39) The series resistance is obtained by plotting a126 a127 a93 a127 a110 or a127 a93 a127a2a207 a46a48a47 a110 a208 against I, the slope is the series resistance, and the intercept is a141 a21 a119a97 from which the ideality factor n can be obtained. 3.1.5 C-V Charateristics of Schottky Diodes Schottky diodes exhibit capacitance associated with their depletion regions. In this regard, they are similar to p-n junction diodes. The capacitance of an n-type Schottky barrier is identical to that of an abrupt a78 a3 a39a246a233 junction, with thea78 a3 side behaving as the metal side of the Schottky diode. One major difference is that there is no minority carrier storage in a Schottky diode. There is also no diffusion capacitance. The capacitance of a Schottky barrier resembles a parallel-plate capacitor with the separa- tion between the plates controlled by the applied reverse bias potential. The differential capaci- tance (a1 a24 a127 a37 a127 a93 ) is usually measured by superimposing a small alternating voltage on a reverse dc bias [50]. The differential capacitance depends on the applied reverse bias. Under favorable circumstances, measurements of the capacitance as a function of time and reverse bias voltage can yield not only the concentration of interfacial traps, but also their time constants and energies 74 Figure 3.6: Capacitance measurements under applied bias [151] relative to the band edges. Applied external stimuli such as light, temperature or forward bias can change the occupation of these traps. Figure 3.6 shows an n-type Schottky diode with doping concentration a73a135a128 and an applied dc bias a13a15a251 . The differential capacitance is a1 a24 a29 a13 a11 a29a79a13a15a251 (3.40) where a29 a13 a11 is the charge increment in the semiconductor and a29a79a13a15a251 is the applied voltage incre- ment. The superimposed small amplitude ac voltage is typically applied at a frequency of 1MHz with an amplitude of 10 to 20mV. The ac voltage changing from zero to small negative value adds a small charge increment a39a140a29 a13 a6 to the metal contact. The semiconductor reacts to the charge increment in the metal with a corresponding charge increment of a29 a13 a11 leading to a slight 75 increase in space charge region (scr) width a29 a231 to maintain a overall charge neutrality [151]. a29 a13 a11a163a24a134a175a30a70a226a73a135a128a215a90 a231 a93a79a29 a231 (3.41) a1 a24 a29 a13 a11 a29a79a13a15a251 a24a184a175a85a70a228a73a135a128a228a90 a231 a93 a29 a231 a29a79a13a18a251 (3.42) For parallel plate capacitor a1 a24a150a149 a11 a144 a139 a70 a231 a193 a29 a1 a29a79a13a15a251 a24 a39 a149 a11 a144 a139 a70 a231 a31 a29 a231 a29a79a13a15a251 a29 a231 a29a79a13 a251 a24 a39 a231 a31 a149 a11a111a144a166a139a58a70 a29 a1 a29a79a13 a251 (3.43) Using equations 3.42 and 3.43, a73a48a128a215a90 a231 a93a203a24 a1 a175a85a70a75a90a91a29 a231 a27a85a29a79a13a15a251a89a93 a24 a39 a1 a34 a175a30a70 a31 a149 a11a111a144a166a139a30a90a91a29 a1 a27a85a29a79a13a15a251a174a93 a24 a102 a175a30a70 a31 a149 a11a111a144a133a139 a15 a127a35a207 a97a142a213 a241 a59 a208 a127 a93 a125 a22 (3.44) The doping concentration is obtained from a C-V measurement using the slope of the plot of a97 a241 a59 against a13a18a251 and the dopant depth obtained from a231 a24a152a151 a106a98a153a92a154 a251 a241 . The width of the space charge region (W) extends only into the semiconductor and is negligible in the metal. The a97 a241 a59 a39a136a13a15a251 plot shows immediately the uniformity of the doping concentration, with a uniformly doped semiconductor characterized by a straight line. The capacitance per unit area of a schottky diode is given by [164], a1 a70 a24 a175 a149 a11a111a144a133a139a66a162a73a135a128a115a39a49a73a48a251a105a162 a102a18a90a160a13a7a14 a108 a158a121a162a229a13a201a162a153a39 a252 a138a161a27a85a175a61a93 (3.45) For an n-type substrate a73a135a128a41a8a136a73a48a251 and a13a157a19a136a50 , a252 a138a161a27a30a175 accounts for the majority carrier tail in the scr. The barrier height is related to the built-in potential a13a7a14 a108 by, 76 a5 a143 a24a21a13a7a14 a108 a158a176a13a15a139 and a13a15a139a161a24 a21 a119 a97 a56a32a31 a36 a54a40a155 a54a40a156 a41 The effective density of states is a73 a241 a24a21a102 a15 a31a76a157 a6 a39 a121 a21 a119 a42 a59 a22 a34 a213 a31 . For a73 a251 a24a184a50 , the capacitance per unit area expression can be linearized as, a211 a70 a1 a212 a31 a24 a102a79a90a132a13a7a14 a108 a39 a252 a138a105a27a30a175a61a93 a175 a149 a11a111a144a133a139a35a73 a128 a158 a102a7a162a229a13 a162 a175 a149 a11a166a144a166a139a54a73 a128 (3.46) Again the plot of a90a91a70a228a27 a1 a93a68a31 against a162a229a13 a162 yeilds a slope and an intercept. a84a20a130a57a72a151a78a83a82a226a24 a31a97 a151 a106a98a153a92a154 a54a95a156 a193 a73a135a128a55a24 a31a97 a151 a106a98a153a92a154 a207 a11a159a158a225a139a64a90 a167 a208 The intercept gives a13 a108 a24a21a13a7a14 a108 a39 a21 a119 a97 However a13a15a14 a108 a24a21a5 a143 a39a115a13a18a139 so that a13 a108 a24a184a5 a143 a39a49a13a15a139a149a39 a252 a138a161a27a85a175 . a5 a143 a24a21a13 a108 a158a176a13 a139 a158 a252 a138a161a27a30a175a215a24a121a13 a108 a158 a252 a138 a175 a245 a46a129a158a160a56a32a31 a73a48a139 a73a48a128 a247 a87 (3.47) 3.2 Physical Analysis 3.2.1 Rutherford Backscattering Spectrometery (RBS) Introduction The model of an atom as a positive nucleus enclosed by cloud of negative electrons was put forward by Ernest Rutherford and confirmed experimetally by Geiger and Marsden (1913). They reported sinlge collision, large angle, scattering of alpha particles by positively charged nuclei. Their experiment not only established Rutherford?s model but also formed the basis for Rutherford Backscattering Spectrometry (RBS) as a modern analytical technique [166]. Rutherford backscattering is simply understood because it is a classical scattering process in a central force field. High energy helium ions undergo close-impact collisions which are 77 Figure 3.7: Close impact collision and backscattering [166] governed by coulomb repulsion between the positively charged incident projectile and the nuclei of the target atoms. Definitions kinematic factor in an elastic collision A very small fraction of incident beam is backscattered on collision with atomic nuclei. Backscattered particles of energy (a103 a97 ) get to the detector. The reduction in the energy of the particles (a103 a97 a19a171a103a105a139 ) is a function of the mass of the incident particle and the target nucleus. Assuming an elastic collision, the principles of conservation of kinetic energy and momentum can be applied to obtain a solution for the kinematics of the collision. The collision is depicted in figure 3.7. 78 Kinetic energy conservation: a46 a102 a57a155a97a68a237 a67 a31 a24 a46 a102 a57a155a97a133a237a32a97 a31 a158 a46 a102 a57 a31 a237 a31 a31 (3.48) Momentum conservation: a57 a97 a237 a67 a24 a57 a97 a237 a97 a99a61a100a86a27a38a161a161a158 a57 a31 a237 a31 a99a45a100a86a27a7a5 a90a118a117a116a39a52a29a66a80a132a71a62a82 a56a54a209 a80a98a72a85a233a181a93 a50a48a24 a57a155a97a68a237a32a97 a27a114a29a32a31a9a161a228a39 a57 a31 a237 a31 a27a114a29a32a31a140a5 a90a57a162a124a39a52a29a66a80a132a71a62a82 a56a54a209 a80a98a72a85a233a181a93 (3.49) Solving the kinetic energy and momentum conservation equation yield, a237a204a97 a237 a67 a24a8a163 a90 a57 a31 a31 a39 a57a155a97 a31 a27a30a29a32a31 a31 a161a61a93 a97a43a213 a31 a158 a57a155a97 a99a61a100a86a27a38a161 a57a155a97 a158 a57 a31 a193 a103 a97 a103 a67 a24a164a136 a90 a57 a31 a31 a39 a57a155a97 a31 a27a30a29a32a31 a31 a161a61a93 a97a43a213 a31 a158 a57a155a97 a99a61a100a86a27a38a161 a57a155a97 a158 a57 a31 a148 a31 (3.50) The ratio of the backscattered energy to incident energy is called the kinematic factor (a252 a24 a103 a97 a27a30a103 a67 ). The eqation aboved shows that k is determined only by the masses of incident particle and target atom and by the scattering angle, a161 . For a fixed incident particle mass and scattering angle, k becomes a function of the target mass a57 a31 only. Therefore, the energy of backscattered particle is a direct signature of the target atom. In practice, when a target contains two types of atoms that differ in massses by a small amount a210 a57 , it is important that this difference produce as large a change in a210 a103 a97 as possible. This gives the largest a210 a252 when a161a135a24a157a46a35a63a61a50 a0 (ideal location for a detector). a161a135a24a157a46a153a101a62a50 a0 is a practical location because of detector size. It is this arrangement that has given the method its name, Backscattering Spectrometry [167]. Quantitatively, a210 a103 a97 a24a21a103a105a139a215a211 a29 a252 a29 a57 a212 a210 a57 (3.51) 79 Every practical detection system has a finite resolution. If a210 a103 a97 falls below this limit, the distinction between two masses is lost. To obtain good mass resolution, it is desirable to [167]: (1) increase the incident particle energy a103a161a139 (2) use a large projectile mass a57a155a97 , but with a57a155a97 a19 a57 a31 (3) measure at scattering angle as close as possible to a46a35a63a62a50 a0 Scattering cross section a107 After an elastic collision, the target atom is identified by the energy lost by the scattered par- ticle. The collision probability between the incident particle and the target nucleus is determined by the number of target nuclei per unit area. The scattering cross section relates the number of target nuclei to the number of particles detected after scattering. For particles scattered through an angle a161 into a differential solid angle, a210 a23 , the differential scattering cross section is defined as a29a204a107 a29a32a23 a24 a46 a73 a209 a245 a29 a13 a27a62a23 a13 a247 (3.52) a13 is the total number of incident particle on the target. a29 a13 is the number of particles getting to the detector. a73 is the volume density of atoms in the target, and t is the sample thickness. Hence, a73 a209 is the number of target atoms per unit area (areal density). a29a204a107a198a27a85a29a32a23 has the dimension of area, and it can be interpreted as the effective area each nucleus presents to the incident beam. In backscattering spectrometry, the detector solid angle a23 is small (a45 a46a35a50 a4a53a31 sterradian or less). The average differential scattering cross section is defined as a107a129a90a159a161a61a93a203a24 a46 a23 a200a166a165 a29a204a107 a29a32a23 a29a118a167 (3.53) 80 Figure 3.8: a23a41a24a121a26a28a27a30a29 a31 = detector solid angle. S = detector area. d = detector-target distance. t = target thickness. The number of particles detected by the detector is called the yield and is given by, a168 a24a134a107a109a90a64a161a61a93a133a23 a13 a73 a209 (3.54) a13 is determined by time integration of the current of charge particles incident on the target. In the backscatteing spectrometry, it is assumed that the incident particle is scattered by the unscreened nucleus of the target atom because the distance of closest approach is well within the inner electron orbits. The unscreened scattering cross section originally derived by Rutherford is a29a204a107 a29a32a23 a24 a211a169a149 a5 a97 a5 a31 a82 a31 a102a62a103 a212 a31 a46 a27a114a29a32a31 a74 a161 a87 (3.55) a5 a97 a82 is the incident particle charge and a5 a31 a82 target particle charge. Rutherfor assumed that the target atom was infinitely heavy while deriving the scattering cross 81 section expression. If the target atoms are considered to have finite mass, then the collision becomes a two-body central force problem which can be analyzed as a one-body collision by replacing a57a155a97 by the reduced mass a77a134a24 a57a155a97a166a57 a31 a27a79a90 a57a155a97 a158 a57 a31 a93 . The modified scattering cross section becomes a29a204a107 a29a32a23 a24a232a211a144a149 a5 a97 a5 a31 a82 a31 a102a62a103 a212 a31 a137 a27a114a29a32a31 a74 a161 a136a32a211a28a46a105a39 a36 a6 a62 a6 a59 a27a114a29a32a31a9a161 a41 a31 a212 a97a43a213 a31 a158a160a99a61a100a86a27a38a161 a148 a31 a211 a46a179a39 a36 a6 a62 a6 a59 a27a114a29a32a31a9a161 a41 a31 a212 a97a43a213 a31 a87 (3.56) For the coulomb potential to govern the backscattering process, distance of closest ap- proach, a29 , must be less than k-shell electron radius, estimated to be a96 a139a35a27a65a5 a31 , a96 a139a186a24a99a50a79a87a100a65a62a64 ?a70 , the Bohr radius. a103a76a24 a5 a97 a5 a31 a82 a31 a29 a96 a233a28a29a109a80a142a146 a29a116a19 a96 a139 a27a65a5 a31 a238 a103a157a8 a5 a97 a5 a31 a31 a82 a31 a96 a139 a87 (3.57) a103 a45 a46a2a50 a252 a82a85a13 for a Si target and a103 a45a248a64a30a137a66a50 a252 a82a85a13 for scattering from Au. Because part of the tra- jectory will always be outside the electron cloud leading to deviation from Rutherford scattering cross section, the screened coulomb cross section is a107 a11 a145 a24a184a107a129a90a159a161a61a93 a196 where a196 a24 a36 a46a105a39 a67a61a170a67a166a74 a106a115a171 a62 a171 a59a76a172a60a173a64a174 a206 a41 and E is in keV. For 1MeV He ions incident on Au, a196 a177 a64a66a92 . For 2MeV He ions, the screening correction can be neglected for most targets [166, 167]. Energy loss (dE/dx) and stopping cross section (a125 ) Composition depth profiles can be obtained from RBS analysis. The depth scale is deter- mined by the energy loss at high energies as the incident particles traversed the sample. Energy is lost by energetic particles mostly through excitation and ionization during inelastic collisions 82 Figure 3.9: Energy loss during inward and outward trajectory of incident particle [166] with atomic electrons. Energy loss due to an interaction with the nucleus, is much less than elec- tronic energy loss. The amount of energy a210 a103 lost per distance a210 a117 by a particular ion depends on the density and composition of target and on the ion energy. For an incident energy a103a105a139 , the ion energy at any depth a117 can be written as a103a122a90a123a117a114a93a203a24a21a103a105a139a149a39 a200 a150 a67 a29a32a103 a29a66a117 a29a66a117 (3.58) The stopping cross section a125 can be defined as a125a215a24 a46 a73 a29a32a103 a29a66a117 a90a160a82a85a13a184a39 a56a54a57 a31 a93 (3.59) where N is the atomic density. In a compound target, the total energy loss is the sum of the losses to each of of the con- stituent elements weighted by the abundance of each element. This postulate is known as Bragg?s 83 rule, and it states that the stopping cross section of a26a89a80a104a85 a31 for example is given by a144 a113 a108 a180 a59 a24a202a90a166a46a153a93a68a144 a113 a108 a158a134a90a132a102a62a93a133a144 a180 (3.60) where a144 a113 a108 and a144 a180 are the stopping cross section of the atomic constituents. a144 a113 a108 a180 a59 is sometimes called the stopping power per molecule, and a127a151a206 a127 a150 a24a184a73a246a144 a113 a108 a180 a59 . N is the number of molecules per volume. Energy width and depth profile In thin films, the total energy loss, a210 a103 , into a depth a117 is proportional to a117 . This is true if a29a32a103a135a27a30a29a66a117 is constant, which is a good assumption if a103a105a139 is large (a8 1MeV) and a117 is the order of a few microns. a210 a103 a108 a141a131a24 a200 a150 a67 a29a32a103 a29a66a117 a29a66a117 a177 a117 a29a32a103 a29a66a117 a162 a108 a141 (3.61) The energy at depth a117 is a103a122a90a123a117a114a93a129a24a184a103a161a139a149a39a183a117 a29a32a103 a29a66a117 a162 a108 a141a53a87 (3.62) After large angle scattering, the ion energy can be written as a103 a97 a90a118a117a12a93a129a24 a252 a103a116a90a118a117a12a93a89a39 a117 a162a76a99a61a100a86a27a24a161a7a162 a29a32a103 a29a66a117 a162a139a142a164 a147 (3.63) where k is the kinematic factor, a161 is the scattering angle and a117a114a27a53a99a45a100a118a27a63a161 is the path length traveled by the ion after scattering. a103 a97 a90a118a117a12a93a129a24 a252 a211a53a103a105a139a149a39a183a117 a29a32a103 a29a66a117 a162 a108 a141 a212 a39 a117 a162a98a99a45a100a118a27a38a161a7a162 a29a32a103 a29a66a117 a162a139a43a164 a147 a103 a97 a90a123a117a114a93a203a24a10a39a161a117 a245 a252 a29a32a103 a29a66a117 a162 a108 a141 a158 a46 a162a76a99a61a100a86a27a38a161a7a162 a29a32a103 a29a66a117 a162a139a142a164 a147 a247 