Citation
Fabrication and Characterization of GaN Based High Electron Mobility Transistors

Material Information

Title:
Fabrication and Characterization of GaN Based High Electron Mobility Transistors
Creator:
Hwang, Ya-Hsi
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (189 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Chemical Engineering
Committee Chair:
REN,FAN
Committee Co-Chair:
TSENG,YIIDER
Committee Members:
GILA,BRENT P
PEARTON,STEPHEN J
Graduation Date:
5/2/2015

Subjects

Subjects / Keywords:
Augers ( jstor )
Dosage ( jstor )
Drains ( jstor )
Electric current ( jstor )
Electric fields ( jstor )
Electric potential ( jstor )
Electrons ( jstor )
Etching ( jstor )
Irradiation ( jstor )
MODFETS ( jstor )
Chemical Engineering -- Dissertations, Academic -- UF
ald -- algan -- gan -- heat -- hemt -- inaln -- irradiation
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Chemical Engineering thesis, Ph.D.

Notes

Abstract:
Aluminum nitride (AlN), deposited by an atomic layer deposition system, was employed as a gate insulator and a passivation layer. By introducing AlN as the gate insulator, the gate modulation can be extended to 4 V, up from 2 V, for a Schottky gate fabricated on the same HEMT. Moreover, the subthreshold drain leakage current was suppressed to 1.13 nA/mm. Since the saturation drain current (IDS) of AlN based HEMTs remained similar to AlGaN/GaN HEMT, the on/off ratio significantly increased to 3.3E8. Besides reducing the leakage current, the effectiveness of passivation was observed. The IDS only showed a 7 percent decrease during the gate lag measurement at a frequency of 100 kHz. In addition, off state drain breakdown voltage (VBR) over 2000 V at drain to gate distance (LDG) equals to 37.5 micrometer was achieved. Specific on state resistance of 1.3 and 10.9 mohmcm2 for the devices with the LDG=7.5 and LDG=37.5 micrometer were obtained respectively. The effects of proton, gamma and electron irradiation on AlGaN/GaN HEMT DC performance were investigated. For proton irradiation, the mechanism of VBR improvement was investigated through backside proton irradiation. The result indicating the increase of VBR was from the reduction of peak electric field on the gate edges due to the extra charges created by irradiated defects. For gamma irradiation, AlGaN/GaN HEMTs were irradiated at doses of 50, 300, 450, or 700 Gy at a fixed energy of 10MeV. After irradiation, IDS proportionally increased with the dose and reached a maximum of 10% with a dose of 700 Gy. The increase was mainly due to the increase of mobility. For electron irradiation, both AlGaN/GaN and InAlN/GaN heterojunctions were studied. It is found out the irradiation hardness of InAlN was higher than AlGaN. Besides, the capacitance-voltage curve shifted positively after irradiation, which was due to the increase of deep acceptor traps in the barrier/interface region. In AlGaN/GaN/Si transistors, the increases of deep barrier/interface traps with activation energy of 0.3, 0.55, 0.8 eV were observed. These increases correlated with the current dispersion at gate lag measurement after irradiation with 1.3E16 cm-2 electrons. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2015.
Local:
Adviser: REN,FAN.
Local:
Co-adviser: TSENG,YIIDER.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2015-11-30
Statement of Responsibility:
by Ya-Hsi Hwang.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Embargo Date:
11/30/2015
Classification:
LD1780 2015 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

1 FABRICATION AND CHARACTERIZATION OF GAN BASED HIGH ELECTRON MOBILITY TRANSISTORS BY YA H SI HWANG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCT OR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2015

PAGE 2

2 © 201 5 Ya Hsi Hwang

PAGE 3

3 To my family

PAGE 4

4 ACNOWLEDGMENTS First of all, I would like to appreciate my advisor and supervisory committee chair, Dr. Fan Ren. Thanks for his broad and profound knowledge, patience and encouragement through my Ph.D. study. The completion of this work would not have be en possible without the guidance from Dr. Fan Ren. He has always been there to provide technical and personal support whenever I had proble ms. He has treated me like his kid, and provided an excellent academic atmosphere for me to perform my research. His education and dedication in the past three years have been priceless and will benefit my future life. I also would like to express my sinc ere gratitude to the rest of my committee members, Dr. Stephen J. Pearton, Dr. Brent P. Gila and Dr. Yiider Tseng. Thanks for their suggestions and efforts to improve my proficiency in my research area. The reason I was able to finish this work was the sup port and assistance of former group members Dr. Lu Liu, Camilo, Yuyin, Brian , Eugene , Shun, Lei as well as c urrent group members Shihyun, Dong, Byung Jae, Lincong, Weidi . Thanks for all your help and accompany in the past few years. Also, I would like to t hank the staff Bill, David , Andres and Al in NRF who give me a lot of technical support . Special thanks to Department of Chemical Engineering at the University of Florida for giving me this opportunity to pursue my PhD degree in the Sunshine S tate, and m y great thanks to department st a ff Carolyn Miller in payroll for purchasing, Craig Smith and Jim Hinnant in mechanical and electronic shops for tool maintenance and troubleshooting . I would like to thank Santiago for helping me out on GPIB and LabView as we ll. Last, I would like to thank my family and friends. Thanks for always being by my side and supportive. Without your support, I can t finish this dissertation and achieve this much.

PAGE 5

5 TABLE OF CONTENTS page ACNOWLEDGMENTS ................................ ................................ ................................ ................. 4 LIST OF TABLES ................................ ................................ ................................ .......................... 9 LIST OF FIGURES ................................ ................................ ................................ ...................... 10 LIST OF ABBREVIATIONS ................................ ................................ ................................ ....... 14 ABSTRACT ................................ ................................ ................................ ................................ .. 18 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ................. 20 1.1 Background ................................ ................................ ................................ ..................... 20 1.1.1 Introduction to High Electron Mobility Transistors ................................ ........ 20 1.1.2 Substrate ................................ ................................ ................................ .......... 21 1.1.3 Review of HEMTs Performance ................................ ................................ ..... 22 1.1.4 Issues ................................ ................................ ................................ ............... 22 1.1.4.1 Self heating effect ................................ ................................ ................ 22 1.1.4.2 Breakdown voltage ................................ ................................ .............. 23 1.1.4.3 Curren t dispersion ................................ ................................ ................ 25 1.2 Dissertation Outline ................................ ................................ ................................ ........ 25 2. METHODOLOGY REVIEW ................................ ................................ ................................ 33 2.1 Semiconductor Fabrication Processes ................................ ................................ ............ 33 2.1.1 Isolation ................................ ................................ ................................ ........... 33 2.1.2 Photolithography ................................ ................................ ............................. 34 2.1.3 Lift off ................................ ................................ ................................ .............. 35 2.1.4 Electron beam evaporation ................................ ................................ .............. 36 2.1.5 Rapid thermal annealing ................................ ................................ .................. 36 2.1.6 Plasma enhanced chemical vapor deposition ................................ .................. 37 2.1.7 Atomic layer deposition ................................ ................................ ................... 38 2.1. 8 Bosch etching ................................ ................................ ................................ .. 38 2.2 Material Characterization ................................ ................................ ............................... 39 2.2.1 Hall measurement ................................ ................................ ............................ 39 2.2.2 Transmission electron microscopy ................................ ................................ .. 40 2.2.3 Scanning electron microscopy and energy dispersive X ray spectroscopy ..... 40 2.2.4 X ray photoelectron spectroscopy ................................ ................................ ... 41 2.2.5 Auger electron spectroscopy ................................ ................................ ........... 42 2.2.6 Ellipsometer ................................ ................................ ................................ ..... 42 2.2.7 Four point probe measurement ................................ ................................ ........ 43

PAGE 6

6 2.3 Device Characterization ................................ ................................ ................................ .. 43 2.3.1 DC performance ................................ ................................ .............................. 43 2.3.2 Transmission line method ................................ ................................ ................ 45 2.3.3 Gate lag measurement ................................ ................................ ..................... 46 2.3. 4 Off state drain breakdown voltage ................................ ................................ .. 46 2.3.5 Deep level transient spectrum ................................ ................................ .......... 47 2.3.6 Admittance spectrum ................................ ................................ ....................... 47 2.3.7 Capacitance voltage measurement ................................ ................................ .. 48 2.4 Finite element method ................................ ................................ ................................ .... 50 3. GAN METAL INSULAT OR SEMICONDUCTOR HEMTS WITH PLASMA ENHANCED ATOMIC LAYER DEPOSITED ALN AS GATE DIELECTRIC AND PASSIVATION ................................ ................................ ................................ ..................... 66 3.1 Introduction to Metal Oxide Semiconductor High Electron Mobility Transistors ......... 66 3.2 Experimental ................................ ................................ ................................ ................... 67 3.2.1 Material Growth ................................ ................................ .............................. 67 3.2.2 HEMTs Fabricatio n ................................ ................................ ......................... 67 3.2.3 Device Characterization ................................ ................................ .................. 68 3.3 Results and Discussion ................................ ................................ ................................ ... 69 3.4 Summary ................................ ................................ ................................ ......................... 77 4. HIGH BREAKDOWN VOLTAGE IN ALN/GAN MISHEMTS ................................ ......... 78 4.1 Overview of High Breakdown Voltage HEMTs ................................ ............................ 78 4.2 Experimental ................................ ................................ ................................ ................... 79 4.2.1 Material Growth ................................ ................................ .............................. 79 4.2.2 HEMTs Fabrication ................................ ................................ ......................... 80 4.2.3 Device Characterization ................................ ................................ .................. 81 4.3 Results And Discussion ................................ ................................ ................................ .. 81 4.4 Summary ................................ ................................ ................................ ......................... 89 5. DEGRADATION MECHANISMS OF Ti/Al/Ni/Au BASED OHMIC CONTACTS ON AlGaN/GaN HEMTs ................................ ................................ ................................ ............. 90 5.1 Introduction to Ti/Al/Ni/Au Based Ohmic Con tacts ................................ ...................... 90 5.2 Experimental ................................ ................................ ................................ ................... 91 5.3 Results and Discussion ................................ ................................ ................................ ... 92 5.4 Summ ary ................................ ................................ ................................ ....................... 104 6 EFFECT OF BACKSIDE PROTON IRRADIATION ON ALGAN/GAN HEMT ON OFF STATE DRAIN BREAKDOWN VOLTAGE ................................ ............................ 105 6.1 Introduction to Pro ton Irradiation ................................ ................................ ................. 105 6.2 Experimental ................................ ................................ ................................ ................. 106 6.2.1 Material Growth ................................ ................................ ................................ ... 106

PAGE 7

7 6.2.2 HEMTs Fabrication ................................ ................................ ............................. 106 6.2.3 Device Irradiation Simulation And Experiment ................................ ................. 109 6.2.4 Electrical Simulation ................................ ................................ ........................... 109 6.2.5 Device Characterization ................................ ................................ ...................... 109 6.3 Results and Discussion ................................ ................................ ................................ . 110 6.4 Summ ary ................................ ................................ ................................ ....................... 115 7 EFFECT OF LOW DOSE GAMMA IRRADIATION ON DC PERFORMANCE OF ALGAN/GAN HIGH ELECTRON MOBILITY TRANSISTORS ................................ .... 116 7.1 Introduct ion to Gamma Irradiation ................................ ................................ ............... 116 7.2 Experimental ................................ ................................ ................................ ................. 117 7.2.1 Material Growth ................................ ................................ ................................ .. 117 7.2.2 Device Fabrication ................................ ................................ .............................. 117 7.2.3 Gamma Irradiation and Device Characterization ................................ ................ 120 7.3 Results and Discussion ................................ ................................ ................................ . 120 7.4 Summary ................................ ................................ ................................ ....................... 129 8 EFFECT OF ELECTRON IRRADIATION ON ALGAN/GAN AND INALN/GAN HETEROJUNCTIONS ................................ ................................ ................................ ........ 130 8.1 Introduction to Electron Irradiation ................................ ................................ .............. 130 8.2 Experimental ................................ ................................ ................................ ................. 133 8.2.1 HEMTs Fabrication ................................ ................................ ............................. 133 8.2.2 Device Characterization ................................ ................................ ...................... 133 8.3 Results and Discussion ................................ ................................ ................................ . 134 8.4 Summary ................................ ................................ ................................ ....................... 146 9 A NOVEL APPROACH TO IMPROVE HEAT DISSIPATION OF ALGAN/GAN HEMTS WITH A CU FILLED VIA UNDER DEVICE ACTIVE AREA ......................... 147 9. 1 Introduction to Thermal Performance of HEMTs ................................ ........................ 147 9.2 Simulation Approach ................................ ................................ ................................ .... 149 9.3 Results and Discussion ................................ ................................ ................................ . 153 9.4 Summary ................................ ................................ ................................ ....................... 163 10 A NOVEL STRUCTURE TO IMPROVE BREAKDOWN VOLTAGE BY BACKSIDE GATE ON ALGAN/GAN HEMTS ................................ ................................ ..................... 164 10.1 Introduction to Multiple Gate Device and Field Plate ................................ ................. 164 10.2 Experimental ................................ ................................ ................................ ................ 165 10.2.1 HEMTs Fabrication ................................ ................................ ........................... 165 10.2.2 Front Side Device Fabrication ................................ ................................ ........... 16 5 10.2.3 Backside Device Fabrication ................................ ................................ ............. 166 10.3 Results and Discussion ................................ ................................ ................................ . 166 10.4 Summary ................................ ................................ ................................ ....................... 168

PAGE 8

8 11 CONCLUSIONS ................................ ................................ ................................ ................. 176 LIST OF REFERENCES ................................ ................................ ................................ ............ 179 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ...... 189

PAGE 9

9 LIST OF TABLES Table page 1 1 Summary of material properties of Si and AlGaN/GaN heterojunction. .......................... 27 1 2 Summary of material properties of Si, sapphire and SiC. ................................ ................. 29 7 1 Summary of drain current at Vg = 0 V, on state resistance, device total resistance, sheet resistance and mobility pr irradiation. ......................... 125 8 1 Initial characteristics of 2DEG in studied HJs before irradiation: 2DEG concentration N (2DEG) (cm 2 ) and 2DEG mobility µ (2DEG) (cm 2 /Vs), also shown are threshold voltages V TH (V) as deduced from room temperature C V characteristics .................... 139 9 1 Material thermal conductivity. ................................ ................................ ........................ 152 9 2 Absolute thermal resistance of HEMT with different configurations. ........................... 158 10 1 Threshold voltage change with backside gate voltage ................................ .................... 174

PAGE 10

10 LIST OF FIGURES Figur e page 1 1 A) Origin of 2DEG 1 B) Energy band gap diagram of AlG aN and GaN heterojunction. 48 ................................ ................................ ................................ ................................ ........... 28 1 2 Drain I V and transfer characteristics of AlGaN / GaN HEMTs on sapphire and on silicon. 8 ................................ ................................ ................................ .............................. 30 1 3 Calculated electron mobility as a function of temperature. 14 ................................ ........... 31 1 4 2D simulations of the electrostatic field distributions between source and drain contacts for HEMT with and without source field plate. 24 ................................ ............... 32 2 1 Schematic view of isolation by A) mesa etching and B) ion implantation. ...................... 51 2 2 Schematic view of photolithog raphy process of A) positive or B) negative photoresist. 13 ................................ ................................ ................................ ..................... 52 2 3 Effect of photoresist profile A) ideal overhang structure and B) tapered angle profile to lift off performance. ................................ ................................ ................................ ...... 53 2 4 Schematic view of an evaporation system. 39 ................................ ................................ .... 54 2 5 Schematic view of ALD process. ................................ ................................ ...................... 55 2 6 Schematic view of Hall measurement system setup. 44 ................................ ..................... 56 2 7 Generation of Auger electrons, backscattered electrons by the incident electron. 48 ........ 57 2 8 Schematic view of four point probe measurement. 50 ................................ ........................ 58 2 9 Typical drain I V characteristics of a field effect transistor. ................................ ............ 59 2 10 Transfer characteristics of A) transconductance characteristics and B) threshold voltage determination. 52 ................................ ................................ ................................ .... 60 2 11 Threshold voltage determined by the saturation ex trapolation technique. 53 ..................... 61 2 12 The configuration of TLM test key and data analysis. 52 ................................ ................... 62 2 13 The system setup for gate pulse me asurement. ................................ ................................ . 62 2 14 High frequency C V curve for a Schottky diode on a HEMT. 68 ................................ ...... 63 2 15 Effect of a fixed oxide charge and interface traps on the C V characteristics of MOS diode. 40 ................................ ................................ ................................ .............................. 64 2 16 Location of the charges associated with a MOS diode. 40 ................................ ................. 65

PAGE 11

11 3 1 Optical mi croscope picture of A) reference B) treated with UV ozone C) treated with UV ozone and NH 4 OH. ................................ ................................ ................................ ..... 72 3 2 XPS analysis of AlN thin film deposited by plasma enhanced mode ALD. A) Surface survey of th e as deposited 10 nm AlN by PEALD. B) High resolution scan of Al 2p spectrum. C) High resolution scan of N 1s spectrum. ................................ ...................... 73 3 3 Depth profile of 10 nm AlN thin film deposited by plasma enhanced mode ALD. ......... 74 3 4 A) Drain current versus drain bias for different gate bias voltages. B) Transconductance characteristics measured with the MIS HEMT in saturation .............. 75 3 5 Gate lag measured on the AlN MISHEMT ................................ ................................ ....... 76 4 1 Drain current as a function of drain bias for different gate bias voltages for devices with drain source dist ance of (A) 10 µm or (B) 40 µm ................................ .................... 84 4 2 Transfer characteristics, drain and gate current for devices with drain source distance of A) 10 µm or B) 40 µm measured with the MISHEMTs in saturation. ......................... 85 4 3 Saturation drain source current as a function of drain source distance (L DS ). .................. 86 4 4 A) Off state drain source characteristi cs as a function of gate voltage for individual devices and B) off state drain breakdown voltage as a function of L GD at V G = 6 V. .... 87 4 5 Specific on state resistance as a function of breakdo wn voltage. ................................ ..... 88 5 1 Sheet resistance of alloyed Ti / Al / Ni / Au based Ohmic metallization as a function of BOE treatment time measured with the four point probe technique. ........................... 96 5 2 Percentages of metal elements in alloyed Ti / Al / Ni / Au metallization prior to and after BOE treatment, measured by energy dispersive x ray spectroscopy (EDX). .......... 97 5 3 OM picture A) before and B) after treatment; SEM pictures from SD C) before and D) after treatment; SEM pictures from BSD E) before and F) after treatment. ..................... 98 5 4 EDX elemental mappings of alloyed Ti / Al / Ni / Au metallization prior to and after 180 sec of BOE treatment. ................................ ................................ ................................ 99 5 5 Auger surface scan analyses of island, ring and field regions for A) untreated sample and B) BOE treated sample. ................................ ................................ ............................ 100 5 6 Auger depth profile of the island area on A) untreated sample and B) BOE treated sample. ................................ ................................ ................................ ............................ 101 5 7 Auger depth profile of the ring area on A) untreated sample and B) BOE treated sample. ................................ ................................ ................................ ................................ ......... 102

PAGE 12

12 5 8 Auger depth profile of the field area on A) untreated sample and B) BOE treated sample. ................................ ................................ ................................ ............................ 103 6 1 Cross sectional TEM image of AlGaN/GaN HEMT structure. ................................ ...... 108 6 2 Schematic of backside proton implantation through via hole and proton irradiation induced vacancy distributions as a function of proton penetration depth. ...................... 112 6 3 Drain I V characteristics of the HEMT before and after irradiated with a proton energy of 330 keV and a dose of 5 × 10 12 cm 2 . ................................ ................................ ......... 113 6 4 Electric field distributions around the gate edge for the proton irradiated and the pre irradiated HEMTs. ................................ ................................ ................................ .......... 114 7 1 Device configuration of the circular HEMTs. ................................ ................................ 119 7 2 A) Drain I irradiation at a dose of 450 Gy. B) Drain current irradiation dose. ...................... 124 7 3 Source resistance, drain resistance, channel resistance and total resistance of a circular irradiation dose. ................................ .............. 126 7 4 A) Sub irradiation with a dose of 450 Gy. B) Gate I irradiation with a dose of 450 Gy. .... 127 7 5 irradiation with different doses. ................................ ................................ ................................ ........ 128 8 1 Mobility as a functi on of electron fluence for type 1 AlGaN / AlN / GaN / sapphire, type 2 AlGaN / GaN / sapphire, type 3 AlGaN / GaN / Si, and type 4 InAlN / GaN / sapphire HJ. ................................ ................................ ................................ ..................... 140 8 2 C V characteristics measure d on the AlGaN / GaN / Si HJ before irradiation (square line) and after electron irradiation with the fluence of 1.3 × 10 16 cm 2 (circle line). ...... 141 8 3 85K C V characteristics measured on the A lGaN / GaN / Si HJ A) before irradiation and B) after irradiation. ................................ ................................ ................................ ... 142 8 4 DLTS spectra measured on the AlGaN / GaN / Si HEMT before irradiation (dashed curve) and after irradiation with 1.3 × 10 16 cm 2 10 MeV electrons (solid curve). ........ 143 8 5 The temperature dependence of capacitance and AC conductance (G). Black curves refer to pre irradiation, red curves refer to after irradiation. ................................ ........... 144 8 6 A) Gate lag measured on the AlGaN / GaN / Si HEMT before irradiation B) gate lag measured on the AlGaN / GaN / Si HEMT after irradiation. ................................ ......... 145 9 1 A) 3 D mesh structure of the AlGaN / GaN HEMT used in the simulation. B) 3 D frame structure with a via under source contact and via under the active area. ............. 151

PAGE 13

13 9 2 A) C ross sectional temperature contours of the reference HEMT. B) Cross sectional temperature contours of the proposed HEMT structure. ................................ ................ 157 9 3 Vertical temperature distributions directly under the gate finger for the reference HEMT and the proposed HEMT structure with copper filled via. ................................ . 159 9 4 A) 2 dimensional temperature contours at 2DEG plane of the reference HEMT. B) 2 dimensional te mperature contours at 2DEG plane of the proposed HEMT. .................. 160 9 5 Maximum junction temperature as a function of power density for the reference, the proposed HEMT, and the proposed HEMT with 1 µm or 2 µm filled copper. .............. 161 9 6 Effect of the through Si substrate via hole location on the maximal junction temperature. ................................ ................................ ................................ .................... 162 10 1 Schematic view of the proposed structure with a backside via. ................................ ..... 170 10 2 Drain I V modulated by A) front side gate or B) backside gate. ................................ .... 171 10 3 A) Drain I V modulated by both front side and backside gate. B) Transfer characteristics modulated by both front side and backside gate. ................................ .... 172 10 4 A) Transfer characteristics of ref sample and sample with backside voltage at 10 V. B) On/off ratio and subthreshold swing change versus different backside gate voltage 173 10 5 A) Gate I V with backside gate biased at differe nt voltages. B) Breakdown voltage change at different backside gate voltages. ................................ ................................ ..... 175

PAGE 14

14 LIST OF ABBREVIATIONS 2DEG 2 Dimensional Electron Gas AES Auger Electron Spectroscopy ALD Atomic Layer Deposition AlGaN Aluminum Gallium N itride AlN Aluminum Nitride BOE Buffer Oxide Etchant BSD Back Scattering Electron Detector CI Contact Inclusion C V Capacitance Voltage DLTS Deep Level Transient Spectroscopy DRIE Deep Reactive Ion Etching E c Conduction Band Edge E DX Energy Disper sive X Ray Spectroscopy E V Valence Band Edge FET Field Effect Transistor

PAGE 15

15 FLOODS FLorida Object Oriented Device Simulator GaN Gallium Nitride HEMT High Electron Mobility Transistor HJ Hetero J uncio n ICP Inductively Coupled Plasma InAlN Indium Alumi num Nitride IR Infrared I V Current Voltage L DG Drain to Gate Distance L DS Drain to Source Distance L GS Gate to Source Distance LOR Lift Off Resist MBE Molecular Beam Epitaxy MISHEMT Metal Insulator Semiconductor Hi gh Electron Mobility Transistor MOCVD Metal Organic Chemical Vapor Deposition MOSFET Metal Oxide Semiconductor Field Effect Transis tor

PAGE 16

16 Ni Nitrogen Interstitial N SS Surface States Density ODEPR Optically Detected E lectron Paramagnetic Resonance OM Optical Microscope PAC Photoactive Compound PEALD Plasma Enhanced Atomic Layer Deposition PECVD Plasma Enhanced Chemical Vapor Deposition PMGI Poly(methyl glutarimide) PMMA Poly (methyl methacrylate) PPC Persistent Photocapacitance PR Photo Resist R c Contact Resistance RI Refracti ve Index R s Sheet Resistance R T Transfer Resistance SBH Schottky Barrier Height

PAGE 17

17 SD Secondary Electron Detector SEM Scanning Electron Microscopy SiC Silicon Carbide SRIM Stopping and Range o f Ions In Matter SS Subthreshold Swing TCAD Technology Computer Aided Design TEM Transmission electron microscopy TLM Transmission line measurement TMA T ri methyl Aluminum UD Under Development UV Ultra Violet V Ga Gallium Vacancy V H Hall voltage V N N itrogen Vacancy V TH Threshold voltage XPS X Ray Photoelectron Spectroscopy

PAGE 18

18 1 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy FABRICATION AND CHARACTERIZATION OF GAN BASED HIGH ELECTRON MOBILITY TRANSISTORS By Ya Hsi Hwang May 201 5 Chair: Fan Ren Major: Chemical Engineering Aluminum nitride (AlN) was employ ed as a gate insulator and a passivation layer. By introducing AlN as the gate insulator , the gate modulation c an be extended from 2 V to 4 V, for a Schottky gate HEMT . Moreover , the subthreshold leakage current was suppressed to 1.13 n A/mm and thus the on/off ratio was increased to 3.3 E8 . Besides reducing the leakage current, the effectivenes s of passivation was o bserved. The I DS only showed a 7% dispersion at 100 kHz. In addition , off state drain breakdown voltage (V BR ) over 2000 V and specific on resistance of cm 2 at drain to gate distance of 3 7.5 µ m w ere achieved. T he effects of proton, gamma and electr on irradiation on AlGaN / GaN HEMT D C performance were investigated . For proton irradiation, the mechanism of V BR improvement was investigated through backside proton irradiation. The result indicating the increase of V BR was from the reduction of peak ele ctric field on the gate edges due to the extra charges created by irradiated defects. For gamma irradiation, AlGaN/GaN HEMTs were irradiated at doses of 50, 300 , 450, or 700 Gy at a fixed energy of 10MeV. After irradiation, I DS proportionally increased wit h the dose due to the increase of mobility and reached a maximum of 10% with a dose of 700

PAGE 19

19 Gy. For electron irradiation, the capacitance voltage curve shift ed positively after irradiation due to the increase of deep acceptor traps in the barrier/interface region. I n AlGaN/GaN/Si transistors, the increase s of deep barrier/interface traps with activation energy of 0.3, 0.55, 0.8 eV w ere observed. Th ese increase s correlated with the current dispersion at gate lag measurement. Novel structure with a backside m etal via under the active area of the HEMT was also proposed. It is found out by simulations that t he thermal resistance decreased 17% by removing the thermal resistive nucleation layer and filling the via with Cu. Besides , the maximum junction temperature could be decreased from 146 o C to 120 o C at a power density of 5 W/mm. Furthermore , V BR could be improved by 10% upon connecting the via with the front side gate. By biasing t he backside gate at 25 V, V BR could be improved by 40%.

