<%BANNER%>

Investigation of Electrical Bias, Mechanical Stress, Temperature and Ambient Effect on AlGaN/GaN Hemt Time-Dependent Deg...

MISSING IMAGE

Material Information

Title:
Investigation of Electrical Bias, Mechanical Stress, Temperature and Ambient Effect on AlGaN/GaN Hemt Time-Dependent Degradation
Physical Description:
1 online resource (143 p.)
Language:
english
Creator:
Gupta, Amit
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Electrical and Computer Engineering
Committee Chair:
Nishida, Toshikazu
Committee Co-Chair:
Thompson, Scott E
Committee Members:
Law, Mark E
Gila, Brent P

Subjects

Subjects / Keywords:
algan -- degradation -- diffusion -- gan -- hemt -- reliability -- stress -- time-dependent
Electrical and Computer Engineering -- Dissertations, Academic -- UF
Genre:
Electrical and Computer Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
AlGaN/GaN HEMT technology is promising for RF and high power applications. However commercial usability of this technology is currently hindered because of its limited electrical reliability which still remains a major concern. AlGaN/GaN HEMTs have been shown to degrade irreversibly under typical device operation and there is widespread disagreement on the underlying fundamental physics for the observed device degradation. Electrical degradation in AlGaN/GaN HEMTs due to DC stressing is studied typically by performing electrical step stress tests and a critical voltage is determined. Device degradation is characterized by changes measured in electrical parameters, such as increase in Rs and RD, decrease in IDsat, decrease in gm, Vt shift and sub-threshold change. The widely accepted theory attributes such degradation to the inverse piezoelectric effect. Electric field due to applied bias generates biaxial tensile stress which together with intrinsic stress from lattice mismatch increases the elastic energy of AlGaN layer. AlGaN layer undergoes crystallographic defect formation when total elastic energy reaches a critical value. Another theory associates such degradationto thermally activated chemical reaction occurring at metal-semiconductor interface 15 resulting in gate metal diffusion into semiconductor layer or consumption of interfacial oxide present. Inverse piezoelectric effect model does not highlight the time dependent device breakdown in AlGaN/GaN HEMTs. The electrical degradation is also observed at voltages below critical voltage during constant voltage stressing in similarly fabricated devices. However time to breakdown is longer for applied lower voltages than for higher voltages. Heterostructures under excessive strain undergo relaxation to lower system energy either through attaining homogeneity at the interface via diffusion or generation of misfit dislocations at the interface. Both mechanisms tend to compete and the one with more favorable kinetic conditions takes place. These kinetic conditions may vary with temperature, electrical bias, external mechanical stress and concentration of chemical species (or ambient). Both diffusion and trap generation are rate dependent processes and hence can explain the time-dependent degradation in AlGaN/GaN HEMT. Strain relaxation models are investigated and validated experimentally in AlGaN/GaN HEMTs by monitoring device degradation under external conditions such as electrical bias, temperature, mechanical stress and varying ambient.
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.
Statement of Responsibility:
by Amit Gupta.
Thesis:
Thesis (Ph.D.)--University of Florida, 2013.
Local:
Adviser: Nishida, Toshikazu.
Local:
Co-adviser: Thompson, Scott E.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2015-08-31

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0045920:00001

MISSING IMAGE

Material Information

Title:
Investigation of Electrical Bias, Mechanical Stress, Temperature and Ambient Effect on AlGaN/GaN Hemt Time-Dependent Degradation
Physical Description:
1 online resource (143 p.)
Language:
english
Creator:
Gupta, Amit
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Electrical and Computer Engineering
Committee Chair:
Nishida, Toshikazu
Committee Co-Chair:
Thompson, Scott E
Committee Members:
Law, Mark E
Gila, Brent P

Subjects

Subjects / Keywords:
algan -- degradation -- diffusion -- gan -- hemt -- reliability -- stress -- time-dependent
Electrical and Computer Engineering -- Dissertations, Academic -- UF
Genre:
Electrical and Computer Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
AlGaN/GaN HEMT technology is promising for RF and high power applications. However commercial usability of this technology is currently hindered because of its limited electrical reliability which still remains a major concern. AlGaN/GaN HEMTs have been shown to degrade irreversibly under typical device operation and there is widespread disagreement on the underlying fundamental physics for the observed device degradation. Electrical degradation in AlGaN/GaN HEMTs due to DC stressing is studied typically by performing electrical step stress tests and a critical voltage is determined. Device degradation is characterized by changes measured in electrical parameters, such as increase in Rs and RD, decrease in IDsat, decrease in gm, Vt shift and sub-threshold change. The widely accepted theory attributes such degradation to the inverse piezoelectric effect. Electric field due to applied bias generates biaxial tensile stress which together with intrinsic stress from lattice mismatch increases the elastic energy of AlGaN layer. AlGaN layer undergoes crystallographic defect formation when total elastic energy reaches a critical value. Another theory associates such degradationto thermally activated chemical reaction occurring at metal-semiconductor interface 15 resulting in gate metal diffusion into semiconductor layer or consumption of interfacial oxide present. Inverse piezoelectric effect model does not highlight the time dependent device breakdown in AlGaN/GaN HEMTs. The electrical degradation is also observed at voltages below critical voltage during constant voltage stressing in similarly fabricated devices. However time to breakdown is longer for applied lower voltages than for higher voltages. Heterostructures under excessive strain undergo relaxation to lower system energy either through attaining homogeneity at the interface via diffusion or generation of misfit dislocations at the interface. Both mechanisms tend to compete and the one with more favorable kinetic conditions takes place. These kinetic conditions may vary with temperature, electrical bias, external mechanical stress and concentration of chemical species (or ambient). Both diffusion and trap generation are rate dependent processes and hence can explain the time-dependent degradation in AlGaN/GaN HEMT. Strain relaxation models are investigated and validated experimentally in AlGaN/GaN HEMTs by monitoring device degradation under external conditions such as electrical bias, temperature, mechanical stress and varying ambient.
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.
Statement of Responsibility:
by Amit Gupta.
Thesis:
Thesis (Ph.D.)--University of Florida, 2013.
Local:
Adviser: Nishida, Toshikazu.
Local:
Co-adviser: Thompson, Scott E.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2015-08-31