a158 a252 a103 a139 (3.64) 84 The energy width a210 a103 is a210 a103a76a24 a252 a103a161a139a163a39a49a103 a97 a90a118a117a12a93a129a24a41a117 a245 a252 a29a32a103 a29a66a117 a162 a108 a141a191a158 a46 a162a76a99a45a100a118a27a38a161a7a162 a29a32a103 a29a66a117 a162a139a142a164 a147 a247 a24a41a117a114a26 (3.65) a103 a108 a141 and a103a105a139a142a164 a147 are the energies at which a29a32a103a135a27a30a29a66a117 is evaluated. a26 is the backscattering energy loss factor. In evaluating a29a32a103a135a27a30a29a66a117 , two possible approximations can be adopted (i) The surface energy approximation SEA is a good approximation for very thin films (a117a41a19 a50a79a87a100a65a85a77 a57 ). In this case, a29a32a103a135a27a30a29a66a117 is evaluated at a103 a108 a141a94a24a184a103a105a139 and a103a105a139a142a164 a147 a24 a252 a103a105a139 . (ii) The mean energy approximation MEA is used when path length traversed by the ions is a8a136a50a79a87a100a65a85a77 a57 , a29a32a103a135a27a30a29a66a117 is evaluated at at mean energies a103 a108 a141a94a24 a97 a31 a110 a103a105a139a109a158a55a103a116a90a118a117a12a93a43a120 and a103a105a139a142a164 a147 a24 a97 a31 a110 a103 a97 a158 a252 a103a116a90a118a117a12a93a43a120 Assuming also that the measured a210 a103 is divided equally between the inward and outward paths, we have approximately a103a116a90a118a117a12a93a129a24a184a103a161a139a149a39 a97 a31 a210 a103 . Then a103 a108 a141a94a24 a110 a103a105a139a149a39 a97 a74 a210 a103a228a120 and a103a161a139a142a164 a147 a24 a110 a103 a97 a158 a97 a74 a210 a103a228a120 . The energy a103 a97 of the detected particle can be related to the depth a117 at which backscattering occurs. For elemental samples with incident particles striking the sample surface perpendicu- larly, along the inward and outward paths, respectively, a117a227a24 a200 a206 a206 a154 a29a32a103 a90a91a29a32a103a135a27a30a29a66a117a12a93 (3.66) a117 a99a45a100a86a27a38a161 a24a10a39 a200 a206 a62 a21 a206 a29a32a103 a90a91a29a32a103a135a27a30a29a66a117a12a93 (3.67) Since a103a161a139 and a103 a97 are the energies in these expressions that can be measured experimantally, it is necessary to obtain a117 in terms of a103a161a139 and a103 a97 . There are three ways of doing this 85 (1) Assume a29a32a103a135a27a30a29a66a117 is constant over each path, then the equations can be integrated and E el- liminated. This assumption leads to a linear relationship between a210 a103 and a117 . Constant a29a32a103a135a27a30a29a66a117 is an approximation, and the resulting depth scale is subsequently approximate. (2) Use tabulated values of a29a32a103a135a27a30a29a66a117 and carry out the integration numerically to find corre- sponding sets of a103 and a117 , and therefore a252 a103 and a103 a97 . The numerical calcultaion can be done by dividing the sample into many slabs of equal width a210 a117 small enough that a29a32a103a124a27a85a29a66a117 can be as- sumed constant. Alternatively, the sample can be divided into slabs of differing thicknesses such that particles scattered from two boundaries of all slabs have fixed energy difference at detector. (3) Assume some functional dependence for a29a32a103a135a27a30a29a66a117 . Matching pairs of a103 and a117 and of a117 and a103 a97 can then be obtained analytically. The linear accelerator facility at Auburn The accelerator at Auburn University (figure 3.10) is a 6SDH-2 pelletron tandem machine purchased from the National Electrostatics Corporation. It has two ion sources, allowing for a wide range of applications. Helium ion beams are produced from an rf exchange ion source, and a wide range of heavy ions are available from the SNICS II (Source of Negative Ions by Cesium Sputtering). For example, nitrogen, aluminum, silicon, phosphorous and gold can be accelerated with energies that range from 100keV to 8MeV. The accelerator is currently used for Rutherford backscatering Spectrometry (RBS), light ion channeling (LIC), nuclear reaction analysis (NRA) and heavy ion implantation (HII). [168]. Ion sources(figure 3.11) Helium RF Source - Helium gas is bled into a quartz bottle, and the gas is ionized by an 100MHz RF oscillator. The resulting a219a159a82a86a3 is separated by applying about 6kV potential to push the ions 86 Figure 3.10: Pelletron tandem accelerator at Auburn University 87 Figure 3.11: (a) Sources section of the accelerator. Illustrations of (b) SNICS source (c) He ion source [168] through an exit aperture. The a219a246a82a86a3 ions then enter the charge exchange chamber where they are neutralized by Rubidium (Rb) vapor. Some neutralized ions undergo further charge exchange to become a219a246a82 a4 . The output of RF source is about a46a2a50a61a50a135a39a25a102a62a50a62a50a30a77a28a70 of neutral a219a159a82 and a46a72a39a55a137a62a77a28a70 of a219a159a82 a4 . Rubidium is used in the charge exchange process because of its high cross section for a219a246a82 a4 production [169, 170]. To maintain a good charge exchange, the Rb is heated in an oven kept at a102a62a65a62a50 a0a2a1 and circulated in the charge exchange chamber that is warmed to a65a61a65 a0a35a1 [21]. 88 SNICS Source - Cesium ions (a1 a84 a3 ) used for sputtering are produced by immersing a tantalum ionizer in a Cs vapor. The ionized Cs is accelerated towards a cold cathode biased at -15kV. The cathode is made of material whose ions are to be accelerated. a1 a84a153a3 ions sputter the cathode ma- terial, and cathode ions which become negatively charged as they moves through the Cs vapor, repeled by the cathode dias. The negative ion current is a function of many parameters including the cathode composition, cathode potential, a1 a84 a3 flux, etc. It is possible to generate currents of up to several hundred microamps for most heavy elements [21]. Charging system Pelletron charging chain was developed in the mid 1960s as an improvement over the older Van de Graff charging belt [171]. The chains are made of metal pellets connected by insulating nylon links. The chain is more durable and produces greater terminal stability compared to the belt. The chain eliminates dust from the belt and does not limit ultimate terminal potential. Pelletron charging system also offers significant advantages over solid state charging sys- tems that require fragile electronics in the high voltage column. A solid state system also takes long time to condition to voltage in order to avoid terminal sparking. Pelletron chains are charged by a induction scheme instead of corona discharge [171]. To generate a positive terminal voltage, the inductor is negatively charged and pushes electrons off the pellet while they are in contact with the grounded driven pulley. The chain subsequently transports positive charge to the high voltage terminal where the reverse charging process occurs. The chain passes through a suppressor (negatively biased for positively charge chain) to prevent arcing as the pellets make contact with the terminal pulley. Figure 3.12 shows a pelletron charg- ing system. This system can provide a two-way charging (up-charging and down-charging), the difference between ?up? and ?down? is in the reversal of the polarity of the inductor/suppressor. 89 Figure 3.12: Pelletron charging system [171] The two-way charging system effectively doubles the charging current capacity of the chain. Depending on the design, this system can deliver charging currents of a46a2a50a61a50a149a39a155a102a30a50a61a50a85a77a28a70 or more per chain to the high voltage terminal. Acceleration and focus Ions from the ion source are accelerated to the high voltage terminal. Inside the terminal, the ions collide with a70a72a71 or a73 a31 molecules from a low pressure source. These collision strip electrons to create positive ions that experience a second acceleration due to repulsion from the high voltage terminal. This double acceleration mode of operation where the two ends of the accelerating column are grounded, is called a ?tandem? configuration. Positive ions of different charge state leave the terminal region, and it is imporant to be able to select ions of single charge state and species. An analyzing magnet is used for this purpose. Passing ions through an arc of radius R, with the acceleration terminal voltage V, in a magnetic field B, particular ions with a 90 charge-to-mass ratio (q/m) will be able to successfully traverse the arc. a40a21a24 a57a227a237 a175a69a73 a24 a46 a73 a211 a102 a57 a175 a13 a212 a97a142a213 a31 (3.68) The RBS beam leg attached to the analyzing magnet (figure 3.10) is a46a35a102 a0 left off the direction of the ion beam as it enters the magnet [21]. Ions of specific species, charge state and energy can be selected for the beam leg where the ions are passed through two small apertures for collimation. A faraday cup (biased Ta cone) is used to measure the beam current. The cup is moved in or out of the beam path by a pneumatic cylinder. When the cup is out of the beam path, the beam can get to the samples mounted on a goinometer with computer control of three degrees of freedom - up/down, rotation about a vertical axis and rotation about a horizontal axis. Detector RBS and channeling experiments use a solid state surface barrier detector (sbd) mounted at a46a153a101a30a50 a0 wih respect to the incident beam direction. The detector is a silicon diode with a very thin p-type surface layer [172]. The detector is reverse biased and incident charge particles incident create electrons and holes that are swept in opposite directions to collection electrodes by the electric field in the depletion region. The amount of collected charges is proportional to the incident particle energy. Charge pulses are subsequently converted electronically to voltage pulses. The beam current integration (BCI) systems counts the number of charged particles incident on the target. About a60a61a60a79a87a225a60a61a60a61a92 of the particles incident on target come to rest in he target, and these induce current in the sample [21]. Detector calibration is performed using a31a68a74 a97 a70 a57 which gives an alpha particle of energy 5.486MeV. This energy is used to calibrate a precision pulse generator 91 that is in turn used to simulate detector signals at the input of the detector pre-amplifier. Typically a pulse height spectrum with known energies between 0 and 2MeV is accumulated [21]. 3.2.2 Auger Electron Spectroscopy (AES) When an inner shell vacancy is created in an atom by a photon, energetic electron or proton, the excited atom relaxes to equilibrium either radiatively or non-radiatively. For radiative de- excitation, a higher shell electron make a transition to occupy the inner shell vacancy, releasing radiation (photons) in the process. For non-radiative de-excitation, the higher shell electron makes the transition to the inner shell vacancy, and the energy released causes the ejection of secondary electrons (Auger electrons) without emission of photon radiation. The detection of Auger electron emission dates to the work of P. Auger in 1925 [173]. Auger electron characteristic energies uniquely identify the element from which electrons are ejected. The energy of an Auger electron is determined by the differences in binding energy associated with the de-excitation as the atom rearranges its electron shells by emitting electrons with characteristic energy [166]. An illustration of Auger de-excitation is shown in figure 3.13. The transition nomenclature (a249 a102 a97 a102 a97 ) implies an initial vacancy in the K shell. An outer a102 a97 electron fills the vacancy, and the energy released is given to another a102 a97 electron which is ejected from the atom. The Coster-Kronig transition in which the primary vacancy and one of the final state vacancies lie in the same shell has a higher probability than the normal Auger transition, and so affects the relative Auger line intensities. Auger energies and intensities Radiationless process consist of transitions involving holes (vacancies). The initial hole can be characterized by the following quantum numbers: principal, angular momentum, magnetic 92 Figure 3.13: Illustration of auger electron de-excitation [166] and spin a90a123a233a78a130 a57 a158 a57 a11a151a93 a108 , and the final two vacancies by a90a123a233a78a130 a57 a158 a57 a11a151a93 and a90a118a233a79a130 a57 a158 a57 a11a151a93 a128 [174]. The transition is a90a118a233a79a130 a57 a158 a57 a11a151a93 a108 a39a63a35 a90a118a233a79a130 a57 a158 a57 a11a151a93a54a90a118a233a79a130 a57 a158 a57 a11a58a93 a128 Enery calculations are carried out via theoretical techniques, a semi-empirical approach or a totally empirical technique based on simplified equations. Empirical Methods The simplest equation for the energy of Auger electron is a103 a164 a178a176a175a149a90a159a177a204a93a203a24a184a103 a164 a90a159a177a204a93a114a158a176a103a179a178a32a90a159a177a204a93a114a158a55a103 a175a11a178 a90a159a177a66a93 (3.69) where a103a179a164a83a90a64a177a204a93 is the energy level in which the initial hole is located, a103 a178 a90a159a177a204a93 is the energy level from which an initial electron falls to fill the initial hole, a103 a175a11a178 a90a64a177a204a93 is the energy appropriate to an atom singly ionized after the Auger electron has been ejected from energy level a179 . a103a179a164 a178a176a175 a90a159a177a204a93 is 93 the auger energy of the transition uvw of element z. a103a105a164a53a90a159a177a204a93 and a103 a178 a90a159a177a66a93 can be approximated as the atomic binding energies of electrons in the u and v energy levels. a103 a175a11a178 a90a159a177a204a93a203a24a184a103 a175 a90a64a177a204a93a15a158 a210 a103 a175 a90a64a177a204a93 where a210 a103 a175 a90a159a177a204a93a203a24 a97 a31 a110 a103 a175 a90a159a177a163a158a55a46a153a93a83a39a246a103 a175 a90a159a177a204a93a15a158a185a103 a178 a90a64a177a149a158a55a46a35a93a12a39a159a103 a178 a90a159a177a66a93a43a120 . Therefore, a103a105a164 a178a115a175 a90a159a177a204a93a203a24a184a103a179a164a53a90a159a177a204a93a198a39a49a103 a178 a90a159a177a66a93a198a39a115a103 a175 a90a64a177a204a93a198a39 a46 a102 a110 a103 a175 a90a64a177a140a158a184a46a35a93a89a39a49a103 a175 a90a159a177a66a93a114a158a176a103 a178 a90a159a177a72a158a184a46a153a93a198a39a115a103 a178 a90a159a177a204a93a132a120 (3.70) Transition probabilities The Auger transistion probability a231 a251 in hydrogenlike atom (KLL transition) can be writ- ten, according to first-order perturbation theory, as [166] a231 a251a185a24 a102a85a197 a180 a243 a82a83a90 a252 a93a153a162 a181 a165 a90 a39a35 a71 a97 a93a76a182 a165 a90 a39a35 a71 a31 a93 a82 a31 a162 a39a35 a71 a97 a39 a39a35 a71 a31 a162 a181 a108 a90 a39a35 a71 a97 a93a76a182 a108 a90 a39a35 a71 a31 a93a68a29 a39a35 a71 a97 a29 a39a35 a71 a31 a162 a31 a82a12a90 a252 a93a203a24 a57 a211 a13 a63a30a197 a34 a180 a243 a31 a212 a252 a27a114a29a32a31a9a161a203a29a84a161a174a29a32a5 (3.71) where a82a83a90 a252 a93 is the density of states associated with normalization. Practical considerations Auger electron transitions generally appear as small features superimposed on the large back- ground of secondary electrons. A derivative technique a29a32a73a115a90a91a103a124a93a133a27a30a29a32a103 is used to minimize the slowly varying background and to enhance observation of the Auger electron signals. Electronic differentiation of the energy distribution function a73a49a90a160a103a135a93 is readily achieved with a velocity anal- yser by superimposing a small ac voltage on the energy selector voltage and synchroniously detecting the power output of the electron multiplier. Auger peak-to-peak height of the differen- tiated signal is a direct measure of the surface concentration of element that produces the Auger electrons. 