PAGE 20

20 CHAPTER 1 1 INTRODUCTION 1.1 B ackground 1.1.1 Introduction t o High Electron Mobility Transistors Transistor, a three terminal device which can be used as a switch or an amplifier, is viewed as the most important invention in the 20th century. It is composed of a semiconductor material and at least three electrodes, which are source (S), drain (D) and gate (G) to connect to the external circuit. Source and drain electrodes are Ohmic c ontacts which conduct current. Gate is either a Schottky contact or a metal insulator semiconductor contact whi ch controls the conductivity of the channel. The drain source current (I DS ) can be modulated by varying the change of gate voltage (V G ), and thus amplify the signal changing at gate electrode. The most common semiconductor material is silicon. Although Si technology is relatively cheap and matured, it can meet the requirements of high power electronics such as high breakdown voltage. III V composite materials such as AlGaN / GaN or InAlN / GaN heterojunction s start to get more attention because of their s uperior material properties such as thermal stability, chemical stability, high electron m obility, high breakdown electr ic field, high saturated drift velocity, and high radiation tolerance. Thus, GaN based HEMTs are extremely suitable for high power and h igh frequency application s such as space, s atellite or military technologies . Table 1 1 s hows the summary of material properties of Si and AlGaN / GaN heterojunction . The formation of AlGaN / GaN high electron mobility transistors (HEMTs) is by employing AlGaN as barrier layer on GaN as buffer layer. The energy band gap of Al 0.25 GaN and GaN are 4.05 eV and 3.4 eV, respectively. The band gap difference between AlGaN and GaN allows the 2 dimensional electron gas ( 2DEG) channel to be formed at the AlGaN/GaN interface. Because of the piezoelectric polarization and spontaneous polarization, the 2DEG

PAGE 21

21 electron sheet density could be up to 10 13 cm 2 even without doping. Figure 1 1 shows the sche matic view of the 2DEG and the energy band gap diagram of Al 0.25 Ga 0.75 N and GaN heterojunction. In other words, unlike traditional Si FETs, the carriers in 2DEG channel will not be scattered by the impurity because no impurity in 2DEG channel is needed to generate carriers. This characteristic guarantees the high electron mobility (~1400 cm 2 / Vs) as compared to other semiconductor material such as Si (~400 cm 2 / Vs). The performance of HEMT can be further improved by optimizing the material structure. Incr ease the Al mole fraction in the AlGaN layer will lead to higher carrier concentration, but mobility will drop due to alloy disorder scattering. Besides, AlGaN barrier layer with a higher Al content is accompanied with enhanced lattice mismatch effects, wh ich degrade the crystalline quality of barrier layer leading to a lower carrier mobility. 1 1 Therefore, the improvement of output power density by increasing the Al co ntent is limited. Instead of increasing the Al content to enhance the carrier concentration, Shen et al. inserted a 1 nm AlN layer between AlGaN and GaN layer and achieved an increase in carrier mobility by 25% and 10% increase in carrier concentration. 2 Besides increasing Al content in AlGaN / GaN based HEMT, In 0.18 Al 0.82 N can be adopted as an alternative barrier layer as well. In InAlN / GaN system, the polarization charge is completely determined by spontaneou s polarization since the structure is free from strain so the piezoelectric polarization is nearly zero. 1 Kuzmik has predicted the 2DEG density to be around 2.7 × 10 13 cm 2 when In 18 Al 0.82 N is nearly lattice mat ched to GaN. The higher 2DEG density of InAlN / GaN system is due to the large spontaneous polarization difference and the large conduction band discontinuity. 2 2 1.1.2 Subs trate Due to the lack of availability of thick GaN substrate, HEMTs are usually grown on substrates such sapphire, SiC or Si. Sapphire is the most common one but its poor thermal

PAGE 22

22 conductivity hinders the electrical performance and the reliability especiall y for high power application. Devices fabricated on GaN grown on SiC usually exhibit the best performance because of the least lattice mismatch between GaN and SiC. However, SiC is a lot more expensive than that of Si or sapphire substrate. 3 Due to the readiness of Si technology, HEMT grown on Si have attracted more attentions recently. HEMTs grown on 8 inch Si (111) substrate have been demonstrated by Arulkumaran et al. recently. 3 To grow GaN on Si, AlN nucleation was generally grown on Si substrate first and AlGaN grading transition layer was grown subsequently to release the lattice mismatch. 3 , 5 Table 1 2 summarizes the material properties of Si, GaN, and sapphire. 1.1.3 Review o f HEMTs Performance The first AlGaN / GaN high electron mobility transistor (HEMT) was demonstrated by Khan et al. in 1993 . 6 The current density was just 50 m A / mm and the transconductance was 28 mS/mm . After that, many studies have been devoted to improve the quality of the epitaxy layers, process techniques or even propose new structures. 7 Till now, the record current density for AlGaN/GaN HEMT is 2.9 A / mm fabricated by Cao et al. 5 Additionally , a current gain cutoff frequency (f T ) of 22 5 GHz with a gate length (L G ) of 55 nm 8 and a power gain cutoff frequency (f max ) of 300 GHz with a gate length of 60 nm 9 have been demonstrated in AlGaN / GaN HEMTs. Sour ce drain breakdown voltage of 2200 V was obtained at L GD = 20 µ m. 10 1.1.4 Issues 1.1.4 .1 Self heating effect Despite good performance of GaN based device, the long term stability and reliability remain major concerns especially at higher power densities. 6 During device operation, the power is generated at device active area and causes local Joule self heating. This phenomenon usually becomes even worse due to the poor thermal co nductivity of sapphire (~30 W / mK). Although

PAGE 23

23 Si substrate has better thermal conductivity (~149 W / mK) as compared to sapphire, there is a huge thermal resistance layer of AlN nucleation layer between the Si substrate and GaN. This nucleation layer is ve ry defective and thermally resistive with a thermal resistance of 7 × 10 8 m 2 K / W. 11 Figure 1 2 show s the typical drain I Vs and transfer characteristics of AlGaN / GaN HEMT degraded with self heating effect; negative drain output resistance. Not only does the output current reduce because of self heating effect, the carrier mobility drop because of phonon scattering. 8 Kuzmik et al . reported the channel temperature of HEMTs on sapphire substrate will reach ~320 o C at an output power of 6 W/mm. 12 Chang et al . simulated the thermal behavior e relationship between electron mobility and temperature, as shown in Figure 1 3 . 14 Several methods including simulation and experimental methods have been used to estim ate the junction temperature of AlGaN / GaN interface including finite element method , 15 DC characterization , 16 IR microscope , 15 scanning thermal microscopy and Micro Raman spectrum . 17 Simms et al. utilized Micro Raman to measure the channel temperature along source and drain contact. 18 Kuzmik et al. reported the channel temperature of AlGaN/GaN on Si substrate was around 75ºC at a power density of 5 W / mm by electrical method while it was 250ºC on sapphire substrate. 16 Singhal et al. studied the thermal performance by Infrared (IR) microscopy on AlGaN / GaN grown on Si substrate. It was found out the channel temperature was around 200ºC at a power density of 22.5 W / mm. 19 Although there are severa l studies about the thermal performance of AlGaN / GaN on different substrates, there are few effective solutions to reduce the junction temperature yet. 1.1.4 .2 Breakdown voltage Based on the GaN breakdown field, the off state drain breakdown voltage of an AlGaN/GaN HEMT with a drain to source distance (L DS ) of 10 µm should be 1.01 MV / cm ×10

PAGE 24

2 4 µm × 10 4 cm / 1µm = 1010 V. Till now, the reported breakdown voltage of L DS = 10 µm is 500 V, which is still below the limit of GaN. 11 , 21 However, breakdown voltage of 600 V or higher for applications i n switching mode power supplies and inverter systems is needed. 20 , 22 23 By increasing the drain to source distance, the breakdown voltage could be increased but the device on resistance would also be sacrificed. As a result, methods t o improve the breakdown voltage , other than just increas ing the drain to source distance have to be developed. The reason that the breakdown voltage is lower than theoretical value is the existence of local peak of electric field at the corner of gate close to drain electrode. One way t o solve this issue is to implement a field plate (FP), which is a metal plate placed on top of an insulator protruded from gate 12 or source 24 electrode. The length of the field plate (L FP ) is an important parameter which influences the extent of breakdown voltage improvement. 12 Lu et al . reported electric field distribution along gate to drain by a 2 dimensional simulation. The device structure and electric field distribution along gate to drain is shown in Figure 1 4 . The breakdown voltage was improved from 55 V to 150 V after implementing a 1 µ m length source field plate at drain t o source distance equals to 3.9 µ m. Ando et al. reported an improvement of breakdown voltage from 50 V to 150 V after employing 1 µ m wide gate field plate. 12 Besides, the breakdown voltage improvement reached a ma ximum at L FP = 1 µ m but decreased after that. Another important factor which influence s the breakdown voltage is the material quality of GaN. Residual impurities, presumably oxygen and silicon, 25 in a unintentio nally n doped GaN buffer have been identified as a main source of buffer leakage. 26 These residual donors could be compensated by deep acceptor states such as Fe or C. 27 , 28 Carbon doped (C dop ed) or iron doped (Fe doped) GaN buffer layer provides traps that closed to the mid band gap. These traps can help to grab charges and made the buffer layer more resistive. By growing t hicker buffer layer, the

PAGE 25

25 dislocation defects could be reduced and thus improved the breakdown voltage. Selvaraj et al. grew 1.25 to 7 µ m GaN buffer layer by Metal Organic Chemical Vapor Deposition (MOCVD) and measured the dislocation densities and breakdow n voltage. By increasing the buffer layer thickness, screw dislocation density decreased while the breakdown voltage increased from 100 V to 400 V. 29 Yu et al. reported breakdown voltage of 500 V with a gate to drain distance of 3 µ m and 2.7 µ m thick GaN buffer layer on a 4 inch Si substrate. 30 1.1. 4 .3 Current dispersion Alternating (AC) drain current of AlGaN/GaN HEMT is usually less than the direct current (DC) for high frequency application. This phenomenon, commonly defined as the current collapse, is caused by the formation of a virtual gate between gate to drain electrode because of the existence of surface states. 31 W hen V DS is biased at a higher voltage, hot electrons can be trapped by the surface states and formed the virtual gate. 31 These charged traps form an additional depletion region between gate and drain electrode. Under high frequency operations, those trapped charges cannot respond to the AC signal. To eliminate the current dispersion, passivation layer such as SiN or Al 2 O 3 was utilized to passivate the surface states. 32 33 Liu e. al. reduced current dispersion at frequency of 5 MHz using PECVD grown SiN / ALD deposited A l 2 O 3 composite material. 33 Besides, Romero et al. reported the passivation effect could be improved by in situ lower power N 2 plasma pretreatment prior to SiN deposition. 34 1.2 Dissertation Outline This dissertation covers three main topics; the fabrication of AlN based metal insulator semiconductor high electron mobility transistors (HEMTs), the irradiation effects on GaN based HE M T s and the novel structure to improve self heating effect and increase off state drain breakdown voltage.

PAGE 26

26 Background knowledge on the III V compound semiconductor transistors, especially, the properties and current status of gallium nitride (GaN) based high electron mobility transistors, as well as the reliability issues hindering the further development of GaN HEMTs technology are presented in Chapte r 1. The methodology including fabrication process, material and device characterization that are used in this thesis is reviewed in Chapter 2. The next several chapters cover research studies on the topics of fabrication of HEMTs and characterization a nd solution to the reliability issues of GaN HEMTs outl ined in Chapter 1. Chapter 3 presents the DC performance and high frequency performance of the f abricated AlN MISHEMT. Chapter 4 examines the breakdown voltage of the fabricated AlN MISHEMT. The degrad ation of buffer oxide etchant (BOE) to Ohmic co ntacts is discussed in Chapter 5 in order to understand the mechanism of BOE to device performance and optim ize our process flows. Chapter 6 examines the mechanism of the breakdown voltage improvement after pr oton irradiation through back side proton irradiation. Chapter 7 discusses the effect of low dose gamma irradiation on DC performance and high frequency response. Chapter 8 reviews the effect of defects generated after electron irradiation on the device pe rformance of AlGaN and InAlN based heterojunctions. Chapter 9 reports the thermal response of proposed AlGaN / GaN HEMTs with a back side Cu via under the device active area. Chapter 10 discusses the DC performance and reliability improvement of the propos ed AlGaN / GaN HEMTs structures. Chapter 1 1 provides a summary and conclusion for all the topics discussed in this dissertation.

PAGE 27

27 Table 1 1 . Summary of material properties of Si and AlGaN/GaN heterojunction. Material Si AlGaN / GaN B and gap (eV) 1.1 3.49 Electron mobility (cm 2 / V s) 400 1400 S aturated electron velocity ( × 1 0 7 cm / s) 1 2.7 C ritical breakdown voltage (MV / cm) 0.3 1.3

PAGE 28

28 zzz Figure 1 1 . A) Origin of 2DEG 1 B) Energy band gap diagram of AlGaN and GaN heterojunction. 50 A B

PAGE 29

29 Table 1 2 . Summary of material properties of Si, sapphire and SiC. Si S apphire SiC Thermal conductivity (W / m K) 149 30 300 Lattice mismatch (%) 17 3.5 14 Substrate cost (US $ / piece) ~ 4 ~100 ~200 Integra tion capability High Low Moderate

PAGE 30

30 Figure 1 2 . Drain I V and transfer characteristics of AlGaN / GaN HEMTs on sapphire and on silicon . 8

PAGE 31

31 Figure 1 3 . Calculated electron mobility as a function of temperature . 14

PAGE 32

32 Figure 1 4 . 2D simulations of the electrostatic field distributions between source and drain contacts for H EMT with and without source field plate. 24 Source Gat e Drain Insulator Field plate

PAGE 33

33 CHAPTER 2 2. METHODOLOGY REVIEW 2.1 Semiconductor Fabrication Processes 2.1.1 Isolation Isolation is the process to reduce the leakage current between devices through GaN buffer layer. Isolation could be either achieved by mesa etching or ion implantation. Poor isolation might induce issues such as high leakage current and low breakdown voltage. A good isolation requires the leakage current ac ross a 5 µm gap to be in around 10 9 A scale. Mesa etching technique involves defining regions surrounding the active devices with a mask layer, and subsequently etching away the exposed area to form isolated islands or mesa For AlGaN / GaN HEMTs, mesa etching is generally performed by rem oving AlGaN layer and some of the underneath GaN buffer layer . Mesa etching needs to be well calibrated, serious undercut might create leakage path and fail the isolation. 35 The schematic view of mesa etching is shown in Figure 2 1 A . Another common method of isolation is ion implantation. The isolation i s achieved by implanting high energy ions such as H, He, N, O, Zn to the device. By a Monte Car l o based simulation software, Stopping and Range of Ions in Matter (SRIM), vacancy concentrations created by the implantation and the stopping range of the vacancies could be estimated. As a result, by varying ion energy or dose, specific stopping range or vacancies could be achieved. Multiple implantation s were used to achieve uniform vacancies distribution in the device . 37 The schematic view of mesa etching is shown in Figure 2 1 B. The major concern of implantation isolation is the therma l stability. Lo. et al. reported that t he isolation effect will be removed after annealing at temperature above 600 o C. 37

PAGE 34

34 2.1.2 Photolithography Photolithography is one of the most critical processes in semiconductor fabrication. It is a process which allows patterns to be transferred to the device accurately. The process requires photoresist, mask, exposure tool, and developer solution. The photolithography process could be further divided into spin coating, exposure and resist developing. Photoresist (PR) is usually composed of a base resin, photo active compound (PAC), and organic solvent. The resin could be p oly (methyl methacrylate) (PMMA), p oly (methyl glutarimide) (PMGI) , novolac resin, or epoxy. Based on the s olubility of PR after exposure, the PR could be further divided into positive PR or negative PR. For positive PR, the dissolution rate in development solution increased after exposure. For negative PR, the dissolution rate in development solution decreased after exposure. By putting a mask, which has opaque and transparent pattern s , on top of the photoresist, the pattern on the mask can be transferred to the photoresist. Figure 2 2 shows the difference of positive a nd negative photoresist. The first step of the photolithography process is spin coating. The liquid PR is usually dispensed at the center of the wafer and spread out by the spinner. The range of the thickness of a PR can achieve is mainly determined by i ts viscosity but it can be further adjusted by the parameters of spin coating. The main rotation speed and time c an be used to adjust the thickness of the PR within its specific thickness range . Other fact ors su ch the acceleration speed , the exhaust of the spinner, and the pretreatment of the wafer c an affect the uniformity of the thickness. P rebake, which is also called soft bake, is a thermal treatment to vaporize some of the organic solvent in the resist after spin coating . I t affects the profile of the pattern , the adhesion of the PR to the substrate , and the solubility after exposure . The prebake temperature also affects the thickness of the PR. The second step of the photolithography process is exposure. The exposure is usually

PAGE 35

35 performed under the wav elength ranging from 193 to 465 nm. The common exposure tools are in the form of contact aligner, stepper or scanner . Only contact aligner is discussed here because it is the tool used in this thesis. By setting the power of the mercury lamp and exposure t ime , PR is exposed with a specific wavelength ultra violet (UV) light . If the UV power or exposure time is not enough, the resist might result in underdevelopment (UD). If the light intensity or exposure time is too high or too long, respectively, the feat ure size on the resist might be bigger than the size on the mask. To keep mask clean is also very important. If there w ere particles on the sample or mask, the contact between mask and wafer would have a gap to induce light diffraction and affect resist pr ofile and pattern resolution . 2.1.3 Lift off Lift off is a technology that can pattern a target material on the surface of the substrate. The gate, Ohmic me tal, and final metal pattern were all fabricated through the standard lift off process in this thesis . Sta ndard photolithography is patterned first and metal contacts are deposited on top of patterned PR. After metal deposition, the samples are soaked in solvent such as acetone to lift off the unwanted area of PR together with the metal on top of it. The thick ness of metal is preferred to be thinner tha n 1/3 of the thickness of PR to assist lift off process . 39 Lift off process is not generally employed in the silicon process due to the following issues. First, the PR pattern is very critical in lift off process . An overhang structure of the top surface layer of PR as illustrated in Figure 2 3 A is preferable. However, it is very difficult to fabricate an overhang structure by a simple photoli thography process. As a result, toluene soak ing or lift off resist (LOR) is generally employed to assist the lift off process. If the side wall of the PR is tapered, the deposited metal on the side wall is relatively thick. The continuous metal would be ea sily lifted off with no pattern be en generated. Another disadvantage for the tapered angel side wall is the generation of the ear structure, which is a very tall metal at the edge of the pattern . The ear

PAGE 36

36 structure might induce bad contact in the following photolithography processes. Figure 2 3 B shows the effect of side wall profile to lift off process. Last, the removed metal might be re deposited on to the surface. It is very difficult to remove these particles after the wafer ha s been dried. 2.1.4 Electron beam evaporation One of the most common metal dep osition method s is e beam evaporation. The schematic view of an evaporation system is shown in Figure 2 4 . 39 The system is usually consisted of a metal target, a crucible , a shutter and a vacuum system. The source metal, which usually sits in a thermal stable carrier called crucible , is heated above its melting point by electron beam bombardment. The chamber is usually connected to either diffusion pump or cryo pump to achieve ultra high vacuum. The samples are located on a dome shaped sample holder directly on top of the target . Because the deposition chamber is under an ultra high vacuum ( ~10 8 torr ), the evaporated atoms can travel at line trajectories all the way to the sample s . Electron beam evaporation is a very useful process for pure metal deposition, but it is difficult for alloy deposition. 2.1.5 Rapid thermal annealing Rapid thermal anneali ng (RTA) is usually used to decrease the resistance of metal alloys contact to semiconductor such as Ohmic contacts or remove the defects generated after irradiation or etching process during the device fabrication. The system usually includes a heating la mp or laser which allows a short time temperature ramping, gas purging line, and thermal couple to monitor the temperature. For Ohmic contacts annealing, the annealing temperature and ramping rate are very critical to achieve low resistance metal contact o n semiconducto r . Besides , to ensure low resistan ce of the Ohmic contacts, the oxygen

PAGE 37

37 concentration inside the RTA system needs to be monitored. The low oxygen concentration is achieved by constantly purging N 2 in the system. 2.1.6 Plasma enhanced c hemical vapor deposition Plasma enhanced c hemical vapor deposition ( PE CVD) is the most common process for low temperature passivation layer deposition such as SiO 2 or SiN. T he plasma enhanced mode can reduce the thermal budget a lot by introducing plasma energy . In this dis sertation, PECVD is chosen to prevent high temperatu re deposition to avoid degradations such as loss of isolation and metal diffusion in the gate electrode . For SiN deposition , silane, NH 3 and N 2 are introduced simultaneously first and plasma is ignite d subsequently. The reaction s are as follows: A RF voltage is biased on the top electrode, while the bottom electrode is grounded. The RF voltage causes a plasma discharge between the electrodes. Wafers are place d on the bottom electrode , which is heated between 100 to 400 o C. There are several parameters needed to be adjusted to optimize the film properties including refractive index (RI) , density and etching resistivity. The composition of silane , NH 3 , and N 2 is a critical parameter because it affects the film properties such as etch rate in buffered oxide etchant and refractive index. The deposition rate generally increased with increasing temperature and RF power. The advantage of PECVD is also its disadvantage. Because of the low temperature deposition, low tensile stress (~2 × 10 9 dyne/cm 2 ) film can be prepared. However, the SiN film deposited by PECVD usually contains 20 25% hydrogen. 40

PAGE 38

38 2.1.7 Atomic layer deposition Atomic layer deposition (ALD) is a thin film deposition that is based on the sequential gas phase chemical process es on a solid surface . Figure 2 5 shows the schematic view of the plasma enhanced ALD (PEALD) process. For the plasma enhanced AlN deposition mode, trimethylaluminum ( TMA ) and N 2 /H 2 are used as precursors for Al and nitrogen, respectively . In one cycle of the deposition , TMA is first flowed into the system and attached to sample surface with Ar gas constantly purging the excessive TMA. As a result, there is just one atomic TMA layer uniformly adsorbed on the wafer surface. Following the TMA deposition, N 2 and H 2 gas es are introduced into the deposition chamber and the plasma is ignited . The nitrogen will react with Al and break the Al C bond. AlN is thus formed after this cycle. The thickness of the AlN for each deposition cycle is around 1 Ã… . Because the deposition is self limiting and a mono layer by mono layer process , it can achieve a very conformal deposition even on wafer s with high aspect ratio pattern s . 41 The temperature of ALD process is usually relatively low b ecause the metal organic precursors might decompose at high temperature. For example, t he temperature for ALD process using TMA as a precursor is usually below 300 o C. 44 This eliminates the potential thermal degradation to the device such as Ohmic metal degradation . 2.1.8 B osch etching B osch process is a dry etching p rocess which features in high aspe ct ratio Si etching. It is named after a German company Robert B o s c h Gmbh which patented the idea. 45 This process uses a two step process which alternates between deposition and etching. The deposited layer is usually a chemical inert passivation layer such as the dielectric deposited by CF 4 . This passivation layer can provide protection to the side wall for the following etching process. The result is an anisotropic etch that ca n have a vertical sidewall regardless of the orientation of the silicon crystal. Although a noticeable rippling of the surface can be found on the sidewalls , t he