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0045920:00001


This item has the following downloads:


Full Text

PAGE 1

1 INVESTIGATION OF ELECTRICAL BIAS, MECHANICAL STRESS, TEMPERATURE AND AMBIENT EFFECT ON AlGaN/GaN HEMT TIME DEPENDENT DEGRADATION By AMIT GUPTA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA I N PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013

PAGE 2

2 2013 Amit Gupta

PAGE 3

3 T o my loving family Mom, Dad, sisters: Abhilasha and Preeti and brother: Ashish

PAGE 4

4 ACKNOWLEDGMENTS Foremost I want to extend my most sincere gratitude to my advisor, Dr. To shikazu Nishida, for providing the opportunity and necessary resources to pursue my doctoral research under his mentorship at the University o f Florida. His const ant guidance and valuable motivation have inspired my professional and intellectual growth as a researcher. I am also grateful to my co chair Dr. Scott Thompson, Dr. Mark Law and Dr. Brent Gila for serving on my PhD committee and giving me technical guidan ce from time to time during the course of my research I would also like to thank Air Force Office of Scientific Research (AFOSR) for funding this research work as well as providing necessary test samples for the experimental study. T heir intellectual as w ell as commercial insight through regular review meetings and telecom s were helpful in guiding this re search in the right direction. I want to extend special thanks to my research colleagues, Dr. Andrew D. Koehler and Dr. Min Chu for their vital and fruit ful collaboration and technical insight during a portion of this project. I am also thankful to my other research colleagues for their assistance and technical discussions and for providing a pleasant working environment: Dr. Hyunwoo Park Dr. Srivatsan P arthasarathy, Dr. Antonio G. Acosta, Dr. Mehmet O. Baykan, Dr. Nicole Rowsey, Dr. Erin Patrick, Dr. Robert Dieme, Viswanath Sankar, Shancy Augustine, Peng Z hao and all fellow IMG (Interdisciplinary Microsystems Group) members. Lastly but not least, I want to express my deepest gratitude to my family mom, dad, brother and sisters and m y friend, Dr. Robert A. Cating, for their constant affection, support an d encouragement towards the completion of this work.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 13 ABSTRACT ................................ ................................ ................................ ................... 14 CHAPTER 1 INT RODUCTION ................................ ................................ ................................ .... 16 1.1 Overview and Motivation ................................ ................................ ................... 16 1.1.1 Brief History ................................ ................................ ........................ 17 1.1.2 Structural Details and Basic Operation of AlGaN/GaN HEMT ............ 18 1.1.3 Applications of AlGaN/GaN HEMT ................................ ...................... 21 1.2 Dissertation Outline ................................ ................................ ........................... 22 1.2.1 Reliability Considerations in AlGaN/GaN HEMT ................................ 22 1.2.2 Research Goals and Contributions ................................ ..................... 23 1.2.3 Dissertation Organization ................................ ................................ .... 24 1.3 Summary ................................ ................................ ................................ .......... 25 2 FAILURE MECHANISMS IN AlGaN/GaN HEMT ................................ .................... 28 2.1 Introduction to Failure Mechanisms in AlGaN/GaN HEMT ............................... 28 2.2 Degradation due to Hot Electron Effect and Trapping ................................ ...... 28 2.3 Degradation due to Strain from Inverse Piezoelectric Effect ............................. 30 2.4 Degradation due to Diffusion and Chemical Reaction ................................ ....... 31 2.5 Summary ................................ ................................ ................................ .......... 33 3 EFFECT OF MECHANICAL STRESS ON CHANNEL RESISTANCE AND GATE LEAKAGE CURRENT IN ALGAN/GAN HEMT ................................ ............ 36 3.1 Mechanical Wafer Bending ................................ ................................ ............... 36 3.1.1 Four Point Bending Setup ................................ ................................ ... 36 3.1.2 Procedure for Stress Measurements in Small Samples ...................... 37 3.2 Effect of Mechanical Stress on Channel Resistance of AlGaN/GaN HEMT ...... 38 3.2.1 Effect of Trapped Charges ................................ ................................ .. 39 3.2.2 Experimental Procedure to Measure Gauge Factor ............................ 40 3.2.3 Results on Gauge Factor Measurement ................................ ............. 42 3.3 Effec t of Mechanical Stress on Gate Leakage Current in AlGaN/GaN HEMT .. 43

PAGE 6

6 3.3.1 Experimental Procedure to Measure Gate Current Stress Se nsitivity ................................ ................................ ................................ ...... 44 3.3.2 Results on Gate Current Stress Sensitivity Measurement .................. 44 3.3.3 Gate Current Measurement as a Function of Temperature ................. 45 3.3.4 Discussion on Gate Leakage Mechanism ................................ ........... 46 3.4 Summary ................................ ................................ ................................ .......... 50 4 TIME DEPENDENT ELECTRICAL DEGRADATION OF ALGAN/GAN HEMT ....... 63 4.1 Background ................................ ................................ ................................ ....... 63 4.1.1 Literature Review on Device Degradation by Electrical Step Stress ... 63 4.1.2 Models for AlGaN/GaN HEMT Device Degradation ............................ 64 4.1.3 Role of Threading Dislocations in AlGaN/GaN HEMT ......................... 65 4.2 Electrical Step Stress with Varying Time Step ................................ .................. 67 4.2.1 Experimental Procedure ................................ ................................ ...... 67 4.2.2 Results and Discus sion ................................ ................................ ....... 68 4.3 Summary ................................ ................................ ................................ .......... 69 5 STRAIN RELAXATION IN AlGaN/GaN HEMT ................................ ....................... 74 5.1 Review of Strain Relaxation in Heterostructures ................................ ............... 74 5 .2 Strain Relaxation through Trap Generation ................................ ...................... 74 5.2.1 Rate of Dislocation multiplication ................................ ........................ 75 5.2.2 Dislocation Motion and Dislocation Glide Velocity in Heterostructures ................................ ................................ ............................ 76 5.3 Strain Relaxation through Diffusion ................................ ................................ ... 77 5.3.1 ................................ ................................ ......... 77 5.3.2 Self Diffusion and Inter Diffusion in Heterostructures ......................... 78 5.3.3 Effect of Traps on Diffusion Rate ................................ ........................ 79 5.4 Summary ................................ ................................ ................................ .......... 80 6 EFFECT OF ELECTRICAL BIAS, TEMPERATURE, MECHANICAL STRESS AND AMBIENT ON AlGaN/GaN HEMT TIME DEPENDENT DEGRADATION ...... 82 6.1 Introduction and Research Objective ................................ ................................ 82 6.1.1 AlGaN/GaN HEMT Sampl e Description ................................ .............. 84 6.1.2 Sample Requirement for Experimental Study ................................ ..... 84 6.2 Effect of Electrical Bias on AlGaN/GaN HEMT Degradation ............................. 86 6.2.1 Experimental Details ................................ ................................ ........... 86 6.2.2 Results and Discussion ................................ ................................ ....... 87 6.3 Effect of Electrical Bias and Temperature on AlGaN/GaN HEMT Degradation ................................ ................................ ................................ ......... 89 6.3.1 Experimental Details ................................ ................................ ........... 89 6.3.2 Results and Discussion ................................ ................................ ....... 89 6.4 Effect of Electri cal Bias and Mechanical Stress on AlGaN/GaN HEMT Degradation ................................ ................................ ................................ ......... 91 6.4.1 Experiment Details ................................ ................................ .............. 91