94 The sensitivity of Auger technique depends on the transition probability a231 a251 , the incident beam current and energy, and the collection efficiency of the analyser. The limit of detection for the elements varies between approximately 0.02 and 0.2 atomic percent [175]. Qualitative and quantitative analysis Principal Auger peaks of KLL, LMM and MNN Auger transitions allow identification of all elements above He by scanning the 0-2keV energy range. Major Auger peaks can be identified by using the Principal Auger Electron Energies Chart to reduce the number of possibilities before attempting a detailed analysis of a standard Auger spectrum. A peak shift of a few eV can be considered insignificant if the element is in a different chemical environment from that used for the standard spectrum. The relationship between Auger electron signal and atomic concentration can be determined as a function of instrumental parameters such as the primary electron beam current (a126a114a90 ), the pri- mary beam energy (a103a183a90 ) and for the modulation energy (a103a105a6 ) [175]. There is a linear relationship between Auger peak-to-peak amplitude and a126a114a90 as long as the incident electron beam diameter does not exceed the analyser source diameter, and the current density is not too large to damage the sample surface. The Auger signal amplitude is proportional to the modulation energy when the modulation energy is small compared with the Auger peak width. For large modulation amplitudes, the peak is broadened and peak-to-peak amplitude becomes nonlinear with a103 a6 . Auger peak-to- peak amplitudes vary with the primary beam energy a103a184a90 because of the energy dependence of the electron impact ionization cross section for the core level involved in Auger transition. The Auger yield rises abruptly from zero as a103a184a90 crosses the ionization threshold a103 a145 and increases to a maximum for a103 a90 a27a62a103 a145 a59a41a65 [175]. 95 A first order approximation for quantitative analysis can be accomplished through comaprison of the Auger signal from the sample to the Auger signal from a pure standard. To avoid the need for a large number of elemental standards, another method comparing the signal from the sample with that from a pure silver target has been adopted. Setting a103a184a90 to a standard value and a103a105a6 low enough to prevent significant distortion due to excessive modulation, the atomic concentration of element a74 is a1 a150a48a24 a126a54a150 a126 a251a7a104 a26a12a150a62a188a124a150 (3.72) a126a54a150 is the peak-to-peak amplitude of element a74 from the test specimen, and a126a54a251a53a104 is the peak-to- peak amplitude from the Ag standard. The relative scale factor between spectra for the sample and the silver standard is a188a75a150a191a24 a102 a150a62a103a105a6 a242 a150a62a126a114a90 a242 a150 a102 a251a53a104 a103a179a6 a242 a251a7a104 a126a114a90 a242 a251a53a104 (3.73) a102 a150 is the lock-in amplifier sensitivity, a103a179a6 a242 a150 is modulation energy and a126a114a90 a242 a150 is the primary beam current. Without elemental or silver standards, it is possible to express the atomic concentration as a1 a150a191a24 a126a54a150 a26a83a150a62a29a66a150 a27a186a185 a2a45a187 a126 a187 a26 a187 a29 a187a189a188 (3.74) The sumation is over one peak per element and a29a66a150a48a24 a102 a150a61a103a179a6 a242 a150a62a126a114a90 a242 a150 is the scale factor. This simple quantitative technique has some inherent errors which include: (1) matrix effects on electron escape depth and backscattering factors (2) chemical effect on peak shapes (3) surface topography Auger electron escape depth dependence on electronic sttructure of the host material may alter 96 the depth measurement in the specimen relative to that in a standard. Chemical effect can cause a change in the peak shape and lead to error using peak-to-peak heights of differentiated spectrum. Highly polished surfaces produce larger Auger signals than rough surfaces [175]. The combination of RBS and AES is quite useful in depth profile analysis because RBS gives quantitative information on depth and heavy mass constituents without the complications of intermixing during sputtering. Ion sputtering causes a change in the composition of the surface layer due to surface segregation and preferential sputtering. Compared to RBS, AES depth profiling provides better depth resolution and is sensitive to both heavy and light elements. 3.2.3 X-ray Photoelectron Spectroscopy (XPS) [166] Photoelectron spectroscopy is one of the major surface analytical tools. This technique arises directly as a result of the interaction between photon and atoms. The source of photons incident on the sample can be ultraviolet light (UPS) or X-ray (XPS). This technique is other- wise called ESCA (Electron Spectroscopy for Chemical Analysis), especially if the focus is on chemical bonding in the sample. A wide range of photon energies can be be used for analysis. Photons with energy as low as 10eV can interact with valence electrons and provide information about chemical bonding in the sample. For elemental identification however, much higher pho- ton energy is required. Photons with energy as high as 0.1MeV can penetrate solids and interact with inner shell electrons. A major advancement for this technique came with the advent of electron synchrotron facilities that provide intense monochromatic photon beams over a broad range of energies. Most laboratory instruments produce X-rays in the 1 - 10keV region. When a sample is exposed to photons of energy a180a243a84a167 , it absorbs the quantized energy and electrons are ejected from the material. The kinetic energy of an ejected electron is related to the binding energy of an electron in the atom. The energy of the electron also gives the identity 97 of the element from which the electron is ejected. Photoelectron spectroscopes require both a source of monochromatic radiation and an electron spectrometer. Sources UPS generally uses a resonant light source such as a a219a246a82 discharge lamp with energies in the a46a2a69a75a39a176a137a53a46a2a82a85a13 range. This energy is sufficent for the analysis of the valence band density of states for most solids. Depending on X-ray energy of interest, characteristic X-ray for XPS analysis is provided by electron bombardment of targets. Magnesium and aluminum targets are used predominantly, and they produce soft X-rays (a45a22a46 a252 a82a85a13 ). Hard X-rays can be produced with a Cu target (a45a10a63 a252 a82a85a13 ), and the associated energy width is about 2.0eV. Molybdenum also gives hard characteristic X-ray energies (a45a10a46a153a101 a252 a82a85a13 ) with an energy width of about a69a61a82a85a13 . Because of the poor energy resolution, most of these sources are not suitable for high resolution studies. Synchrotron radiation from electron storage rings can provide a continuous spectrum with intensities much higher than X-ray or resonance light sources. The use of polarized tunnable radiation from the synchrotron is a distinct advantage in experimental investigations, but access to synchrotron facilities is limited. Detection system The electrostatic analyzer is used primarily to determine the energy of photoelectrons. De- flection and reflection (mirror) are the two general operating modes for analyzers. The cylin- drical mirror analyzer (CMA) is a common type. Deflection is caused by the potential differ- ence between the inner and outer cylinders. CMAs maintain a constant energy resolution. The 98 electron analyzer measures the number of photoelectrons at different kinetic energies, and the information is displayed as a spectrum of photoelectron intensity as a function of binding energy. Quantitative analysis In quantitative analysis, the area under a photoelectron peak or the line intensity is used. Line intensity depends on factors such as photoelectric cross section a107 , the electron escape depth within the sample a190 , the spectrometer transmission, surface roughness or inhomogeneities and the presence of satellite structures. Chemical analysis of a sample with component elements A and B can give the relative concentration, a233a12a251a203a27a86a233 a143 , in terms of the peak intensities and cross sections if photopeaks from A and B have about same energy and same detection efficiency, a233 a251 a233 a143 a24 a126 a251 a126 a143 a107 a143 a107a15a251 (3.75) Photoelectron peak area can also be compared to known standard for quantitative analysis. 3.2.4 X-ray Diffraction (XRD) X-ray diffraction is one of the most common tools for material characterization because of its simplicity, reliability and nondestructive nature [176]. The basis of X-ray diffraction is the Bragg equation which describes the condition for constructive interference for X-ray scattering from atomic planes of a crystal. a102a85a29a71a27a114a29a32a31a12a161a191a24a41a233a95a190 (3.76) X-ray diffraction is useful in many technological applications. It is used to identify the crystalline phases present in materials and to measure structural properties such as strain state, grain size, 99 Figure 3.14: X-ray diffraction geometry [177] defect structure, phase composition and preferred orientation (figure 3.14). It is also used in thin films and multilayer films to determine thicknesses and atomic arrangement in amorphous materials. In packaging materials evaluation, XRD can be used to investigate diffusion and phase formation at interfaces[177] For ultra thin film analysis (thickness a19a218a46a35a50a61a50a85a233 a57 ), the conventional Bragg-Brentano (a161a28a39a94a102a69a161 ) X-ray geometry (figure 3.15a) is not very useful because of interference effects from the sub- strate. A better geometry is a low incidence angle (glancing angle) geometry because it allows more material of the thin film to be sampled and provides information about the crystallinity of the sample surface with reduced interference from the substrate. The grazing incidence X-ray diffraction (GIXRD) geometry has been used to characterize monolayer films [178]. With the incident angle small enough, the X-ray can penetrate a45 a46a2a50a48a39 a102a62a50a85a233 a57 into the specimen. The exit angle of the diffracted X-ray is also small and structural information is obtained about (a243 a252 a130 ) planes perpendicular to the specimen surface. However if 100 Figure 3.15: (a) Bragg-Brentano diffraction geometry for thin films (b) Seemann-Bohlin diffraction geometry [177] the incidence angle is too small, the radiation will not penetrate the sample, but will result instead in evanescent wave (due to total external reflection) at the surface [179]. The Seemann-Bohlin diffraction geometry (figure 3.15b)is suitable for thin film XRD. It provides good sensitivity for thin films, due to parafocusing and large diffracting volume because the angle of incidence a182 is small (a45a21a65a198a39a159a46a2a50 a0 ), and X-ray path length in the film is large because it is proportional to a46a35a27a191a27a114a29a32a31a105a182 . The angle of incidence a182 is normally fixed and the angle between the diffracting planes and the incident X-ray moves through a102a69a161 as the detector sweeps the angle. The geometry is most useful for polycrystalline films that have random or nearly random crystallite orientation [177, 180]. 101 Diffraction patterns from amorphous materials do not have sharp peaks that characterize crystalline materials, but rather broad features qualitatively indicate amorphous material. Quan- titative analysis of XRD data from amorphous materials is complicated, but can provide impor- tant information on local atomic structure (short range order), including bond lengths, number of neighbors and the extent of atomic correlations [177]. 3.3 Mechanical Analysis 3.3.1 Introduction Mechanical reliability for composite contact metallizations is an essential part of metalliza- tion evaluation during and after device fabrication. The primary concern is adhesion between layers in the multilayer structure that constitutes the composite contact. Monitoring the degrada- tion of adhesion properties is important, and long-term reliability testing is generally performed under temperature-humidity cycling. Wirebond pull and shear (push) testing and brazing for die attach are normally used to study adhesion properties. Vertical pull force and horizontal push force on bond points can cause sep- aration within the metallization structure, since the weakest interface in the multilayer structure will give way first. The pull or push strength at which this failure occurs can be measured exper- imentally. The bond can be either a ball or a wedge depending on the type of wire bonder used. The bonding scheme is a wire loop terminating at two bond points seperated by a horizontal distance (d) (figure 3.16). The length of the loop and the bond-to-bond distance determine the angle of pull on the bond points. A simple analysis of the force distribution is shown below. 102 Wire Bonding and Die Attach [181] Bonding between two dissimilar metals (e.g. Au-Al) leads to the formation of intermetallics. The property of the intermetallic formed (e.g brittleness) will mostly determine the strength of the bond. Excessive formation of intermetallics is manifested in the form of an increase in resis- tance, intermittent conduction, or an open circuit. Intermetallic formation is normally accompa- nied by Kirkendall voiding, leading to the brittle fracture of the wire bond. Because dissimilar metals diffuse into each other at different rates, an excess of vacancies is formed on the side of the fast diffusing metal. This vacancies coalesce into voids, known as Kirkendall voids, which act as a stress concentrator. Wirebonding with same metal (e.g., Au-Au, Al-Al) elliminates the possibility of interface corrosion, intermetallic formation, Kirkendall voiding and other bond degrading conditions. However, the cost of a Au-Au system, for example, may be a draw back when considering monometallic wire bonding, apart from additional and/or difficult processing steps that may be involved. Wirebond pull testing An applied pull force F on the hook pulling on the loop and the tension on the wire bond T are related as shown in figure 3.16. a196 a24a21a102a85a138a192a27a114a29a32a31a12a161 a193 a138a41a24 a196 a102a189a27a114a29a32a31a9a161 (3.77) If the bond-to-bond distance is a29 , and the length of wire loop between the bond points is a130 , we define a fractional distance between these lengths as, a193a116a24 a158 a4 a127 a158 i.e. a46a179a39a58a193a94a24 a127 a158 103 Figure 3.16: Wire bond pull and shear testing a27a114a29a32a31a9a161a191a24 a102a61a243 a130 a96 a233a114a29 a90a159a130a91a27a62a102a61a93 a31 a24a202a90a91a29a79a27a61a102a62a93 a31 a158a176a243 a31 a27a114a29a32a31a9a161a48a24 a102a62a243 a130 a24a195a194 a102a196a193a75a39a58a193 a31 a193 a19a76a46 (3.78) Therefore, a138a134a24 a196 a102 a194 a102a196a193a124a39a58a193 a31 (3.79) This equation implies measuring a29 and a130 or the difference between them and the pull force F is enough to estimate the pull strength on the bond. Wirebond shear testing An horizontal force can be used to push on the wire bond to its shear point. The shear force is proportional to the bond contact area. Part of figure 3.16 illustrates the bond shear testing structure. Die attach shear testing is similar to the bond shear testing (figure 3.17). The difference is that a larger area chip is brazed to a ceramic carrier. Between the ceramic and the 104 Figure 3.17: Chip brazed shear testing chip is a preform used to solder or barze the ceramic and the chip together. The chip is then push horizontally while the ceramic substrate is held in place. 105 CHAPTER 4 EXPERIMENTAL PROCEDURES 4.1 Standard Procedures Research grade 4H-SiC for this work was purchased from Cree, Inc. The profiles of the a102 a128 a128 diameter wafers and their intended uses are listed in the table 4.1. The wafers were diced into 5mma47 5mm pieces for fabrication. 4.1.1 Sample Cleaning Substrate cleaning is required to remove surface particulates as well as traces of organic, ionic and heavy metallic impurities. The two stage substrate cleaning process includes an organic clean and RCA clean. 