PAGE 39

39 Bosch process is currently the most popular process for deep silicon etching. It provides a ve ry consistent etch ing with a near 90 degree sidewall. 2.2 Material Characterization 2.2.1 Hall measurement Hall measurement is a method to estimate the carrier type, carrier concentration and carrier mobility. T he system includes a magnetic field, current supply an d a voltage monitor. Figure 2 6 shows the schematic view of a Hall measurement system setup. As shown in Figure 2 6 , the current is applied in x direction and the magnetic field is applied in z direction. 46 Because of Lorentz force, the majority carrier in the semiconductor will be swept to +y direction. Thus, by determining the sign and the value of the voltage drop between +y an d y direction, the carrier type and carrier concentration can be calculated. The equation to calculate carrier concentration is as sh own in Equation 2 1. If the sheet resistance of the semiconductor is known, the mo bility can be estimated from Equation 2 2 as well. ... Eq. 2 1 w here n s is sheet density of carrier concentration I is the current applied B is the magnetic field applied q is elementary charge V H is the Hall voltage me asured ................................................................... Eq. 2 2 w here V H is the Hall voltage I is current applied B is magnetic field applied

PAGE 40

40 Rs is the sheet resistance of the semiconductor 2.2.2 Transmission electron microscopy Transmission electron microscopy (TEM) is a method which uses high energy focused electron beam to examine a very thin (below ~100 nm) structure. T he electron beam passes through the thin sample and interact s with the specimen and forms the image. To get high resolution image, the TEM system is usually operated under pressure range at 10 7 to 10 9 Pa. The resolution of TEM is around 1 2 Ã… and can provide information such as crystallographic phas e, crystallograp hic orientation . 47 2.2.3 Scanning electron microscopy and energy dispersive X ray spectroscopy Scanning electron microscopy (SEM) is a method which uses high energy focused electron beam to examine the surface topogra phy of the sample. Under the incident of focused electron beam, secondary electron s or back scattering electron s emit from the excited atoms. By analyzing the intensities of backscattering electron s or secondary electron s , the atomic number s and the topogr aphy of the analyzed surface can be determined , respectively . There are generally two types of detectors for SEM system. O ne of the detectors is called backscattering electron detector (BSD) which collects the back scattering electron. The BSD is usually located on top of the sample. The probability of backside electron generated is in proportional to the atomic mass of the element. In other words, the numbers of backside electron been collected are in proportional to the atomic mass. As a result, by the c ontrast of the image, the relative atomic mass distribution can be estimated. The other detector is secondary electron detector (SD). The energy of secondary electron s is generally lower than 50 eV. A s a result, the electrons that are collected by the SD o riginate within a few nanometers from the sample surface. Besides, t he brightness of the signal depends on the number of secondary electrons reaching the detector. As a result, the number of secondary electrons reaching the SD is very sensitive to the

PAGE 41

41 topo graphy . The image generated from SD thus provides a well defined, 3 dimensional appearance . 49 Use secondary detector, t he resolution of SEM image can be down to 1 nm. In SEM system, there is usually embedded an energy dispersive X ray spectroscopy (EDX, EDS, or XEDS). It is a technique generally used in semi quantitative elemental analysis. The analysis is performed by analyzing the X ray intensity and energy generated from the interaction of incident electro n beam and sample. Each element has its own specific spectrum based on its atomic number and chemical state. EDX can provide elemental analysis on areas as small as nanometers in diameter and thickness with micrometer scale . 49 With the assistance of SEM, it can provide an elemental and spatial analysis on the sample which is especially good for defect analysis . 2.2.4 X ray photoelectr on spectroscopy X ray photoelectron spectroscopy (XPS) is a surface sensitive quali tative and quantitative spectroscopic technique that obtains the elemental composition , chemical state and electronic state of the elements that exist within a material. D ifferent from EDX, X ray is used instead of focused electron beam to excite the atoms . The detector analyzes the electron excited from the top 10 nm of the sample surface. X ray incident knocks the 1s electron, so electrons at other higher energy states such as 2s or 2p state jump from excited states to ground state. During this process, s pecific energy will be released. Because of this, e ach element except hydrogen and helium thus produces a characteristic set of XPS peaks at characteristic binding energy . The characteristic spectrum thus identifies each element in different electron state s. The number of electrons is directly related to the amount of elements in the sample. Generally speaking, t he resolution of the atomic concentration is around 1000 ppm. 49

PAGE 42

42 2.2.5 Auger electron spectroscopy Auger Ele ctron Spectroscopy (AES) is a surface sensitive spectroscopic technique used for elemental analysis of surfaces. I t offers high sensitivity of all elements except of H and He. AES can be further divided into three steps which are atom ic ionization, electro n relaxation and analysis of the emitted auger electrons. The atomic ionization is the process of incident electron to remove the core 1s electron. After the removal of core electron, the electrons at higher energy states jump to fill the cavity of the 1s vacancy. To compensate for the energy change from this transition, an Auger electron or an X ray is emitted. By analyzing the energy of the emitted Auger electron, the characteristics spectrum is generated. For light elements, the probability is greatest f or emission of an Auger electron, which accounts for the light element sensitivity for AES . Similar to XPS, the chemical state of the atoms can also be determined from energy shifts and peak shapes. Because of the energy of Auger electron is relatively low , only the Auger electrons from top 1 nm surface will have sufficient energy to escape the surface and reach the detector. as shown in Figure 2 7 . 50 Utilizing this fact wi th the ion milling process, elemental depth profile can be generated. 2.2.6 E llipsometer Ellipsometer is a way to measure thin film thickness and refractive index (RI) by an optical method. The thickness range for measurement is around 1 nanometer to few hund red nanometers. The technique uses plane polarized monochromatic light source w ith wavelength ranging 200 to 1050 nm to illuminate the thin film/ s ubstrate at an angle. Based on the angle and intensity of the reflective light measured by the detector, the R I and thickness can be back calculated. Multi stack of thin films can also be measured if the model is constructed appropriately. 50

PAGE 43

43 2.2.7 Four point probe measurement Four point probe measurement is a technique to mea sure the material sheet resistance. The schematic view of four point probe measurement is illustrated in Figure 2 8 . 52 The outer two probes are connected a current suppl y to supply current (I) while the inner two probes connect to a high impedance voltage meter to monitor the voltage drop (V). Each probe has the same spacing which is around 1mm . Depends on the sample geometries, edge effect, and probe spacing (s) , correc tion factor (F) needs to be considered. The correction factor will be usually provided by the tool vendor. The sheet resistance , , can be calculated by . 2.3 Device Characterization 2.3.1 DC performance DC performance is one of the most important c haracteristics of a field effect transistor . It is generally composed of drain I V, gate I V and transfer characteristics. The DC performanc e is obtained by connecting a FET to a three channel DC parameter analyzer. Source is grounded and drain, gate is co nnected to channel which supplies voltage and monitors c urrent simultaneously . Drain I V is a plot shows modulation of gate voltage (V G ) to the relationship of drain source current (I DS ) versus drain source voltage ( V DS ). Figure 2 9 shows a typical drain I V characteristic of a n AlGaN/GaN FET. From drain I V characteristics, several parameters can be obtained , such as drain saturation current, drain to source resistance, and the mobility . Saturated drain source current (I DSS ), whic h is the maximum I DS of a FET , can be obtain ed . The drain current is usually normalized to gate width , leading to a unit of mA/mm . From the low field drain I V curve s , the resistance of drain to source and the mobility can be obtained . The resistance of dr ain to source (R DS ) is the reciprocal of the slope of the low field drain I V curve.

PAGE 44

44 If the resistance of drain contact (R D ), source contact (R S ), and threshold voltage (V TH ) is known, the mobility can be estimated by Equation 2 3. 53 . 2 3 where V DS is drain source voltage I DS is the drain source current R S is the resistance between gate to source R D is the resistance between dra in to gate L is the gate length d is the thickness of AlGaN layer µ is the mobility W is the gate width is the permittivity of AlGaN layer V G is the gate voltage V TH is the threshold voltage Gate I V is the rel ati onship of gate current to gate voltage (V G ). To measure gate I V, gate electrode is swept from deep depletion region to forward turn on voltage while drain electrode is floated. From gate I V , the reverse leakage current, S chottky barrier height (SBH) c an be obtained. When the V G is biased at deep depletion, the gate current is called the leakage current. By plotting the log (I G ) with V G , the SBH can be obtained by Equation 2 4. Eq. 2 4 W here k is Boltzmann constant 1.38 × 10 23 [J/K] T is absolute temperature [K] q is elementary charge 1.6 × 10 19 [ q ] A is gate area

PAGE 45

45 A* is Richard constant which is 31 AlGaN and 26 for GaN, respectively I 0 is t en to the power of y intercept of regression line on forward gate I V part Another important performance is transfer characteristics. To measure transfer characteristics, the drain voltage is biased at fixed voltage at device saturation region and gate voltage is swept from off region to on region with both drain current and gate current being monitored. From transfer characteristics, transconductance (Gm), subt h reshold swing (SS), and threshold voltage (V TH ) can be obtained. Transconductance, which represents the amplification of a FET, is defined as the derivative of I DS to V G at fixed V DS . As shown in Figure 2 10 A, the Gm peaks at around V G =1.2 V. To obtain SS, semi log curve of I DS versus V G is required. In the subt h reshold curve as shown in Figure 2 10 B, the slope is the subthreshold slope. T he reciprocal of subt h reshold slope is subt h reshold swing, which shows the effectiv eness of the gate control. The unit of SS is mV/dec, and SS represents the voltage required to change one decade of the drain current. 54 As shown in Figure 2 11 , by taki ng square root of the I DS and linear regression at forward turn on region , the threshold voltage can be obtained. 55 2.3.2 Transmission line method Trans mission line method (TLM) is a set of test key used to estimate the contact and sheet resistance of a FET. The common TLM test key is designed as shown in Figure 2 12 . 55 Rec tangular metal pads were deposited and separated in differe nt gaps. The resistance across different gaps are measured by an hp parameter analyzer. By plotting the resistance versus gap spacing, t he sheet resistance (Rsh) , transfer resistance (R T ) and transfer length (L T ) can be extracted from the linear regression of the curve. Utilizing this technique, the source resistance (R S ) and drain resistance of a FET can be calculated from Equation 2 5. Eq . 2 5

PAGE 46

46 w here Rs is source resistance Rsh is sheet r esistance of semiconductor L is gate length W is the distance between gate to source Z is the length of Ohmic contact 2.3.3 Gate lag measurement The gate lag measurement is a technique that can evaluate the drain current dispersion and characteriz e the defects in the semiconductor. The system setup includes a pulse generator, power supply, resistor and an oscilloscope. Gate is connected to a pulse generator, and source is connected to ground. A DC power supply is connected to the resistor which is connected in series with the drain electrode. By reading the voltage drop across the resistor by the osc illoscope, the current can be calculated . Figure 2 13 shows the setup of gate lag measurement. When the device is pulsed from off to on at high frequency , the trapped charges can t respond due to large time constant of the defects. As a result, the drain current measured in AC mode is lower than that measured in DC , which is known as current dispersion. Usually th e drain current dispersion from AC to DC is more severe when the drain voltage or the frequency is higher. The amount of current dispersion from DC to AC current can represent the amounts of the defects. 56 2.3.4 Off s tate drain b reakdown voltage Off state drain breakdown voltage is a very important characteristic to evaluate the performance of FET in high power application. The off state drain breakdown voltage is usually performed at a fixed V G which sets the device b e operate d at deep depletion state with V DS gradually increases. When the I DS or I G suddenly increases dramatically , the voltage is recorded as off state drain breakdown voltage (V BR ).

PAGE 47

47 2.3.5 Deep level transient spectrum Deep level transient spectrum (DLTS) is a useful technique to characterize and identify the defects. DLTS establishes fundamental defect parameters and measures their concentration in the material. Some of the parameters are considered as defect "finger prints" used for their identifications and analysis. It utilizes the fact that the RF capacitance of the sample depends on the charge state of deep levels in the space charge region. The device capacitance is monitored with time with the gate constantly being pulsed. T 1 and t 2 is chosen to estimat e the recovery of capacitance after the pulse. The rate window (e) , which is the reciprocal of time constant, can then be calculated by Eq. 6. By plotting the rate window versus temperature , the activation energy can be obtained by Equation 2 6 as well. 57 .................................................. ............................. Eq. 2 6 w here e is the rate wind ow is the time constant N c;v is effective density of states in the conduction band or the valence band C n;p is capturing coefficient g is th e degeneracy factor for the lev e l, usually assumed to be 1 kT is the thermal energy E t is the acti vation energy 2.3.6 Admittance spectr um Admittance spectrum is another method to obtain the thermal activation energy of the defect . It mainly measures the change of admittance in the semiconductor when the carriers are captured or emitted from the defects. In admittance spectrum measurement, the diode is treated as an equivalent circuit which has a capacitor and resistor in series. When the angular frequency is , the capacitance an d admittance are as shown in Equation 2 7.

PAGE 48

48 Eq . 2 7 w here Cp is the measured capacitance C is the real capacitance is the angular frequency Gp is the measured conductance is the time constant which equals to R × C When equal to 1, the G p / reaches a maximum and there exists a n inflection point in the capacitance versus temperature plot . Substitute , T where local maximum Gp / happens into Equation 2 8 , the thermal activation energy can be obtained . 58 . Eq . 2 8 where is the coefficient which is independent of temperature E A is activation energy T is temperature k is Boltzmann constant 2.3.7 Capacitance voltage measurement Capacitance voltage measurement is a very powerful technique which can gives users information about doping type, carrier concentration, and interfa ce state densities. The C V curve is usually obtained with a C V meter, which applies a DC bias voltage and a small sinusoidal signal (1 kHz 10 MHz) to the capacitor and measures the capacitive current with an AC ammeter. When the Schottky contact on AlGa N/GaN HEMT is biased at 0 V , the depletion region barely touches 2DEG. As a result, the capacitance measured is mainly contributed by AlGaN

PAGE 49

49 barrier layer. Utilizing this fact, the thickness of AlGaN layer can be calculated within 10% accuracy by Equation 2 9 . 2 9 With the bias being more negative , the depletion region penetrates through 2DEG. In the point when the depletion region completely penetrates through the 2DEG, the capacitance drops dramatically. At deep depletion, the capacitance measured can be viewed as two parallel capacitors connected in series. Figure 2 14 shows the typical C V curve of AlGaN / GaN HEMT. 70 Besides the bar rier layer, t he carrier concentration distribution can also be generated by C V measurement. At each capacitance and voltage, the space charge region, W , can be generated by Eq. 9 and the carrier concentration can be generated by Equation 2 10 . C arrier con centration distribution can thus be obtained. Eq. 2 10 Last, the existence of traps can also be characterized by the capacitance voltage curve. Typically, the defects can be categorize into four types including mobile charges mobile charges (Q m ), fixed charges at the surface (Q f ), or fixed charges distributed in oxide layer (Q ot ) , and interface charges (Q it ). F or an ideal C V curve, the stretching at the transition region is almost zero. However, the C V curve might be parallel shifted because of Q m , Q f or Q ot and the transition region might be stretched because of Q it . In Figure 2 15 , the non ideal CV curve is shown. Figure 2 15 B curve is the ty pical C V curve resulted from the existence of Q m , Q f or Q ot . As shown in Figure 2 15 , both shifting and stretching happens on C curve. The mechanism of shifting is the same as B curve while the stretching is due to Qit. By readin g the voltage shift, the interface densities can be estimated. Figure 2 16 shows the typical location of mobile charges

PAGE 50

50 (Q m ) , fixed charges at the surface (Q f ) , fixed charges distributed in oxide layer (Q ot ) , interface traps (Qit) in a MOS diode. 2.4 Finite element method Finite element method (FEM) is a numerical method to solve differential equations. It uses variation method and small elements, meshes, to approach the solution numerically. If the meshes are small enough, the c hange of the parameters inside this mesh can be viewed as linear. The process of solving is an iterative process until the error function is minimized. With proper boundary condition and mesh size, the solution can be fairly accurate.

PAGE 51

51 Figure 2 1 . Schematic view of isolation by A) m esa etching and B) ion implantation. A B

PAGE 52

52 Figure 2 2 . Schematic view of photolithography process of A) positive or B) negative photoresist. 13 A B

PAGE 53

53 Figure 2 3 . Effect of photoresist profile A) ideal overhang structure and B) tapered angle profile to lift off perf ormance. B A

PAGE 54

54 Figure 2 4 . Schematic view of an evaporation system. 39 e

PAGE 55

55 Figure 2 5 . Schematic view of ALD process.

PAGE 56

56 Figure 2 6 . Schematic view of Hall measurement system setup . 46 y x z

PAGE 57

57 Figure 2 7 . Gener ation of Auger electrons, backscattered electrons by the incident electron. 50

PAGE 58

58 Figure 2 8 . Schematic view of four point probe measurement. 52

PAGE 59

59 Figure 2 9 . Typical drain I V characteristic s of a field effect transistor.

PAGE 60

60 Figure 2 10 . Transfer character istics of A) transconductance characteristics and B) threshold voltage determination. 54 B

PAGE 61

61 Figure 2 11 . Threshold voltage deter m ined by the satur ation extrapo lation technique. 55

PAGE 62

62 Figure 2 12 . T he configuration of TLM test key and data analysis. 54 Figure 2 13 . T he system setup for gate pulse measurement .

PAGE 63

63 Figure 2 14 . High frequency C V curve for a Schottky diode on a HEMT. 70 C 2DEG

PAGE 64

64 Figure 2 15 . Effect of a fixed oxide charge and interface traps on the C V characteristics of MOS diode . 40

PAGE 65

65 Figure 2 16 . Location of the ch arges associated with a MOS diode . 40

PAGE 66

66 CHAPTER 3 3. GAN METAL INSULATOR SEMICONDUCTOR HEMTS WITH PLASMA ENHANCED ATOMIC LAYER DEPOSITED ALN AS GATE DIELECTRIC AND PASSIVATION 3.1 Introduction t o Metal Oxide Semiconductor High Electron Mobility Transistors AlGaN/GaN high electron mobility transistor (HEMT) performance has made remarkable progress in recent years, showing great promise for applications such as militar y radar and satellite based communications systems . 1 , 22 Conventional Schottky metal gate HEMTs a re expected to offer the highest frequency performance , however, issue s such as current collapse, high gate leakage and questionable long term reliability have limited the potential for high power applications. To resolve the gate leakage problem, SiO 2 , 23 Si 3 N 4 , 59 , 60 Al 2 O 3 , 61 67 Sc 2 O 3 , 68 HfO 2 , 69 , 70 La 2 O 3 , 72 and TaO x N y 73 based metal oxide semiconductor (MOS) or metal insulato r semicondu ctor (MIS) HEMTs have been employed . These gate dielectrics may also be used for device passivation to alleviate the issue of current collapse . 69 However, inserting a gate dielectric has typically resulted in su ppression of transconductance, g m, and a negative threshold voltage shift . 66 , 74 Us e of dielectrics with high permittivity could help solve these problems, because a l arger dielectric constant translate s to more efficient gate modulation and thus a smaller decrease in g m and threshold voltage shift . 69 73 AlN with a band gap of 6 .2 eV and a relatively high permittivity of ~8.9 is another candidate for the gate dielectric on nitride based HEMTs . AlN deposited by a molecular beam epitaxy system (MBE) was used to replace an AlGaN for an AlN / GaN based HEMT to achieve higher sheet ca rrier concentration due to a larger conduction band difference between AlN and GaN . 66 , 72 73 Selvaraj et al. used metal or ganic chemical va por deposition (MOCVD) to grow an AlGaN / GaN HEMT structure capped with a 2 nm AlN layer as the gate insulator and the passivation layer. 74 The HEMTs showed a very low gate leakage current of 5 × 10 5 mA / mm ;

PAGE 67

67 however, the DC performance suffered from a relative ly mm due to the incorporation of the AlN cap layer. Atomic layer deposition (ALD) is a surface controlled layer by layer deposition process. Each atomic layer formed in sequence is the result of saturated surface controlled chemical reactions . 60 , 75 F ilm s deposited by ALD have been demonstrated to have low defect densi ty, high uniformity and p recisely controlled thickness at the nanometer scale . 67 ALD deposited AlN has also been used as a passivation layer for AlGaN/G a N HEMT s, which significantly reduc ed current collapse . 76 To date, there is no report of ALD deposited AlN as a gate insulator for AlGaN / GaN HEMTs. In this study we report the DC and power performance of AlGaN/GaN MIS HEMTs employing ALD AlN as the gate dielectric an d surface passivation layer. X ray photoelectron spectroscopy ( XPS ) and Auger electron spectroscopy (AES) were us ed to analyze the AlN films. DC performance and gate pulse measurement s were conducted to investigate the effectiveness of AlN as the gate insu lator and passivation layer. 3.2 Experimental 3.2.1 Material Growth HEMT structures were grown on 3 inch c plane Al 2 O 3 substrates using a metal organic chemical vapor deposition system, starting with a thin AlGaN nucleation layer followed with a 2 defect , car bon doped GaN buffer layer, a 23 nm undoped GaN layer, a 21.2 nm undoped AlGaN layer with a 25.6 % Al mole fraction and capped with a 2.5 nm undoped GaN layer. Hall measurements on the as grown structures showed sheet carrier densities of 1.2 × 10 13 cm 2 an d the corresponding electron mobility of 1050 cm 2 / V s. 3.2.2 HEMTs Fabrication Device fabrication began with Ohmic contact deposition with a standard lift off e beam

PAGE 68

68 evaporated Ti / Al / Ni / Au based metallization, and the samples were subsequently annealed a t 850°C for 30 s under N 2 ambient. A typical transfer resistance of 0.4 mm was obtained using the t ransmission line method (TLM). Multiple energy and dose nitrogen implantation was used for the device isolation, and photoresist AZ1045 was used as the mas k to define the active region of the devices. Isolation currents were less than 10 nA at 40 V of bias voltage across two 100 µm × 100 µm square Ohmi c contact pads separated by a 5 µm implanted gap . 75 Prior to Al N based gate insulator deposition, the sample was treated with UV ozone for 3 minutes and followed by 30% NH 4 OH for 1 min. to remove the surface contaminations and improve the adhesion . 10 nm AlN was deposited in a Cambridge Nano Fiji 200 remote RF plasma e nhanced atomic layer deposition (ALD) system at 300 ºC. 30 , 44 T ri methyl aluminum (TMA) was used as Al precursors and N 2 / H 2 was used as N precursors. Ar was used as ca rrier/purging gas. For plasma enhance ALD, N 2 / H 2 flowed into the chamber first and t hen 300 W of RF power was used to ignite t he plasma and kept on for 40 s. The growth rate of AlN film was 0.73 Å / cycle . The exposure time for TMA w as set at 0. 06 second s and the growth temperature w as kept at 300°C. After ALD AlN, 1 µm × 200 µm gates were achieved via standard lift off e beam deposited Pt / Ti / Au metallization . Ti / Au based metallization was utilized for the interconnect metals. The source to gate dis tance was 1 µm . 3.2.3 Device Characterization The AlN surface and bulk film analys e s we re carried out with an X ray photoelectron spectroscopy ( XPS ) utilizing a Perkin Elmer 5100 system and by Auger electron spectroscopy utilizing a Perkin Elmer 660 system. A J. A. Woolam EC110 ellipsometer was used to measure the refractive index (RI) and the thicknesses of the AlN films. The DC characteristics of the AlN MIS HEMTs were measured with a Tektronix curve tracer 370A and an HP 4156 parameter

PAGE 69

69 analyzer. For g ate lag measure ment , the VDS was kept at 5 V . T he gate pulse was created by A gilent 81104A pulse generated with 10% duty cycle and frequency of 100 kHz. 3.3 Results a nd D iscussion Figure 3 1 shows the optical microscope picture of AlN deposited on Ohmic metal. As shown in Figure 3 1 , the AlN pe eled off from the edge of the Ohmic metal. The peeling might be due to the stress induced poor adhesion from the AlGaN surface to AlN film. UV ozone treatment was performed to remove the possible carbon contamination to clean the surface. However, as shown in Figure 3 1 B, there is no noticeable improvement after ozone treatment. NH 4 OH treatment was known to remove surface oxidation such as Al and Ga oxide. 41 , 42 As a result, the sample was treated with NH 4 OH prior to AlN deposition. As shown in Figure 3 1 C, there is no peeling at the Ohmic edge after O 3 UV and NH 4 OH treatment. The refractive index (RI) of a 40 nm Al N film grown on a Si substrate was 1.92 at a wavelength of 632.8 nm, which was slightly lower tha n the reported value, 2.1 . 30 , 44 The atomic concentration on the AlN s urface layer was oxygen measured with XPS, as shown in Figure 3 2 A . To further investigate the content of the AlN film, high resolution scans of the Al 2p peak and N 1s peak were used as illustrated in Figure 3 2 B and Figure 3 2 C . The Al 2p photoelectron peak at 73 eV was attributed to Al N bonds . The N 1s photoelectron peak was deconvoluted into 2 subpeaks with binding energies o f ~396.07 eV and ~397.9 eV, respectively. The peak at 396.07 eV was attributed to Al N bonds while the 398 eV peak was attributed to Al O N bonds . 44 Based on the XPS surface analysis, Al 2 O 3 was definitely prese nt on the surface of the ALD AlN film. The theoretical RI of AlN and Al 2 O 3 were 2.1 and 1.72, respectively. However, t he measur ed RI of the ALD AlN was 1.92. T he lowering of the RI indicated the existi ng of oxygen in the AlN film . Auger electron spectrosco py of AlN film depth profiling was also performed to determine the oxygen throughout the ALD AlN film, as shown in Figure 3 3 . The oxygen signal

PAGE 70

70 disappeared at 1 2 nm depth from the top of AlN surface. This Al 2 O 3 l ayer was formed because of the exposure of AlN to the ambient. Leskelä et al . also reported a similar result that a 5 10 nm thick Al 2 O 3 layer formed after the AlN film was exposed to air . 75 Figure 3 4 show s the drain I V and transfer characteristics of the ALD AlN / AlGaN / GaN MISHEMTs. As illustrated in Figure 3 4 A , the drain current could be modulated to + 4 V of the gate voltage with a sharp channel pinch o ff at 3 V of the gate voltage, and the maximum saturation drain current was around 600 mA / mm. The drain current of conventional metal gate HEMTs can typically only be modulated up t o around +2 V of gate voltage. In other words, by introducing AlN, the g ate voltage modulation was e xtended around 2 V. This degree of extension was limited by the breakdown fi e l d of gate insulator . The breakdown field of unannealed ALD AlN was reported around 1.7 2.2 MV/cm , 44 , 76 thus 10 nm of ALD AlN should provide an additional 2 V of gate voltage to the ty pical Schottky barrier height. The breakdown field of the crystalline AlN have been reported as much higher values around 10 12.8 MV / cm . 78 , 79 As a result, i t is possible to change the process sequence by depositing the ALD AlN right after the Ohmic metal deposition and annealing both the Ohmic c ontacts and ALD AlN at the same time to improve the ALD AlN crystallinity and breakdown strength. A maximum extrinsic transconductance of 127 mS / mm was obtained; this was comparable to 133 mS / mm for the Schottky gate based HEMT. The reduction in transc onductance was due to the increased gate channel distance from inserting the dielectric layer under the gate contact . 61 , 73 There was a negative threshold voltage shif t of 0.23 V observed for ALD AlN MISHEMT as compared to that of the Schottky gate HEMT. Figure 3 4 B also show s the sub threshold drain leakage current, which strongly depend ed on the reve rse bias gate leakage current. The ALD AlN MISHEMT, with a 10 nm ALD AlN gate, exhibited a very low sub threshold drain leakage current of 1.13 × 10 9 (A / mm) at V G = 3.5 V and resulted in a large

PAGE 71

71 drain current on off ratio of 3.3 × 10 8 . Below the threshold voltage, the drain current increased e xponential ly with the gate voltage. The slope of this exponential increase on a logarithmic scale was defined as the drain current sub threshold slope, which has been used to quantify trap densities in the gate modulated region of metal insulator semicondu ctor field effect transistors . 79 , 80 The drain current sub threshold slope was also dominated by the reverse bias gate leakage current . 79 , 80 A drain current sub threshold slope of 76 mV / decade was obtained, indicating that a good quality AlN / AlGaN interface was achieved. To investigate the effectiveness of ALD AlN for surface p assivation in these structures, we employed gate lag measurements. Figure 3 5 show s the drain current response of an ALD AlN / AlGaN / GaN MISHEMT to a pulsed gate voltage as well as measured in DC mode. In this figure, V G was pul sed from 5 V to different gate voltages ranging from 2.2 to 0 V at different freq uencies with a 10% duty cycle. There was less than a 7% reduction in drain current measured in DC mode compar ed to the pulsed measurements. By sharp contrast, the u npassivat ed HEMTs exhibited a 50% decrease in drain current. This is clear evidence that the A L D AlN effectively mitigated the surface states to avoid current collapse.