PAGE 7

7 6.4.2 Results and Discussion ................................ ................................ ....... 93 6.5 Effect of Electrical Bias and Ambient on AlGaN/GaN HEMT Degradation ........ 94 6.5.1 Experimental Details ................................ ................................ ........... 94 6.5.2 Results and Discussion ................................ ................................ ....... 96 6.6 Summary ................................ ................................ ................................ .......... 97 7 MODELLING AND ANALYSIS OF EFFECT OF ELECTRICAL BIAS, TEMPERATURE, MECHANICAL STRESS AND AMBIENT ON AlGaN/GaN HEMT TIME DEPENDENT DEGRADATION ................................ ........................ 109 7.1 Resultant Mechanical Stress in AlGaN layer due to Applied Bias ................... 109 7.2 Metal to Semiconductor Current Transport ................................ ..................... 112 7.3 Summary ................................ ................................ ................................ ........ 116 8 CONCLUSION AND FUTURE WORK ................................ ................................ .. 124 8.1 Conclusion ................................ ................................ ................................ ...... 124 8.2 Recommendations for Future Work ................................ ................................ 127 LIST OF REFERENCE S ................................ ................................ ............................. 129 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 143

PAGE 8

8 LIST OF TABLES Table page 3 1 Best fit material coefficient value s for AlN and GaN [57]. ................................ ........ 52 7 1 Best fit of parameters used in I G simulation by metal semiconductor tunneling model. ................................ ................................ ................................ ............... 118

PAGE 9

9 LIST OF FIGURE S Figure page 1 1 Schematic of the cross section of a basic AlGaN/GaN HEMT structure. ................ 26 1 2 Schematic of wurtzite GaN crystal structure with G a face and N face polarities [13]. ................................ ................................ ................................ .................... 26 1 3 Schematic of AlGaN/GaN heterostructure showing directions of spontaneous and piezoelectric polarizations and sheet charge density induced due to result ant polarization [13]. ................................ ................................ .................. 27 2 1 EL intensity with respect to changing V G and V DS indicating that hot electron effects are maximized in the "semi on" mode of device operation [20]. .............. 34 2 2 Degradation in various electrical quantities above critical voltage (Vcrit = 26V) in a step stress experiment at V DS = 0V state in AlGaN/GaN HEMT [30]. .............. 34 2 3 Improved electrical stability of Pt gated AlGaN/GaN HEMT over Ni gated AlGaN/GaN HEMT [48]. ................................ ................................ ..................... 35 2 4 TEM images of cross section of Ni gate/AlGaN layer interface in AlGaN/GaN HEMT before and after electrical stress. ................................ ............................ 35 3 1 Schematic of four point mechanical wafer bending set up. As indicated, top surface is under tensile stress while bottom surfa ce is under compres sive stress. ................................ ................................ ................................ ................. 53 3 2 Four point mechanical wafer bending fixture ................................ ......................... 53 3 3 Calibration of stress applied by four point flexure based bending set up using a strain gauge mounte d on top of wafer, under stress ................................ .......... 54 3 4 A 15% reduction in channel resistance is observed during 1200 seconds of measurement under optical illumination with incandescent microscope light. .... 54 3 5 Experimental setup to measure stress sensitivity of channel resistance (R ch ), i.e. gauge factor, while illuminating with sub bandgap photon energie s using a filtered UV light ................................ ................................ ................................ .. 55 3 6 Measured channel resistance (R ch ) stabilized to less than 0.02% variation over 1500 seconds of measurement under illumination with filtered UV light. ............ 55 3 7 Change in normalized R ch measured while incrementally applying longitudinal uniaxial stress. ................................ ................................ ................................ .... 56

PAGE 10

10 3 ch measurements averaged at each increment of applied stress. Error bars represent 99% confidence interval of uncertainty in the measurement. ................................ ................................ ................................ ..... 56 3 9 Measured J G stabilized befo re applying external mechanical stress. ...................... 57 3 10 Normalized change in J G for incrementally increasing and decreasing longitudinal uniaxial mechanical stress for gate biases V G = 0.25V and V G = 4 V. ................................ ................................ ................................ .................... 58 3 11 Normalized change in J G averaged at each increment of applied stress, for gate biases V G = 0.25V and V G = 4 V. Error bars represent 99% confidence interval of uncertainty in the measurement. ................................ ........................ 59 3 12 Experimental and simulated mechanical stress sensitivity of J G (per 100 MPa of stress), including T and r variation with mechanical stress, as a function of reverse gate bias. ................................ ................................ ........................... 60 3 13 Measured J G vs. V G for different chuck temperatures from T C = 300K to 400K. ... 61 3 14 PF plot showing linear fit to th e measured J G vs. V G data for T C = 300 K to 400 K. ................................ ................................ ................................ ........................ 61 3 15 Schematic of PF emission through a dislocation band and conceptual reverse current in AlGaN/GaN HEMTs. ................................ ................................ ........... 62 4 1 Changes in normalized I DMAX R D R S I Gstress and I GOFF versus voltage step at V DS =0V state in a step stress experiment [30]. ................................ ................... 71 4 2 Output power degrada tion during RF stress of AlGaN/GaN HEMT with field plated gate [83]. ................................ ................................ ................................ .. 71 4 3 SSRM images of undeformed and 5% deformed samples respectively [104]. ....... 72 4 4 I G degradation of AlGaN/GaN HEMT at V DS =0V during electrical step stress test with 100sec/step time step. Inset shows I G G ................................ 72 4 G breakdown) on time step (length of time for each applied voltage step) in electrical step stress test. ........ 73 4 square root of the time step s indicates potential role of diffusion phenomenon in determining critical voltage. ................................ ................................ ................................ ............... 73 5 1 Plot of activation energy of dislocation motion (E A ) v ersus lattice energy in III V materials [121]. ................................ ................................ ................................ ... 81