4H-SiC profile (cma4 a34 ) Purpose N sub/a50a32a87a88a101a85a77 a57 epi a78a89a90a132a101a75a47a183a46a35a50 a97a43a214 a93 / p-ohmic contact a50a79a87a225a64a30a77 a57 epi a78a83a3a179a90a91a64a94a47a183a46a35a50 a97a43a106 a93 N sub/a50a32a87a88a101a85a77 a57 epi a78a89a90a132a101a75a47a183a46a35a50 a97a43a214 a93 / p-ohmic contact a50a79a87a225a64a30a77 a57 epi a78a83a3a179a90a91a64a94a47a183a46a35a50 a97a43a106 a93 /impl. a78a83a3a179a90a133a46a191a47a183a46a2a50 a31 a97 a93 P sub/epi a233 a3 a90a132a65a75a47a52a46a2a50 a97a132a106 a93 n-ohmic contact a73a246a3 sub/epi a233 a4 a90a160a60a32a87a100a63a131a47a183a46a35a50 a97 a33 a93 n-schottky contact Semi-Insulating SiC Metal sheet resistance measurements Low quality SiC Samples for RBS, AES and XPS analysis Table 4.1: Profile of 4H-SiC purchased from Cree research Inc. 106 Organic clean The purpose of this cleaning procedure is to remove organic and oily contaminants on the substrate surface. The samples were: a44 immersed in acetone and agitated in an ultrasonic bath for 5mins. a44 immersed in trichloroethylene and agitated in an ultrasonic bath for 5mins. a44 immersed in acetone and agitated in an ultrasonic bath for 5mins. a44 immersed in methanol and agitated in an ultrasonic bath for 5mins. a44 immersed in fresh mehtanol and agitated in an ultrasonic bath for 5mins. a44 rinsed in de-ionized water (DI water) for 3mins. a44 immmersed in buffer oxide etch (BOE) for 1min. a44 rinsed in DI water for 2mins and dried with N a31 gas. RCA clean (Radio Corporation of America) The purpose of this cleaning procedure was to remove ionic and heavy metal contaminants on the samples. The samples were: a44 immersed in a 1:1 (solution) of H a31 O a31 :H a31 SO a74 for 15mins. The mixture of H a31 O a31 and H a31 SO a74 is exothermic. a44 rinsed in DI water for 2mins. a44 immersed in a 20:6:6 mixture of DI H a31 O:H a31 O a31 :NH a74 OH warmed on hot plate to boil gently for 15mins. 107 a44 rinsed in DI water for 2mins. a44 immersed in a 20:6:6 mixture of DI H a31 O:H a31 O a31 :HCl warmed on hot plate to boil gently for 15mins. a44 rinsed in DI water for 2mins. a44 immmersed in a buffered oxide etch (BOE) for 2mins. a44 rinsed in DI water for 3mins and dried with N a31 gas. These cleaning procedures are the same irrespective of the type of device that is being fabricated. 4.1.2 Photolithography Photolithography is the process by which a mask aligner is used to transfer features on a mask onto a sample. A Karl Suss MJB3 UV400 mask aligner was used for this work. Square TLM mask features and circular mask features for diodes and MOS structures are shown in figure 4.1. The light source on the mask aligner is Karl Suss UV lamp with an output power of 160W. The mask aligner is manually controlled and a102a48a39a55a64a30a77 a57 features can be aligned with the help of optical microscope having magnifications of 5, 10 and 20. The 5mma47 5mm sample was held on a a64 a128 a128 Si wafer, and AZ 5213-EIR photoresist (posi- tive photoresist) was spun onto the sample for 30s at a rotor speed of 4000rpm. The sample was then soft baked on a hot plate at a46a62a46a35a50 a0a54a1 for about 90s to dry the photoresist and make it more photosensitive. Under baking or over baking can reduce or destroy the sensitivity of the photoresist. The sample on a64 a128 a128 Si wafer was mounted on the mask aligner, and X, Y, and a161 adjustment of the sample table was made to bring the sample into alignment with the mask features. By 108 Figure 4.1: (a) TLM masks (b) Diode/MOS Mask 109 exposing the sample to UV light through the mask, the features on the mask were transfered on the sample. UV exposures was made for about 25s. Following exposure, samples were developed in a mixture of DI water and Microposit 351 developer or AZ 400K developer. The volume ratio DI water to developer was 4:1. The samples were immersed in this diluted developer for few seconds. During the time of development, the part of the photoresist on the sample exposed to UV light was washed off, and the sample was subsequently rinsed in DI water and dried in N a31 gas. Complete development of the sample was confirmed under a microsope. Reversing the effect of UV light exposure on photoresist, keeping the photoresist exposed to UV light stays on the sample after development while removing the part not exposed can be achieved by using a negative photoresist or simply reversing the positive photoresist as follows. Spin-on the positive photoresist and soft bake it for 30s, then expose the sample to UV light through the mask for 25s. Then bake the sample after exposure for 60-90s at a46a61a46a2a50 a0a2a1 . Expose the whole sample to UV light through a clear mask or no mask for 60s. Develop the sample in 4:1 DI water and developer yeild the same result as obtained using a negative photoresist. 4.1.3 Sputter-deposition All metals in this study were sputter-deposited in an Argon plasma. The sputter-system is a high vacuum system. It has three a102 a128 a128 diameter Polaris magnetron sputtering sources or guns. Three different targets can be sputtered consecutively without breaking vacuum in this system. A sputter target can operate at a maximum dc power of 1000W or maximum RF power of 600W, and cathodes potential can be between 200-1500V. The sputter guns are cooled by a46a35a64 a0a35a1 water circulating beneath the cathode surface. The deposition rate is limited by the type and quality of target materials. Large area, uniform films can be obtained if the samples were located at height 110 a137 a128 a128 or so from the target. Figure 4.2 shows the sputter-system used for this study. The metals or metallic alloys to be sputtered were mounted on the magnetrons, and sprayed cylindrical glass chimneys were placed on the gun to confine and focus the sputtered materials. Samples on a64 a128 a128 Si wafers were mounted on a plain disk carrier or a disk carrier with capability of heating the sample during sputter-deposition. The system was pumped down for about 2 hours to a base pressure of about a65a75a47a52a46a2a50a32a4 a216 torr. To start sputter-deposition, cooling water circulation to the gun to be used was checked. Argon gas flow was set on an MKS 247 mass flow controller at 106.4sccm for most of the de- position in this work. At this flow rate, the chamber pressure rises to about 5mtorr but with a butterfly valve of the turbo-pump slightly closed, the pressure rises to 20mtorr for the depo- sitions. Power was turned-on and the dc voltage to the cathode incresed until an Ar plasma was generated and the pre-determined sputter current attained and stabilized. Off-sample, pre- sputtering was carried out for few minutes, then the sample was rotated to a position over the sputter-gun, and sputtering onto the sample carried out for as long as required to yield a desired thickness. For every target, calibration of the sputtering rate must have been done a priori. A Tecnor profilometer is used for a quick thickness measurement, following deposition. 4.1.4 Annealing System The high temperature annealing system used for ohmic contact anneals is also used for implant activation annealing. Temperatures up to a46a2a69a61a50a61a50a124a39a76a46a153a101a30a50a61a50 a0 a1 can easily be attained with this furnace. Heating elements in the annealing system are two carbon strips, and the current through the strips is controlled manually with a variac while temperature is measured with an Omega pyrometer focused on the carbon strip close to the samples. The samples were held on 111 Figure 4.2: Sputter-deposition system with disk sample holders having heating capability 112 Figure 4.3: Annealing system with enlarged carbon strips showing sample 113 the carbon strip with tungsten clips. The annealing chamber is pumped to a vacuum of about a46a191a47a183a46a2a50a32a4 a216 torr for vacuum annealing. Power is turned-on when the chamber pressure is about a46a48a47a115a46a35a50 a4 a216 torr, and current through the carbon strips increased slowly. The annealing temperature (a60a62a50a61a50a131a39a10a46a2a50a61a50a61a50 a0a2a1 ) is reached in 2-3mins with a variac setting of 40-45%. Most of the ohmic contact anneals were carried out for 1-2mins in vacuum. During annealing, the chamber pressure rises to about a102 a47a187a46a35a50a32a4a7a51 torr. It is also possible to anneal samples in Ar or N a31 ambients with this system. After annealing the variac is set to zero, and the power switched off. The annealing chamber temperature drops to room temperature in vacuum in about 3 hours. Samples were removed in Ar at atmospheric pressure. 4.2 Device Fabrication and Measurements 4.2.1 Ohmic Contact Fabrication Ohmic contact to a78a12a3 -4H-SiC Ohmic contacts were fabricated on implanted and epitaxial a78a83a3 -4H-SiC. The profiles of these materials are listed in table 4.1. The sequence for ohmic contact fabrication is shown in figure 4.4. An aluminum-titanium alloy (Al-Ti 70-30 wt%) and nickel (Ni-V 93-7wt%) were sputter-deposited. Contact anneals for the Al-Ti contact on the a78 a3 -implanted and a78 a3 -epilayers was carried out at a46a35a50a62a50a61a50 a0a2a1 for 2 minutes in vacuum. These conditions were the optimum for low resistance contacts [11]. Nickel contacts were annealed ona78 a3 -implanted anda78 a3 -epilayer SiC at a60a61a50a62a50 a0 a1 for 1 minute in vacuum. 114 Figure 4.4: Sequence for ohmic contact fabrication 115 Ohmic contact to a233 a3 -4H-SiC Ohmic contacts were fabricated on epitaxial a233a181a3 -4H-SiC (see table 4.1). The fabrication sequence is the same as for p-ohmic contacts. A nickel-chromium alloy (Ni-Cr 80-20 wt%) and nickel (Ni-V 93-7wt%) were deposited. Contact anneals for Ni-Cr contacts on a233 a3 -epilayers were carried out at a46a2a50a61a50a61a50 a0a35a1 for 2 minutes in vacuum. Nickel contacts were annealed on a233a28a3 -epilayers at a60a66a65a30a50 a0 a1 for 2 minutes in vacuum [5]. 4.2.2 Schottky Contact Fabrication Schottky contacts were fabricated on a46a2a50a30a77 a57 a233a174a4 -epilayers on a233 a3 -4H-SiC substrates using the sequence shown in figure 4.5. The samples were cleaned using standard procedures after which a 30nm thermal oxide was grown at a46a61a46a153a65a30a50 a0 a1 in dry a85 a31 for about 4 hours. These oxide layers were used as either sacrificial or passivating layers. For Schottky diodes, the sacrificial oxide was striped just before photolithography that defined the Schottky diode feaures. After photolithography, nickel silicide was sputtered, and the samples subsequently sintered at a65a62a50a62a50 a0a2a1 for 24 hours. A second set of Schottky diodes was also fabricated. The back side oxide (oxide on the a233a28a3 side) of the sample was etched in buffer oxide (BOE) for about 1 minute while the front side oxide was protected with crystal bond. Nickel ohmic contacts were fabricated on the back side by sputter-deposition and high temperature annealing at a60a66a65a30a50 a0a2a1 for 2 minutes. After the fabrication of back contact, the Schottky contacts were formed by nickel silicide (Ni a31 Si) sputter deposition through windows opened in the front side oxide layer. The contact thickness was about a46a35a50a62a50a61a50 ?a70 . The Schottky diodes were formed with passivating SiO a31 surrounding each device. This set of samples was not sintered. 116 Figure 4.5: Sequence for Schottky contact fabication 117 4.2.3 Protective Stack Fabrication and Device Mesurements Devices operating at a64a66a65a30a50 a0a2a1 in air have to be protected from oxidation and cap layer/contact layer inter-diffusion. The protective metallization fabrication sequence is shown in figure 4.6. Tantalum silicide (TaSi a31 ) was sputter-deposited on ohmic and Schottky contacts as diffusion/oxidation barrier. About a46a35a65a62a50a62a50 ?a70 of tantalum silicide was sputter-deposited in Ar and Ar/a73 a31 gas mixtures. Then a a46a2a50a61a50a61a50 ?a70 platinum layer was deposited followed by cap layers of Au (1200 ?a70 ) or Au/Sn/Au (500 ?a70 /25 ?a70 /700 ?a70 ). Cross-section of ohmic and Schottky contacts with protective stacks are shown in figure 4.7. During processing, it was observed that sputtering Pt at a102a61a65a62a50 a0 a1 improved the adhesion with the Ta-Si-N layer considerably. Good adhesion is essential for wire bonding and die attach. The metallization sequence in figure 4.6 was subsequently modified to allow sputter-deposition of Pt at a102a62a65a62a50 a0a2a1 . Since photoresist starts to carbonize at this temperature etch back of Pt in an acid solution as well as lift-off with Mo was used. Figure 4.8 shows the modified sequence. Pt etch back After sputter-deposition and lift-off of Ta-Si-N layer, the samples were loaded back into the sputter system on the sample holder and Pt was depostited at a102a61a65a30a50 a0a2a1 . Photolithography, to define the contact areas, was conducted with the thick photoresist, AZ P4620, UV light exposure time of 1 minute. The photoresist over the defined contact regions was baked for about 7 minutes at a46a61a46a2a50 a0a2a1 . The baking reduces undercutting of the Pt during acid etch back. The etchant for Pt is a46a35a219a159a73a197a85a140a34 /a64a62a219 a1 a130 /a137a204a219 a31 a85 at a60a66a65 a0a2a1 . Etching a46a35a50a62a50a61a50 ? a70 of Pt was completed in about 6 minutes. The etching process is more effective for diodes/MOS features where the separation between adjacent devices is large. It is a little difficult to etch the fine features in a TLM pattern, especially the 118 Figure 4.6: Sequence for protective metallization on (a) ohmic contact (b) Schottky contact 119 Figure 4.7: Device cross-section of (a) ohmic (b) initial Schottky (c) new Schottky contacts 120 Figure 4.8: Modified sequence for protective metallization (a) ohmic contact with Mo lift-off (b) Schotky contact with Pt etch-back 121 last two smallest gaps. To be able to etch the fine features (a19a55a137a61a77 a57 gaps) effectively, the samples stays longer in the etchant and significant undercutting on lager features will occur. Thin Au cap layers (about a46a153a102a62a50a62a50 ? a70 ) were deposited on samples with contact pads re-defined by photolithography. The Au cap layer was then lifted-off everywhere except on the contact pads. Pt lift-off with Mo To avoid etching samples in acid, molybdenum was used to lift-off Pt. After Ta-Si-N was deposited and lifted-off in acetone, normal photolithography was carried out to leave photoresist on the contact pads after development. Molybdenum was then deposited on the samples, and the Mo on photoresist above the contact pads was lifted-off in acetone, leaving Mo everywhere except on the contact pads. As a result, Mo serves the function of photoresist if the Pt is not sputter-deposited at a102a62a65a62a50 a0 a1 . At this temperature Mo does not diffuse into SiC or the metallization on the samples. The lift-off of Mo with the Pt layer on top of it is done in a219 a31 a85 a31 in about 10 minutes for a Mo of about a64a61a50a61a50a62a50 ?a70 . A thin Au cap layer (about a46a153a102a62a50a62a50 ?a70 ) can be deposited over the Pt before lifting-off the Mo, so that a single process defines the Pt and Au cap layers. For thick Au (about a69a62a50a61a50a61a50 ?a70 ) needed for wire bonding, another photolithography process may be necessary. For higher Au thicknesses, photoresist spin-on may be repeated 3 or more times to accumulate thicker AZ 5214 photoresist on the sample for easier lift-off in acetone. Alternatively, Au can be etched back. 4.2.4 Wirebonding and Chip Brazing Large area samples were fabricated using low quality SiC for wire bonding (1.8 a47 1.8cm sample size) and for die attach (0.5 a47 0.5cm sample size) testing. The samples were cleaned 122 using standard procedures. For indirect wire bonding, adhesion of the protective stack to SiO a31 was investigated. This structure was also brazed on ceramic carriers. Cleaned SiC was first oxidized, then Ta-Si-N, hot Pt and thick Au layers were sputter-deposited onto the oxide layer. Adhesion of the protective stack to ohmic contacts and Schottky contacts for the case of direct bonding was also investigated. The protective metallization and cap layers were deposited without breaking the vacuum. The top Au cap layer for wire bonding was about a69a61a50a62a50a61a50 ?a70 thick. The wire bonder used is a computer controlled ultrasonic model. This equipment is located in the Center for Advanced Vehicle Electronics (CAVE) in the department of Electrical and Computer Engineering. The bonder is capable of using wire diameters in the range a46a153a102a62a101a18a39a191a65a30a50a61a63a30a77 a57 . Parameters like bond force, bond power, bond time, loop value, hop distance and so on were pre- set on the system. This bonder operates by pressing the bonding wire thermosonically onto the sample cap layer and thus forming a wedge bond (not a ball bond) on the sample. Careful selection of the input parameters leads to good wirebonding. For this work a102a61a65a85a137a62a77 a57 (10mil) diameter Au wire was used. The Au wire was wedged thermosonically at about a102a62a64a61a50 a0a35a1 to the a69a61a50a62a50a61a50 ? a70 Au cap layer. Chips were brazed to a direct-bond copper (DBCu) substrate which is simply a Cu-coated ceramic (a70a81a130 a31 a85a140a34 ) substrate with Ni film on the Cu and a and top thin Au film on the Ni. DBCu substrates were purchased from Stellar industries, then plated with thick Au (few microns) be- cause the flash Au layer is too thin for brazing. The substrates was diced slightly bigger than the SiC chips to allow room for shear testing. The brazing of the chip to the substrate was carried out with the help of a Au-Sn alloy (80-20 wt%) preform. This preform was selected because a Au-Sn alloy eutetic occurs at a19a184a64a66a102a30a50 a0a2a1 which is very much lower than the melting point of Au. The preform was sandwiched between the chip and the substrate, kept under a load and place in the brazing machine. The brazing was carried out at a pressure of about a46a2a50 a4 a33 torr. The braze 123 Figure 4.9: (a) Large area wire bonded sample for pull and bond shaer testing (b) Chip brazed sample for die shear testing temperature profile ramped from room temperature to a64a62a50a61a50 a0a35a1 in 5 minutes, and then from a64a61a50a62a50 a0a2a1 to a137a204a50a62a50 a0a2a1 over 3 hours. Figure 4.9 shows brazed and wirebonded samples for shear and pull testing. For both, a Dage micro-mechanical test system was used. The Dage-PC2400 can apply a 5kg force for bond shear testing, up to 100kg for die shear testing and up to 10kg for wire-pull testing. 