PAGE 72

72 Figure 3 1 . Optica l microscope picture of A) reference B) treated with UV ozone C) treated with UV ozone and NH 4 OH. A B C 100 µm 100 µm 100 µm

PAGE 73

73 Figure 3 2 . XPS analysis of AlN thin film deposited by plasma enhanced mode AL D. A) Surface survey of the as deposited 10 nm AlN by PEALD. B) High resolution scan of Al 2p spectrum. C) High resolution scan of N 1s spectrum. A B C

PAGE 74

74 Figure 3 3 . Depth profile of 10 nm AlN thin fil m deposited by plasma enhanced mode ALD.

PAGE 75

75 Figure 3 4 . A) Drain current versus drain bias for different gate bias voltages. B) T ransconductance characteristics measured with the MIS HEMT in saturation A B

PAGE 76

76 Figure 3 5 . Gate lag measured on the AlN MISHEMT

PAGE 77

77 3.4 Summary In conclusion, an ALD AlN film was employed as the gate insulator in nitride HEMTs to suppress the gate leakage current and t o increase the gate voltage modulation range as well as the passivation layer to reduce drain collapse. GaN MISHEMTs with an AlN gate insulator thickness of 10 nm and 1 off ratio of 3.3 1 0 8 , a maximum transconductance of 127 mS / mm and a saturation current of 600 mA / mm at V G = + 4 V. The drain current reduction was less than 7% reduction when AlN passivation was introduced. These results indicated that ALD AlN can be used to improve th e performance of AlGaN / GAN HEMTs for high power applications.

PAGE 78

78 CHAPTER 4 4. HIGH BREAKDOWN VOLTAGE IN ALN/GAN MISHEMTS 4.1 Overview of High Breakdown Voltage HEMTs There is a strong interest in GaN based switching devices with breakdown voltages of 600 V or hig her for applications in switching mode power supplies and inverter systems where the higher figure of merits obtained relative to Si could reduce power losses . 22 , 23 T he realization of efficient GaN power devices with high breakdown could lead to smaller, more efficient power supplies. The on state resistance of GaN can be more than an order of magnitude less than Si, and combined with the higher breakdown voltage this should lead to smaller, simpler cooling systems that result in significant energy savings. These characteristics have made GaN devices potential candidates in power conditioning in large industrial motors, pulsed power for avionics and electric ships, in s olid state drivers for heavy electric motors and in advanced power management and control electronics. There are now commercially available 600 V class GaN power devices, but even higher values are desirable. The lower on resistances of AlGaN / GaN heteros tructure devices such as high electron mobility transistors (HEMTs) relative to Si transistors are due to the high carrier mobility and larger sh e et carrier concentration in the two dimensional electron gas (2DEG) channels. High breakdown voltage AlGaN / G aN devices have been demonstrated and have shown on resistances below those achieved in Si. A typical breakdown voltage for HEMTs with a common layer thickness of 2 µ m is roughly 800 V. I ncreas ing this voltage requires methods such as acceptor doping of th e buffer layer, increasing the active layer thickness, use of a recessed gate or increasing the band gap of the buffer layers through use of AlGaN. Another approach to increasing gate breakdown voltage is to use a metal insulator semiconductor HEMT (MISHEM T) structure rather than the conventional Schottky gate . 81 97 While Schottky metal gate

PAGE 79

79 HEMTs offer the highest frequency performance, issues such as current collapse , high gate leakage, and reliability problems related to the stability of the metal gate have limited the potential for high power applications. Many different dielectric layers have been demonstrated for GaN based metal oxide semiconductors (MOS) or MISHE MTs, including SiO 2 , Si 3 N 4 , Al 2 O 3 , Sc 2 O 3 , HfO 2 , La 2 O 3 , and TaO x N y . 20 , 81 97 Use of a gate dielectric with high permittivit y produces more efficient gate modulation and thus a smaller decrease in transconductance and threshold voltage shifts. Arulkumaran et al. ha s employed AlN (band gap 6.2 eV and permittivity of 8.9) to demonstrate GaN MISHEMTs . 98 Finally, the use of field plates over dielectric passivation layers can increase the device breakdown voltage. The use of two field plates, combined with a gate drain distance of 24 µm, produced a breakdown voltage of 900 V . 99 This compares to a value of 250 V for devices without field plates . 99 In this chapter , we report on the breakdown performance of AlGaN / GaN MISHEMTs employing ALD AlN as the g ate insulator and surface passivation layer, as a function of gate to drain distance without the need for field plates. 4.2 Experimental 4.2.1 Material Growth plane Al 2 O 3 substrates using a metal organic chemical vapor deposition system, starting with a thin AlGaN nuc leation layer followed with a 2 m low defect carb on doped GaN buffer layer, a 23 nm undoped GaN layer, a 21.2 nm undoped AlGaN layer with a 25.6% Al mole fraction and capped with a 2.5 nm undoped GaN layer . Hall meas urements on the as grown structures showed sheet carrier densities of 13 and the corresponding electron mobility of 1050 cm 2 / V s.

PAGE 80

80 4.2.2 HEMTs Fabrication Device fabrication began with Ohmic contact deposition with a standard lift off e beam evapor ated Ti / Al / Ni / Au based metallization, and the samples were subsequently ann ealed at 850°C for 30 s under N 2 mm / were obtained using the transmission line method (TLM). Mu ltiple energy an d dose nitrogen implantation w ere used for the device isolation, and photoresist AZ1045 was used as the mask to define the active region of the devices. Isola tion currents were less than 10 nA at 40 V of bi m 2 Ohmi c contact pads separated by a 5 m implanted gap. Prior to the AlN gate insulator deposition, device was treated with UV ozone for three minutes and subsequently treated with 30% NH 4 OH for 1min to remove surface contamination. It is imperative that t he nitride surface be as clean as possible to achieve a high quality interface with the gate dielectric . 73 The AlN was deposited in a Cambridge Nano Fiji 200 remote RF plasma enhanced ALD system at 300°C. The gr / cycle . T ri methyl aluminum (TMA) and N 2 / H 2 plasma were used as the precursors for Al and N, respectively, while Ar was used as the carrier / purging gas. For plasma enhance ALD, N 2 / H 2 flowed into the chamber first and kept on for 40 s. The exposure time for TMA was set at 0.06 s and the growth temperature was kept at 300°C. The AlN had a refractive index of 1.92 and no detectable oxygen was measured in the film by Auger Electron Spectroscopy. The breakdown field of the AlN was of the order of 2 MV / cm. Aft er 10 nm ALD AlN deposition, 1 m × 200 m gates were achieved via standard lift off e beam deposited Pt / Ti / Au metallization and Ti / Au based metallization was utilized for the interconnect metals. The source drain distance was varied from 5 40 µm, while the gate source distance was kept at 1.5 µm.

PAGE 81

81 4.2.3 Device Characterization The DC characteristics of the AlN MISHEMTs were measured with a Tektronix curve tracer 3 70A and an HP 4156 parameter analyzer. 4.3 Results a nd D iscussion Figure 4 1 sho ws drain current versus source drain voltage characteristics and off state drain breakdown voltage as a function of drain bias for differe nt gate bias voltages for devices with drain source distances of either 10 µm or 40 µm. The AlN layer not only served as the passivation layer but also as a gate insulator . The gate was able to be modulated up to +2 V and both devices showed a pinch off ar ound 3 V of the gate voltage. Under pulsed conditions, the drain current was typically ~7% less than under DC conditions , showing the effectiveness of the AlN as a surface passivation layer as a gate d ielectric. The suppression of DC to RF dispersion or s o called current collapse is a key issue for the GaN based devices and the fact that the AlN can perform as both gate dielectric and surface pass ivation layer is advantageous. Off state drain breakdown voltage was measured at a gate voltage of 6 V. The of f state drain breakdown voltage increased from 500 to 2000 V, when the drain to source distance of the devices increased from 10 to 40 µm. With a lager gate drain distance, parasitic resistance an d knee voltage would increase. Owing to the low sheet resist ance of AlGaN / GaN MISHEMT, 500 / , t he HEMT on resistance only increased from 1.3 to 10.9 m / cm 2 for the devices with the drain source distance increasing from 10 to 40 µm. The knee voltage increased from 4.8 to 9.5 V for the device with t he larger gate drain distance. Usually the devices with a larger knee vol tage suffer with a low output radiofrequency (RF) power efficiency. A higher drain off state breakdown voltage allows device to be op erated a larger voltage range. Since the drain off state bre akdown voltag e of the 40 µm devices improved significantly from 500 to 2000 V, the impact of the knee voltage could be neglected.

PAGE 82

82 Figure 4 2 show s transfer characteristics, drain and gate for MISHEMTs with drain so urce distance of 10 µm or 40 µm measured at drain voltage of +5 and +10 V, respectively. The peak transconductances were 119 and 90 mS / mm for MISHEMTs with drain source distance of 10 µm and 40 µm, respectively. The reduction of the peak trans c o nductance was due to the increase of gate channel distance by inserting the dielectric layer under the gate contact. A lthough the MISHEMT peak trans c o nductance was lower as compared to the peak transocnductance of Schottky gate based HEMT, a much broader transc o ndu ctance curve was achieved with MISHEMTs which resulted from a higher gate turn on voltage. Both peak transconductance and saturation drain current were lower for the device with 40 µm drain source distance owing to larger parasitic resistance. Figure 4 3 show s the saturation drain current as a function of the drain to source distance. The saturation drain current is inversely linear proportional to the drain to source distance and the slope of the fitting line depen ds on the sheet resistance of the HEMT structure. As a result of 10 nm AlN gate insulator, these MISHEMTs exhibited a very low gate current in the range of 1.2 × 10 9 A / mm at V G = 3.5 V and V D = + 5 or + 10 V for MISHEMTs with drain source distance of 10 µm and 40 µm, re spectively. Since the sub threshold drain leakage current is dominated by the gate leakage current, the lower gate leakage current of the AlN based MISHEMTs achieved a large drain current on off ratio of 3 × 10 8 . Lower sub threshold drain l eakage not only increases the power added efficiency, linearity, and noise figure of the power amplifiers, it also significantly improves the drain current on off ratio of the HEMT. Figure 4 4 A show s off state drain source chara cteristics as a function of gate voltage for individual devices with different drain source distance, while Figure 4 4 B show s the off state drain breakdown voltage as a function of gate drain distance (L GD ) at a fixed gate vol tag e of 6

PAGE 83

83 V. The latter reache d a value of 2000 V a t a source drain distance of 40 µm without the use of field plates. This show s that the combination of a MIS gate and increasing the source drain distance is capable of producing high breakdown voltage devic es with relatively low on state resistances. The off state drain breakdown voltage was linearly proportional to the gate drain distance and the off state drain breakdown strength derived from Figure 4 4 B was 0.5 MV/cm. Figure 4 5 show s specific on state resistance, R on , as a function of breakdown voltage . 11 , 79 , 100 106 T he th eoretical limit line of GaN w as calculated using the equation R on = 4 V BR 2 / ( c 3 ) , where is the dielectric constant, is the electron mobility and E c is the electrical strength of GaN . 107 The specific on resistance of these devices was calculated based on the total resistance extrac ted from the low field I V characteristics in Figure 4 1 and the active area between the source and drain contacts including a 1 m transfer length from the contact pads. The specific on state resistances were 1.3, 3.7, 6.6 and 10 cm 2 for devices with gate drain distances of 7.5, 17.5, 2 7.5 and 37.5 µm respectively. The increase of specific on resistance for the larger gate drain devices was resulted from the parasitic res istance of the HEMT structure. The sheet resistance of the AlGaN / GaN used in this work and typical InAlN / GaN based HEMT structure are 500 and 200 / . By employing InAlN/GaN HEMT structure, the specific on resistance can be reduced more than 60%.

PAGE 84

84 Figure 4 1 . Drain current as a function of drain bias for different gate bias voltages for devices with drain source distance of (A) 10 µm or (B) 40 µm A B

PAGE 85

85 Figure 4 2 . Transfer characteristics, drain and gat e current for devices with drain source distance of A) 10 µm or B) 40 µm measured with the MISHEMTs in saturation. A B

PAGE 86

86 Figure 4 3 . Saturation drain source cur rent as a function of drain source distance (L DS ).

PAGE 87

87 Figure 4 4 . A) Off state drain source characteristics as a function of gate voltage for individual devices and B) off state drain breakdown voltage as a functio n of L GD at V G = 6 V. A B

PAGE 88

88 Figure 4 5 . Specific on state resistance as a function of breakdown voltage.

PAGE 89

89 4.4 Summary While use of field plates is an effective method to increase HEMT breakdown voltage, there are penalties in terms of increased process complexity and degradation of high frequency performance because of the additional gate capacitance and thus limitations in terms of power switching above the 1 GH z range. Use of a gate dielectric to produce a MISHEMT structure and then increasing the gate drain spacing avoids the additional processing complexity of field plate formation and can produce high breakdown voltages. The fact that the gate dielectric can also act as a surface passivation layer is critical because while unpassivated devices show large breakdown due to the reduced field resulting from negatively charged surface traps, the current dispersion is large . The ability of AlN to increase gate break down voltage and also passivate these surface traps is an attractive approach. Breakdown voltages up to 2000 V were achieved with use of gate to drain spacing of 37.5 µm.

PAGE 90

90 5. CHAPTER 5 5. DEGRADATION MECHANISMS OF Ti / Al / Ni / Au BASED OHMIC CONTACTS ON AlGaN / GaN HEMTs 5.1 Introduction to Ti / Al / Ni / Au Based Ohmic Contacts The performance of AlGaN / GaN high electron mobility transistors (HEMT) has made remarkable progress in recent years, showing great promise for applications such as military radar and sat ellite based communications systems . 22 , 108 Due to the large energy band gap of GaN (3.26 eV) and high breakdown field, these devices are well suited to high power applications. Howev er, due to this large band gap, it is difficult to create a low resistance Ohmic metal on nitride HEMTs. The typical Ohmic contact used is a Ti / Al / Ni / Au metal stack annealed at 800 850 in N 2 ambient. By changing the annealing temperature or time , 81 a contact resistance of 3.22 × 10 7 / cm 2 was achieved for Ti / Al / Ni / Au metal stacks to AlGaN / GaN HEMT. This data is reported for measurement right after annealing and without exposure to chemicals that are part of subsequent processing steps . During the fabrication process, these contacts might be exposed to a number of chemicals, including Buffered Oxide Etchant (BOE) which is an acidic buffered solution. As a result, it is important to know the perfor mance change of Ohmic metal and possible degradation mechanisms after such BOE exposure. Buffered oxide etchant is a typical etchant to remove the native oxide or passivation layers such as silicon oxide or silicon nitride . 82 It is a mixture of HF and NH 4 F, which gives a stable etching performance by acting as a buffering agent to maintain pH. Based on their standard electrode potentials, most of the metals including Ti, Al, Ni are anodic and will react with ac id solution to form hydrogen gas . 83 , 84 The etching rate of HF for sputtered Ti wa s reported to be faster than 10 kÅ / min . 84 For another common contact metal, Al, the etching rate in 5:1 BOE solutions was reported to be 11 nm/min . 85 The etching rate of 5:1 BOE solution for Ni and Au is almost zero . 85 During the opening of the passivation layer on HEMTs, the Ohmic metal is

PAGE 91

91 exposed to the BOE solution and it is inevitable it will react with HF. As a result, care must be taken to open the passivation window but not damage the Ohmic metal. T here is no study on changes in Ohmic metal performance of Ti / Al / Ni / Au contacts on nitride HEMTs after BOE treatment. Therefore, it is important to study this performance change after BOE treatment. In this work, we monitored the resistance change of Ti / Al / Ni / Au Ohmic metal when exposed to BOE by four point probe measurements and the transmission line method (TLM) . After treatment, energy dispersive X ray spectroscopy (EDX), Auger analysis was used to quantitatively determine the composition chan ge. Possible mechanisms are discussed to explain the performance changes after BOE treatment. 5.2 Experimental HEMT structures were grown on Si substrates using a metal organic chemical vapor deposition (MOCVD) s ystem, starting with a thin Al N nucleation laye r followed with a 2 low defect carbon doped GaN buffer layer, a 23 nm undoped GaN layer, a 21.2 nm undoped AlGaN layer with a 18% Al mole fraction and capped with a 2.5 nm undoped GaN layer. Ohmic contact was deposited with a standard lift off e beam evaporated Ti / Al / N i / Au (45 nm / 125 nm / 45 nm / 100 nm) based metallization, and the samples were subsequently annealed at 850°C for 1 min under N 2 ambient. After annealing at 850 °C for 1 min under a N 2 ambient, the sheet resistance of the Ohmic metal from four point pro / . The Ti / Al / Ni / Au O hmic metal was treated in BOE for 3 min , with the sheet resistance being monitored every 15 sec by f our point probe measurements . Scanning electron microscope (SEM) with either backscattering elect ron detector or secondary electron detector was used to examine the surface morphology change before and after BOE treatment. EDX was used to quantify the composition change before and after BOE treatment. Lastly, Auger depth profiling was used to analyze the composition within the metal film.

PAGE 92

92 5.3 Results a nd D iscussion Figure 5 1 show s the sheet resistance (Rs ) of alloyed Ti / Al / Ni / Au Ohmic metallization as a function of BOE treatment time. T he metal sheet resistance gradually inc rease d / after 180 sec of BOE treatment. The BOE solution was comp osed of a 6:1 volume ratio of 40% NH 4 F in water to 49% HF in water. Au should be inert in BOE. However, since the standard reduction potentials of Ti, Al and Ni are all n egative, oxidation and dissolution processes are favorable for these three metals . The reaction of these three metals with BOE might lift Au metal and result in the increase of metal sheet resistance after treatment in BOE. Broad area EDX scan was used to estimate the composition change of alloyed Ohmic metallization after BOE treatment , as shown in Figure 5 2 , and the composition of each element was averaged for individu al content across an area of 30 µm by 30 µm . Unexpectedly, Au decreased by aro und 7% and Ga increased by 4%. The content of Ti decreased by 0.1% and the contents for the rest of elements, Al, Ni and N, increased by 0.2, 1.3 and 0.6%, respectively. For EDX, t he X rays are generated in a region about 2 µm in depth of the sample. The thickness of the Ti / Al / Ni / Au Ohm ic metallization was around 0.3 µm, and thus the EDX signals of Ga and N should be dominated by the bulk GaN even with some Ga or N out diffusing through the alloyed Ohmic metallizati on. Thus, it is reasonable to assume that the Ga and N contents did not change significantly. The percent increases of Ga and N content after BOE treatment were due to t he decrease of the Au content. By comparing the increases of percent content of Al, it was less than th e percent increase of Ga or N. Therefore, besides Ti and Au, the content of Al also decreased after BOE treatment. However the increase of Ti / Al / Ni / Au Ohmic metallization sheet resistance should be mainly caused by the decrease of Au content in the BOE treated sample.

PAGE 93

93 Figure 5 3 show s optical microscope (OM) pictures of alloyed Ohmic metallization before and after BOE treatment, and pictures for the same samples taken with SEM using a secondary electron detector as well as a bac kscattering electron detector. There were no clear differences for the pictures taken with the optical microscope between Figure 5 3 A the reference and Figure 5 3 B BOE treated sample, and both pictures exhibited very rough surface morphology across the entir e alloyed Ohmic metallization. However, the SEM pictures using the secondary electron detector clearly exhibited 3 distinct regions o n the alloyed Ohmic metal lization. There were islands bulging up and surrounded by a ring, and there was a flat surface between these islands. The SEM pictures also revealed a couple of distinct differences between Figure 5 3 C the reference and Figure 5 3 D BOE treated samples. First, the surface of the islands was etched and becam e rougher after BOE treatment. Besides etching the surface of islands, those rings surrounding the islands also got etched and became narrower. The height of islands was reduce d from around 300 nm to 200 250 nm after BOE treatment, as measured with an Alpha Step profiler. Further, SEM pictures taken with backscattering electron detector also exhibi ted 3 distinct regions on the alloyed Ohmic metal lization; dark color black islands surrounded by a brighter ring, and islands connected with a gray color area, as shown in Figure 5 3 E and Figure 5 3 F . T he signal intensit ies detected by the backscattering electron detector are proportional to the atomic number s; the brighter areas are dominated by heavier elements, such as Au in our system, and the darker areas should comp rise lighter elements, such as Al . Thus, those darker color islands ranging from 3 to 5 µ m must have Al, Ni and Ti bounded by a ~ 1 µm wide ring containing Au . The gray color region, defined as the field region, could have both a heavier element, Au, as we ll as li ghter elements, Al, Ti and Ni. It was reported that Ni Al intermetallic phases react to form bumps and are bounded by Au Al intermetallic phases. 86

PAGE 94

94 Based on these results, some of the Al and Ga were etch ed off by BOE. Au in the ring area could be removed by the etching of Al underneath this layer. Figure 5 4 illustrates the EDX mapping of Au, Ga, Al, Ti, and Ni for reference and BOE treated samples. Al and Ni are mainly located in those islands, and Au and Ga mostly contained in the ring and field areas, respectively. Ti, on the other hand, is uniformly dispensed in the ring and field areas and scattered in the islands. After BOE treatment, the contrast between Au based ring s and the island area s bec a me less obvious, which were consistent wit h the result s obtained by SEM. The intensities of the Ga signal in the ring areas also decreased after BOE treatment. The Al signal generally decreased for the entire areas, whi ch was in line with Auger surface scan result shown in Figure 5 5 . For the as alloyed sample, Ti tanium (Ti) , Aluminum ( Al ) , carbon (C) and oxygen (O) were present on the surface for all three regions. After BOE tre atment, Al was removed and Ga was detected in the island areas. To investigate the depth dependent composition of alloyed Ti / Al / Ni / Au Ohmic metallization in island, ring and field regions between islands, Auger depth profiling analys e s were performed . Figure 5 6 A shows the Auger depth profiling of the island area on the untreated sample . The surface region consisted of C, O, Ti and Al, followed with a 40 nm Ti layer and a layer of 200 nm Ni Al alloy. Ga clear ly diffused throughout the Ni Al alloy and Ti layer. It has previously been reported that a Ni A l intermixing layer in these types of contacts was form ed after annealing to minimize the interfacial energy. 87 , 88 After BOE treatment, the surface Al was removed, as shown in Figure 5 6 B . There was no clear effect of BOE treatment on both Ti and Ni Al alloy layer. However, there wa s a Ti peak appearing at the Ni Al alloy and GaN interface for the reference sample, which was not obse rved for the untreated sample. Zhou et al. reported that TiN based contact inclusion (CIs) form at the GaN surface. 86 The diameter and density of

PAGE 95

95 CIs were around 100 nm and 2×10 7 / cm 2 , respectively. The Auger beam spot size used i n this study was around 35 nm. Thus, it is possible that the CI region might be missed during the Auger depth profiling for the untr eated sample. Figure 5 7 A show s Auger depth profile of the ring area for the untreated sample . There was a surface layer containing C, O, Ti and Al, followed with a Ti layer mixed with Au Al alloy layer , a thicke r Au Al layer , and a n interface layer between Au Al Ti alloy layer and GaN. After BOE treat ment, not only was the surface Al was removed, but also some Ti and Au Al alloyed layers were etched off, as shown in Figure 5 7 B . This is consistent with EDX data which Au decreased around 7%. Although Au is quit e stable with acid solution, t he etching of surface Al and Al in Au Al alloy layers might take away Au. The removal of Au on these surface layers increased the Ohm ic metallization sheet resistance. Figure 5 8 shows the Auger depth profiles of the field region of an untreated sample and BOE treated sample . There was a similar surface layer as the one on island and ring areas, c ontaining C, O, Ti and Al, foll owed with a Ti layer mixed with Au Al alloy layer , a Ti layer on the top of GaN. A fter treatment, the surface Al was etched off. There were no changes on the other layers.