PAGE 11

11 6 1 Schematic of the cross section of AlGaN/GaNHEMT hterostructure used for time dependent degra dation study. ................................ ................................ .... 99 6 2 Time dependent I G degradation of AlGaN/GaN HEMT on SiC substrate by varying reverse gate bias (V GS = 25V, 30V, 35V) at V DS =0V state and at room temperature. ................................ ................................ .............................. 99 6 3 I Gfailure_avg estimated over time of degradation increases with increase in applied reverse gate bias indicating that degradation is accelerated with increase in applied bias in AlGaN/GaN HEMT. ................................ ................................ ... 100 6 4 Strong linear dependence between I Gfailure_avg (estimated over time of degradation) and square root of time to degrade for the time dependent degradation test at varying reverse gate biases. ................................ .............. 100 6 5 Peltier heater set up to vary chuck temperature. ................................ ................... 101 6 6 Time dependent I G degradation of AlGaN/GaN HEMT by varying the ch uck temperature (T=375K, 400K, 450K) at V GS = 25V and V DS =0V. ....................... 101 6 7 Plot of log of time to degrade versus chuck temperature. Activation energy of E A =0.28 0.06 eV is obtained. ................................ ................................ ........... 102 6 8 Linear dependence between I Gfailure_avg and square root of time to degrade for the time dependent degradation test at varying temperatures. ........................ 102 6 9 Experimental set up to investigate effect of mechanical stress on rate of I G degradation in AlGaN/GaN HEMT. ................................ ................................ ... 103 6 10 Varying two different mechanical stresses alternatively to study e ffect of mechanical stress on rate of AlGaN/GaN HEMT degradation. ......................... 103 6 11 Effect of uniaxial tensile stress on rate of I G degradation in AlGaN/GaN HEMT on Si substrate which is electrically st ressed at V DS =0V, V GS = 65V at room temperature. ................................ ................................ ................................ ..... 104 6 12 % Change in I G during time dependent degradation in AlGaN/GaN HEMT at V DS =0V, V GS = 65V at room temperature during two cycles of varying e xternal uniaxial tensile stress ................................ ................................ ...................... 104 6 13 Effect of uniaxial compressive stress on rate of I G degradation in AlGaN/GaN HEMT on Si substrate which is electrically stressed at V DS =0V, V GS = 35V at room temperature. ................................ ................................ ............................ 105 6 14 % Change in I G during time dependent degradation in AlGaN/GaN HEMT at V DS =0V, V GS = 35V at room temperature during two cycles of varying external uniaxial compress ive stress. ................................ ................................ ............. 105

PAGE 12

12 6 15 Lakeshore cryogenic probe station chamber used for the study of effect of ambient variation on time dependent degradation in AlGaN/GaN HEMT. ........ 106 6 16 Varying two different ambients alternatively in cyclic manner while device is being electrically stressed under constant voltage to study the effect of ambient variation on time dependent degradation in AlGaN/GaN HE MT. ........ 106 6 17 Flow chart of the steps to purge cryostat chamber with desired ambient gas completely. ................................ ................................ ................................ ....... 107 6 18 Effect of varying amb ient between Oxygen and Nitrogen at 400K on I G degradation while device is electrically stressed at V GS = 25V and V DS =0V. ..... 108 6 19 Effect of varying ambient between Forming gas and Nitrogen at 4 50K on I G degradation while device is electrically stressed at V GS = 25V and V DS =0V. ..... 108 7 1 Simulated E field contours under the gate in AlGaN layer for varying reverse gate biases at V DS =0V. ................................ ................................ ..................... 119 7 2 Simulated and extrapolated vertical E field values (E xx ) under the gate for varying reverse gate bias at V DS =0V. ................................ ............................... 119 7 3 Si mulated in plane mechanical stress generated in AlGaN layer due to applied gate bias as a result of inverse piezoelectric effect. ................................ ......... 120 7 4 Schematic of energy band diagram in AlGaN/GaN HEMT unde r reverse gate bias showing electron tunneling from metal into semiconductor. ...................... 120 7 5 Schematic showing concentration of gate current near localized areas of gate metal diffusion in AlGaN/GaN H EMT. ................................ ............................... 121 7 6 Flow chart showing steps to simulate I G in AlGaN/GaN HEMT under reverse gate bias including diffusion under the gate at Schottky interface. ................... 121 7 7 Simulated and measured I G indicating time dependent degradation in AlGaN/GaN HEMT at varying reverse gate biases. ................................ .......... 122 7 8 Simulated and measured I G indicating tim e dependent degradation in AlGaN/GaN HEMT at varying temperatures. ................................ .................... 122 7 9 Simulated I G during AlGaN/GaN HEMT degradation at varying reverse gate biases due to gate metal diffusion into the AlG aN layer. ................................ .. 123 7 10 Linear dependence between I Gfailure_avg and square root of time to degrade. ....... 123

PAGE 13

13 LIST OF ABBREVIATIONS 2DEG 2 Dimensional e lectron g as CVS C onstant voltage stressing DC Di rect current FN Fowler Nordheim GF Gauge factor HEMT High electron mobility transistor MBE Molecular beam e pitaxy MEMS Micro electro mechanical systems MOCVD Metal organic chemical vapor deposition MOSFET Metal oxide semiconductor field effect transistor MTTF Mean time to f ailure PF Poole Frenkel SEM Scanning electron microscope TLM Transmission line m ethod UV Ultraviolet