4.2.5 Thermal Aging of Samples All the samples fabricated were first characterized physically, electrically and mechanically. These pre-aging measurements are referred to as 0-hr data. The samples were then thermally aged at a64a61a65a62a50 a0 a1 in air using two furnaces - an old furnace operated at a64a66a65a30a50 a0 a1 24 hours a day, and a new programmable box furnace with a temperature cycling capability. This furnace was 124 programmed to cycle from a102a62a65 a0 a1 to a64a66a65a62a50 a0 a1 in 1 hour, remain at a64a66a65a62a50 a0 for 8 hours, then cycle back to a102a62a65 a0 a1 in 1 hour, remain at a102a61a65 a0 a1 for another 2 hour and then repeat the cycle. At intervals (100, 500, 1000hr, etc.) the samples were removed from the furnaces and characterized again. This process continued for thousands of hours of annealing in air at a64a66a65a62a50 a0a2a1 . 4.2.6 Measurements The set-up for TLM measurements is shown in figure 4.10. The total contact-to- contact resistance was measured by passing a 1mA current between adjacent TLM pads using a Keith- ley 220 programmable current source. The potential (millivolts) developed between the pads was measured with a Hewlett Packard 3478A multimeter. Thus, what was measured directly by the multimeter is the total contact-to-contact resistance as a function of the inter-contact spac- ing. The contact resistance, specific contact resistance and semiconductor sheet resistance were then obtained from a TLM analysis. Samples without and with protective metallizations were measured. TLM measurements were also performed for some protected stacks as a function of temperture up to a64a66a65a30a50 a0a2a1 . Current-voltage (I-V) and Capacitance-voltage (C-V) measurements were carried out for the Schottky diodes. The set-ups for the I-V and C-V measurements are shown in figure 4.11. I-V measurements up to 2V forward bias and reverse bias out to -300V were performed. 125 Figure 4.10: (a) TLM measurement set-up (b) measurement configuration 126 Figure 4.11: (a) Current-Voltage measurement set-up (b) Capacitance-Voltage measurement set-up 127 CHAPTER 5 RESULTS AND DISCUSSION 5.1 The Choice of Barrier Layer Material 5.1.1 Silicide Barrier Layer In support of very large or ultra large scale integrated (VLSI or ULSI) microelectronics, metal-silicon-nitrogen (M-Si-N) amorphous films have been studied as candidates for diffusion barrier layers between interconnect metals (Cu, Al) and silicon [112, 106, 136]. In this study, four silicide materials (MoSi a31 , WSi a31 , TaSi a31 and Taa33 Sia34 ) were considered for barrier layers on SiC devices for high temperature applications. The effeciveness of TaSi a31 , WSi a31 and MoSi a31 as diffusion/oxidation barrier layers at a64a61a65a62a50 a0 a1 in air was investigated by sputter- depositing the silicides on SiC. Platinum cap layers were deposited. RBS data of these samples as a function of annealing time in air at a64a66a65a62a50 a0a35a1 are shown in figure 5.1. Based on the RBS spectra the following conclusions can be made. 1. Platinum permitted oxygen to diffuse in air at a64a66a65a62a50 a0a2a1 . 2. Mo-Si and W-Si films under Pt cap layer were oxidized when the samples were annealing in air. Mo-Si and W-Si layers were almost totally oxidized after 120 hours and 780 hours of annealing in air, respectively. 3. Significant mixing of W-Si with Pt cap layer was observed. 4. The Ta-Si layer showed high resistance to oxidation and little mixing with the Pt over- layer. Based on these results, Mo-Si and W-Si were droped in preference to Ta-Si as candidate for the diffusion/oxidation barrier. 128 Figure 5.1: RBS spectra of (a) SiC/Mo-Si/Pt (b) SiC/W-Si/Pt and (c) SiC/Ta-Si/Pt 129 TaSi a31 and Taa33 Sia34 The choice between TaSi a31 and Taa33 Sia34 for the deposition of Ta-Si-N was made in favor of TaSi a31 . Figure 5.2(a) shows that nitrogen was actually incorporated in the tantalum silicide films. The amount of nitrogen incorporated increases with increasing percentage of nitrogen flow during sputter-deposition. Van der Pauw sheet resistivity measurements [figure 5.2(b)] showed that above 2% nitrogen flow the resistivity of the Ta-Si-N layer obtained from the Taa33 Sia34 target increses rapidly and is significantly higher than the resistivity of Ta-Si-N layer deposited using TaSi a31 . Figure 5.2(c) is the plot of nitrogen content in the Ta-Si-N layer as a function of nitrogen flow percentage. Films deposited from Taa33 Sia34 contain more nitrogen and hence are more resistive as indicated by the van der Pauw resistivity measurements. Nitrogen atomic percentages in the films were computed using fit to RBS spectra taken with and without nitrogen in the a138 a96 a26a89a80 a31 films. The atomic ratios used in the fits were Ta:Si:N = 1:1.5:x, and the nitrogen atomic fraction was computed as x/(1+1.5+x). The sheet resistance of Ta-Si-N deposited on semi-insulating SiC using TaSi a31 is shown in figure 5.3(a) as a function of nitrogen content and time of annealing in air at a64a61a65a62a50 a0a54a1 . The sheet resistance increases with increasing nitrogen content in the film. For any particular amount of nitrogen however, the sheet resistance is very stable with anneal time up to 5000 hours. However, it became more difficult to measure the sheet resistance after 2500 hours because of a thin surface oxide layer as indicated by a gradual color change of the surface from metallic gray to light gold. Figures 5.3(b) and (c) show the RBS spectra used to monitor the oxidation kinetics of the Ta-Si-N films. In these spectra, oxygen incorporated during film deposition was observed throughout the films; however, oxygen in the surface oxide layer increases with annealing time. 130 Figure 5.2: (a) RBS spectra of SiC/Ta-Si/Ta-Si-N with 0, 2, 5, 10% nitrogen content by flow rate (b) van der Pauw sheet resistivity of Ta-Si-N from TaSi a31 and Taa33 Sia34 targets (c) Nitrogen aomic percentage against nitrogen flow 131 Figure 5.3: (a) Sheet resistance of Ta-Si-N on semi-insulating SiC against anneal time in air at 350a0a2a1 (b) RBS spectra of Ta-Si-N on semi-insulating SiC for 0 and 5000hrs (b) 0% and (c) 2% nitrogen flow 132 5.1.2 Structure of Ta-Si-N X-ray diffraction scans in figure 5.4 show that the Ta-Si-N films with different nitrogen content were amorphous (or possibly nanocrystalline) as deposited. The XRD scans also show that the films are more amorphous as the nitrogen content increases. There are no observable changes in the XRD scans after 750 hours of annealing in air or in an evacuated quartz tube at a64a66a65a30a50 a0 a1 . Annealing was performed to determine whether the films would become polycrystalline. It is known that amorphous materials become polycrystalline with time at elevated temperature [112]. Since amorphous layers are generally good diffusion barriers because of the absence of grain boundaries, the observed amorphous structure and its stability is a plus for Ta-Si-N. 5.2 Results on Ohmic Contacts 5.2.1 Contacts Without Protective Metallization TLM results on ohmic contacts ona78a83a3 -4H-SiC and a233a28a3 -4H-SiC are presented in this section for samples without protective metallization. These samples were not annealed in air after fab- rication. The results are for as-fabricated ohmic contacts but with different contact metal, metal thicknesses and contact anneal conditions. Total contact-to-contact resistance was measured on samples annealed between a60a62a50a61a50a32a39a124a46a2a50a61a50a62a50 a0 a1 for 1 - 2 minutes. Typical plots of total resistance versus inter-contact spacing is shown in figure 5.5. Using the equation, a40a72a119a155a24 a102 a224 a71 a145 a40 a11a43a42 a231 a158 a40 a11a43a42a69a102 a231 (5.1) the specific contact resistance and sheet resistace of the implanted or epitaxial samples were 133 Figure 5.4: XRD scan of Ta-Si-N with 0, 2, 5, and 10% nitrogen flow (a) as-sputtered samples (b) samples annealed in air at 350a0 a1 for 750hrs (c) samples annealed in evacuated quartz tube at 350a0a2a1 for 750hrs 134 Figure 5.5: Total contact-to-contact resistance against inter-contact spacing for (a) nickel ohmic contact on pa3 implanted material (b) nickel-chromium (80-20 wt%) ohmic contact on na3 epitaxial material 135 Target 4HSiC t(nm) Anneal condition a198a61a199a176a200a57a201a203a202a108a204a104a205a207a206a176a208 a209a169a210a60a211a71a200a57a201a184a212a20a213a115a214a20a208 M-a209a215a210a159a211a86a200a57a201a184a212a20a213a176a214a20a208 a216a144a217a147a218a98a219a114a220a222a221a64a223a98a219 a224a166a225 -impl. 240 a226a176a227a20a227a20a227a196a228a55a229 /2min/Vac a226a65a230a231a226a115a227a81a232a233a226a176a227a69a234 a223 a226a65a230a235a226a176a227a81a232a233a226a115a227 a223 16.4/1.8 a216a144a217a147a218a98a219a114a220a222a221a64a223a98a219 a224 a225 -impl. 110 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a236a191a230a238a237a65a239a12a232a233a226a176a227 a234a71a240 a237a191a230 a241a18a236a12a232a233a226a115a227a196a206 40.2/7.2 a216a144a217a147a218a98a219a114a220a222a221a64a223a98a219 a224a166a225 -epi. 100 a226a176a227a20a227a20a227a196a228a55a229 /2min/Vac a236a191a230a134a242a61a227a81a232a233a226a176a227a69a234a71a240 a226a65a230 a241a191a242a9a232a233a226a115a227 a223 67.8/13.5 a216a144a217a147a218a98a219a114a220a222a221a64a223a98a219 a224 a225 -epi. 190 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a241a118a230a238a243a45a227a81a232a233a226a176a227 a234 a223 a239a191a230 a227a18a242a9a232a233a226a115a227a196a206 15.7/1.9 a216a144a217a32a244a76a219a114a220a222a221a76a245a92a219 a224 a225 -impl. 300 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a243a191a230a231a226a115a227a81a232a233a226a176a227 a234a71a240 a226a65a230a48a246a196a242a9a232a233a226a115a227 a223 -/0.9 a247 a221 a202a248a242a65a249a81a250 a251 a225 -epi. 200 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a226a65a230a48a227a196a242a9a232a233a226a176a227 a234a71a240 43.6 11.9/2.0 a247 a221 a202a248a242a65a249a81a250 a224 a225 -impl. 140 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a252a191a230a238a236a65a236a12a232a233a226a176a227 a234a166a253 a237a191a230a48a239a20a239a12a232a233a226a115a227a196a206 52.1/4.4 a247 a221 a202a248a242a65a249a81a250 a224 a225 -impl. 100 a237a65a227a20a227 a228 a229 /1min/Vac a239a191a230a134a242a45a237a12a232a233a226a176a227 a234a166a253 a226a65a230 a227a196a237a12a232a233a226a115a227 a223 43.5/5.4 a247 a221 a202a248a242a65a249a81a250 a251 a225 -epi. 100 a237a20a239a65a227 a228 a229 /2min/Vac a226a65a230a238a237a65a239a12a232a233a226a176a227 a234a166a253 42.6 - a247 a221a92a254a98a219 a229a215a198 a206 a219 a251 a225 -epi. 50 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a252a191a230a231a226a115a227a81a232a233a226a176a227 a234a166a253 34.3 -/13.6 a247 a221a92a254a98a219 a229a215a198 a206 a219 a251 a225 -epi. 100 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a246a191a230a238a237a45a227a81a232a233a226a176a227 a234a84a255 53.6 -/5.2 a239a65a227a20a227 a228 a229 /120min/a0 a206 a242a18a230a48a227a65a227a81a232a233a226a176a227 a234a84a255 53.6 5.2/4.2 a247 a221a92a254a98a219 a229a215a198 a206 a219 a224 a225 -impl. 130 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a226a65a230a238a243a45a227a81a232a233a226a176a227 a234a71a240 a237a191a230a238a242a191a226a124a232a233a226a115a227a196a206 48.9/5.9 a247 a221a92a254a98a219 a229a215a198 a206 a219 a224 a225 -impl. 100 a237a65a227a20a227 a228 a229 /1min/Vac a243a191a230a48a227a20a237a12a232a233a226a176a227 a234a71a240 a226a65a230a48a243a20a236a12a232a233a226a115a227 a223 60.3/7.4 a247 a221 a206 a219 a229a215a198 a254a76a219 a224 a225 -impl. 90 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a243a191a230a48a227a65a227a81a232a233a226a176a227 a234a71a240 a237a191a230a48a246a20a243a12a232a233a226a115a227a196a206 - a247 a221 a206 a219 a229a215a198 a254a76a219 a251 a225 -epi. 100 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a236a191a230a231a226a115a227a81a232a233a226a176a227 a234a71a240 57.3 -/10.0 a247 a221 a255 a219 a229a215a198 a240 a219 a251 a225 -epi. 100 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a237a191a230a238a239a45a227a81a232a233a226a176a227 a234a166a253 41.2 -/3.7 a247 a221 a255 a219 a229a215a198 a240 a219 a251 a225 -epi. 50 a226a176a227a20a227a20a227 a228 a229 /2min/Vac a241a118a230a238a252a45a227a81a232a233a226a176a227 a234a166a253 39.6 -/8.3 Al1% Si/ a224 a225 -impl. 10/40 a237a20a243a20a239 a228 a229 /2min/Vac a226a65a230a238a243a45a227a81a232a233a226a176a227 a234a71a240 a226a65a230a235a226a55a243a12a232a233a226a115a227 a223 -/1.9 a247 a221 a202a248a242a65a249a81a250 Ti/Al1% Si/ a224 a225 -impl. 1/1/4 a237a20a243a20a239 a228 a229 /2min/Vac a226a65a230a238a239a191a226a124a232a233a226a176a227 a234a71a240 a226a65a230 a227a196a246a12a232a233a226a115a227 a223 - a247 a221 a202a248a242a65a249a81a250 Table 5.1: Various ohmic contacts fabricated in this study. The metal in contact with the SiC is on the left for the case of two or more layers. The metal sheet resistance (a38 a39a155a40 a11a43a42 ) column has values before/after contact annealing. t is the contact metal thickness. determined. Table 5.1 shows the list of various ohmic contacts that were fabricated during the course of this study. RBS spectra of the as-deposited and annealed nickel contacts are shown in figure 5.6. These plots indicated the reaction of Ni with SiC at high temperature (a8a136a60a61a50a62a50 a0a2a1 ) to form nickel silicide which is ohmic on SiC. Nickel silicide sputter-deposited on SiC from a silicide target does not form an ohmic contact to SiC. A high temperature reaction is necessary, as reported previously [5]. 