PAGE 96

96 Figure 5 1 . Sheet resistance of alloyed Ti / Al / Ni / Au based Ohmic metallization as a function of BOE treat ment time measured with the four point probe technique.

PAGE 97

97 Figure 5 2 . Perce ntages of metal elements in alloyed Ti / Al / Ni / Au metallization prior to and after BOE treatment, measured by e nergy dispersive x ray spectroscopy (EDX).

PAGE 98

98 Figure 5 3 . OM picture A) before and B) after treatment; SEM pictures from SD C) before and D) after treatment ; SEM pictures from BSD E) before and F) after treatment. B 100 µm ( A) 100 µm E F A C D 5 m 5 m 10 m 10 m

PAGE 99

99 Reference After BOE treatment Au Ga Al Ti Ni Figure 5 4 . EDX elemental mappings of alloyed Ti / Al / Ni / Au met allization prior to and after 18 0 sec of BOE treatment. 10 m

PAGE 100

100 Figure 5 5 . Auger surfa ce scan analyses of island, ring and field regions for A) untreated sample and B) BOE treated sample . A B

PAGE 101

101 Figure 5 6 . Auger depth profile of the island area on A) untreated sample a nd B) BOE treated sample . A B

PAGE 102

102 Figure 5 7 . Auger depth profile of the ring area on A) untreated sample and B) BOE treated sample . A B

PAGE 103

103 Figure 5 8 . Auger depth profile of the field area on A) untreated sample and B) BOE treated sample . A B

PAGE 104

104 5.4 Summary The degradation of Ohmic metal lization dipped in BOE w as studied. The sheet resistance increase d significantly after treatment in BOE for 3 minutes. Moreover, after annealing, there were island like structure s surrounded by Au Al alloy rings and a field area between the islands . The BOE etching occurred mainly at the island and ring area s instead of the field area between the islands. The incr ease of sheet resistance was due to the etching of surface Al and Ti and the loss of Au in the island and ring areas .

PAGE 105

105 CHAPTER 6 6. EFFECT OF BACKSIDE PROTON IRRADIATION ON ALGAN / GAN HEMT ON OFF STATE DRAIN BREAKDOWN VOLTAGE 6.1 Introduction to Proton Irradia tion AlGaN/GaN devices successfully proved that they could outperform conventional silicon based devices in high power, high frequency and high temperature applications (ex. high temperature gas sensors, base station in wireless communication, weather fore casting system, space communication system etc.) . 1 , 109 111 Since the lattice constants of wurtzite GaN is a = 3.19Å c = 5 .19 Å, which is much smaller than conventional Silicon (5.43 Å) and GaAs (5.65 Å), III nitrides can demonstrate exceptional tolerance under high energy proton radiations . 112 Typically, proton irradiated AlGaN/Ga N high electron mobility transistors (HEMTs) exhibited degradation of Schottky barrier height, gate leakage, saturation drain current and extrinsic transconductance (g m ), whose magnitude depends on the proton energy and dose . 113 117 However, it was recently reported that the off state drain breakdown voltage and the critical voltage during the off state drain voltage step stress improved in the proton irradiated HEMTs . 112 , 118 , 119 The defects created by the high energy proton irradiation were evenly distributed through out the entire HEMT structure. Since the proton irradiation created defects in the GaN buffer layer or in the 2DEG , these effects could not be separated and, it wa s very difficult to identify the cause of drain breakdown voltage and step stress critical voltage improvement th rough the proton irradiation. In this work, protons were irradiated from the backside of the HEMT active area through via holes et ched through the Si substrate. Proton energies of 275 and 330 k eV with the doses of 4 × 10 12 cm 2 to 5 × 10 12 cm 2 , respect ively, were used to place the defects created by proton irradiation at different positi ons within the HEMT structure. Drain current and off state drain breakdown voltage were measured before and after proton irradiations to examine the impact of

PAGE 106

106 proton irr adiation on device dc performance. The F l orida O bject O riented D evice and P rocess S imulator ( FLOODs) Technology Computer Aided Design (TCAD) finite element solver was employed to simulate the electrical field around the gate edge under the influence of pro ton irradiation induced defects placed at specific locations inside the AlGaN / GaN HEMT structure. 6.2 E xperimental 6.2.1 Material Growth The AlGaN / GaN HEMT structure was grown by metal organic chemical vapor deposition (MOCVD) on Si wafers with conventional pr ecursors in a cold wall, rotating disc reactor designed from flow dynamic simulations. The epitaxi al layer structure included an AlGaN transition layer , 15 and an ~ 800 nm thick Ga N buffer layer capped with a 16 nm unintentionally doped Al 0.26 Ga 0.74 N barrier layer. 6.2.2 HEMT s Fabrication HEMT fabrication began with Ti / Al / Ni / Au Ohmic metallization and rapid thermal annealing in flowing N 2 at approximately 825ºC. Device isolation was achieved with multiple energy and multiple dose of N + ion implantations. Plasma enhanced chem ical vapor deposited (PECVD) 70 nm silicon nitride was used for device passivation. Schottky gate were defined on the SiN x layer by patterning the gate and selectively removing the SiN x passiva tion layer. After SiN x etching, wider gate patterns were redefined with another photolithography step and Ni / Au based gate metallization was deposited as the gate metallization . The wafers were then passivated with another 400 nm layer of PECVD SiN x at 3 00ºC. There was an additional metal deposition for the HEMT with the source field plate. The field plate was connected to the so urce terminal and extended by 1 µm out over the gat e to the gate to drain region. R ectangular via holes, which covered the entir e device active area, were fabricated from the back side of the sample (the Si substrate side), by etching through the Si substrate and stopping on the AlGaN

PAGE 107

107 transition layer with a standard B osch process using a Surface Technology Systems (STS) deep react ive ion etching (DRIE) system. Figure 6 1 shows the TEM cross section view of the device used in this study.

PAGE 108

108 Figure 6 1 . Cross sectional TEM image of AlGaN/Ga N HEMT structure .

PAGE 109

109 6.2.3 Device Irradiation Simulation a nd Experiment Proton irradiations were performed with proton energy at 275 and 330 keV and proton doses of 4 × 10 12 cm 2 and 5 × 10 12 cm 2 , respectively, to achieve the same peak vacancy concentration from the back side of the samples. The implanted samples were intentionally tilted 7° away from the normal direction of the proton beam to prevent the channeling effect. The Stopping and Range of Ions in Matter (SRIM) simulator was used to estimate the penetra tion depth of the protons into the AlGaN / GaN HEMT structure and the distributions of the Ga and N vacancies generated through proton irradiations. 6.2.4 Electrical Simulation The FLOODS TCAD finite element solver was employed to simulate the electric field aro und the gate edge under the influence of proton irradiation induced defects placed at specific locations inside the AlGaN / GaN HEMT structure. To model irradiation induced changes, the Ga and N vacancy concentrations, Trap , estimated with the SRIM simulat ion were included in the Poisson equation as shown in Equation 6 1 . Eq. 6 1 n is the density of el ectrons, p is the hole density, Doping is the ionized pre irradiation acceptor or donor densities, Trap is the ionized post irradiation trap concentration, q were simultaneo usly solved with the Poisson equation. 6.2.5 Device Characterization D evice DC performance was measured with an Agilent 4156C parameter analyzer, and the off state drain breakdown voltage measurements was conducted with a Glassman high voltage power supply.

PAGE 110

110 6.3 Res ults a nd D iscussion Figure 6 2 shows the schematic view of the vacancy distribution among the device a fter irradiation at 275 and 330 keV. As shown in Figure 6 2 , the tails of the vacancy distribution pro files dropped very sharply . Utilizing this feature, by irradiating the device from the backside of the wafer, the defects could be placed at specific region, i.e. transition layer, buffer layer or 2DEG layer. As a result, 275 keV and 330 keV were chosen to put the defects at buffer layer and 2DEG layer respectively. The purpose of selecting doses at 5 × 10 12 cm 2 or 4 × 10 12 cm 2 was to create a similar peak vacancy density as the vacancies created with the conventional MeV implantation from the front side of the sample. Figure 6 3 show s the drain I V characteristics before and after irradiation with 330 keV protons from the backside of the samples. The drain voltage was swept from 0 V to 3 V, and the corresponding drain current was measured at different gate voltage ranging from 0 to 2 V with a step of 0.5 V. As illustrated in Figure 6 3 , the drain current suffered a 13% reduction fo r the HEMTs irradiated with 330 keV prot ons, demonstrating that the proton irradiation induced defects placed in the 2DEG region and AlGaN barrier layer degraded the drain current. On the contrary, no drain current degradation was observed fo r the HEMTs irradiated with 275 keV protons fr om the b ackside of the sample. This indicates that the proton irradiation induced defects placed in the GaN buffer do not affect the drain current. Figure 6 3 also show s the off state drain breakdown volta ge of the un irra diated and 330 keV proton irradiated HEMTs. These proton irradiated HEMTs exhibited an improvement of off state drain breakdown voltage, similar to the previously reported MeV proton irradiated HEMTs. 112 , 117 There was no such improvement of the off state drain breakdown voltage observed for the HEMTs ir radiated with 275 keV protons. Thus, the defects placed in the GaN buffer layer by 275 keV irradiation only decreased the residual conductivity of the u n intentional doped GaN buffer. On the contrary, the

PAGE 111

111 vacancies placed in the 2DEG cha nnel and AlGaN barrier with 330 keV proton irradiations not only removed carriers and decreased the drain current but also could form a v irtual gate resulti ng from the charged vacancies. Such a virtual gate could change the electric field distribution around the gate edges. By reducing the maximum electric field under the drain side of the gate edge, the off state drain breakdown would be i mproved. Figure 6 4 show s the simulated electric field distributions around the gate edge for the proton irradiated and the pre irradiated HEMTs. By introducing a virtual gate in the GaN buffer layer assuming 4% of the vacancies created by the proton irradiation with a dose of 4 × 10 12 cm 2 become negatively charged traps, the simulated peak electric field at the gate edges only lessened around 1% for the post irradiated HEMT as compar ed to the pre irradiated HEMT. Therefore, it seemed unlikely that the increase of off state drain breakdown voltage could be attributed to the electric field reduction at the gate edges by charged defects in the GaN buffer. By sharp contrast, by introducing the same amount of negative trap charges in troduced by the irradiation at the Al GaN / GaN interface, the simulated electric field at t he gate edges was reduced 50%. Thus the improvements of the off state drain breakdown voltage for 330 keV proton irradiations are consistent with the simulated resul ts.

PAGE 112

112 Figure 6 2 . Schematic of backside proton implantation through via hole and proton irradiation induced vacancy distributions as a function of proton penetration depth. Si

PAGE 113

113 Figure 6 3 . Drain I V characteristics of the HEMT before and after irradiat ed with a proton energy of 330 k eV and a dose of 5 × 10 12 cm 2 .

PAGE 114

114 Figure 6 4 . E lectric field distributions around the gate edge for the proton irradiated and the pre irradiated HEMTs.

PAGE 115

115 6.4 Summary In summary, we studied the effect of proton irradiations on DC performance and off state drain breakdown voltage of AlGaN / GaN HEMTs grown on Si substrates. Proton irradiations were performed from the backside of the samples through via holes fabricated on the Si substrate. There was no degradation observed for the proton irradiated AlGaN / GaN HEMTs in which the d efects created by the proton irradiations were intentionally placed in the GaN buffer. On the contrary, the irradiated HEMTs with the defects placed in the 2DEG channel region and AlGaN barrier using higher energy protons showed degradation of drain curre nt and ex trinsic transconductance. FLOODS TCAD finite element simulations were performed to confirm the hypothesis of a virtual gate formed around the 2DEG region to reduce the peak electric field around the gate edges and increase the off state drain brea kdown voltage.

PAGE 116

116 CHAPTER 7 7. EFFECT OF LOW DOSE GAMMA IRRADIATION ON DC PERFORMANCE OF ALGAN / GAN HIGH ELECTRON MOBILITY TRANSISTORS 7.1 Introduction to Gamma Irradiation AlGaN / GaN high electron mobility transistors (HEMTs) have received increasing attention because of their great promise for applications such as military radar and satellite based communications systems . 1 , 22 Due to their potential use in extreme radiation environments, it is important to characterize the hardness of the AlGaN / GaN heterostructure under these environments. There are many studies reporting the performance under proton , 112 neutron , 113 and electron irradiation 120 for AlGaN / GaN HEMTs. However, comparable studies for irradiation are limited. The effect of irradiation on these devices is both qualitatively a nd quantitatively different from other forms of radiation, eg. some studies report drain saturation current increases 121 , 122 while others report decreases after ir radiation . 123 125 In addition, the mechanism for the current changes is not well understood. Detailed studies are needed to characterize the performance of HEMTs afte r irradiation. The bulk of experiments with AlGaN / GaN HEMTs 121 124 and GaN 126 diodes have been carrie d out using 60 Co irradiation of various energies and doses. Compton electrons induced from radiation create electron hole pairs, thus changing occupancy of traps. Unlike proton irradiation , 112 some studies claim that these de fects can improve device performance such as increasing drain saturation current . 121 , 122 These defects are believed to be nitrogen vacancies that have electrical act ivation energies about 216 meV from the conduction band. Nitrogen vacancies act as donors and increase the effective channel doping and thus increase I DS . 121 These types of defects have been reported after low e nergy proton, electron, and irradiation . 122 , 127 129 It was also reported that low dose irradiations partially relaxed the AlGaN/GaN hetero structure

PAGE 117

117 elastic strains and enhanced the electron mobility by around 7 8% . 130 This improvement of the electron mobility could also increase the drain current. In contrast to these resul ts, Schwartz et al. showed that the drain current was reduced by about 60% after irradiation at around 700 Gy . 124 These defects in that case must reduce the carrier concentration in the irradiated devices, and there was more degradation with increasing irradiation dose. The defects produ ced by irradiation may be structure sensitive which would be one reason for the discrepancies between different reports. Also, dose clearly plays a role in the performance of the devices after irradiation. Vintusevich et al. found out that at 10 5 rad, th e Ids increased but started to deteriorate after higher doses of 10 6 rad . 129 In this study , we report on the effect of low dose irradiations on DC characteristics of AlGaN / GaN HEMTs. Drain and gate I V char acteristics, sheet resistance, sub threshold drain and gate current as well as gate lag measurements were conducted for the HEMTs prior to and after irradiations . 7.2 Experimental 7.2.1 Material Growth AlGaN / GaN HEMTs layer structures were grown on c plane Al 2 O 3 substrates by molecular beam epitaxy. The layer structure consisted of a thin AlGaN nuc leation layer followed with a 2 µm thick un doped GaN buffer topped by a 25 nm thick unintentionally doped AlGaN layer. The mobility was determined to be 1080 cm 2 / V s with a sheet carrier concentration of ~1×10 13 cm 2 by Hall measurements conducted at room temperature. 7.2.2 Device Fabrication Device fabrication started with mesa isolation. Mesa isolation was achieved by using an inductively coupled plasma system with Ar / Cl 2 based discharges. The Ohmic contacts were form ed by lifting off e beam evaporated Ti / Al / Ni / Au (200 Å / 1000 Å / 400 Å / Au 800 Å).

PAGE 118

118 The contacts were annealed at 850°C for 45 s under a flowing N 2 ambient in Heatpulse 610T system. Standard lift off of e beam deposited Ni / Au wa s used for gate met allization. 1000 Å plasma enhanced chemical vapor deposited SiNx was used for device passivation and the crossover between the gat e finger and gate contact pad. The final step was the deposition of e b eam evaporated Ti / Pt / Au (300 Å / 200 Å / 2000 Å) me tallization for intercon nection contacts. Figure 7 1 show s a micrographic image of the circular AlGaN / GaN. The diameter of t he circular gate finger was 100 µm and the gate dimension was 1.5 µm × 314 µm. The gate to source and gate to drain distance were kep t at 2 and 4 µm, respectively. Transmission line method (TLM) testers were also fabricated along with the HEMTs for monitoring sheet irradiations.

PAGE 119

119 Figure 7 1 . Device configuration of the circular HEMTs. 100 µm G D S

PAGE 120

120 7.2.3 Gamma Irradiation a nd Device Character ization Devices were exposed to 60 rays with different dos es of 50, 300, 450, or 700 Gy. Irradiations were performed at temperatures <50°C. During irradiation, the samples were held in nitrogen ambient and the electrodes of the HEMTs were floated. The device DC performance was characterized using a n Agilent 4156 parameter analyzer. 7.3 Results a nd D iscussion Figure 7 2 illustrates the drain I Vs of a typical circular HEMT before and after 450 Gy irradiation as well as drain cu i rradiation dose, respectively. In Figure 7 2 A , the drain I V curves were modulated by sweeping the gate voltage from 0 to 3 V with a step of 1 V. The drain saturation current increa sed from 293 to 339 mA/mm irradiation. As shown in Figure 7 2 B , the saturation drain current irradiations with the doses used in this experiment. As mentioned in the introduction, se irradiation. However, different mechanisms for such drain current increases were proposed. Aktas et al. reported around 10% increase of drain current at V G = 0 V and 6 V of drain volta ge irradiation, and there was no change of sheet resistance or mobility observed after the irradiation. 128 irradiation contributed el ectrons to the channel under the gate. The density of these defects are small enough not to effect the linear band bending induced by the polarization charge , 131 thus there is no effect on the conduction channel under the silicon passivated layer to vary the sheet resistance. Bert hl et et al. studied the effect of 4 irradiation on drain current, Schottky gate characteristics, sheet resistance measured with TLM testers, and light sensitivity of the drain cur rent under white light illumination . 121 radiation produced a decrease of the electrical traps and resulted in an increase of drain current, with a lower drain I -

PAGE 121

121 V on resistance . The HEMT also became less sensitive to the white light due to the decrease of the traps. There was no change for the Schottky gate characteristics or the sheet resistance of the TLM testers after irradiation. Besides the commonly observed drain current e nhancement after a radiation, Kurakin et al. reported a decrease of sheet resistance resulting from an increase of electron mobility through the relaxation of the elastic strain in the HEMT hetero structures . 130 Magneto transport spectroscopy exhibited a significant decrease in short range carrier scattering, which was also consistent with the increase of mobility measured by Hall measurements. X ray diffraction spectra revealed a shift of the peak cor responding to the (0004) plane of the AlGaN layer to a lower angle indicating a strain relaxation in the AlGaN layer after irradiation. They also employed surface curvature measurements to confirm the relaxation of the elastic strain in the HEMT hetero str uctures. radiation on HEMT performance were generally different from previously reported results. As illustrated in Table 7 1 , drain current, electron mobility, on state resistance, radiation. The electron mobility was assessed using the drain I Vs at different gate voltages in the low field region, the on state resistances were determined by dividing the device total resistance of the saturated drain I V in the low field region with the active area of the device. The source and drain resistances were estimated with sheet resistance, TLM transfer resistance and device layout configuration. Unlike the wo rk from Bert hl et et al . , in our case, the channel resistance reduction and drain current increase were responsible for the reduction of the on resistance . 121 In contrast to the work from Kurakin et al. , the shee t resistance reduction in our case can be attributed to the mobility increase in the channel . 130 In our study, the on state resistance or device total resistance reduction resulted from the decreases of source, drain and

PAGE 122

122 channel resistances, as shown in Figure 7 3 . In other words, the drain current increase and on resistance reduction in this study could be due to the combined effects of strain relaxation induced electron radiation generated donor type defects. radiation on the drain current sub threshold characteristics were also in vestigated, as shown in Figure 7 4 . The drain irradiated samples at a fixed drain voltage of 5 V, and the gate current was simul taneously recorded during the drain current measurement. The sub threshold leakage current, sub threshold slope and on/off drain current ratio are essential to the power added efficiency, linearity, noise figure and reliability of power amplifiers. 132 The current sub threshold characteristics were insensitive to radiation. The typical drain current on off ratio and the sub threshold slope for the reference HEMTs were in the range of 105 and 95 mV / dec, respectively and there was no radiated HEMTs. Figure 7 4 B

PAGE 123

123 Figure 7 5

PAGE 124

124 ; Figure 7 2 . A) Drain I V irradiation at a dose of 450 Gy. B) irradiation dose. A B

PAGE 125

125 Table 7 1 . Summary of drain current at V G = 0 V, on state resis tance, device total resistance, irradiation.

PAGE 126

126 Figure 7 3 . Source resistance, drain resistance, channel resistance and total resistance of a circular AlGaN / irradiation dose.

PAGE 127

127 Figure 7 4 . A) Sub thre shold drain and gate cu rrent of irradiation with a dose of 450 Gy. B) Gate I V irradiation with a dose of 450 Gy. A B

PAGE 128

128 Figure 7 5 . Gate lag measurement s of circular AlGaN / irradiation with different doses. D A B C

PAGE 129

129 7.4 Summary The effect of relatively low dose irradiation doses of 50, 300, 450 or 700 Gy on circular AlGaN / GaN HEMTs has been studied with dc and pulse measurements. T he increases in drain currents and the reduction of on irradiated samples could be due to the combined effects of strain relaxation induced electron mobility enhancement and additional radi ation generated donor type defects. The sub threshold drain and gate current were insensitive to the irradiation. The Schottky barrier height and ideality factor of the HEMTs were not influenced by the lose dose radiation either. There was a minimal dr radiated HEMTs due to more defects generated between gate and drain, where an virtual gate was formed resulting from the injection of hot electrons into the surface between gate and drain electrodes.