PAGE 14

14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of t he Requirements for the Degree of Doctor of Philosophy INVESTIGATION OF ELECTRICAL BIAS, MECHANICAL STRESS, TEMPERATURE AND AMBIENT EFFECT ON AlGaN/GaN HEMT TIME DEPENDENT DEGRADATION By Amit Gupta Chair: Toshikazu Nishida Co chair: Scott E. Thompson Major: Electrical and Computer Engineering AlGaN/GaN HEMT technology is promising for RF and high power applications However commercial usability of this technology is currently hindered because of its limited electrical reliability which still remains a major concern AlGaN/GaN HEMTs have been shown to degrade irreversibly under typical device operation and there is widespread dis agreement on the underlying fundamental physics for the observed device degradation. Electrical degradation in AlGaN/GaN HEM Ts due to DC stressing is studied typically by performing electrical step stress tests and a critical voltage is determined. Device degradation is characterized by changes measured in electrical parameters, such as increase in R s and R D decrease in I Dsat decrease in g m V t shift and sub threshold change. The widely accepted theory attributes such degradation to the i nverse piezoelectric effect. Electric field due to applied bias generates biaxial tensile stress which together with intrinsic stress from la ttice mismatch increases the elastic energy of AlGaN layer. AlGaN layer undergoes crystallographic defect formation w hen total elastic energy reaches a critical value Another theory associates such degradation to thermally activated chemical reaction occu rring at metal semiconductor interface

PAGE 15

15 resulting in gate metal diffusion into semiconductor layer or consumption of interfacial oxide present. Inverse p iezoelectric effect model does not highlight the time dependent device breakdown in AlGaN/ GaN HEMTs The electrical degradation is also observed at voltages below critical voltage during constant voltage stressing in similarly fabricated devices. However time to breakdown is longer for applied lower vol tages than for higher voltages. Heterostructures under e xcessive strain undergo relaxation to lower system energy either th r ough attaining homogeneity at the interface via diffusion or generation of misfit dislocations at the interface. Both mechanisms tend to compete and the one with more favorable kinetic con ditions takes place. These kinetic conditions may vary with temperature, electrical bias, external mechanical stress and concentration of chemical species (or ambient). Both diffusion and trap generation are rate dependent processes and hence can e xplain t he time dependent degradation in AlGaN/ GaN HEMT. Strain relaxation model s are investigated and validated experimentally in AlGaN/GaN HEMTs by monitoring device degradation under external conditions such as electrical bias, temperature, mechanical stress an d varying am bient.

PAGE 16

16 CHAPTER 1 INTRODUCTION 1.1 Overview and Motivation High Electron Mobility Transistors (or HEMTs) are the most advanced of a new generation of III V compound semiconductor devices incorporating a hetero junction for their operation such as InGaAsP/InP AlGaN/GaN etc H etero junction devices are formed between two semiconductor materials of different composition and different band gaps and utilize the high mobility and high velocity of sheet of charge formed at the interface. This is contra ry to the conventional Si or GaAs based field effect transistors or bipolar devices, where a junction is formed between similar materials for e.g. bipolar transistors with n or p type Si forming emitter/base/collector regions In hetero junction devices designer s have an extra degree of freedom to vary both band structure and type of doping in various portions of the device. This gives HEMTs the advantage of significant improvements in charge transport properties and resultant device performance over co nventional FETs AlGaN/GaN HEMTs have gained importance over its other III V counterparts because of their material and structural benefits that make them attractive for high power and high frequency operation capabilities Temperature accelerated life tes ts for this technology have reported mean time to failure MTTF > 10 7 hr at T=150 o C [1] and MTTF~ 3.5x10 9 hr at T=125 o C [2] While the progress on AlGaN/GaN HEMT performance improvement is sustained, electrical instability and reliability issues observed during typical operation has limited the achi evement of full potential of these devices. A lGaN/ GaN HEMT s lack relatively process independent uniformity of results and widespread agreement on underlying basic physics for observed device failure. The

PAGE 17

17 device degradation still remains a challenge and hin ders the reliable long term implementation of AlGaN/GaN HEMTs. Although there is a continuous evolution of adopted proc esses and technologies to improve performance, understanding the fundamental mechanisms causing device degradation is necessary for impro ving AlGaN/GaN HEMT reliability. 1.1.1 Brief History The idea of a heterojunction was first proposed by William Shockley in 1951 [3] and A. I. Gubanov around the same time, who developed theor ies illustrating heterojunctions [4] However it was in 1960 when R.A. Anderson predicted an accumulation of a sheet of charge carriers for ming at the interface of Ge/ GaAs interface based on a simple energy band model [5] Alternative models for heterojunction behavior have also been proposed by Adams and Nussbaum [6] by von Roos [7] and by Hil m i and Nussbaum [8] Although the idea of a heterojunction had been proposed way back in 1950s, it could never be implemented because of difficulty to grow high quality extremely abrupt interfaces between dissimilar materials with avail able fabrication techniques at that time However the development of advanced epitaxial growth techniques in 1970s such as Molecular Beam Epitaxy (MBE) and Metal Organic Chemical Vapor Deposition (MOCVD) enabled the fabrication of high quality semiconducto r heterostructures [9] In 1978, R. Dingle et. al. first observed the formation of two dimensional electron gas or sheet of charge at the GaAs/ AlGaAs superlattice interface with improved carrier mobility [10] Eventually in 1980 T. Mimura et. al. of Fujitsu, Japan; first fabricated and demonstrated GaAs/n Al x Ga 1 x As heterojunction field effect transistor, u tilizing high mobility of 2D electron gas [11] The first fabrication and