136 Figure 5.6: Experimental and RUMP simulated RBS spectra of (a) as-deposited nickel contact on SiC (b) nickel ohmic annealed on SiC 137 5.2.2 Ohmic Contact with Protective Metallization Ohmic contact on a78a12a3 -4HSiC The first sets of ohmic contacts on a78a83a3 4H-SiC were formed from 70-30wt% Al-Ti and 93- 7wt% Ni-V. The protective stack was Ta-Si-N/Pt-N and the cap layer was Au/Sn/Au. Tin was sandwiched between Au layers to simulate the 80-20wt% Au-Sn preform that was later used for chip brazing. The percentages of N a31 flow in the sputter gas was 0, 1, 1.5 and 2. These percentages are the fractions of N a31 flow rate to the total flow rate (Ar plus N a31 ) during sputter- deposition. TLM results for the as-fabricated ohmic contacts with protective metallization stacks were analyzed to obtain the specific contact resistance and sheet resistance of the contacts. At certain times during the anneals of in air, the measurements were repeated. Figure 5.7 shows the plot of specific contact resistance and SiC sheet resistance as a function of annealing time up to 4000 hours for 0 and 2% nitrogen content in the stack. The specific contact resistance and the sheet resistance for nickel contact were stable after 4000 hours of annealing. However, the Al-Ti contact without nitrogen in the protective stack showed significant changes in the specific contact resistance and sheet resistance values after about 2500 hour of anealing. The variations most probably mark the begining of the failure of Al-Ti contact due to the oxidation of Al. The Al-Ti contact with 2% nitrogen content in the protective stack was still very stable after 4000 hours of annealing. The observed electrical stability of the ohmic contacts testifies to the effectiveness of the protective stack on the p-type ohmic contact. Mixing/diffusion and oxidation of the composite ohmic multi-layered metallization stack were monitored with RBS and AES. Little oxidation and inter-diffusion were observed in the RBS spectra of figure 5.8(a) for the broad area Ni contacts. However, mixing is apparent in 138 Figure 5.7: Specific contact resistance and SiC sheet resistance against annealing time for (a,c) SiC/Ni(ohmic)/Ta-Si-N/Pt-N/Au/Sn/Au (b,d) SiC/Al-Ti(ohmic)/Ta-Si-N/Pt-N/Au/Sn/Au 139 Figure 5.8: RBS spectra of p-ohmic contact after 24, 1500 and 4000hrs annealing (a) nickel ohmic contact with 2% nitrogen in the stack (b) Al-Ti ohmic contact with 2% nitrogen in the stack. 140 the Al-Ti contact after 4000hrs. AES data shows better resolution of the very light elements. Figure 5.9(a) is for an unannealed, as-deposited nickel contact. Oxygen at the nickel surface was observed, with less oxygen throughout the nickel and Ta-Si-N layers. The surface oxygen was picked-up when the sputter system was opened to change targets. Oxygen is also observed in the ohmic annealed nickel contact [figure 5.9(c)], but does not affect the electrical characteristics of the contacts adversely. During ohmic contact annealing, Ni reacts with Si in SiC leaving a carbon rich surface between SiC and the resulting Ni-Si contact. The opposite surface of the Ni-Si contact is also rich in C, and surface oxygen is clearly visible as shown in figures 5.9(b) and (d). With or without nitrogen in the stack, the cap layers and the Pt layer do not diffuse through Ta-Si layer. Tantalum silicide effectively prevents the diffusion of Pt and Au, though limited mixing is observed at the Pt-Au and Pt-TaSiN interfaces after 1500 hours. The Sn layer sand- wiched between Au layers diffuse into the Au layers with increasing annealing time. Figures 5.10 shows that carbon diffuses more readily through the Ta-Si barrier layer if there is no nitrogen in the layer. 2% (by flow rate) nitrogen content in Ta-Si-N reduces the rate of C diffusion appreciably. Sputter-deposition of the tantalum silicide target in an Ar/a73 a31 gas mixture improves the barrier property of Ta-Si-N significantly. It is also true that after a 3000hr anneal in air the amount of a85 a31 in the stack does not increase significantly above the level observed in the as-deposited samples. As a result the electrical properties of the composite contacts remain stable after 4000hr of annealing in air at a64a66a65a62a50 a0a2a1 . Ohmic contact on na3 -4HSiC Ohmic contact on a233 a3 -epilayers were first fabricated with 80-20wt% Ni-Cr. The specific contact resistances with or without nitrogen content in the protective stack (Ta-Si-N/Pt-N/Au/Sn/Au) 141 Figure 5.9: AES spectra for (a) as-deposited nickel contact with no nitrogen in the metallization stack (b) as-deposited nickel with 2% nitrogen in the stack (c) nickel ohmic contact with no nitrogen in the stack (d) nickel ohmic with 2% nitrogen in the stack 142 Figure 5.10: AES spectra of nickel ohmic contacts annealed in air at 350a0a2a1 for 1500hrs (a) nickel ohmic contacts with no nitrogen in the stack (b) nickel ohmic contacts with 2% nitrogen in the stack 143 were very low prior to annealing the samples in air at a64a66a65a30a50 a0 a1 . For contacts without nitrogen con- tent and with 1% (by flow rate) of N a31 in the stack, it was impossible to accurately measure the total contact-to-contact resistance after 24 hours of annealing in air at a64a66a65a30a50 a0a2a1 . For stacks with 1.5% and 2% N a31 rate, the contact resistance increases after 24 hours of annealing by about a factor of four. After the initial increase, the specific contact resistance remains almost constant for up to 4000 hours of annealing. Figure 5.11 shows the plots of specific contact resistance and sheet resistance against the time of annealing. The reason for the sudden increase in contact resistance is not apparent from the RBS data where no appreciable oxidation or diffusion was ob- served. AES spectra, however, showed that there is high diffusion of carbon into the barrier layer as the time of annealing increases, so the increase in resistance may be due to the increase of carbon content and its accumulation at the interface between the ohmic contact and the Ta-Si-N layer. Figure 5.12 shows RBS and AES plots. 5.2.3 Metallization with Hot Platinum Adhesion problems between Ta-Si-N and Pt-N in the metallization stack were identified during the brazing of the stack to a ceramic carrier. To overcome this problem, Pt was sputter- deposited on SiC/.../Ta-Si-N kept at a102a62a65a62a50 a0a35a1 during the period of Pt deposition. This modification increases adhesion between Pt and Ta-Si-N dramatically. The process of ohmic contact fabrica- tion with the protective stack was subsequently modified because the photoresist normally used to pattern the metallization will carbonize at a102a61a65a30a50 a0 a1 and become difficult to remove. Molybde- num was found to be suitable for lifting-off hot Pt in H a31 O a31 . An acid etch back of the hot Pt also worked well, except that very small features of TLM were difficult to etch. 144 Figure 5.11: (a) Plots of specific contact resistance against annealing time for Ni-Cr (80-20 wt%) ohmic contacts with 2% nitrogen in the stack (b) Plots of SiC sheet resistance against annealing time for Ni-Cr (80-20 wt%) ohmic contacts with 2% nitrogen in the stack 145 Figure 5.12: (a) RBS spectra of Ni-Cr (80-20 wt%) ohmic contacts after 24 and 4000hrs anneal- ing; AES spectra of Ni-Cr (80-20 wt%) ohmic contacts with 2% nitrogen in the stack (b) before annealing in air (c) afer annealing in air for 1500hrs 146 All nickel ohmic contact with hot Pt metallization Because of the success with nickel ohmic contacts on a78a83a3 -implanted 4H-SiC, the decision was made to use nickel ohmic contacts on both a78a83a3 and a233a28a3 material with Pt in the protective metallization stack sputtered on hot substrates. With this modified metallization scheme, TLM measurement on nickel ohmic contacts to p- and n-SiC were carried out. Nickel ohmic con- tacts on a78 a3 -implanted 4H-SiC was very similar to those obtained before with cold Pt in the metallization stack. This implied that sputtering Pt at a102a61a65a30a50 a0 a1 does not affect the contact resis- tance significantly (Figure 5.13). For the ohmic contact on a233a28a3 -epilayers, the contact resistance increases generally with annealing in air at a64a66a65a30a50 a0 a1 . However, the total contact-to-contact re- sistance for several pads increased sharply out of the normal range (a45a124a19 a46a2a50a61a23 ) and the direct proportionality of the resistance to inter-contact spacing was lost after a few hundreds of hours annealing. It was difficult to determine the problem with the RBS (figure 5.14). Both RBS and AES did not indicate oxidation of the contacts as the cause of the problem. Plots of total contact-to-contact resistance, specific contact resistance and SiC sheet resis- tance as a function of annealing time for nickel contacts on p- and n-SiC are shown in figures 5.15 and 5.16. Contact resistance at elevated temperatures TLM results at temperatures up to a64a66a65a30a50 a0a2a1 for implanted samples and the corresponding spe- cific contact resistance and sheet resistance are shown in figure 5.17. Both a71 a145 and a40 a11a43a42 approache minimum values at elevated temperatures. The contact properties measured at room temperature were reproduced after measurements at elevated temperatures. 147 Figure 5.13: Total contact-to-contact resistance of (a) cold and hot Pt in the metallization stack with no nitrogen content (b) cold and hot Pt in the metallization stack with 2% nitrogen content (c) nickel ohmic contact on na3 epilayer SiC with hot Pt and 2% nitrogen in the metallization stack after different anneal times in air 148 Figure 5.14: RBS spectrum of a nickel ohmic contact on na3 epitaxial 4H-SiC. 149 Figure 5.15: Specific contact resistance and SiC sheet resistance for nickel ohmic contacts on na3 epitaxial 4H-SiC with hot Pt and 2% nitrogen content in the metallization stack. 150 Figure 5.16: Specific contact resistance and SiC sheet resistance of nickel ohmic contacts on pa3 implanted 4H-SiC with hot Pt in the metallization stack. 151 Figure 5.17: Plots of (a) total contact-to-contact resistance against inter-contact spacing (b) Specific contact resistance against temperature and (c) SiC sheet resistance against temperature for nickel ohmic contact on pa3 implanted 4H-SiC with 2% nitrogen in the metallization stack. 152 5.3 Results for Schottky Contacts 5.3.1 I-V and C-V Results The metal contact on the first set of Schottky diodes were sintered nickel silicide (Ni a31 Si) with 0 - 2% nitrogen (by flow rate) in the stack metallization. A significant decrease in the number of good diodes with 0% nitrogen content was observed after only 24 hrs annealing. After 1500 hrs anneal, almost all devices with 0% and 1% nitrogen content were bad, and less than 50% of the diodes with 2% nitrogen in the stack were still functional. The first indication of problems with the diodes was the drop of 3 to 6 orders of magnitude in current at 2V forward bias. Because the structure of the diodes was such that the protective stack does not cover the whole of the silicide contact, it is believed that at a64a66a65a30a50 a0a2a1 the exposed silicides at the circumference of the diode oxidized and that the oxide grew laterally under the protective stack, thus causing an increase in the resistance of the contact metal. The barrier height or turn-on voltage of the devices incresed to more than double the initial value after 24 hrs anneal, but subsequently remained stable subsequently as shown in figure 5.18. The ideality factor also shows that the diodes behave more ideally after stabilizing. However, very few devices survived the 4000 hour anneal. New Schottky devices The new sets of Schottky devices were fabricated so that the protective stack overlayed the contact layer and a portion of the passivating oxide layer around devices. Two sets of diodes were fabricated, devices with unsintered nickel silicide contacts and devices with Ta-Si-N used as the contact metal. 0, 2, 5 and 10% by nitrogen flow rate was used in the stacks. The Pt in the metallization was sputter-deposited hot and etched-back in an acid solution. 153 Figure 5.18: Nickel silicide Schottky devices (a) forward J-V dependence on annealing time (b) reverse J-V dependence on annealing time (c) barrier height and ideality factor against annealing time. 154 The yeild and distribution of the devices according to the leakage current at -10V reverse bias are shown on bar charts in figure 5.19. The Ta-Si-N Schottky have higher leakage than the nickel silicide Schottky, but Ta-Si-N devices are more stable as a function of annealing time. The purpose of the Schottky device was to evaluate the effectiveness of the protective stack in air at a64a66a65a62a50 a0a2a1 , since Schottky diodes are more sensitive to whatever is happening at the metal- semiconductor interface. The current density - voltage characteristics of the nickel silicide Schot- tky devices are shown in figures 5.20 and 5.21, while figures 5.22 and 5.23 show similar result for the Ta-Si-N devices. The forward characteristics up to 2V and the reverse characteristics out to -300V are shown for various nitrogen percentages. Very slight shifts to higher barrier height were generally observed for all devices, but more significantly for devices with 0% nitrogen con- tent. Theoretical fits to the forward characteristics using the thermionic emission equations were plotted with the experimental data to monitor the barrier height and ideality factor as a function of anneal time. The barrier heights were generally stable with time of annealing for both set of devices. The ideality factors of the diodes also remain stable and near unity with annealing time. These results implied that the stack layer effectively protects the contacts for high temperature exposures in air. Figure 5.22 show plots of barrier height and ideality factors as a function of annealing time for devices with different nitrogen content in the stack metallization. Capacitance measurements were made with the diodes under reverse bias to -4V, in order to extract barrier parameters. Plots of a90a91a70a215a27 a1 a93a133a31 against the magnitude of reverse potential are shown in figures 5.25 and 5.26 for various annealing times and for different nitrogen content in the stack. The linearity of a90a91a70a215a27 a1 a93 a31 indicated uniform doping concentration and the fact that the mi- nority carrier concentration in these devices are negligible. More than 70% of the devices had linear characteristics that remained so after 1500 hr anneals in air. The barrier heights were 155 Figure 5.19: Bar chart of current density at -10V reverse bias for nickel silicide and Ta-Si-N Schottky devices (a) 0hr nickel silicide contacts (b) 0hr Ta-Si-N contacts (c) 1500hr nickel silicide contacts (d) 1500hr Ta-Si-N contacts 156 Figure 5.20: Nickel silicide Schottky devices (a) forward characteristics with 0% nitrogen (b) reverse characteristics with 0% nitrogen (c) forward characteristics with 2% nitrogen (d) re- verse characteristics with 2% nitrogen. 157 Figure 5.21: (e) Forward characteristics with 5% nitrogen (f) Reverse characteristics with 5% nitrogen (g) Forward characteristics with 10% nitrogen (h) Reverse characteristics with 10% nitrogen. 158 Figure 5.22: Ta-Si-N Schottky devices (a) forward characteristics with 0% nitrogen (b) re- verse characteristics with 0% nitrogen (c) forward characteristics with 2% nitrogen (d) reverse characteristics with 2% nitrogen. 159 Figure 5.23: (e) Forward characteristics with 5% nitrogen (f) reverse characteristics with 5% nitrogen (g) forward characteristics with 10% nitrogen (h) reverse characteristics with 10% nitrogen. 160 Figure 5.24: Nickel silicide Schottky devices (a) barrier height against annealing time (b) ideal- ity factor against annealing time. Ta-Si-N Schottky devices (c) barrier height against annealing time (d) ideality factor against annealing time. 161 Figure 5.25: Plots of (A/C)a31 against applied bias for nickel silicide diodes with (a) 0% nitrogen (b) 2% nitrogen (c) 5% nitrogen (d) 10% nitrogen content in the metallization stack. 162 Figure 5.26: Plots of (A/C)a31 against applied bias for Ta-Si-N diodes with (a) 0% nitrogen (b) 2% nitrogen (c) 5% nitrogen (d) 10% nitrogen content in the metallization stack. 