PAGE 130

130 CHAPTER 8 8. EFF ECT OF ELECTRON IRRADIATION ON ALGAN / GAN AND INALN / GAN HETEROJUNCTIONS 8.1 Introduction to Electron Irradiation AlGaN / GaN high electron mobility transistors (HEMTs) based on AlGaN / GaN heterojunctions (HJs) have applications in radar and communications systems . 133 They must sustain their characteristics when subjected to irradiation in space and military applications, and nuclear safety. For space applications, the most important types of particles are protons with rays, and heavy particles with very high energies . 134 Early experiments performed on single layers of GaN and on AlGaN / GaN HEMTs showed that, in general, the radiation tolerance of GaN and GaN based devices to all types of radiation is much higher than for their AlGaAs / GaAs counterparts . 135 137 For single layers of n GaN, the main defects introduced were the nitrogen vacancy V N related donors with levels near E c 0.06 eV, the nitrogen interstitials related acceptors Ni with levels near E c 1 eV, and relatively shallow elect ron traps with levels near E c (0.15 0.18) eV and E c 0.2 eV . 136 , 138 The latter defects were variously attributed to complexes involving gallium vacancy nitrogen inters titials V Ga N i 139 or nitrogen vacancies nitrogen interstitials V N N i 138 , 140 . Among the defects on the gallium sublattic e deep Ga vacancy acceptors V Ga with the triply negatively charged state level predicted to be near E v +1 eV and Ga interstitials donors Ga i with levels close to E c 0.9 eV . 141 The bulk of experiments with AlGaN / GaN HJs and HEMTs have been carried out using protons of various energies . 142 147 The main effects observed were a decrease of 2DEG mobility and the decrease of the negative threshold voltage of transistor structures V TH . For reactor neutron irradiation 113 and for 10 MeV electron irradiation of MBE grown AlGaN / GaN HJs , 148 simila r effects were observed. A higher radiation tolerance of AlN/GaN HJs compared to AlGaN

PAGE 131

131 / GaN HJs subjected to electron irradiation was attributed to lower energy deposited into lattice defects formation in the thinn er AlN barriers . 53 There has been much less research on the effects of electron irradiation of G aN. Look et al. showed that electron irradiation with energies 0.7 1 MeV introduced new donors with ionization energy of~ 0.06 eV and introduction rates of 1 cm 1 . 149 151 The net electron concentration was not affected while the electron mobility decreased with dose. Acceptor centers were introduced at a rate similar to t he rate of the 0.06 eV donors. Deep centers in as grown, electron irradiated n GaN on sapphire using 1 MeV electrons were ascribed to N vacancy related centers (the 0.06 eV donors) and a deeper center with a thermal activation energy of 0.85 eV. Electron irradiated n GaN showed the presence of deep electron traps with activation energies of 0.9 eV that could be attributed to N i acceptors. In undoped n GaN samples irradiated with 10 MeV electrons, deep acceptor traps with activation energy ~1 eV were attrib uted to N i 2.5 MeV electron irradiation at 4.2K of GaN produced a strong defect photoluminescence band near 0.95 eV, for which optically detected electron paramagnetic resonance spectra could be obtained. The transition was attributed to Ga vacancies , V Ga , with a level near E v +1 eV and the optically detected electron paramagnetic resonance ( ODEPR ) process was interpreted as interaction with two different Ga i interstitial centers with levels close to E v +2.6 eV. The quenching of the electron resonance signal for annealing to room temperature was then attributed to moving of the Ga i defect away from V Ga . In terms of the effects of electron irradiation on GaN devices, McClory et al. reported the effects of 0.45 MeV electron radiation on the gate and drain cur rents of Al 0 27 Ga 0.73 N / GaN High Electron Mobility Transistors (HEMTs). 152 Following irradiation, the gate and drain currents increased at low temperatures and reached a saturation level. Following a subsequen t

PAGE 132

132 room temperature anneal, the gate and drain currents returned to pre irradiation levels. These results were explained by the buildup of positive charge in the AlGaN layer at low temperature and traps formed in the AlGaN layer. The positive charge increas ed the carrier concentration in the two dimensional electron gas (2DEG) and hence the drain current. Studies of AlGaN / AlN / GaN, AlGaN / GaN, and InAlN / GaN HJs after fast reactor neutron irradiation showed an increase in the density of deep negatively charged traps in the barrier or at the interface with GaN . 153 155 These deep centers have thermal activation energy of 0.6 0.8 eV, optical ionization energy of 1.5 1. 7 eV and a high barrier for capture of electrons, resulting in persistent changes in the charged state of the traps after illumination at low temperature. They are present in high areal concentration of >10 12 cm 2 in all types of HJs and HEMTs . 153 They account for deviation to more positive voltages of room temperature V TH values in AlGaN / AlN / GaN HJs as compared to the results of modeling based solely on the strength of the piezoelectric and spontane ous polar ization field . 156 The filling of these centers by electrons due to tunneling from the Schottky metal at high reverse voltage is responsible for the shift of low temperature V TH after cooling at high reverse volt age as compared to cooling at 0 V , 157 whilst persistent V TH decreases after illumination are also caused by these centers. In this paper we perform similar analysis for the effects of 10 MeV electrons irradiati on on electrical properties of MOCVD grown HJs. We compare results for AlGaN / GaN, AlGaN / AlN / GaN, InAlN / GaN HJs on sapphire substrates with those for AlGaN / GaN HJs and HEMTs on Si substrates and show that, as with neutron irradiation, the presence of deep acceptor centers can explain the results of electron irradiation in GaN based HJs and HEMTs. Since the radiation effects in GaN based heterojunctions grown on Si have not been studied in any detail, the main thrust in these experiments was on the behavior of AlGaN / GaN / Si HJs and HEMTs.

PAGE 133

133 8.2 Experimental 8.2.1 HEMT s Fabrication HEMT device fabrication involved Ohmic contact deposition with standard lift off e beam evapora ted Ti / Al / Ni / Au annealed at 800 °C for 30 s under N 2 ambient. Multiple energy and dose nitrogen implantations were used for device isolation and AZ1045 resist used as the mask to define the active region of the devices. Isolation currents were <10 nA at 40 V bias across two 100 µm × 100 µm Ohmic pads separated by a 5 µm implanted g ap. 1 µm gates were defined by lift off of an e beam evaporated metal stack of Pt / Ti / Au (total gate length of the multi finger gate was 8000 µm). Ti / Au metallization was used for the interconnect metals for source, gate, and drain electrodes. The tra ns istors were passivated with 400 nm of plasma enhanced chemi cal vapor deposited SiNx at 300 °C, followed by opening of contact windows using plasma etching. 8.2.2 Device Characterization Static current voltage (I V) characteristics were measured for the HEMTs be fore and after the electron irradiation, including transconductance (g m ) and sub threshold characteristics at var ious doses using an Agilent 4156 C. Gate pulse measurements were employed to evaluate traps creat ed during electron irradiation . In this techni que, the drain current (I DS ) response to a pulsed gate source voltage (V GS ) was measured. The V GS was pulsed from 5 V at frequencies of 100 Hz and 10 k Hz with 10% duty cycle, while drain voltage was kept constant at +5 V. The reduction of the drain curren t in the pulsed mode as compared to the drain current in the dc mode was due to the presence of traps located in the access region between gate and drain contact. In addition, gate source C V and I V characteristics of transistors were measured at differen t temperatures for cooling in the dark at various gate voltages and after illumination at low temperature, as for the

PAGE 134

134 large area Schottky diodes on HJs. These measurements were complemented by deep level transient spectroscopy (DLTS) and admittance spectra measurements. The measurements conditions for DLTS was bias ed at 1.5 V, forward bias pulse of 1 V, time windows 1 ms / 10 ms. The 2DEG concentration and mobility at room temperature before and after electron irradiation was measured by Hall / van der Pau w using evaporated In contacts in the standard geometry. For HJs, capacitance voltage (C V), current voltage (I V) and admittance spectra were measured on Ni Schottky diodes with area 5×10 3 cm 2 . Low temperature C V characteristics were observed after cool ing down in the dark at various reverse biases and after illumination with high power GaN based light emitting diodes (LED s) with wavelength 365 660 nm . 156 , 158 8.3 Res ults and D iscussion The results of Hall / van der Pauw measurements performed on the four studied types of HJs before irradiation are presented in Table 8 1 . The trends in these results are expected. The 2 DEG conc entration in AlGaN / GaN / sapphire structure is the lowest and increases when a thin AlN interlayer is introduced due to increased strain. The main effect is the increased 2DEG mobility in the AlGaN / AlN / GaN HJ caused by suppression of scattering on io nized donors and on composition fluctuations in AlGaN by the undoped AlN insert . For the AlGaN / GaN / Si structure, the 2DEG concentration and mobility are not too different from the structure on sapphire indicating a good quality GaN buffer on Si and AlG aN/GaN interface, although the lower threshold voltage V TH as determined from C V characteristics indicates a higher strain level in the structure. For the InAlN / GaN HJ the 2DEG concentration is high due to the high spontaneous polarization field while t he 2DEG mobility is relatively low, which points to a comparatively high defect density characteristic cur rently for such heterojunctions .

PAGE 135

135 After electron irradiation the 2DEG concentration is only sligh tly affected and the main effect is the gradual decre ase of the 2DEG mobility in all HJs, as illustrated by Figure 8 1 . The rate of mobility decrease is comparable for AlGaN / AlN / GaN / sapphire, AlGaN / GaN / sapphire, and AlGaN / GaN / Si HJs suggesting that the radiation tolerance of the structures on sapphire is not significantly lo wer than for those on s apphire. The onset of mobility degradation is near a fluence of 5 × 10 15 cm 2 , close to that reported for AlGaN / GaN / sapphire HJs gr own by MBE . 157 However, the rate at which the 2DEG mobility of MOCVD grown AlGaN / GaN HJs decreases for further higher doses is much lower than previously reporte d for the MBE grown structures . For InAlN / GaN HJs the degradation of 2D EG mobility with electron fluence proceeds much faster than for other HJs . For the highest electron fluence of 3.3 × 10 16 cm 2 , no 2DEG conductance could be observed. This is in line with the results reported for neutron irradiated HJs . 155 F or AlGaN / AlN / GaN HJs with Al mole fractions in the AlGaN barrier of 40% and 50% , the rate of 2DEG mobility decrease was also much higher than for the barriers with 20% and 30% Al. The conclusion, as in t he case of neutron irradiation , 155 is that the lower the crystalline quality of the barrier layer and the higher the expected level of contamination, the lower is the ra diation tolerance of the heterojunctions. Based on previous neutron irradiation data, changes in the threshold voltage of transistors and in 2DEG mob ility might depend on the density of deep acceptor traps in the barrier or at t he barrier interface with GaN. S uch deep acceptors are present in h igh density in all st udied HJs . 155 156 The deep acceptors densities in the AlGaN / AlN / GaN / sapphire, AlGaN / GaN / sapphire, AlGaN / GaN / Si, and InAlN / GaN / sapphire are, respecti vely, 1.7 × 10 12 cm 2 , 1.5 × 10 12 cm 2 ,

PAGE 136

136 (1 1.4) × 10 12 cm 2 , and 3.4 × 10 12 cm 2 . T hese concentrations increase w ith electron irradiation, as will be seen below. Figure 8 2 compares the room temperature C V characteristics of the AlGaN / GaN / Si HJ before and after irradiation with the fluence of 1.3 × 10 16 cm 2 of 10 MeV electrons . T he threshold voltage V TH (determined as the voltage at which the 2DEG electrons are driven away from the triangular quantum well near the interface a nd the space charge region boundary moves into the GaN buffer) decreases by 0.2 V after irradiation and the capaci tance in accumulation decreases. This is similar to neutron irra diated HEMT structures 155 and can be attributed to increased concentration of deep acceptor s in the barrier. The increase in concentration to account for such shift would be 7 × 10 11 cm 2 . T he decreased capacitance after irradiation could be due to increased non uniformity of N ss causing local total or partial pinching off of the 2DEG thus reducing the effective 2DEG area. These local variations can also change the effective 2DEG Hall mobility if the Fermi level at the surface is pinned . 155 The reason is the switching from uniform two dimensional conductivity to percolation type conductivity because of the fluctuations of local band bending caused by high density of deep acceptors. The concentration of N ss before and after irradiation can be est imated from the difference in V TH values obtained from low temperature C V characteristics measured after cooling down in the dark and after subsequent illumination. Figure 8 3 compares th ese C V characteristics be fore and after irradiation. 2 V in the dark. er cooling down in the dark at 2 V and illumination with 365 nm wavelength LED; it shows the PPC increase after illumination and was used for calculating the density of deep traps in the barrier / interface. Before irradiation

PAGE 137

137 the calculated N ss value is 1.4 × 10 12 cm 2 , after irradiation it is 2.1 × 10 12 cm 2 and the difference fits the r oom temperature data . Further irradiation to 3.3 × 10 16 cm 2 increased the N ss value to 3.1 × 10 12 cm 2 , i.e. the N ss change is approximately linear with fluence , with an introduction rate of about 5 × 10 5 . S tandard DLTS measurements on large area Schottky diodes on HJs are unreliable because of the shift of the threshold voltage with temperature during the measurements and the effects of the series resistance of the str uctu re. Such measurements are possible for HEMT structures with the total Schottky gate contact area sui table for capacitance transient detection, as for the structures used in this study . 157 Figure 8 4 presents spectra measured for quiesent bias 1.5 V corresponding to partial depletion of the 2DEG region. The space charge region boundary in stationary conditions is near the AlGaN / GaN interface while the 1 V filling pulse can also fil l the traps in the barrier . 155 T he inter face trap spectra in the AlGaN / GaN / Si HEMTs varied from sample to sample. In the case of the HEMT structure in Figure 8 4 , the interface spec trum before irradiation showed only a prominent peak near 0.8 eV. After irradiation with 1.3 × 10 16 cm 2 10 MeV electrons we observed an increased concentration of 0.17 eV, 0.3 eV, 0.45 eV, 0.55 eV, and 0.8 eV interfacial traps. A dmittance spectra for the AlG aN / GaN / Si HJ before and after irradiation with 1.3 × 10 16 cm 2 electrons are compared in Figu re 8 5 . The data are shown for only one frequency of 2 kHz and an applied bias of 1.5 V below the threshold voltage, so that only the traps near the interface or in the barrier could be detected. A strong increase in the magnitude of the signals from traps with activation energy 0.3 eV and 0.55 eV occurs after irradiation, in agreement with DLTS results for transistors. Similar resu lts were observed for admittance spectra measured on transistors. Thus, both DLTS and admittance spectra measurements indicate that electron

PAGE 138

138 irradiation increases the concentration of interface/barrier traps with activation energy 0.3 eV and 0.55 eV, with 0.8 eV a lso increased . Among these traps the 0.8 eV centers seem to be the ones related the threshold voltage shift after irradiation since only they are observed in high densi ty before irradiation . Among these barrier / interface traps the traps with activation energy 0.55 eV have been detected previously in AlGaN / GaN HEMTs and their concentration observed to increase after high voltage stress experiments. The gate lag measur ements before and after irradiation with 1.3 × 10 16 cm 2 10 MeV electrons fluence are presented in Figure 8 6 . After irradiation the pulsed signal compared to DC signal decreased by about 2 times pointing to a much more prominent contribution of deep traps. From DLTS and admittance spectra results it seems reasonable to associate this decrease with the increased density of barrier/interface traps with activation energy 0.3 eV, 0.45 eV, 0.55 eV, and 0.8 eV introduced by irradiation. For other types of HJs, the shift of C V characteristics increasing with electron fluence was observed and is compatible with increased concentration of deep traps in the barrier/interface. At that, the AlGaN / GaN / sapphire and AlGaN / AlN / GaN / sapp hire HJs show an introduction rate of deep acceptors similar to the one observed for AlGaN / GaN / Si HJs, whereas, for InAlN / GaN HJs, the introduction rate was about 5 times higher.

PAGE 139

139 Table 8 1 . Initial cha racteristics of 2DEG in studied HJs before irradiation: 2DEG concentration N (2DEG) (cm 2 ) and 2DEG mobility (2DEG) (cm 2 / Vs), also shown are threshold voltages V TH (V) as deduced from room temperature C V characteristics HJ type N(2DEG) ( cm 2 ) (2DEG), ( cm 2 / Vs ) V TH ( V ) AlGaN / AlN / GaN / sapphire 1.2 × 10 13 1490 4 AlGaN / GaN / sapphire 9 × 10 12 1250 3 AlGaN / GaN / Si 1. 1 × 10 13 1 150 2 InAlN / GaN / sapphire 1.9 × 10 13 1210 3.5

PAGE 140

140 Figure 8 1 . M obility as a function of electron fluence for type 1 AlGaN / AlN / GaN / sapphire , type 2 AlGaN / GaN / sapph ire, type 3 AlGaN / GaN / Si , and type 4 InAl N / GaN / sapphire HJ .

PAGE 141

141 Figure 8 2 . C V characteristics measured on the AlGaN / GaN / Si HJ before irradiation ( square line ) and after electron irra diation with the fluence of 1.3 × 10 16 cm 2 ( circle line) .

PAGE 142

142 Figure 8 3 . 85K C V characteristics measured on the AlGaN / GaN / Si HJ A) before irradiation and B) after irradiation . A B

PAGE 143

143 Figure 8 4 . DLTS spectra measured on the AlGaN / GaN / Si HEMT before irradiation (dashed curve) and after irradiation with 1.3 × 10 16 cm 2 10 MeV electrons (solid curve) .

PAGE 144

144 Figu re 8 5 . The temperature dependence of capacitance and AC conductance ( G ) . Black curves refer to pre irradiation, red curves refer to after irradiation.

PAGE 145

145 Figure 8 6 . A ) Gate lag measured on the AlGaN / GaN / Si HEMT before irradiation B ) gate lag measured on the AlGaN / GaN / Si HEMT after irradiatio n. A B

PAGE 146

146 8.4 Summary The main effect of electron irradiation of AlGaN / AlN / GaN, AlGaN / GaN , and InAlN / GaN heterojunctions grown by MOCVD on sapphire and of AlGaN / GaN HJs grown by MOCVD on Si (111) is the decrease of 2DEG mobility of heterojunctions with consequent increase of the sheet resistivity, while the 2DEG concentration was effected only slightly. The decrease in threshold voltage of AlGaN / GaN / Si HJs and HEMTs can be explained by an increase of the density of deep acceptor traps. DLTS and admittance measurements on AlGaN / GaN / Si HEMTs and admittance spectra measurements on AlGa N / GaN / Si HJs show that the traps involved have activation energies of 0.3 eV, 0.45 eV, 0.55 eV, and 0.8 eV. The radiation tolerance of AlGaN / GaN on Si HEMTs is similar to that of more established AlGaN / GaN and AlGaN / AlN / GaN structures on sapphi re, while the radiation tole rance of InAlN / GaN structures are considerably lower.

PAGE 147

147 CHAPTER 9 9. A NOVEL APPROACH TO IMPROVE HEAT DISSIPATION OF ALGAN/GAN HEMTS WITH A CU FILLED VIA UNDER DEVICE ACTIVE AREA 9.1 Introduction to Thermal Performance of HEMTs

PAGE 148

148 nucleation This thermally resistive nucleation layer contains high densities of disloca tions and impurities, which degrade device performance due to the self heating effect induced by this defective interfacial layer . 12 Micro nucleation layers and showed values

PAGE 149

149 9.2 Simulation A pproach Table 9 1 The model used for the thermal simulation is based on the steady state energy balance of the 3 D unit chip using rectangular coordinates (x , y and z axes).

PAGE 150

150 where k is the thermal conductivity in units W / m K, T is the temperature and P D is the heat source density in W / m 3 . A boundary condition wa s set with a temperature of 300 K at the bottom of the Si substrate.

PAGE 151

151 Figure 9 1 . A) 3 D mesh structure of the AlGaN / GaN HEMT used in the simulation. B) 3 D frame structure with a via under source contact and via under the active area . A B

PAGE 152

152 Table 9 1 . Material thermal conductivity . Material Thermal c onductivit y (W / m K) Cu 401 Si 149 AlN defective layer 0.538 AlGaN 25 @ 300K GaN 130 @ 300K Ohmic metal 200 Metal contact 381 Gate metal 381 SiN 30

PAGE 153

153 9.3 Resul ts and D iscussion Figure 9 2 A illustrate s the cross sectional temperature contours for the reference HEMT with a conventional through wafer source contact via hole. Figure 9 2 B show s the HEMT with an additional via hole, in which the Si substrate is etch off, formed directly under the active area filled with copper simulated at a power density of 5 W / mm. For the reference HEMT, the simulated maximum junction temperature was 146 o C, while the maximum junction temperature of the HEMT with additional Cu filled Si substrate via hole under the device active area was 120 o C. The reduction of the maximum junction temperature for the HEMT with additional Cu filled Si substrate via hole under the device active area was achieved by removing the highly defective and thermally resistive nucleation layer under the device active area, as well as filling the through Si substrate via hole wit h thermally conductive copper. To evaluate the ef fectiveness of thermal dissipation, absolute thermal resistance, R, is typically used, which is defined as where is T J T S , T J is the maximum junction temperature, T S is the temperature of the heat sink and W is the total power dissipated b y the device in Watts. To study the effects of the defective nucleation layer and copper filled through Si substrate via hole on the absolute thermal resistance and maximum junction temperature, the temperature distributions of the devices with and without the removal of defective AlN nucleation layer as well as the devices with through Si substrate via hole filled with Cu were simulated. The maximum junction temperature and absolute thermal resistance for different conditions is presented in Table 9 2 and the nucleation layer exhibited the lowest thermal cond uctivity among all the layers. In other words, the total therma l resistance could be reduced 18 % if the nucleation layer was etched off. On the other sides, the difference of the absolute thermal resistance between the device with and without the

PAGE 154

154 copper filled through Si substrate via hole under the active area of the device was around 10 K / W. Thus the combination of the removal of the nucleation layer and the implementation of the copper filled via hole provided a reduction of 18 % of the total thermal resistance of the device. Figure 9 3 show s vertical temperature distributions in the region of epitaxial layers directly under the gate finger for the reference HEMT with a through wafer source contact via hole and the HEMT with both a through wafer source contact via hole and an additional through Si substrate via hole under the active area filled with copper. The dimensio ns of the top AlGaN layer and 2DEG channel are too thin and the temperature changes across these regions are also too small to be observed in Figure 9 3 . For the reference sample, the temperature drops across the G aN buffer layer, AlGaN transition layer and AlN nucleation layer wer e 9, 64 and 6°C, respectively. The thermal conductivity of the 800 nm GaN buffer layer is 130 W / K m, which is around 6 times larger than the thermal conductivity of the 1.4 µm AlGaN tr a nsition layer, 25 W / K m. Thus, the temperature drop in the AlGaN transition layer is around 6 times larger than the one simu lated in the GaN buffer layer. Although the thickness of the very defective AlN nucleation is only 28 nm, there is an obvious 6°C temperature drop due to its very high thermal resistivity, as shown in Figure 9 3 . In the case where the defective AlN nucleation layer under the active layer of the HEMT with the through Si via hole was removed an d the via hole was filled up with plated copper, not only was the junction temperature much lower than the reference HEMT, but also the temperature at the bottom of the AlGaN transition layer was around 10°C lower th an that of the reference HEMT. This indi cates the effectiveness of heat removal from the copper filled via hole. Figure 9 4 show s 2 dimensional temperature contours on the plane of two dimensional electron gas (2DEG) channel for the reference HEMT with a through wafer source contact via

PAGE 155

155 hole and the HEMT with both a through wafer source contact via hole and an additional Si substrate via hole under the active area filled with copper, respectively, simulated at a power density of 5 W / mm. The gradient of the temperature distribution was much larger, and the temperature dropped to less than 35°C for the region 30 µm away from the gate area. Since the through wafer source contact via hole was placed 50 µm away from the gate finger, it could not effectively as sist in the heat dissipation. By contrast, when the through Si via hole was placed directly under the act ive area with a distance of 2.2 µm between the gate finger and via hole, the heat generated in the 2DEG channel could be effectively dissipated from under the active area. As shown in Figure 9 5 , the maximum junction temperature is directly proportional to the power consumption of the HEMT. For the reference HEMT, the maximum junction temperature increased ar ound 28°C per watt of power consumption. On the other hand, the maximum junction temperature only increased 23°C per watt of power consumption for the HEMTs with a 6 µm × 100 µm through Si via hole filled with plated copper under the active area. The effec t of the plated copper thickness inside the via hole on the maximum junction temperature was also investigated, as illustrated in Figure 9 5 . The heat transfer mechanism inside the via hole without filling any meta l is dominated by free convection, which is a couple of orders less efficient than that of heat conduction. As a result, the via holes have to be filled with plated metal to achieve better step coverage. However, as shown in Figure 9 5 , with 1 µm thick copper around the via hole, the maximum junction temperature changing rate can be reduced to 25°C per watt of heat dissipation. When the copper thickness increased to 2 µm, the maximum junction temperature increase ra te was 23.8°C per watt of power consumption, which is very close to the rate of 23°C per watt of heat consumption for the via hole completely filled with plated copper.

PAGE 156

156 The effect of the dimensions of the through Si substrate via holes on the maximum junct ion temperature was also studied. The simulation results presented so far for the HEMT with a through Si substrate via hole were based on the via hole covering the regions of 4 µm between source and drain contacts as well as 1 µm of the transfer length on each s ide the Ohmic metal electrode. The majority of the heat generation of the HEMT is in these two regions. Thus the maximum junction temperature increases dramatically when the opening of the through Si substrate via hole is symmetrically decreased arou nd the gate finger, as shown in the area designated in Figure 9 6 as region I. The maximum junction temperature increases f rom 120°C for the HEMT with a 6 µm wide through Si substrate via hole to 146°C for the HEMT without the t hrough Si substrate via hole. Region II of Figure 9 6 shows the impact of expanding the width of the through Si substrate via hole under the source Ohmic metal contact on the maximum junction temperat ure. The maximum junction temperature continuously decreases as the width of the via hole increases due to larger areas of thermal resistive layer being removed and replaced with a less resistive copper layer. Since this through Si substrate via hole is el ectrically connected to the conventional source via hole, this via hole can be treated as a backside source field plate. Thus this via hole can not only reduce the maximum junction temperature but also reduce the maximum electric field around the gate edge s to increas e the drain breakdown voltage. The gate source capacitance for HEMTs with the through Si substrate via hole in the region II of Figure 9 6 should be the same as the HEMT with 6 µm wide through Si substra te via hole. The maximum junction temperature can be further reduced by extending the width of the through Si substrate via hole to under the drain Ohmic contact, as shown in region III of Figure 9 6 , but the gate source and source drain feed back capacitance will significantly increase and degrade the RF performance of the HEMTs.

PAGE 157

157 Figure 9 2 . A) Cross sectional temperature contours of the re ference HEMT . B) Cross sectional temperature contours of the proposed HEMT structure. x (µm) x (µm) Temp ( O C) B A z (µm) z (µm) Temp ( O C)

PAGE 158

158 Table 9 2 . Absolute thermal resistance of HEMT with different c onfigurations . Type of HEMT Reference HEMT HEMT I Proposed HEMT Source via Cu filled Cu filled Cu filled Si via No Y es Y es Removal of AlN defective layer No No Yes Maximum junction temperature at 5 W / mm ( o C) 152 14 6 120 Absolute heat resistance [K / W] 304 2 92 240 Percentage change compared to reference H EMT (%) --3.45 17.24

PAGE 159

159 Figure 9 3 . Vertical temperature distributions directly under the gate finger for the reference HEMT and the proposed HEMT structure with copper f illed via.