PAGE 18

18 dc characterization of wide band gap GaN/Al x GaN 1 x N HEMT grown on s apphire substrate using low pressure MOCVD, was presented by M. A Khan et. al. in 1993 [12] 1.1. 2 Structural Details and Basic Operation of AlGaN/GaN HEMT AlGaN/GaN HEMTs are depletion mode field effect transistors with a high density of charge carrier (2DEG) forming channel at the interface, as a result of large spontaneous and piezoelectric polariz ations in AlGaN and GaN layers. AlGaN/GaN HEMTs consists of a heterojunction formed by AlGaN and GaN layers grown epitaxially on a semi insulating substrate using advanced fabrication techniques, notably MBE or MOCVD Figure 1 1 shows a schematic of a AlGa N/GaN HEMT cross section. One of the key issues in HEMT structure is the lattice mismatch between differing materials forming epi taxial layers. Si, SiC and Sapphire are the most widely used substrate material s for AlGaN/GaN HEMTs. Substrate should have hig h resistivity, low interconnect capacitance, high thermal conductivity and lattice and thermal expansion coefficient match with the subsequent III V layer. After depositing GaN on the substrate, AlGaN barrier layer is pseudomorph ically grown on top of GaN which implies that relatively a very thin layer of AlGaN is deposited so that its crystal lattice stretches itself on relaxed thicker GaN layer to achieve lattice match This induces a biaxial strain in the AlGaN layer. A stress mitigating transition layer (such as AlN) is often deposited between GaN layer and substrate to achieve lattice match as shown in Figure 1 1 The high charge density in 2DEG results from polarization induced bound interface charges. The wurtzite GaN and AlN have very high spontaneou s and piezoelectric polarization, about ten times larger than other III V semiconductor compounds [13] The direction of polarization can be determined by the polarity of AlGaN/GaN heterostructure (N face or Ga face). GaN (similarly AlN) ha s a

PAGE 19

19 noncentrosymmetric compound crystal with closely spaced hexagonal bilayers, one formed by cations and other by anions, ar ranged in {0001} basal plane. This lead s to two possible polar faces or polarities whose direction is given by a vector pointing from Ga atom to nearest neighbor N atom. A basal surface with Ga atoms (or N atoms) on top is Ga faced (or N faced) and corresponds to [0001] (or [000 1 ]) polarity as shown in Figure 1 2 The Ga face samples have higher carrier density and smaller interface roughness, hence higher carrie r mobility than N face samples. Polarization and Formation of 2DEG Spontaneous polarization in a crystal depends on its structural parameters such as lack of inversion symmetry and ionic nature of covalent bonds between the atoms, similar to wurtzite structure of GaN. In GaN crystal, the ratio of distance between two adjacent atomic layer and lattice constant (c/a) which results in a built in dipole and induces spontaneous polarization (P SP GaN ) a long the direction of polarity. In AlGaN, spontaneous polarization is expressed in terms of Al mole fraction x using Ve g as (1 1) It should be noted that spontaneous polarization in AlN ( 0.081 C/m 2 ) is larger than in GaN ( 0.029 C/m 2 ) [13] GaN and AlN are piezoelectric materials and ex hibit electrical polarization under mechanical strain. Due to lattice mismatch, psedomorphic AlGaN layer is under biaxial tensile stress resulting in piezoelectric polarization (P PE AlG aN ) along c axis given by [13] (1 2 )

PAGE 20

20 Here e 31 and e 33 are piezoelectric coefficients, c 13 and c 33 are elastic coefficients and a o is unstrained lattice constant Typical values of these coefficients for GaN and AlN are listed in Table 3 1. GaN layer is relaxed and he nce does not exhibit piezoelectric polarization. T he piezoelectric polarization is negative for layers under tensile strain and positive for layers under compressive strain A negative spontaneous polarization is directed towards the substrate for a Ga fac e layer Hence piezoelectric and spontaneous polarizations are aligned parallel in tensile strain ed layer and anti parallel in a compressively strained AlGaN layer as shown in Figure 1 3 Both piezoelectric and spontaneous polarization s change sign for N f ace layer The difference between the total polarization in AlGaN and GaN layers induces a positive fixed sheet charge density at AlGaN/GaN interface (on AlGaN side) (1 3) Free electrons will accumulate at AlGaN/GaN interface (on GaN side) to compensate th is positive fixed charge an d get quantum mechanically confined forming a sheet of charge or 2DEG near the interface Since intrinsic carrier density for GaN is low ( n i = 10 10 cm 3 ), e lectrons forming 2DEG origina te from surface states as described by [14] Other theories attribute source/drain Ohmic contacts to provide necessary electrons forming 2DEG. The maximum 2DEG sheet carrier concentration located at the interface is [13] (1 4)

PAGE 21

21 Here t AlGaN is AlGaN barrier thickness, b is the Schottky barrier height at the gate, E F is Fermi level relative to GaN conduction band edge a E C is the conduction band offset at AlGaN/GaN interface. The ability to achieve high density sheet carrier concentration at the interface with superior transport properties due to its quantum confinement contributes to the performance of AlGaN/GaN HEMT s. 1.1.3 Applications of AlGaN/GaN HEMT AlGaN/GaN HEMTs have emerged as the most suitable technology for high voltage, high power operation s at microwave frequencies. The ability to achieve high density sheet carrier concentration at the interface with sup erior transport properties due to its quantum confinement contributes to the performance of AlGaN/GaN HEMTs. Historically, AlGaN and GaN devices were used in optoelectronics application, such as lasers and LEDs, because of its direct and tunable band gap [15] However benefits of GaN over other III V materials namely high breakdow n field, high er carrier mobility large r saturation velocity higher thermal stability and 3.4eV wide band gap makes GaN suitable for high power, high frequency applications [16], [17] Consequently, superior device properties such as P out = 30Wmm 1 at 8 GHz has been reported for such devices [18] Hence, AlGaN/GaN HEMT technology is currently emerging as a promising candidate for extremely relevant applications such as high power and high frequency applications, ultra wide band communication systems, robust low noise applications and high temperature scaled digital devices. AlGaN/GaN HEMTs are currently employed in expanding markets in communications, radar, sensors, and automotive for both military and commercial applications. Integrati ng with Si(111) substrates improves its device performance while reducing cost S everal commercial vendors, such as Cree, Fujitsu,