163 calculated from the intercept on the potential axis using the equation, a5 a143 a24a21a13 a108 a158 a252 a138 a175 a90a166a46a109a158 a56a32a31a198a90 a73 a145 a73a135a128 a93a166a93 (5.2) where for 4H-SiC, a73 a145 a177 a137a15a87a225a63a66a102a75a47a52a46a2a50 a97 a33 a38a248a90 a57 a145 a27 a57 a139a35a93 a34 a213 a31 a138 a34 a213 a31 a56a54a57 a4 a34 , a57 a145 a24a184a50a79a87a225a64a66a101 a57 a139 , the number of equivalent conduction band minima is M = 3. Then for T = 300K, a73 a145 a177 a46a30a87a100a69a62a60a116a47a52a46a35a50 a97a132a106 a56a58a57 a4 a34 , and with the measured doping concentration of about a102a204a87a88a65a75a47a52a46a2a50 a97 a51 a56a54a57 a4 a34 , a103 a145 a39a115a103a161a16a183a24a202a90 a252 a138a161a27a85a175a61a93a2a90a57a56a32a31a181a90a160a73 a145 a27a2a73a48a128a161a93 a177 a50a32a87a234a46a153a101a30a82a85a13 and a5 a143 a177 a13 a108 a158a55a50a32a87a88a102a62a50 Figure 5.24 also shows the plot of the C-V barrier height versus the annealing time. The stability of the barrier height with annealing time supports the observed J-V characteristics though the barrier heights from the C-V measurements are higher compared to those from the J-V measure- ment. C-V data was obtained with the devices in depletion under reverse bias, while for the J-V characteristics, the barrier height was determined with the devices forward biased. RBS and AES spectra for the Schottky contacts are shown in figure 5.27. Neither signifi- cant inter-diffusion nor oxidation of the contacts was observed. However, moverment of carbon within the stack was observed from the AES data. 5.4 Wirebonding and Brazing Results 5.4.1 Chip Shear Testing Adhesion within the metallization stack as well as adhesion of the stack to thermal SiO a31 (indirect bonding) and to ohmic and Schottky devices (direct bonding) are essential parts of this study. Indications of adhesion problems were first noticed while brazing the Ta-Si-N/Pt- N/Au/Sn/Au stack to a thermal oxide. Though Ta-Si-N has good adhesion to SiO a31 , the adhesion 164 Figure 5.27: Nickel silicide Schottky contacts with 2% nitrogen in the metallization stack (a) RBS spectra (b) AES spectra after 1500hr anneal in air. 165 Figure 5.28: Chip shear strength with cold Pt in the metallization stack against annealing time. of the brazed a65 a57a227a57 a47a115a65 a57a227a57 samples on their ceramic holders degraded significantly in air at a64a66a65a30a50 a0 a1 . Figure 5.28 shows the initial brazing results that indicates that high nitrogen content in the stack may be detrimental to adhesion. RBS spectra of a sheared sample showed the interface between Ta-Si-N and Pt is the weak- est link in the metallization stack. This adhesion was significantly increased when Pt was sputter- deposited on SiC/SiO a31 /Ta-Si-N at a102a61a65a62a50 a0a35a1 . With hot Pt, adhesion was very good, and the stack did not fail under maximum shear force (100 kg). The adhesion remain excellent with annealing time up to 1000hrs. Samples that passed the shear testing were subjected to additional shear testing after further annealing in air. The stress incurred during the first shear testing seems to heal with further annealing. 166 Figure 5.29: Chip shear strength with hot Pt and 2% nitrogen in the metallization stack against annealing time. 167 5.4.2 Wirebond Pull and Shear Testing Bond pull and shear testing were carried out on the following metallization structures: SiC/SiO a31 /Ta-Si-N/Pt-N/Au (indirect wirebonding), SiC/Ni-Si(ohmic)/Ta-Si-N/Pt-N/Au, SiC/NiSi(Schottky)/Ta-Si-N/Pt-N/Au. Metallizations with hot Pt on thermal SiO a31 and NiSi Schottkys held up very well. 100% of the time, it was the Au wire that broke while doing pull testing, and for the shear testing, the separation of the wirebond from the samples was above the Au cap layer on the samples. Figure 5.30(a) shows the bond pull and shear strengths against the time of annealing for the samples. The breaking mode of the Au wire during pull testing and shear testing are shown in figure 5.30(c). Wirebond pull and shear tests results were not as good for hot Pt stacks deposited directly on annealed nickel ohmic contacts. Many bonds failed at the interface between the ohmic contact and the Ta-Si-N layer. The reason for the failure was attributed to carbon liberated during the contact anneal and left on the surface of the nickel silicide layer. The RBS and AES results in figures 5.31 and 5.32 show that a thin layer of carbon was left on the silicide surface for a starting nickel thickness of about 120nm. The carbon was primarily surface carbon, but there is considerable carbon thoughout the nickel silicide layer as well. Figures 5.31 and 5.32 also show that the surface carbon can be removed through either oxygen plasma etching or by ion milling with a low energy Ar beam (1keV, 5-8 minutes). The ratio of nickel to carbon at the surface of as-annealed contacts was about 1:2. On Ar ion cleaned samples the ratio was about 5:4 and on the oxygen etched samples it was about 3:1. The result of cleaning showed that both cleaning procedures remove surface carbon; how- ever, the Ar ion clean also removed a thin layer of left over nickel during incomplete ohmic 168 Figure 5.30: Wirebond pull and shear strengths against annealing time for (a) metallization with 2% nitrogen and hot Pt on a thermal oxide (b) metallization stack with 2% nitrogen and hot Pt on a nickel silicide Schottky contact (c) Au wire failure mode and sheared wedged bond. 169 Figure 5.31: RBS and AES results for nickel ohmic contacts (a) RBS of ohmic annealed contacts (b) AES of ohmic annealed contacts (c) RBS of RIE cleaned ohmic contacts (d) AES of RIE cleaned ohmic contacts 170 Figure 5.32: RBS and AES results for nickel ohmic contacts (a) RBS of Ar ion cleaned ohmic contacts (b) AES of Ar ion cleaned ohmic contacts. 171 contact annealing. Oxygen RIE could not remove the nickel which is sometimes used as pro- tective mask during oxygen RIE. The result of wirebond pull and shear testing showed that RIE sample did not show much improvement in adhesion because of the unreacted nickel. However, for the samples cleaned using Ar, where both carbon and nickel surface layers were removed, there is significant improvement in the adhesion as shown in figure 5.33. At 1000hr annealing, the adhesion is still good. Pictures of failed metallization on nickel ohmic contacts are included in figure 5.33. Direct wirebond pull and shear test results on annealed 80-20 wt% Ni-Cr or as- deposited Ni with hot Pt were also not good as a result of the same problems described for the annealed nickel contacts. 172 Figure 5.33: (a) wirebond pull and shear strength of Ar ion cleaned nickel ohmic contacts (b) picture of failed wirebond and (c) picture of good pull and shear bonding. 173 CHAPTER 6 CONCLUSIONS Various ohmic contacts were fabricated in this study, and fabrication conditions were op- timized to obtain very low specific contact resistance. Low contact resistance is essential to minimize power losses in high voltage/high current devices. Specific contact resistances in the range a46a30a87a234a46a215a47a183a46a35a50 a4 a34 to a69a79a87a225a60a131a47a52a46a2a50 a4a7a51 a23a187a39 a56a58a57 a31 were obtained. RBS results showed that of four silicides studied for use as diffusion/oxidation barriers, tantalum silicide sputter-deposited from TaSi a31 was more resistant to oxidation and more effective as a diffusion barrier compared to Taa33 Sia34 , molybdenum silicide and tungsten silicide. Even with Pt cap layers, Mo-Si and W-Si oxidize because of oxygen diffusion through the cap layers. Tantalum silicide sputter-deposited from TaSi a31 in an Ar/N a31 gas mixture has lower sheet resistance compared to tantalum silicide from Taa33 Sia34 . The sheet resistance of Ta-Si-N increases with increasing nitrogen content in the films. Ta-Si-N films with 0, 2, 5 and 10% by nitrogen flow rate in the sputter-gas were amorphous as-deposited, and they remain amorphous after annealing for 750 hours in air or vacuum at 350a0a2a1 . It was observed from XRD that increasing nitrogen content in the films made them more amorphous. The Ta-Si-N films are resistant to oxidation and amorphous with good diffusion barrier characteristics, so that they are suitable for the job of protecting ohmic and Schottky contacts operating at elevated temperature in air. The protective metallization stack, Ta-Si-N/Pt-N/Au, was found to be very effective on ohmic or Schottky contacts annealed in air at 350a0a54a1 . Physical characterization (AES and RBS) showed that the protective metallization stack prevents rapid oxidation of the contacts and min- imizes inter-diffusion/mixing within the contact/stack combination. RBS data after 4000 hour 174 anneals in air at 350a0 a1 and AES data after 1500 hours were analyzed for signs of oxidation and diffusion. For ohmic contacts however, carbon generated during the high temperature ohmic contact anneal is very mobile, and it diffuses through the barrier layers, especially those with zero ni- trogen content. Barrier layers with 2% nitrogen are much more resistant to carbon diffusion. For pa3 implanted 4H-SiC, the movement or accumulation of carbon does not affect the elec- trical characteristics (specific contact resistance and SiC sheet resistance) significantly. But for na3 epitaxial SiC, the total contact-to-contact resistance between TLM adjacent pads increased significantly with increasing anneal time in air. The conclusion was that carbon accumula- tion/diffusion increased the resistance of the contact. The effect noticeable because initially, the contact-to-contact resistances for the n-epilayers were much lower than for the pa3 implanted material. For Schottky devices, J-V characteristics and C-V characteristics were largely stable with annealing time. Ta-Si-N Schottky devices had higher reverse leakage currents than the nickel silicide Schottky devices. Barrier heights and ideality factors were determined from the J-V and C-V measurements by fitting the experimental data to theoretical transport equations. The barrier height from the J-V data were lower compared to the C-V data, and the barrier heights of Ta-Si-N Schottky devices were lower than those of the nickel silicide devices. This accounts for the higher leakage currents of the Ta-Si-N devices. Increasing nitrogen content in Ta-Si-N Schottky contact improves the stability of the barrier height and ideality factor with annealing time in air. Without the protective stack, oxidation of nickel silicide Schottky contact and Au diffusion into the contact was obvious after just 24 hours of annealing in air at 350a0a2a1 . Adhesion tests carried out on the samples resulted in the following observations. 1. Initial die attach shear testing showed degraded adhesion with annealing time in air and with 175 increased nitrogen content in the metallization stack. 2. Sputter-deposition of Pt at 250a0 a1 increased adhesion, with die shear strengths above the max- imum possible of 100kg, for up to 1000 hour annealing in air. 3. Hot Pt metallization stacks on SiO a31 (for the purpose of indirect wirebonding) were wire- bonded with 10mil Au wire. Wirebond pull and shear testing showed that the adhesion with the stack was excellent, with no degradation for up to 2000 hour of annealing and testing. 4. Direct wirebonding on nickel silicide Schottky contacts with hot Pt stacks showed that ad- hesion was very good. Bond pull and shear test results were very stable for anneal times up to 1000 hours. 5. Direct wirebonding on annealed nickel ohmic contacts with hot Pt stacks resulted in poor adhesion. However, Ar ion cleaning the samples just after the ohmic contact anneal improved adhesion dramatically and produced bond pull and shear results that are very stable for 1000 hour anneals (longer anneals are currently underway). It was concluded that the problem of ad- hesion on the as-annealed ohmic contacts was due to the surface carbon and unreacted surface nickel left after the contact anneal. More detailed studies of the differences between ohmic contact characteristics for implanted and epitaxial layers is essential. Also, further study of the ability of the Ta-Si-N barrier layer to prevent diffusion and oxidation at higher temperatures, say 400 - 500a0 a1 , would be informative. Modification of the ohmic contact anneal process by the introduction of a protective layer on the contact metal during the anneal or by the addition of a sticking layer between the ohmic contact and the barrier layer may prevent the need for Ar ion cleaning of the contacts before stack metallization. These options need to be investigated. This study has been based on the application of a comprehensive set of characterization techniques - electrical (I-V, C-V, LTLM), physical (RBS, AES) and mechanical (wirebond pull 176 and shear testing). Based on the application of these techniques for the contact metallizations studied, the following recommendations can be made for the formation of composite, high tem- perature contacts to 4H-SiC: a44 A protective stack, Ta-Si-N/Pt-N/Au, with 2% nitrogen content and Pt sputter-deposited at 250a0 a1 , is an effective diffusion/oxidation barrier for ohmic and Schottky contacts op- erating in air at 350a0 a1 . a44 Sputter-deposited nickel ( a45 100nm thick) annealed at 900a0a2a1 for 1 minute in vacuum makes an acceptable ohmic contact for heavily doped p-4H-SiC (a73 a251 a59a136a64a131a47a52a46a2a50 a97a132a106 a56a54a57 a4 a34 ). a44 Sputter-deposited 80-20 wt% Ni-Cr ( a45 100nm thick) annealed at 1000 a0 a1 for 2 minutes in vacuum can be used for ohmic contacts on na3 epitaxial SiC. a44 Sputter-deposited nickel silicide forms a stable Schottky contacts on n a4 epitaxial SiC. a44 Nickel silicide Schottky contacts with the protective stack can be wirebonded directly (on silicide) or indirectly (on SiO a31 ) a44 Argon ion cleaned nickel and nickel-chromium ohmic contacts with the protective stack can be wirebonded directly. Without ion cleaning, indirect wirebonding is necessary. 177 BIBLIOGRAPHY 178 [1] Materials for high temperature semiconductor devices, National Research Control Report NMAB-747 (National Academy Press, Washington, DC, 1995). [2] J.R. Williams and R.W. Johnson, Contact Metallization and Packaging Technology Devel- opment for SiC Bipolar Junction Transistors, PiN Diodes and Schottky Diodes Designed for Long-term Operation at a64a61a65a62a50 a0a2a1 (Proposal submitted to Wright-Patterson Air Force Base, AFRL, 2001). [3] G.L. Harris (ed.), EMIS Data Reviews Series, No 13, (Short Run Press, Exeter, England, 1995). [4] T. Uemoto, Jpn. J. Appl. Phys. 34 L7 (1995). [5] J. Crofton, P.G. McMullin, J.R. Williams and M.J. Bozack, J. Appl. Phys. 77 (3) 1317 (1995). [6] S. Liu, K. Reinhardt, J. Scofield and C. Severt, Workshop on High Temperature Power Electronics for Vehicles (April 1995). [7] T. Nakata, K. Koga, Y. Matsushita, Y. Ueda and T. Niina, Amorphous and Crystalline SiC II (Springer Verlag, Berlin, 26 1989). [8] A. Suzuki, Y. Fujii, H. Saito, Y. Tajima and K. Furukawa, J. Cryst. Growth 115 623 (1991). [9] J. Crofton, L. Beyer, J.R. Williams, E.D. Luckowski, S.E. Mohney and J.W. DeLucca, Solid-State Electron. 41 (1) 1725 (1997). [10] N. Braslau, J. Vac. Sci. Technol. 19 803 (1981). [11] J. Crofton, S.E. Mohney, J.R. Williams and T. Isaacs-Smith, Solid-State Electron. 46 (1) 109-113 (2001). [12] J. Feitknecht, ?Silicon carbide as a semiconductor? Springer Tracts in Modern Physics G. H a1a72 hler (ed.), 58 48 (1971). [13] A.G. Acheson, Brit. Pat. 17911 (1892). [14] O.W. Lossew, Telegraphy and Telephony 18 61 (1923). [15] G. Busch, Helv. Phys. Acta 19 463 (1946). [16] http://www.ecn.purdue.edu/WBG/introduction/index.html [17] D. Nakamura, I. Gunjishima, S. Yamaguchi, T. Ito, A. Okamoto, H. Kondo, S. Onda and K. Takatori, ?ultrhigh-quality silicon carbide single crystal?, Nature 430 1009 (2004). [18] W.F. Knippenberg and G. Verspui, Mat. Res. Bull. 4 S.45 (1969). 179 [19] W.F. Knippenberg and G. Verspui, Mat. Res. Bull. 4 S.33 (1969). [20] V.A. Izhevskyi, L.A. Genova, J.C. Bressiani, A.H.A. Bressiani, ?Review article: SiC. structure, properties and processing?, Ceramica 46 (297) (2000) [21] T.N. Oder, ?Fabrication and Characterization of Ohmic Contact to P- and N-type SiC with applications to P-N Junction diodes?, Auburn University Doctoral thesis (1999). [22] R.S. Ramsdell, Am. Min. 32 64-82 (1947). [23] G.R. Zhadanov, Compte Rende Acad. Sci. URSS 48 39-42 (1945). [24] K. J a1a96 rrendahl and R.F. Davis, SiC Materials and Devices Y.S. Park (ed), 52 1-20 (1998). [25] P.G. Neudeck, Silicon Carbide Electronic Devices, in Encyclopedia of materials: Science and Technology K.H.J. Bushchow, R.W. Cahn, M.C. Flemings, B. Ilschner, E.J. Kramer and S. Mahajan, (eds) (Oxford: Elsevier Science) 9 8508-8519 (2001). [26] W.F. Knippenberg, Philips Res. Rept. 18 16 (1963). [27] P.T.B. Shaffer, Mat. Res. Bull. 4 S.13 (1969). [28] D.J. Larkin, P.G. Neudeck, J.A. Powell and L.G. Matus, ?5th Int. Conf. on SiC and related Matls.? (Washington, DC, USA, 1993); Inst. Phys. Conf. Ser. No 137 51 (1994). [29] M.S. Shur, SiC Materials and Devices Y.S. Park (ed), 52 161-193 (1998). [30] Y. Wang, W. Xie, Jr.J.A. Cooper, M.R. Melloch and J.W. Palmour, Silicon Carbide and Related Materials S. Nakashima, H. Matsunami, S. Yoshida, H. Harima (eds), (IOP Pub- lishing, Bristol, UK, 809-812, 1995). [31] D.M. Brown, E. Downey, J. Kretchmer, V. Krishnamurthy, W. Hennessy and G. Michon, Phys. Status Solidi (a) 162 459-479 (1997). [32] http://www.ecn.purdue/WBG/Device Research/Power Devices/ Index.html [33] J. Spitz, M.R. Melloch and J.A. Cooper, Jr., IEEE Device Research Conference (Ft. Collins, CO, June 23-25, 1997). [34] J. Spitz, M.R. Melloch, J.A. Cooper, Jr., and M.A. Capano, IEEE Electron Device Lett. 19 100 (1998). [35] F. Dahiquist, C.M. Zetterling, M. Ostling, and K. Rottner, Silicon Carbide, III-Nitrides, and Related Materials (Materials Research Forum 264-8, Trans Tech, Switzerland) G. Pensl, H. Morkoc, B. Monemar, E. Janzen (eds), 1061-1064 (1998). [36] R. Held, N. Kaminski, E. Niemann, Silicon Carbide, III-Nitrides, and Related Materials (Materials Research Forum 264-8, Trans Tech, Switzerland) G. Pensl, H. Morkoc, B. Monemar, E. Janzen (eds), 1057-1060 (1998). 