PAGE 160

160 Figure 9 4 . A) 2 dimensional tempe rature contours at 2DEG plane of the reference HEMT. B) 2 dimensional temp erature contours at 2DEG plane of the proposed HEMT. x (µm) A B Temp ( o C) Temp ( o C) y (µm) y (µm) x (µm)

PAGE 161

161 Figure 9 5 . Maximum junction temperature as a function of power de nsit y for the reference , the proposed HEMT, and the proposed HEMT with 1 µm or 2 µm filled copper.

PAGE 162

162 Figure 9 6 . Effect of the through Si substrate via hole location on t he maximal junction temperature .

PAGE 163

163 9.4 Summary In summary, a new approach of implementing an additional through Si substrate via hole under the active area of HEMT is proposed to reduce the maximal junction temperature. For AlGaN / GaN structures grown on Si substrates, the AlN nucleation layer on the Si substrate is a very defective and thermally resistive layer, which causes inefficie nt heat dissipation. The proposed through Si substrate via hole provides access to this AlN nucleation layer. Based on the simulation result, the maximum junction temperature can be significantly decreased by removing this thermally resistive layer and pla ting Cu or Au to fill the via holes. At a 5 W / mm condition, the maximum junction temperatu re of the reference HEMT was 146 °C, while the maximum junction temperature decreased to 120°C for the HEMT with an additional through Si substrate via hole under th e active area of HEMT. Besides reducing the maximal junction temperature, since this through Si substrate via hole is electrically connected to the conventional source via hole, it acts as a backside source field plate. This via hole can also reduce the ma ximum electric field around the gate edges and increas e the drain breakdown voltage. If this through Si substrate via hole is separately connected to the front gate pad, it can behave as a back gate to improve front gate modulation.

PAGE 164

164 CHAPTER 10 10. A NOVEL ST RUCTURE TO IMPROVE BREAKDOWN VOLTAGE BY BACKSIDE GATE ON ALGAN/GAN HEMTS 10.1 Introduction to Multiple Gate Device and Field Plate AlGaN / GaN high electron mobility transistor (HEMT) performance has gained more attention in recent years because of the high bre akdown voltage . 159 , 160 Based on the material property, the breakdown field of GaN could go up to 1.01 MV / cm. 107 However , breakdown voltage was around 1300 V at L GD = 20 µm. 10 Not only the breakdown voltage is lower than theoretical value, but also subthreshold swing (SS) is higher than theoretical value (~60 mV / dec) . 79 The intrinsic defects serve as a tunneling center and increase the leakage current at off state. As a result, the breakdown voltage decreased and SS increased because of the increased leakage current. Thus, it is impor tant to find a way to suppress the leakage current and thus improve the breakdown voltage and subthreshold swing . To suppress gate leakage current, one way is to implement additional gates o n the transistors. There are several forms of m ultigate devices. One example is double gate SOI MOSFETS, which could be used to suppress short channel effect . 161 , 162 . The other example is FinFET, which is also known as tri gate tra nsistor, utilize s three gates to suppress the gate leakage . 163 The body is wrapped around by the gate, which provides a better electrical control over the channel to reduce gate leakage current and prevent short channel effect. There are even transistors which have four gates such as gate all around (GAA) FET 164 and G 4 FET . 165 The additional gate could provide robustness to the device; either induced more current or suppressed leakage current depends on the operation mode. So far, this four gate technology has not been implemented in HEMT technology yet.

PAGE 165

165 The source or gate field plate (FP) can be used to redistribute the el ectric field to decrease peak electric field around the gate edges and reduce hot electrons. Thus, drain breakdown voltage was enhanced and drain current collapse was lessened. 21 , 24 , 108 , 166 170 Zhang et al . extended the gate on top of SiN and get breakdown voltage of 570 V at L GD = 13 m . 171 . Hikita et al . opened a via through the epitaxy layer to Si on the front side of the device to connect the front si . de source to back side grounding electrode and achieve low on state resistance (1.9 m c m 2 ) and breakdown voltage as 350 V . 172 . Lu et a l. also demonstrated the effect of source field plate which improve the breakdown voltage from 55 V to 155 V . 24 10.2 Exper imental 10.2.1 HEMTs Fabrication HEMT structures were grown on Si substrates using a metal organic chemical vapor deposition (MOCVD) system, starting with a thin AlN nucleation layer followed with a 1.4 µm AlGaN graded transition layer, a 0.8 defect ca rbon doped GaN buffer layer, 50 nm undoped GaN layer and a 16 nm undoped AlGaN barrier layer with 26% Al mole fraction. 10.2.2 Front Side Device Fabrication Front side device fabrication started with Ohmic contact formation. Ti / Al / Ni / Au Ohmi c metallization was deposited using electron beam evaporator and the metal contacts were subsequently annealed in a rapid thermal annealing system at 825ºC. Inter device isolation was accomplished by the use of multiple energy N+ implantations to create da mages throughout the top 400 nm of the HEMT structure. After implantation, the devices were passivated with 70 nm thick SiNx in a plasma enhanced chemical vapor deposition (PECVD) system. Schottky gate definition was achieved by patterning the gate and con tact windows to the Ohmic contact pads were also opened at the same time. After SiNx etching, wider patterns for Schottky gates and contact windows to the Ohmic contacts were re patterned with another photolithography step. Ni

PAGE 166

166 / Au based gate metallization was deposited on the gate and Ohmic contact pads simultaneously. The devices were then passivated with another 400 nm layer of PECVD SiNx. The contact windows to Ohmic contact and gate pads were opened by dry etching. There was an additional metal deposit ion for source field plate. The field plate was connected to the source terminal and extended by 1 µm over the gate electrode to the gate to drain region. The source to gate distance and channel length of the HEMTs was kept constant at 1 and 4.7 µm, respec tively. 10.2.3 Backside Device Fabrication After finishing the front side device fabrication, wafers were mounted on another Si substrate for back processing. Prior to drilling via holes, the wafers were polished and thinned to 150 µm. Photoresist AZ9260 were use d to pattern via holes on the backside of the wafer aligning to alignment marks fabricated on the front side of the wafer with a backside aligner. High aspect ratio via holes were fabricated by employing BOSCH process to etch through the Si substrate. The dimensions of via holes are 25 µm × 60 µm × 150 µm. BCl 3 / Cl 2 / Ar based plasma was subsequently employed to etch off AlN nucleation and 1.4 µm graded AlGaN transient layers in an inductively coupled plasma system. After via hole etching, Ti / Au was sput tered to cover the entire backside of the wafer to form back side contacts. A schematic of the HEMT with both front and back gate contact is illustrated in Figure 10 1 . 10.3 Results and Discussion Figure 10 2 A and B show drain IV characteristics of AlGaN / GaN HEMTs by either biasing front gate or back gate. Front gate modulated the two dimensional gas (2DEG) channel well exhibiting a saturation current and a pinch off gate voltage of 574 mA / mm and 1.5 V, respectively. The back gate was formed on the bottom of carbon doped GaN buffer layer, which is 0.8 µm away from the 2DEG channel, thus the modulation of 2DEG channel was not effectively. The back gate modulated around 120 mA of the drain cu rrent by applying gate

PAGE 167

167 voltages ranging from 0 to 25 V, but could not able to pinch off the 2DEG channel. However, these problems can be overcome by etching a part of the GaN buffer layer and placing back gate closer to the 2DEG channel. Figure 10 3 A and B show the drain IV and transfer characteristics of AlGaN / GaN HEMTs by biasing both front and back gate together as compared to the ones with the front gate only. Since the back gate was not effective as the front gate for modu lating 2DEG channel, the applied voltages for the back gate was set 10 times of voltages applied to the front gate. As shown in Figure 10 3 , the drain saturation current increased from 512 mA / mm to 547 mA / mm. The increase migh t be due to the better 2DEG confinement by the additional backside gate. Besides, because of the additional gate, the Gm peak increased from 267 mS / mm to 322 mS / mm. In other words, t he HEMTs biased with both gate biasing exhibited a higher peak extrins ic transconductance resulted from a better 2DEG channel modulation by depleting the 2DEG channel on both sides at the same time. To examine the modulation of back side gate, the drain current was measured by changing the back side gate voltage while kept V G constant. Figure 10 4 A sho w s the transfer characteristics of device with and without applying V BG . As shown in Figure 10 4 A, gate leakage current decreased from 3.9 × 10 5 to 1.2 × 10 6 mA / mm, subthreshold swing improved from 204 to 137 mV / dec while I DS slightly decreased after applying V BG at 10 V. In other words, on / off ratio increased after applying V BG due to reduction of I G . Also, from Figure 10 4 A, the subthreshold swing (SS) became steeper after applying V BG , which mean t SS improved. Besides, the threshold voltage (V TH ) shift positively with increasing backside voltage. Table 10 1 show s the V TH change with different backside voltage. The V TH could shifted positively from 1.38 V to 1.09 V. In other words, V TH could also be adjusted based by applying V BG . Figure

PAGE 168

168 10 4 B illustrat es the increase of on/off ratio and improvement of subthreshold swing with V BG . On/off ratio improved about one order after applying V BG as 40 V and subthreshold swing improved from 2 04 mV / dec to 137 mV / dec. The on / off ratio was improved by the reduction o f I G . In our study, double gate was implemented in D mode device. However, it could also be used in E mode device . By implementing other gate s , it would give better control of the device, i.e. better deple t ion or inversion. Akarvardar et al . utilized four gates in the silicon on insulator device . 165 It was reported that subthreshold swing, mobility, transconductance, high transconductance to current ratio and early voltage were improved with the specific gate bia s. To further examine the effect of backside voltage, front side gate I V was also monitored with backside gate voltage at different voltages. From Figure 10 5 A , it shows the gate leakage current can be suppresse d from 3.8 × 10 4 mA / mm to 2.3 × 10 4 mA / mm by applying the additional backside gate voltage at 25 V . When the back side gate was biased at negative voltage, it provided another depletion at the backside of the device. This depletion region makes the buff er layer more resistive, leading to a smaller leakage current. Because of the suppressed leakage current, the breakdown voltage is expected to improve. As shown in Figure 10 5 B, the breakdown voltage was improved dramatically aft er first applying V BG at 5 V but reached saturation at V BG = 25 V . The depletion depth might touch the channel at V BG = 25 V already so it no longer provided noticeable improvement. In summary, the breakdown voltage can be improved by 40% with V BG = 25 V because of the suppression of leakage current. 10.4 Summary In summary, using the back side gate technology, the device DC performance and breakdown voltage can be improved. The drain saturation current can be improved from 512 to 537 mA / mm. The on / o ff ratio can improve by one order and the subthreshold swing can be improved from 204 to 137 mV / dec . Most important of all, because of the suppression of

PAGE 169

169 leakage current, the breakdown voltage can be improved by 40% when the backside voltage was biased a t 25 V. This technique could not only implement to D mode device but E mode device, making the device more robust and perform better in analog application.

PAGE 170

170 Figure 10 1 . Schematic view of the proposed s tructure with a backside via.

PAGE 171

171 Figure 10 2 . Drain I V modulated by A) front side gate or B) backside gate. A B

PAGE 172

172 Figure 10 3 . A) Drain I V modulated by both front side and backside gate. B) Transfer characteristics modulated by both front side and backside gate. A B

PAGE 173

173 Figure 10 4 . A) Transfer characteristics of ref sample and sample with backside voltage at 10 V. B) On/off ratio and subthreshold swing change versus different backside gate voltage. A B A

PAGE 174

174 Table 10 1 . Threshold voltage change with backs ide gate voltage BG (V) 0 5 10 20 V TH (V) V TH change (%) 1.38 1.27 7.97 1.23 10.87 1.09 21.01

PAGE 175

175 Figure 10 5 . A) Gate I V with backside gate biased at different voltages. B) B reakdown voltage change at different backside gate voltages. A B

PAGE 176

176 CHAPTER 11 11. CONCLUSIONS AlN used as gate insulator and passivation for the high electron mobility transistors deposited by plasma enhanced atomic layer deposition was characterized by X ray p hotoelectron microscopy (XPS) and Auger electron spectroscopy. The refractive index of deposited AlN was 1.9, which was lower than the crystalline AlN due to the top 2 nm AlN oxidized during the device fabrication. DC performance of AlN based MISHE M T was a lso reported. The modulation of gate voltage could be increased to 4 V from 2 V for the same epitaxy structure with Schottky gate. The drain saturation current (I DS ) was 600 mA / mm and the transonductance was 127 mS / mm. Because of the introduction of Al N, the gate leakage current could be suppressed to 1.13 × 10 9 (A / mm) . Due to the reduction of gate leakage current, the drain current on/off ratio was increased to 3.3 × 10 8 . In addition , the drain current dispersion at 100 kHz was only 7%, showing the effective passivation of this AlN passivation layer. The enhancement of the drain breakdown voltage of AlN MISHEMT s was also investigated . The drain breakdown voltage s of 500 and 2000 V w ere achieved for MISHEMT with a drain source distance of 10 µm and 40 µm , respectively. T he mechanism of Ti Al Ni Au based Ohmic contacts degradation upon exposure of buffer oxide etchant (BOE) solution was studied . T he main effect of BOE on Ohmic contacts degradation was the increase of metal she et resistance from 2.7 to / after 180 sec of BOE exposure . Scanning electron microscopy (SEM), energy dispersive X ray spectrum (EDX) and Auger electron microscopy w ere used to characterize the change before and after BOE treatment. Ohmic contacts were composed of 3 to 5 µ m wide Ni Al based islands surrounded by 1 µ m wide Au Al based ring. After BOE treatment, the height of islands reduced from 300 nm to 200 250 nm and the Au based ring became narrower. T he content s of Ni, Al and Au decrease d

PAGE 177

177 after BOE treatment confirmed w ith EDX and Auger electron microscopy. The reduction s of Ni and Al were due to the etching reaction between BOE to metals. The reduction of Au might be resulted from Au removal by etching off Al underneath Au layer. The mechanism of breakdown voltage impr ovement of AlGaN / GaN HEMTs after proton irradiation was investigated. The devices w ere irradiated at 275 keV or 330 keV from the back side of the device. The defects generated after proton irradiation at 275 keV or 330 keV were in GaN buffer layer or AlG aN barrier layer of the HEMT structure , respectively. By putting 2 × 10 18 numbers / cm 3 vacancies in the AlGaN barrier layer, t he drain saturation current decreased from 300 mA / mm to 250 mA / mm. However, the drain breakdown voltage increase d from 55 V t o 150 V. Simulation was perform to identify the degradation mechanism; electric field at the gate edge near the drain electrode decrease d around 50% by placing defects in the AlGaN barrier layer , which would account for the increase of drain breakdown volt age . The effect of low dose gamma irradiation on DC performance of AlGaN / GaN HEMTs was examined . After gamma irradiation, the I DS increased and the resistance between drain and source contacts decreased. T he increase of I DS was in positive correlation w ith doses and reached 11% at dose of 700 Gy. The resistance decrease between drain and source contacts after 700 Gy was around 14%. The increase of the I DS and reduction of source drain resistance w ere mainly due to the increase of the 2DEG electron mobili ty. The gate I V characteristics didn t change much after irradiation. T he effect of electron irradiation on device performance of AlGaN / AlN / Ga N , AlGaN / GaN, and InAlN / GaN HJs were studied. After electron irradiation on AlGaN / GaN / Si HJs, the th reshold voltage shift ed positively because of the increase of the density of deep acceptor traps. Besides, the surface states increased linearly with dose of electron irradiation, with an

PAGE 178

178 introduction rate of about 5 × 10 5 . Deep level transient spectrum ( DLTS ) and admittance measurements on AlGaN / GaN / Si HEMTs and admittance spectra measurements on AlGaN / GaN / Si HJs revealed the trap activation energies of 0.3 eV, 0.45 eV, 0.55 eV, and 0.8 eV. T he drain current dispersion in gate pulsed measurement i ncreased after electron irradiation, which was in agreement with the DLTS and admittance spectrum . T hermal performance of the HEMT structure with an additional Cu via underneath the active area was investigated . The maximum junction temperature could be r educed f or 26 °C from 146 to 120 °C at a power density of 5 W / mm. T he absolute thermal resistance of the device decreased from 292 K / W to 240 K / W due to the removal of thermal resistive nucleation layer and the introduction of Cu via. The effect of backside via structure to DC performance and off state drain breakdown voltage was studied. The drain saturation current can be improved from 512 to 547 mA / mm when V BG and V G was biased simultaneously. At this condition, the transconductanc e can be improved from 267 mS / mm to 322 mS / mm. Besides, the subthreshold swing can be improved from 204 to 137 mV / dec when the backside voltage was biased at 10 V. Because of the suppression of leakage current, the on / off ratio can be improved fro m 3.9 × 10 5 to 1.2 × 10 6 . Last, because of the suppression of leakage current, the breakdown voltage can be improved by 40% when the backside voltage was biased at 25 V.

PAGE 179

179 2 LIST OF REFERENCES 1 J. Xue, Y. Hao, X. Zhou, J.Zhang, C. Yang, X. Ou, L. Shi, H. Wang , L. Yang, and J. Zhang, J. Cryst. Growth 314 , 359 (2011). 2 J. Kuzmik, IEEE Electron Device Lett. 22 , 510 (2001). 3 C. Wang, S. Cho, W. Lee and N. Kim, Int J Adv Manuf Technol 67, 1491 (2012). 4 S. Arulkumaran, G. I. Ng, S. Vicknesh, H. Wang, K. S. Ang, J. P. Y. Tan, V. K. Lin, S. Todd, G. Q. Lo and S. Tripathy, Jpn. J. Appl. Phys. 51 , 111001 (2012). 5 Y. Cao, D. Deen, J. Simon, J. Bean, N. Su, J. Zhang, P. Fay, H. Xing, and D. Jena, in Semiconductor Device Research Symposium (IEEE 2007), pp.1 2. 6 M. A. Khan, A. Bhattarai, J. N. Kuznia, and D. T. Olson, Appl. Phys. Lett. 63 , 1214 (1993). 7 C. Lee, L. Witkowski, H. Q. Tseng, P. Saunier, R. Birkhahn, D. Olson, D. Olson, G. Munns, S. Guo, and B. Albert, Electron. Lett. 41 , 155 (2005). 8 J. W. Chung, T. W. Kim and T. Pala cios, in International Electron Devices Meeting ( IEEE 2010) , pp.30.2.1 30.2.4. 9 J. W. Chung, W. E. Hoke, E. M. Chumbes and T. Palacios, IEEE Electron Dev ices Lett. 31 , 195 (2010). 10 P. Srivastava, J. Das, D. Visalli, M. V. Hove, P. E. Malinowski, D. Marcon, S . Lenci, K. Geens, K. Cheng, M. Leys, S. Decoutere, R. P. Mertens and G. Borghs, IEEE Electron Device Lett. 32 , 30 (2011). 11 G. Simin, X. Hu, N. Ilinskaya, A. Kumar, A. Koudymov, J. Zhang, M. Asif Khan, R. Gaska, and M. S. Shur, Electron. Lett. 36 , 2043 (200 0). 12 J. Kuzmik, S. Bychikhin, D. Pogany, C. Gaquiere, E. Pichonat, and E. Morvan, J. Appl. Phys. 101 , 054508 (2007). 13 14 Y. Chang, Y. Zhang, Y. Zhang, and K. Y. Tong, J. Appl. Phys. 99 , 044501 (2006). 15 J. D. Brown, R. Borges, E. Piner, A. Vescan, S. Singhal, and R. Therrin, Solid State Electron. 46 , 1535 (2002). 16 J . Kuzmik , P. Javorka, A. Alam, M. M arso , M. Heuken , and P. Kordos, IEEE Trans. Electron Devices 49 , 1496 (2002). 17 Andrei Sarua, Hangfeng Ji, K. P. H ilton, D. J. Wallis, Michael J. Uren, T. Martin, and Martin Kuball, IEEE Trans. Electron Devices 54 , 3152 (2007). 18 R. J. T. Simms, J. W. Pomeroy, M. J. Uren, T. Martin, and M. Kuball, IEEE Trans. Electron Device 55 , 478 (2008).

PAGE 180

180 19 S. Singhal, J. D. Brown, R. B orges, E. Piner, W. Nagy, A. Vescan, GAAS2002 Conference Proceedings. Milan, Italy. Sept. 23027, 2002, pp. 37 40. 20 N. Ikeda, S. Yoshida, S. Kato, Y. Sato, L. Jiang, and S. Kaya, 20th International Symposium (2008), pp. 287 290. 21 Y. Ando, Y. Okamoto, H. Miyamoto, T. Nakayama, T. Inoue, and M. Kuzuhara , IEEE E lectron D evice L ett. 24 , 289 (2003). 22 H. Sun, A. Alt, H. Benedickter, and C. Bolognesi, IEEE Electron Dev ices Lett . 45 , 376 (2009). 23 Y. Z. Chiou, S. J. Chan g, Y. K. Su, C. K. Wang, T. K. Lin, and B. R. Huang, IEEE Electron Dev ice Lett . 50 , 1748 (2003) . 24 L. Liu, T. S. Kang, D. A. Cullen, L. Zhou, J. Kim, C. Y. Chang, E. A. Douglas, S. Jang, D. J. Smith, S. J. Pearton, W. J. Johnson, and F. Ren, J. Vac. Soc. Tec hnol. 29 , 032204 (2011). 25 C. G. Van de Wall e , C. Stampfl , and J. Neugebauer , J. Cryst. Growth 189 , 505 10 (1998). 26 W. Gotz, R. S. Kern, C. H. Chen, H. Liu, D. A. Steigerwald, and R. M. Fletcher, Mater. Sci. Eng. B 59 , 211 7 (1999). 27 Y. C. Choi, M. Pophristic, H. Y. Cha, B. Peres, M. G. Spencer, and L. F. Eastma n, IEEE Trans. Electron Devices 53 , 2926 (2006). 28 Y. Niiyama, S. Kato, Y. Sato, M. Iwami, J. Li, H. Takehara, H. Kambayashi, N. Ikeda, and S. Yoshida, Prooceedings of the Material Research Society Sympos ium, 2007, Boston MA, 26 30 November 2007 (Cambridge University Press, Cambridge, England, 1940), P. 369. 29 S. Selvaraj, T. Suzue and T. Egawa, in Device Research Conference (IEEE, 2009), pp. 283 284. 30 X. X. Yu, J. Y. Ni, Z. H. Li, C. Kong, J J. Zhou, X. Dong , L. Pan, Y .C. Kong, and T S. Chen, Chin . Phys. Lett 31 , 037201 (2013). 31 R. Vetury, N. Q. Zhang, S. Keller, and U. K. Mishra, IEEE Trans. Electron Devices. 48 , 560 (2001). 32 H. Kim, A. Vertiatchikh, R. M. Thompson, V. Tilak, T. R. Prunty, J. R. Shealy, and L . F. Eastman, IEEE Electron Device Lett. 24 , 421 (2003). 33 Z. H. Liu, G. I. Ng, H. Zhou, S. Arulkumaran, and Y. K. T. Maung, Appl. Phys. Lett. 98 , 113506 (2011). 34 M. F . Romero, A. Jimenez, J. Miguek Sanchez, A. F. Brana, F. Gonzalez Posada, R. Cuerdo, F. Call e, and E. Munoz, IEEE Electron Device Lett. 29 , 209 (2008). 35 F. Ali and A. Gupta, Hemts And Hbts (Artech House, Boston, 1991). 36 K. Jackson, Compound Semiconductor Devices (Wiley VCH, Weinheim, 1998).

PAGE 181

181 37 C. F. Lo, T. S. Kang, L. Liu, C. Y. Chang, S. J. Pearton, I. I. Kravchenko, O. Laboutin, J. W. Johnson, and F. Ren, Appl. Phys. Lett. 97 , 262116 (2010). 38 Enoch and N. Bonod, Plasmonics (Springer, Berlin, 2012). 39 S. Campbell, The Science And Engineering Of Microelectronic Fabrication (Oxford University Press, New Yo rk, 2001). 40 G. May and S. Sze, Fundamentals Of Semiconductor Fabrication (Wiley, New York, 2004). 41 T. Hashizume, S. Y. Ootomo, R. Nakasaki, S. Oyama, and M. Kihara, Appl. Phys. Lett. 76 , 2880 (2000). 42 S. Ootomo, T. Hashizume, and H. Hasegawa, Phys. Stat. Sol. (a) 188 , 371 (2001). 43 J. W. Elam, D. Routkevitch, P. P. Mardilovich, and S. M. George, Chem. Materials 15 , 3507 (2003). 44 M. Alevli, C. Ozgit, I. Donmez, and N. Biyikli, J . Cryst. Growth 335 , 51 (2011) . 45 Robert Bosch GmbH, 5501893 (1996). 46 Kazumi Kurihara, St eady State And Transient Performance Analysis Of Permanent Magnet Machines Using Time Stepping Finite Element Technique (INTECH Open Access Publisher, 2012). 47 D. Williams and C. Carter, Transmission Electron Microscopy (Plenum Press, New York [U.A.], 2009). 48 S. Sze, Physics Of Semiconductor Devices (Wiley, New York, 1981). 49 J. Goldstein , D. E. Newbury , P. Echlin , D. C. Joy , C. Fiori , and E. Lifshin , Scanning electron microscopy and x ray microanalysis ( Plenum Press , New York [U.A.], 1981). 50 K. Oura, Surface Sci ence (Springer, Berlin, 2003). 51 R.J. Archer, J. Electrochem. Soc. 104 , 619 ( 1957 ) . 52 Kazumi Kurihara, Steady State And Transient Performance Analysis Of Permanent Magnet Machines Using Time Stepping Finite Element Technique (INTECH Open Access Publisher, 2012 ). 53 X. Hu, A.P. Karmarkar, B. Jun, D.M. Fleetwood, R.D. Schrimpf, R.D. Geil, R.A. Weller, B.D. White, M. Bataiev, L.J. Brillson, and U.K. Mishra, IEEE T ran s . Nucl . Sci . 50 , 1791 (2003) . 54 R. Doering and Y. Nishi, Handbook Of Semiconductor Manufacturing Techno logy (CRC Press, Boca Raton, 2008). 55 D. Schroder, Semiconductor Material And Device Characterization (IEEE Press, [Piscataway, NJ], 2006).