PAGE 22

22 Nitronex, RFMD, and Triquint, are actively utilizing AlGaN/GaN HEMT tec hnology for commercial applications. For example a commercially available AlGaN/GaN power transistor by CREE operate s at 2.7 3.5 GHz range (S band) with P out = 240 W and 60% power added efficiency (PAE) 1.2 Dissertation Outline 1.2.1 Reliability Considerat ions in AlGaN/GaN HEMT Despite its attractive high frequency and high power performance, the electrical stability and reliability issues of AlGaN/GaN HEMT restrict its full potential deployment for commercial applications. Point and structural defects can reside in GaN bulk, AlGaN barrier layer at AlGaN/GaN inte rface, gate AlGaN Schottky interface and on ungated AlGaN surface and create trapping centers [19] ,[ 18] These trap ping centers compromise device performance by causing degradation characterized by low frequency noise [21], [22] transconductance frequency dispersion [21], [23] current collapse [24], [25] gate lag and drain lag transients [26], [27] enhanced gate leakage [28] shift in threshold voltage [29] and light sensitivity [24], [29] Generation of new defects can reduce device performa nce or cause permanent failure. Stress is inherent to AlGaN/GaN HEMTs and impacts device performance and reliability. Large mechanical s tr ess profiles are created during device fa brication due to lat tice mismatch between epi t a xial layers potentially creating point defects such as vacancies and interstitials and other structural defects such as threading dislocations. Stress can also be generated during operation by inverse piezoele ctric effect or due to thermal expansion coefficient mismatch La rge E fields generated under gate during typical operation induces additional stress in the active layers, since GaN and AlGaN are piezoelectric materials. AlGaN/GaN HEMTs are found to underg o device

PAGE 23

23 degradation during typical operation. It has been proposed that stress generated from the inverse piezoelectric effect can lead to observed permanent device degradat i on [30] This additional stress together with inbuilt stress in device epi taxial layers resultantly increase its critical limit, strain relaxation occurs through crystallographic defect formation. Another theory explains observed device failure by the degradation of gate/AlGaN Schottky interface via field driven chemical reaction or diffusion. A potential metal diffusion into AlGaN barrier (also called gate sinking) reduces AlGaN barrier and creates leakage path leading to gate current increase [20 ] Device degradation has also be attributed to h ot electron effects that result in current collapse [31] 1.2.2 Research Goals and Contribution s The main objective of this work is to provide a refined understand ing of fundamental mechanisms leading to observed degradation of electrical performance in AlGaN/GaN HEMTs. To accomplish this, a systematic study of AlGaN/GaN HEMT electrical degradation will be conducted through detailed experimentation and study of appr opriate physical models. Time dependent degradation of AlGaN/GaN HEMTs will be investigated through effect s of electrical bias, temperature, mechanical stress and ambient (such as O 2 N 2 and forming gas) on degradation of gate leakage current These extern al parameters will be varied to accomplish a systematic and detailed experimental study of device degradation. Also strain relaxation models will be evaluated to describe observed degradation. Since strain is inherent to AlGaN/GaN HEMTs, it is essential t o understand the effect of mechanical stress on device performance and reliability. The effect of externally applied mechanical stress on AlGaN/GaN HEMT channel resistance and gate leakage will also be investigated.

PAGE 24

24 The potential contribution of this study is to ascertain a time dependent degradation causing permanent failure in AlGaN/GaN HEMTs. This study shall also identify the effects of electrical bias, temperature, mechanical stress and ambient on device electrical degradation through reliable experime ntal techniques suggest most suitable physical model s that qualitatively explain experimental results and gain a better understanding of the physics o f degradation mechanisms causing device failure in AlGaN/GaN HEMT s An accurate value of AlGaN/GaN HEMT g auge factor wi ll be extracted as well as dominant transport mechanism s for gate leakage in AlGaN/GaN HEMT will be suggested. In addition, the potential outcomes of this work can be used as guidelines to improve AlGaN/GaN HEMT design for a reliable long ter m operation. 1.2.3 Dissertation Organization The remaining document is organized as follows. Chapter 2 presents a brief literature review of published physical models that explain the observed degradation in AlGaN/GaN HEMTs. Chapter 3 d iscusses the effect of externally applied uniaxial tensile and compressive stress on channel resistance and on gate current. It describes a reliable experimental technique to estimate gauge factor value accurately for AlGaN/GaN HEMTs and also discusses possible transport mech anisms dominating gate leakage current in these devices. Chapter 4 presents experimental results to ascertain a time dependent degradation causing permanent device failure in AlGaN/GaN HEMT s Chapter 5 briefly discusses models for strain relaxation phenome non in hetero structure devices, including III V hetero structures. Chapter 6 focuses on a detailed experimental study to investigate the effect s of varying external conditions like bias, temperature, mechanical stress and ambient on the observed time depe ndent degradation in AlGaN/GaN HEMT s It presents experimental techniques to

PAGE 25

25 gain fundamental understanding of device failure mechanisms. Chapter 7 investigates some fundamental mathematical models to analyze experimental results on observed time dependent degradation in AlGaN/GaN HEMTs (in Chapter 6) under varying bias, temperature and mechanical stress and gain an insight into physics of failure mechanism s causing device degradation Chapter 8 summarizes the work presented in this dissertation and also pr ovides recommendations for potential future work to further study reliability issues in AlGaN/GaN HEMTs. 1.3 Summary AlGaN/GaN HEMTs have gained tremendous importance for high power and high frequency operation capabilities since its inception This chapt er presents a brief history on evolution of III V heterostructures and advantages of AlGaN/GaN HEMTs over other existing semiconductor technologies Some fundamental structural details of AlGaN/GaN HEMT are discussed including presence of in built strain i n the epitaxial layer and formation of 2DEG. There is a continuous evolution of adopted processes and technologies to improve AlGaN/GaN HEMT performance, however high power and high field operations have exacerbated reliability issues of these devices. Th is chapter also briefly discusses the reliability considerations that hinder achievement of full potential of this technolo gy. Finally, main objectives and potential contribution of thi s work and organization of this dissertation document are presented.