180 [37] R.J. Trew, SiC Materials and Devices Y.S. Park (ed), 52 237-282 (1998). [38] http://www.ecn.purdue/WBG/Device Research/Microwave Devices/ Index.html [39] R.C. Clarke, A.K. Agarwal, R.R. Siergiej, C.D. Brandt and A.W. Morse, IEEE Device Research Conf. (Santa Barbara, CA, June 24-26, 1996) [40] S.A. Lloyd, A. Baranzahi, P. Tobias, and I. Lundstrom, Phys. Status Solidi (a) 162 493- 511 (1997). [41] G. W. Hunter, P.G. Neudeck, L.Y. Chen, D. knight, C.C. Liu, and Q.H. Wu, Silicon Car- bide, III-Nitrides, and Related Materials (Materials Research Forum 264-8, Trans Tech, Switzerland) G. Pensl, H. Morkoc, B. Monemar, E. Janzen (eds), 1093-1096 (1998). [42] M. Mehregany, C. Zorman, N. Narayanan and C.H. Wu, Proc. IEEE 14 1594-1610 (1998). [43] V. Saxena and A.J. Steckl SiC Materials and Devices Y.S. Park (ed), 52 77-151 (1998). [44] R.F. Pierret Semiconductor Device Fundamentals (Addison-Wesley Publishing Company, Inc, 1996). [45] M.J. Bozack Phys. Stat. Sol. (b) 202 549-579 (1997). [46] J. Crofton, L.M. Porter, and J.R. Williams Phys. Stat. Sol. (b) 202 581-603 (1997). [47] J. Bardeen Phys. Rev. 71 717 (1947). [48] http://www.stanford.edu/class216/handouts/6-MS%20contacts.pdf [49] A.M. Cowley and S.M. Sze J. Appl. Phys. 36 3212 (1965). [50] E.H. Rhoderick Metal-semiconductor contacts (Oxford University Press, 1978). [51] J. Crofton, P.G. McMullin, J.R. Williams and M.J. Bozack, Trans. 2nd High Temperature Electronics Conf. Charlotte (NC) XIII-15 (1994). [52] S. liu, S.R. Smith, S. Adams, C. Severt and J. Leonad, Trans. 2nd High Temperature Electronics Conf. Charlotte (NC) XIII-9 (1994). [53] G. Kelner, S. Binari, M. Shur and J.W. Palmmour Electronic Lett. 27 1038 (1991). [54] J. Crofton, J.M. Ferrero, P.A. Barnes, J.R. Williams, M.J. Bozack, C.C. Tin, C.D. Ellis, J.A. Spitznagel and P.G. McMullin, Amorphous Crystalline Silicon Carbide IV C.Y. Yang, M.M. Rahman and G.L. Harris (eds) (Springer-Verlag, Berlin, p.176 1992). [55] J.W. Palmour, H.S. Kong, E.D. Waltz, J.A. Edmond, and C.H. Carter, In: First Interna- tional High Temperature Electronic Conf. D.B. King and F.V. Thome (eds) (Albuquerque, NM, 1991). 181 [56] R.C. Glass, L.M. Spellman and R.F. Davis, Appl. Phys. Lett. 59 2868-2870 (1991). [57] R.C. Glass, L.M. Spellman, S. Tanaka and R.F. Davis, J. Vac. Sci. Technol. A 10 1625- 1630 (1992). [58] L.M. Porter and R.F. Davis, Mat. Sci. and Eng. B34 83-105 (1995). [59] R.N. Hall, J. Appl. Phys. 29 914 (1958). [60] J. Crofton, P.A. Barnes, J.R. Williams and J.A. Edmond, Appl. Phys. Lett. 62 384-386 (1993). [61] J. Crofton and P.A. Barnes, J. Appl. Phys. 69 7660 (1991). [62] J. Crofton, P.A. Barnes and M.J. Bozack, Amer. J. Phys. 60 499 (1992). [63] J.B. Petit, P.G. Neudeck, C.S. Salupo, D.J. Larkin and J.A. Powell, (presented at the Silicon Carbide and Related Materials Conf., Washington DC, 1993) M.G. Spencer, R.P. Devaty, J.A. Edmond, M.A. Khan, R. Kaplan and M. Rahman (eds), Inst. of Phys. Conf. Ser. 137, Inst. of Phys. 679-682 1993. [64] N. Lundberg and M. Ostling, Solid State Electronics 39 1559 (1996). [65] H. Daimon, M. Yamanaka, E. Sakuma, S. Misawa and S. Yoshida, Jpn. J. Appl. Phys. 25 L592-L594 (1986). [66] J.A. Edmond, J. Ryu, J.T. Glass and R.F. Davis, J. Electrochem. Soc. 135 359-362 (1988). [67] A.J. Steckl and J.N. Su, IEDM 695-698 (1993). [68] H.J. Cho, C.S. Hwang, W. Bang, and H.J. Kim, (presented at the Silicon Carbide and Related Materials Conf., Washington DC, 1993) M.G. Spencer, R.P. Devaty, J.A. Edmond, M.A. Khan, R. Kaplan and M. Rahman (eds), Inst. of Phys. Conf. Ser. 137, Inst. of Phys. 663 1993. [69] T.B. Massalski, H. Okamoto, P.R. Subramanian and L. Kacprzak (eds), Binary Alloy Phase Diagrams 2 (ASM International, Materials Park, OH 1990). [70] J.S. Shor, R.A. Weber, L.G. Provost, D. Goldstein and A.D. Kurtz, J. Electrochem. Soc. 141 579-581 (1994). [71] C.A. Mead, and W.G. Spitzer, Phys. Rev. 134(3A) A713-A716 (1964). [72] C.A. Mead, Solid State Electronics 9 1023-1033 (1966). [73] S.H. Hagen, J. Appl. Phys. 39(3) 1458-1461 (1968). [74] S. Yoshida, K. Sasaki, E. Sakuma, S. Misawa and S. Gonda, Appl. Phys. Lett. 46 766-768 (1985). 182 [75] J.R. Waldrop, and R.W. Grant, Appl. Phys. Lett. 56 557-559 (1990). [76] P. Shenoy, A. Moki, B.J. Baliga, D. Alok, K. Wongchotigul and M. Spencer, in:Tech. Digest Intl. Electron Dev. Meeting, IEEE Cat. No. 94CH35706 411-414 (1994). [77] A. Itoh, T. Kimoto and H. Matsunami, IEEE Electron Dev. Lett. 16 280-282 (1995). [78] A. Itoh, T. Kimoto and H. Matsunami, (in Silicon Carbide and Related Materials VI - Kyoto 1995) Inst. of Phys. Conf. Ser. 142 689-692 (1995). [79] A. Itoh, T. Kimoto and H. Matsunami, IEEE Electron Dev. Lett. 17 139-141 (1996). [80] R. Raghunathan, D. Alok and B.J. Baliga, IEEE Electron Dev. Lett. 16 226-227 (1995). [81] C.E. Weitzel, J.W. Palmour, C.H. Carter, K. Moore, K.J. Nordquist, S. Allen, S. Thero and M. Bhatnagar, IEEE Trans. Electron Dev. 43 1732-1741 (1996). [82] V. Saxena and A.J. Steckl, in: International Semiconductor Device Research Symposium 539-542 (1997). [83] V. Saxena and A.J. Steckl, (in: Silicon Carbide, III-Nitrides and Related Materials VII - Stockholm 1997) G. Pensl, H. Morkoc, B. Monemar and E. Janzen (eds) Materials Science Forum 264 937-940 (1998). [84] J.N. Su and A.J. Steckl, (Workshop on High Temperature Power Electronics - Ft. Mon- mouth, NJ) 50-53 (1995). [85] A. Itoh, T. Kimoto and H. Matsunami, (in Silicon Carbide and Related Materials VI - Kyoto 1995) Inst. of Phys. Conf. Ser. 142 697-700 (1996). [86] V. Saxena, A.J. Steckl, M. Vichare, M.L. Ramalingam and K. Reinhardt, in: Third Intl. High Temp. Electronic Conf. - Albuquerque 1 VII/15-20 (1996). [87] J. Crofton, E.D. Lukowski, J.R. Williams, T. Isaacs-Smith, M.J. Bozack, and R. Siergiej, (in: Silicon Carbide and Related Materials VI - Kyoto 1995) Inst. of Phys. Conf. Ser. 142 569-572 (1996). [88] C. Arnodo, S. Tyc, F. Wyczisk and C. Brylinski, (in Silicon Carbide and Related Materials VI - Kyoto 1995) Inst. of Phys. Conf. Ser. 142 577-580 (1996). [89] D. Alok, B.J. Baliga and P.K. Mclarty, IEDM 691-694 (1993). [90] M.M. Anikin, M.G. Rastegaeva, A.L. Syrkin and I.V. Chuiko, Amorphous Crystalline Silicon Carbide III - Springer-Verlag, Berlin G.L. Harris, M.G. Spencer and C.Y. Yang (eds) 56 183-189 (1992). [91] J. Crofton, J.R. Williams, M.J. Bozack and P.A. Barnes, Inst. of Phys. Conf. Ser. 137 719 (1994). 183 [92] J.B. petit, P.G. Neudeck, C.S. Salupo, D.J. Larkin and J.A. Powell, Inst. of Phys. Conf. Ser. 137 679 (1994). [93] S. Liu, K. Reinhardt, C. Severt and J. Scofield (in Silicon Carbide and Related Materials VI - Kyoto 1995) Inst. of Phys. Conf. Ser. 142 589-592 (1996). [94] A.K. Chaddha, J.D. Parsons, and G.B. Kruaval, Appl. Phys. Lett. 66 760 (1995). [95] J.S. Shier, J. Appl. Phys. 41 771-773 (1970). [96] J. Crofton, P.A. Barnes and J.R. Williams, Appl. Phys. Lett. 62 384-386 (1993). [97] L. Spiess, O. Nennewitz and J. Petzoldt, Inst. of Phys. Conf. Ser. 142 585 (1996). [98] N. Nordell, S. Savage and A. Sch a1a72 nder Inst. of Phys. Conf. Ser. 142 573 (1996). [99] M.I. Chaudhry, W.B. Berry and M.V. Zeller, Int. J. Electronics 71 439-444 (1991). [100] J.S. Chen, A. Bachli, M.-A. Nicolet, L. Baud, C. Jaussaud and R. Madar, Mat. Sci. Eng. B 29 185-189 (1995). [101] V. Kumar, L. Zhou, D. Selvanathan and I. Adesida, J. Appl. Phys. 29 (3) 1712-1714 (2002). [102] J.M. Harris, E. Lugujjo, S.U. Capisano, M.-A. Nicolet and R. Shima J. Vac. Sci. Technol. 12 524 (1975). [103] J. Imahori, T. Oku and M. Murakami Thin Solid Films 301 142 (1996). [104] T. Oku, E. Kawakami, M. Uekebo, K. Takahiro, S. Yamaguchi and M. Murakami, Appl. Surf. Sci. 99 265 (1996). [105] J. Kwak, H.-K. Baik, J.-H. Kim and S.-M. Lee, Appl. Phys. Lett. 72 2832 (1998). [106] Y.-J. Lee, B.-S. Shu, S.-K. Rha and C.-O. Park Thin Solid Films 320 141 (1998). [107] E.D. Luckowski, J.M. Delucca, J.R. Williams, S.E. Mohney, M.J. Bozack, T. Isaacs-Smith and J. Crofton, J. Electron. Mat. 27 330-334 (1998). [108] R.S. Okojie, D. Lukco, Y.L. Chen, S. Spry and C. Salupo, Mater. Res. Soc. Symp. Proc. 423 137 (2001). [109] T. Laurila, K. Zeng, J.K. Kivilahti, J. Molarius, T. Riekkinen and I. Suni Microelectronics Engineering 60 71-80 (2001). [110] P.J. Pokela, C.-K. Kwok, E. Kolawa, S. Raud and M.-A. Nicolet, Appl. Surf. Sci. 53 364- 372 (1991). 184 [111] J. Molarius, T. Laurila, T. Riekkinen, K. Zeng, A. Niskanen, M. Leskela, I. Suni and J.K. Kivilahti Advanced Metallization Conference D. Edelstein (ed) (Proceedings of the conference), 355-359 (2000). [112] J.S. Reid, E. Kolawa, R.P. Ruiz and M.-A. Nicolet Thin Solid Films 236 319-324 (1993). [113] J.S. Reid, X. Sun, E. Kolawa and M.-A. Nicolet IEEE Electron Device Letters 15 (8) 298-300 (1994). [114] N.A. Papanicolaou, W.T.Jr. Anderson and A. Christou Inst. Phys. Conf. Ser. 65 407 (1983). [115] I. Suni, M a1a96 en a1a96 a1a96 , M.-A. Nicolet and M. Luomaj a1a96 rvi, J. Electrochem. Soc. 103 1215 (1983). [116] M. a1a85 stling, S. Nygren, C.S. Peterson, H. Norstr a1a72 m, P. Wiklund, R. Buchta, H.-O. Blom and S. Berg, J. Vac. Sci. Technol. A2 281 (1984). [117] J.R. Shappirio, J.J. Finnegan, R.A. Lux and D.C. Fox, Thin Solid Films 119 23 (1984). [118] D. Gupta, Diffusion Phenomena in Thin Films and Microelectronic Materials D. Gupta and P.S. Ho (eds) (Noyes publication, USA, 1 1988). [119] L. Tian, M.O. Thompson, R. Dieckmann, C-Y Hui and Y-Y Lin, J. Appl. Phys. 90 (8) 3799-3808 (2001). [120] L.G. Harrison, Trans. Faraday Soc. 57 1191 (1961). [121] D. Gupta, Phys. Rev. 7 586 (1973). [122] D. Gupta, J. Appl. Phys. 44 4455 (1973). [123] J. Shao and C.A. Angell, Diffusion in Amorphous Materials H. Jain and D. Gupta (eds) (Proceedings of an International Symposium, Pittburgh, USA, 1 1994). [124] Y. Limoge, J.M. Delaye and J.L. Bocquet, Diffusion in Amorphous Materials H. Jain and D. Gupta (eds) (Proceedings of an International Symposium, Pittburgh, USA, 79 1994). [125] Y. Limoge, G. Brebec and Y. Adda, Trans. Tech. Pub., F.J. Kedves and D.L. Beke (eds), 285 1982. [126] Y. Limoge, Diffusion in Materials, L. Laskar, G. Brebec, J.L. Bocquet and C. Monty (eds), (Nato ASI Series, Kluwer Acad. Press, 601 1990). [127] Y. Limoge, Defect and Diffusion Forum, 83 145 (1992). [128] J.W. Haus and K.W. Kehr, Physics Reports, 150 5-6 (1987). 185 [129] T. Laurila, Tantalum-based diffusion barriers for copper metallization, (Dissertation, Helsinki Univ. of Tech., Finland, 2001). [130] L. Darken and R. Gurry, (Physical Chemistry of Metals, McGraw-Hill, 1953). [131] E.A. Guggenheim, (Thermodynamics, Elservier Science, The Netherlands, 1967). [132] E. Kolawa, J.S. Chen, J.S. Reid, P.J. Pokela and M.-A. Nicolet, J. Appl. Phys., 70 (3) 1369-1373 (1991). [133] J.D. Wiely, J.H. Perepezko, J.E. Nordman and K.-J. Guo, IEEE Trans. Ind. Electron., 29 (1982). [134] M.-A. Nicolet, I. Suni and M. Finetti, Solid State Technol., 29 129 (1983). [135] L.S. Hung, F.W. Saris, S.Q. Wang and J.W. Mayer, J. Appl. Phys., 59 2416 (1986). [136] Y.-J. Lee, B.-S. Suh and C.-O. Park, Thin Solid Films, 357 237-241 (1999). [137] E.R. Weber, Appl. Phys. A, 1 (1983). [138] A. Broniatowski, Phys. Rev. Lett., 62 3074 (1989). [139] M.-A. Nicolet, Thin Solid Films, 54 415 (1978). [140] J.W. Mayer and S.S. Lau, Electronic Materials Science: For Integrated circuits in Si and GaAs, (Macmillan publishing company, NY 1990). [141] M.-A. Nicolet, Diffusion in Amorphous Materials H. Jain and D. Gupta (eds) (Proceedings of an International Symposium, Pittburgh, USA, 225 1994). [142] M. Azuma, Y. Nakato and H. Tsubomura, J. Electroanal. Chem., 255 179 (1988). [143] X. Sun, E. Kolawa, J.S. Chen, J.S. Reid and M.-A. Nicolet, Thin Solid Films, 236 347 (1993). [144] M. Thuillard, T.W. Workman, E. Kolawa and M.-A. Nicolet, Jrnl Less-Common Metals, 145 505 (1988). [145] K. Affolter, H. Kattelus and M.-A. Nicolet, Mat. Res. Soc. Symp. Proc., 47 167 (1985). [146] J.A. Cunningham, C.R. Fuller and C.T. Haywood, IEEE Trans. Reliab., 19 182 (1970). [147] M.L. Green, M.E. Gross, L.E. Papa, K.J. Schnoes and D. Brasen, J. Electrochem. Soc., 132 2677 (1985). [148] K. Tominaga, T. Murayama, I. Mori, T. Okamoto, K. Hiruta, T. Moriga and I. Nakabayashi, Vacuum, 59 546 (2000). [149] R.B.H. Tahar, T. Ban, Y. Ohya and Y. Takahashi, J. Appl. Phys., 83 2631 (1998). 186 [150] Y. Abe, K. Kato, M. Kawamura and K. Sasaki, Jpn. J. Appl. Phys., 39 245 (2000). [151] D.K. Schroder, Seminconductor Material and Device Characterization, (John Wiley and Sons, Inc. 1990). [152] L.J. van der Pauw, Phil. Res. Rep., 13 1-9 (1958). [153] L.J. van der Pauw, Phil. Tech. Rev., 20 220 (1958). [154] R. Chwang, B.J. Smith and C.R. Crowell, Solid-State Electron., 17 1217 (1974). [155] E.H. Rhoderick and R.H. Williams, Metal-Semiconductor Contacts, (2nd ed. Clarendon, Oxford, 1988). [156] A.Y.C. Yu, Solid-State Electron., 13 239 (1970). [157] F.A. Padovani, in: Semiconductors and Semimetals, R.K. Willardson and A.C. Beer (eds.), (Academic Press, New York 7A 75 1971). [158] F.A. Padovani and R. Stratton, Solid-State Electron., 9 695 (1966). [159] G.K. Reeves and H.B. Harrison, IEEE Electron Device Letters, EDL-3 (5) 111 (1982). [160] http://ece-www.colorado.edu/ bart/book/book/chapter3/pdf/ ch3 5 3.pdf [161] H.B. Harrison, Proc. IREE Aust., 41 95 (1980). [162] N.T. Tam and T. Chot, Phys. Stat. Sol., 93a K91 (1986). [163] N. Toyama, J. Appl. Phys., 63 2720 (1988). [164] A.M. Goodman, J. Appl. Phys., 54 922 (1963). [165] E.H. Rhoderick, J. Phys. D: Appl. Phys., 5 1920 (1972). [166] L.C. Feldman and J.W. Mayer Fundamentals of Surface and Thin Film Analysis, (Elsevier Science Pub. Co., Inc. 1986). [167] W.-K. Chu, J.W. Mayer and M.-A. Nicolet, Backscattering Spectrometry, (Academic Press, Inc. 1978). [168] http://www.physics.auburn.edu/ condmatt/accellab.html [169] National Electrostatics Corporation. Instruction Manual for Operation and Service: Charge Exchange RF Source, Model 2JA002110. [170] R.J. Girnius and L.W. Anderson, Nuclear Instruments and Methods, 137 373 (1976). [171] http://www.pelletron.com/charging.htm 187 [172] G.F. Knoll, Radiation Detection and Measurement, (Wiley, 1979). [173] D.F. Stein, Auger Electron Spectroscopy, C.L. Briant and R.P. Messmer (eds.) (Academic Press, Inc. 1988). [174] M. Thomson, M.D. Baker, A. Christie and J.F. Tyson, Auger Electron Spectroscopy, (John Willey and Sons, Inc. 74 1985). [175] ?Handbook of Auger Electron Spectroscopy? A reference book of standard data for iden- tification and interpretation of Auger electron spectroscopy data. 2nd ed. L.E. Davis, N.C. MacDonald, P.W. Palmberg, G.E. Riach and R.E. Weber, (Physical Electronic Industries, Inc. 1978). [176] R.D. Tarey, R.S. Rastogi and K.L. Chopra, The Rigaku Journal, 4 (1/2) 11 (1987). [177] M.F. Toney, X-Ray Diffraction in Encyclopedia of Materials Characterization: surfaces, interfaces, thinfilms., C.R. Brundle and C.A. Evans (eds.) 198 (1992). [178] A. Shegmuller, I.C. Noyan and V.S. Sperious, Prog. Cryst. Growth and Charact., 18 21 (1989). [179] D. Rafaja, Adv. in Solid State Phys., 41 275 (2001). [180] B.D. Cullity and S.R. Stock, Elements of X-Ray Diffraction, (3rd ed. Prentice-Hall, Inc. 2001). [181] R.W. Johnson, Professional Development Course, (HiTEC 2004, Santa Fe, New Mexico, 2004). 188