PAGE 182

182 56 O . Mitrofanov, and M . Manfra , Superlattices and Microstructures 34 , 33 (2003) . 57 D. V. Lang, J. Appl. Phys. 45 , 3023 (19 74) . 58 D. V. Singh, K. Rim, T. O. Mitchell, J. L. Hoyt, and J.F. Gibbons, J. Appl. Phys. 85 , 985 (1999). 59 X. Hu, A. Koudymov, G. Simin, J. Yang, M.A. Khan, A. Tarakj i , M. Shur, and R. Gaska, Appl. Phys. Lett . 79 , 283 2 (2001) . 60 M. Ochiai, M. Akita i, Y. Ohno, S. Kishimoto, K. Maezawa, and T. Mizutani, J . Appl . Phys . 42 , 2278 (2003) . 61 T. Hashizume, S. Anantathanasarn, N. Negoro, E.Sano, H. Hasegawa, K. Kumakura, and T. Makimoto, Jpn. J. Appl. Phy s . Pt.2, Lett. & Express Lett. 43 , L777 (2004). 62 T. Hashizume, S. Ootom o, and H. Hasegawa, P hysica S tatus S olidi (C) 38 , 2952 (2003). 63 M. Kanamura, T. Ohki, T. Kikkawa, K. Imanishi, T. Imada, A.Yamada, and N. Hara, IEEE Electron Dev ices Lett . 31 , 189 (2010) . 64 H. Y. Liu, B. Y. Chou, W. C. Hsu, C. S. Lee, J. K. Sheu, and C. S. Ho , IEEE Trans. Electr on Dev ices 60 , 213 (2013) . 65 K.Y. Park, H.I. Cho, J.H. Lee, S.B. Bae, C.M. Jeon, L.J. Lee, D.Y. Kim, C.S. Lee, and J.H. Lee, Phys. Status S olidi (C) 7 , 2351 (2003). 66 P.D. Ye, B. Yang, K. Ng, J. Bude, G. Wilk, S. Halder and J . Hwang, Appl . Phys . Lett . 86 , 063501 (2005) . 67 Y. Z. Yue, Y. Hao, Q. Feng, J. C. Zhang, X. H. Ma, J. Y. Ni , Chin. Appl. Phys. Lett. 24 , 2419 (2007). 68 R. Mehandru, B. Luo, J. Kim, F. Ren, B. Gila, A. Onstine, C. Abernathy, S. Pearton, D. Gotthold and R. Birkhahn, Appl . Phys . Lett . 82 , 2530 (2003) . 69 C. Liu, E.F. Chor and L.S. Tan, Appl . Phys . Lett . 88 , 17350 4 (2006) . 70 P. Javorka, Fabrication And Characterization Of Algan/Gan High Electron Mobility Transistors, Ph. D., Rheinisch westfälischen Technische n Hochschule Aachen Univer sity ( 2004 ) . 71 O. Seok, W. Ahn, M. K. Han, and M. W. Ha, Semicon ductor Science and Technology 28 , 025001 (2013). 72 Q. Feng, Q. Wang, T. Xing, Q. Li, and Y. Hao, Science China Technological Sciences 56 , 629 (2013). 73 T. Sato, J. Okayasu, M. Takikawa and T. K. Su zuki, IEEE Electron Device Lett. 34 , 375 (2013).

PAGE 183

183 74 S. L . Selvaraj and Takashi Egawa , Appl. Phys. Lett . 89 , 193508 (2006). 75 M. Leskelä, and M. Ritala, Thin Solid Films 409 , 1 38 (2002). 76 S. Huang, Q. Jiang, S. Yang, C. Zhou, and K.J. Chen, IEEE Electron Dev ices Lett . 33 , 516 (2012). 77 H. Kamata, Y. Ishii, T. Mabuchi, K. Naoe, S. Ajimura, and K. Sanada, Fujikura Technical Rev. 38 , 41 (2009). 78 G. Liu, E. Deguns, L. Lecordier, G. Sun daram and J. Becker, ECS Trans. 41 , 219 (2011). 79 J. W. Chung, X. Zhao, and T. Palacios, in Proc. 65th Device Res. Conf., Notre Dame, IN, (IEEE, 2007). p. 111. 80 L. Liu, F. Ren, S. Pearton, R. Fitch, D. Walker, K. Chabak, J. Gillespie, M. Kossler, M. Trejo, an d D. Via, J. Vac. Sci. Technol. B 29 , 060603 (2011). 81 Y . Wei, Z . Renping, D . Yandong, H. Weihua, and Y . Fuhua, J. Semiconductors 33 , 064005 (2012). 82 Gary S. May, and Simon M. Sze, Fundamental of Semiconductor Fabrication, 1st ed. (Wiley, New York, USA, 2003) , PP. 87 88. 83 J. Algueperse, P. Mollard, D. Devilliers, M. Chemla, R. Faron, R. Romana, and J. P. Cuer, 84 K . R. Williams, and R . S. Muller, J. of Microelectromechanical System 5 , 2 56 (1996). 85 K . R. Williams, K . Gupta, and M . Wasilik, J.of Microelectromechanical Sys 12 , 761 (2003). 86 L . Zhou, J . H. Leach , X . Ni, H . Morkoc, and D . J. Smith , J. Appl. Phys. 107 , 14508 (2010). 87 Y . Z. Zhang , S W. Feng, C. S. Guo, G. C. Zhang , S. X. Zhuang , R . Su , Y. X. Bai , and C. Z. Lu , Chin. Phys. Lett. 25 , 4083 (2008). 88 C. F. Lo, L. Liu, C. Y. Chang, F. Ren, V. Craciun, S. J. Pearton, Y. W. Heo, O. Laboutin, and J. W. Johnson, J. Vac. Sci. Technol. B 29 , 042201 (2011). 89 S. L. Selvaraj, T. Suzue, and T. Egawa , IEEE Electron Dev L ett . 30 , 587 (2009). 90 L. Bin and T. Palacios. IEEE Electron Devices L ett . 31 , 9 (2010). 91 M. Ochiai, M. Akitai, Y. Ohno, S. Kishimot o, K. Maezawa, and T. Mizutani, Jpn. J. Appl. Phys 42 , 2278 (2003). 92 T. Hashizume, S. Anantathanasarn, N. N egoro, E. Sano, H. Hasegawa , K. Kumakura, and T. Makimoto, Jpn. J. Appl. Phys. 43 , L777 (2004). 93 M. Kanamura, T. Ohki, T. Kikkawa, K. Imanishi, T. Imada, A. Yamada, and N. Hara, IEEE Electron Devices L ett . 31 , 189 (2010).

PAGE 184

184 94 H. Y. Liu, B. Y. Chou, W. C. Hsu, C. S. Lee, J. K. Sheu, and C. S. H o, IEEE Trans. Electron Devices 60 , 213 (2013). 95 P. D. Ye, B. Yang, K. Ng, J. Bude, G. Wilk, S. Halder, and J. Hwang, Appl. Phys. Lett. 86 , 063501 (2005). 96 Y. Z. Yue, Y. Hao, Q. Feng, J. C. Zhang, X. H. Ma, J. Y. Ni, Chine se Phys. Lett. 24 , 2419 (2007). 97 R. Mehandru, B. Luo , J. Kim , F. Ren , B. P. Gila , A. H. Onstine , C. R. Abernathy , S. J. Pearton , D. Gotthold , R. Birkhahn , B. Peres , R. Fitch , J. Gillespie , T. Jenkins , J. Sewell , D. V ia and A. Crespo , Appl. Phys. Lett. 82 , 2530 (2003). 98 S. Arulkumaran, T. Egawa, S., Matsui and H. Ishikawa, Appl. Phys. Lett. 86 , 123503, (2005). 99 H. Xing, Y. Dora, A. Chini, S. Heikman, S. Keller and U.K. Mishra, IEEE Electr. Dev. Lett. 25 ,161 (2004). 100 S. Yagi, M. Shimizu, M. Inada, Y. Yamamoto, G. Piao, H. Okumura, Y. Yano, N. Akutsu, and H. Ohashi, Solid State Electron. 50 , 1057 (2006). 101 J . Shi, Lester F. Eastman, Xiaobin Xin, and M. Pop hristic, Appl. Phys. Lett. 95 , 042103 (2009). 102 T. Nomura, H . Kambayashi, M . Masuda, S . Ishii, N . Ikeda, J . Lee, and S . Yoshida, Proceedings of the 18th International Symposium on Power Semiconductor Devices, 2006, Naples Italy, 4 8 June 2006 (IEEE, Piscataw ay, NJ 2003) P.1 103 S . Yagi, M . Shimizu, H . Okumura, H . Ohashi, Y . Yano and N . Akutsu, Jpn. J. Appl. Phys. 46 , 2309 (2007) 104 Y.C. Choi, M. Pophristic, B. Peres, M. G. Spencer, and L. F. Eastman, J. Vac. Sci. Technol. B 24 , 2601 (2006) 105 S. Arulkumaran, T. Egawa, S. Matsui, and H. Ishikawa, Appl. Phys. Lett. 86 , 123503 (2005) 106 S. Yagi, M. Shimizu, H. Okumura, H. Ohashi, K. Arai, Y. Yano, N. Akutsu, Proceedings of the 19th International Symposium on Power Semiconductor Devices, 2007, Jeju Korea, 27 31 May 2007 (IEEE, Piscataway, NJ 2003) P.261 107 B. J. Baliga, IEEE Electron Device Lett. 10 , 455 (1989). 108 , A. Saxler, M. Moore, R. Smith, S. Sheppard, P. Chavarkar, T. Wisleder, U. Mishra, and P. Parikh, IEEE Electron Dev ices Lett . 25 , 117 (2004). 109 A. Chini, D. Buttari, R. Co ffie, S. Heikman, S. Keller and U. K. Mishra, Electron Lett. 40 , 73(2004). 110 S. J. Pearton , R. Deist , F. Ren , L. Liu , A.Y. Polyakov and J Kim , J. Vac. Sci. Technol. A 31 , 05 0801 (2013).

PAGE 185

185 111 K.Matocha, T. P. Chow and R. J. Gutmann, IEEE Trans. Electron Dev ices 52 , 6 (2005). 112 L. Liu, C. V . Cuervo, Y. Y. Xi, F. Ren, S. J. Pearton, H. Y. Kim, J. Kim and I. I. Kravchenko, J. Vac. Sci. Technol. B 31 , 042202 (2013). 113 M. Osinski, P. Perlin , H. Schone, A. H. Paxton, and E.W. Taylor, Electron. Lett. 33 , 1252 (1997). 114 B. D. White, M. Bataiev, S. H. Goss, X. Hu, A. Karmarkar, D. M. Fleetwood, R. D. Schrimpf, W. J. Schaff and L. J. Brillson, IEEE Trans. Nucl. Sci. 50 , 1934 (2003). 115 H. Y. Kim, T. A nderson, M. A. Mastron, J. A. Freitas Jr., S.W. Jang, J. Hite, C. R. Eddy Jr., and J.Y. Kim, J. Cryst. Growth 326 , 62 (2011). 116 L. Chernyak, A. Osinsky, V. N. Fuflyigin, J. W. Graff, and E. F. Schubert, IEEE Trans. Electron Devices 48 , 433 (2001). 117 G. Sonia, E. Richter, F. Brunner, A. Denker, R. Lossy, M. Mai, and F. Lenk, Solid State Electron. 52 , 1011 (2008). 118 L. Liu, C. F. Lo, Y. Xi, Y. Wang, F. Ren, S. J. Pearton, H. Y. Kim, J. Kim, R. C. Fitch, D. E. Walker, K. D. Chabak, J. K. Gillespie, S. E. Tetlak, G. D. Via, A. Crespo and I . I. Kravchenko, J. Vac. Sci. Technol. B 31 , 022201 (2013). 119 K. H. Cho, I. H. Ji, Y. H. Choi, J. Lim, Y. S. Kim, K. R. Kim and M. K. Han, Proceedings of the 20th International Symposium on Power Semiconductor Devices & IC's, 2008, Orl ando FL, 18 22, May 2008 (IEEE, Piscataway, NJ 2009), P. 241 120 Z. Q. Fang, J. W. Hemsky, D. C. Look, and M. P. Mack, Appl. Phys. Lett. 72 , 448 (1998). 121 F. Berthlet, Y. Guhel, B. Boudart, H. Gualous, J.L. Trolet, M. Piccione and C. Gaquiere, Electron Lett. 48 (2012). 122 M. R. Hogsed, Y. K. Yeo, M. Ahoujja, M. Y. Ryu, J. C. Petrosky, and R. L. Hengehold, Appl. Phys. Lett. 86 , 261906 (2005). 123 W. Gu, C. Chen, H. Duan, Y. Hao, J. Zhang, C. Wang, Q. Feng and X. Ma, J. Semiconductors 30 , 044002 1 (2009). 124 C. Schwarz, A. Yadav, M. Shatkhin, E. Flitsiyan, L. Chernyak, V. Kasiyan, L. Liu, Y. Y. Xi, F. Ren, S. J. Pearton, C. F. Lo, J. W. Johnson, and E. Danilova, Applied Phys. Lett. 102 , 062102 (2013) 125 S. Jha, E . V. Jelenkovic, M. M. Pejovic, G.S. Ristic, M. Pejovic, K. Y. Ton g, C. Surya, I. Bello, and W. J. Zhang, Microelectronic Engineering 86 , 37 (2009). 126 G. A. Umana Membreno, J. M. Dell, G. Parish, B. D. Nener, L. Faraone, and U. K. Mishra 50 , 2326 (2003). 127 O. Lopatiuk, L. Chernyak, A. Osinsky, J. Q. Xie, and P. P. Chow, Appl . Phys. Lett. 87 , 162103 (2005).

PAGE 186

186 128 O. Aktas, A. Kuliev, V. Kumar, R. Schwindt, S. Toshkov, D. Costescu, J. Stubbins, and I. Adesida, Solid State Electron. 48 , 471 (2004). 129 S.A. Vitusevich, N. Klein, A. E. Belyaev, S. V. Danylyuk, M. V. Petrychuk, R. V. Konako va, A. M. Kurakin, A. E. Rengevich, A. Yu. Avksentyev, B. A. Danilchenko, V. Tilak, J. Smart, A. Vertiatchikh, and L. F. Eastman, Phys. Status Solidi A 195 , 101 (2003). 130 A.M. Kurakin, S. A. Vitusevich, S.V. Danylyuk, H. Hardtdegen, N. Klein, Z. Bo ugrioua, B . A. Danilchenko, R. V. Danilchenko, R. V. Konakova, and A. E. Belyaev, J . App l. Phys. 103 , 083707 (2008) . 131 H. Morkoc, A. Di Carlo, and R.Cingolani, Solid State Electron. 46 ,157 (2002). 132 S. J. Pearton, J. C. Zolper. R. J. Shul, and F. Ren, J. Appl. Phys. 86 , 1 (1999). 133 R. Fagotti, A. Cidronali, and G. Manes, IEEE Microw. Wireless Compon. Lett. 21 , 89 (2011) . 134 A. Y. Polyakov, S.J. Pearton, P. Frenzer, F. Ren, L. Liu and J. Kim, J. Mater. Chem. C 1, 877 (2013) 135 S. Pearton, Gan And Zno Based Materials And Devices (Springer, Berlin, 2012). 136 P. Bhattacharya, R. Fornari and H. Kamimura, Comprehensive Semiconductor Science And Technology (Elsevier Science, Amsterdam, 2011). 137 L. Polenta, Z Q. Fang, and D.C. Look, Appl. Phys. Lett. 76 , 2086 (2000) 138 F.D. Auret, S.A. Goodman , F.K. Koschnick, J. M. Spaeth, B. Beaumont, and P. Gibart, Appl. Phys. Lett. 74 , 407 (1999) 139 A.Y.Polyakov, I . H . Lee, N.B. Smirnov, A.V. Govorkov, E.A. Kozhukhova, N.G. Kolin, A.V. Korulin, V.M. Boiko, and S.J. Pearton, J. Appl. Phys. 109 1 23703 (2011) 140 I . H . Lee, A.Y. Polyakov, N.B. Smirnov, A.V. Govorkov, E.A. Kozhukhova, N.G. Kolin, V.M. Boiko, A.V. Korulin, and S.J. Pearton, J. Vac. Sci. Technol. B 29 , 041201 (2011) 141 J. Neugebauer and C.G. Van de Walle, Phys. Rev. B 50 , 8067 (1994) 142 K. Saarinen, T. Suski, I. Grzegory, and D.C. Look, Phys. Rev. B 64 , 233201 (2001) 143 K.H. Chow, L.S. Vlasenko, P. Johannesen, C. Bozdog, and G.D. Watkins, Phys. Rev. B 69 , 045207 (2004) 144 A.Y. Polyakov, N.B. Smirnov, A.V. Govorkov, A.V. Markov, S.J. Pearton, N.G. Kolin, D.I. Merkuris ov, V.M. Boiko, C . R . Lee, and I . H . Lee, J. Va c. Sci. Technol. B 25 , 436 (2007) 145 A.Y. Polyakov, N.B. Smirnov, A.V. Govorkov, A.V. Markov, N.G. Kolin, D.I. Merkurisov, V.M. Boiko, K.D. Shcherbatchev, V.T. Bublik, M. I. Voronova, I H. Lee, and C.R. Lee , J. Ap pl. Phys. 100 , 093715 (2006)

PAGE 187

187 146 A.Y. Polyakov, N.B. Smirnov, A.V. Govorkov, A.V. Markov, E.B. Yakimov, P.S. Vergeles, N.G. Kolin, D.I. Merkurisov, V.M. Boiko, In Hwan Lee, Cheul Ro Lee, and S.J. Pearton, J. Electron. Mater. 36 , 1320 (2007) 147 I . H . Lee, A.Y. Pol yakov, N.B. Smirnov, A.V. Govorkov, E.A. Kozhukhova, N.G. Kolin, V.M. Boiko, A.V. Korulin, and S.J. Pearton, J.El ectrochem. Soc. 158 H866 (2011) 148 A.Y. Polyakov, N.B. Smirnov, A.V. Govorkov, A.V. Markov, S.J. Pearton, A.M. Dabiran, A. M. Wowchak , B. Cui , A.V. Osinsky, P.P. Chow, N.G. Kolin, V.M. Boiko , and D.I. Merkurisov, Appl. Phys. Lett. 93 , 152101 (2008) 149 D.C. Look, D.C. Reynolds, J.W. Hemsky, J.R. Sizel ove, R.L. Jones, and R.J. Molnar, P hys. Rev. Lett. 79 , 2273 (1997) 150 Z Q Fang, J.W. Hemsky, D.C. Lo ok, and M.P. Mack, Appl. Phys. Lett. 72 , 448 (1998) 151 D. C. Look , G. C. Farlow , P. J. Drevinsky , D. F. Blis s , and J. R. Sizelove , Appl. Phys. Lett. 83 , 3525 (2003) 152 J. W. McClory, J. C. Petrosky , J. M. Sattler, and T. A. Jarzen, IEEE Tran. Nucl. Sci 54 , 1946 (2007) 153 A.P. Karmarkar, B. Jun, D.M. Fleetwood, R.D. Schrimpf, R.A. Weller, B.D. White, L. Brillson, and U . Mishra, IEEE Tran. Nucl. Sci. 51 , 3801 (2004) 154 S.J. Cai, Y.S. Tang, R. Li, Y.Y. Wei, L. Wong, Y.L. Chen, K.L. Wang, M. Chen, Y.F. Zhao, R.D. Schrimpf, J.C. Keay, and F. Galloway, IEEE T rans. Electron Dev ices 47 , 304 (2000) 155 A.Y. Polyakov, N.B. Smirnov, A.V. Govorkov, and E.A. Kozhukhova, S.J. Pearton, F. Ren, J.W. Johnson, J. Vac. Sci. Tech nol. B 30 , 061207 (2012) 156 C.F. Lo, L. Liu, T. S. Kang, F. Ren, O. Laboutin, Yu Cao, Wayne J. Johnson, A.Y. Polyakov, N.B. Smirnov, A.V. Govor kov, I.A. Belogorokhov , and S.J. Pearton, J. Vac. Sci. Technol. B 30 , 011205 (2012) 157 A.Y. Polyakov, N.B. Smirnov, A.V . Govorkov, E.A. Kozhukh ova, S.J. Pearton, F. Ren, L. L i u , J.W. Johnson, N.I. Kargin and R.V. Ryzhuk, J. Vac. Sci. Technol. B 31 , 011211 (2013) 158 A.Y. Polyakov, N.B. Smirnov, A.V. Govorkov, E.A. Kozhukhova, S.J. Pearton, F. Ren, S.Yu. Karpov, K. D.Shcherbach ev, N.G. Kolin, W. Lim, J. Vac. Sci.Technol. B 30 , 041209 (2012) 159 S. Yoshida, H. Ishii, J. Li, D. Wang, and M. Ichikawa, Solid State Elect. 47 , 589 (2003). 160 W. Saito, M. Yamaguchi, I. Omura, K. Tsuda, Y. Saito, Y. Kakiuchi and T. Nitta, IEEE Elect. Dev. Let t. 29 , 8 (2008). 161 Wong, H SP, Kevin K. Chan, and Yuan Taur., IEDM Tech. Digest , Washington , DC. PP427 430, Dec 7 10, 199 7 .

PAGE 188

188 162 T. Schulz, W. Rosner, E. Landgraf, L. Risch, U. Langmann, Solid State Electronics 46 , 985 (2002). 163 X. Huang, W C. Lee, C. Kuo, D. Hisam oto, L. Chang, J. Kedzierski, E. Anderson, H. Takeuchi, Y. K. Choi, K. Asano, V. Subramanian, T J. King, J. Bokor, C. Hu, IEDM Tech. Digest, Washington , DC. PP6 70, Dec 5 8, 1999. 164 N. Singh, A. Agarwal, L.K. Bera, T. Y. Liow, R. Yang, S. C. Rustagi, C. H. T ung, R. Kumar, G. Q. Lo, N. Balasubramanian, D. . Kwong, Elect. Dev. Lett. 27 , 383 (2006). 165 K . Akarvardar, S . Cristoloveanu, P . Gentil, R . D. Schrimpf, and B . J. Blalock , IEEE Trans. on Elect. Dev. 54 , 323 (2007). 166 V. Kumar, G. Chen, S. Guo, and I. Adesida, IEEE Trans. Elect. Dev. 53 , 1477 (2006). 167 Y. Hori, M. Kuzuhara, Y. Ando, and M. Mizuta, J. Appl. Phys. 87 , 3483 (2000). 168 S. Karmalkar, M. S. Shur, G. Simin, and M. A. Khan, IEEE Trans. Electron Devices 52 , 2534 (2005). 169 Lu Liu, Y. H. Hwang, Y. Xi, F. Ren, V. Craciun, S. J. Pearton, G. Yang, H. Y. Kim, and J. Kim, J. Vac. Sci. Technol. B 32 , 022202 (2012). 170 X . Gang, E . Xu, N . Hashemi, Z . Bo, F . Y. Fu, and W . T . Ng, Chin. Phys. B 21 , 086105 (2012) 171 N. Q. Zhang, S. Keller, G. Parish, S. Heikman , S. P. Denbaars, U. k. Mishra, Elect. Dev. Lett. 21 , 421 (2000). 172 M. Hikita, M. Yanagihara, K. Nakazawa, U. Hiroaki, H. Yutaka, U. Tetsuzo, U. Yasuhiro, T. Tsuyoshi, U. Daisuke, T. Egawa, IEEE Trans. on Elect. Dev. 52 , 1963 (2005).

PAGE 189

189 3 BIOGRAPHICAL SKETCH Ya Hsi Hwang was bor n in Taipei, Taiwan, on June 9, 1985. She received her B.S. degree from National Taiwan University in 2007. After graduation, she attended graduate school in National Taiwan University and received her Master of S cience degree from National Taiwan Universi ty in 2009. She then worked in Taiwan Semiconductor Manufacturing Company (tsmc) as a Research and Development Engineer for three years. She enrolled Ph.D. program in 0 technical journal papers . She also had a provisional patent with Dr. Fan Ren and the other two professor s . She received her Ph . D. from the University of Florida in the sp r ing of 2015.