PAGE 26

26 Figure 1 1 Schematic of the cross section of a basic AlGaN/GaN HEMT structure Figure 1 2 Schematic of wurtzite GaN crystal structure with Ga face and N face polarities [13]

PAGE 27

27 Figure 1 3 Schematic of AlGaN/GaN heterostructure showing directions of spontaneous and piezoelectric polarizat ions and sheet ch arge density induced due to resultant polarization [13]

PAGE 28

28 CHAPTER 2 FAIL URE MECHANISMS IN AlGaN/GaN HEMT 2.1 Introduction to Failure Mechanisms in AlGaN/GaN HEMT AlGaN/GaN HEMTs exhibit an excellent capability for high power and high frequency applications because of its structural and material benefits of GaN, namely wide band gap, higher saturation velocity and higher breakdown field. However at high field, these devices also electrically degrade, leading to reliability issues that currently limit the attainment of its full poten tial deployment commercially. Epitaxially grown GaN films contain a high density of threading dis locations, about 10 5 10 6 cm 2 [32] The abundance of point and structural defects in the bulk or at the surface act as trapping centers and are considered the main origin of AlGaN/GaN HEMT reliabil ity issues. Charging and discharging of traps can limit device performance. In addition to native defects formed during epitaxial growth, additional defects can be created during device operation that can permanently degrade the device, even to breakdown Several theories have been proposed for degradation mechanisms; however there still remains a widespread disagreement on the fundamental physics of electrical degradation in AlGaN/GaN HEMT s. Some of the more widely accepted theories for AlGaN/GaN HEMT de gradation are reviewed in the subsequent sections. 2.2 Degradation due to Hot Electron Effect and Trapping The h ot electron effect is a current driven phenomenon due to the presence of large electric field densities leading to device failure. During high v oltage operation of AlGaN/GaN HEMTs, high lateral electric field increases the drift velocity of electrons in electrons may gain enough kinetic energy to surmount the c onduction band offset at the

PAGE 29

29 AlGaN and GaN heterojunction and tunnel into AlGaN barrier layer creating traps at AlGaN/GaN or passivation/AlGaN interfaces between drain and gate c ontacts or even under the gate [33], [34] Consequently hot electron degradation is observed as current collapse, increase in series resistance, decrease in transconductance, increase in gate leakage a nd DC to RF dispersion effects [31], [34] The influence of hot electron effect on field accelerated failure is modeled in [33] given as (2 1) This law relates the ratio of (absolute) gate current and drain current to the length of the channel region (L eff ) over which impact ionizatio n occurs which depends on lateral E field. As can be seen, this model pre dicts that hot electron effects respond logar ithmically to the inverse of the difference between V D SAT and V DS However in AlGaN/GaN HEMTs the gate current cannot be used as an indic ator for hot electron effects because gate leakage in these devices is influenced by several other parasitic contributions and tunneling [31] Meneghesso et.al. [2 0] employed e lectroluminescence (EL) to detect hot electrons in AlGaN/GaN HEMTs within the device active area by observing the light intensity emitted from the channel. It was observed that the hot electron effect, hence EL intensity is most pronounced i n the semi on mode of operation of the device when the gate is just biased high enough to negate pinch off and form a conducting channel depicted in Figure 2 1 If a high V DS is now applied, the lateral field across the channel in maximized, thereby maxim izing hot electron density. If the gate is biased past threshold voltage, the difference between the gate and drain potential decreases and the lateral field within the channel becomes too small to induce

PAGE 30

30 a substantial density of hot electrons [35] The h ot electron effec t is also reduced if there is any form of surface passivation on the device, possibly due to reduction of surface traps as a result of passivation. It has also been observed that NH 3 plasma treatment increases AlGaN/GaN HEMT resistance to hot electron effe cts, indicating that the treatment of sample surfaces with hydrogen reduces susceptibility to hot electron degradation [36] 2.3 Degradation due to Strain from Inverse Piezoelectric Effect One of the wide ly accepted the ories for AlGaN/GaN degradation is based on strain generated due to the inverse piezoelectric effect [37] A s discussed in the previou s chapter, strain is inherent in the AlGaN/GaN HEMT Large mechanical strain profiles are created within the structure during epitaxial growth and during high power oper ation. In built stress in AlGaN/GaN HEMTs can arise from lattice mismatch between epitaxial layers, between epitaxial layer and substrate, from the passivation layer and also from thermal expansion coefficient mismatch. A thin AlGaN barrier layer is pseudo morphically grown on relaxed GaN layer which induces a biaxial tensile stress in AlGaN layer as it stretches over GaN to achieve lattice match. For 26% Al mole fraction, this biaxial tensile stress is ~ 3GPa [38] J. Joh et.al. [30] [37] c onducted a series of electrical step stress experiments to study electrical degradation in AlGaN/GaN HEMTs. Electrical bias (V GS or V DS ) was stepped an d various electrical parameters were monitored. It was observed that when a critical voltage is reached, devices exhibited electrical degradation marked by an increase in gate leakage, decrease in drain current and increase in source/drain resistance as sh own in Figure 2 2 Such electrical degradation was attributed to additional mechanical stress generated inside the device due to the inverse piezoelectric effect that can initiate defect formation

PAGE 31

31 leading to irreversible degradation. During device operatio n at high voltages, large vertical electrical fields are generated under the gate in the AlGaN barrier layer This electric field is the largest at the gate edges Since AlGaN is a piezoelectric material, this field creates additional mechanical stress wit hin the layer (500 MPa for V GS = 30V in Al 0.26 Ga 0.74 N/GaN HEMT with 18nm AlGaN layer [38] ). The mechanical stress due to inverse piezoelectric effect adds to the built in stress inside AlGaN layer and increases the When this elastic energy reaches its critical limit (at applied critical vol tage) ; the epitaxial layer undergoes relaxation through cr ystallographic defect creation in the form of cracks and pits, near the gate edges where field is the largest These defects may act as trapping centers for electrons, degrading device performance a nd reliability 2.4 Degradation due to Diffusion and Chemical Reaction Gate sinking or gate metal in diffusion in to the semiconductor bulk is a well known effect in III V HEMTs causing degradation [39], [40] and resultant variation in the Schot tky interface characteristics [41] Gate contacts to AlGaN/ GaN HEMTs are applied as Schottky contacts using Ni, Pt, Re as well as high refractory intermetallic compounds suc h as Silicides (e g. NiSix) and N itrides (e g. WSiN) [42] Temperature activated stability tests such as annealing are often used for reliability study of Schottky interfaces i n III V devices especially GaAs and AlGaAs based Schottky diodes. For a Al/AlGaAs diode, Schottky barrier height was found to decrease from 0.96eV to 0.80eV while ideality factor i ncreased from 1.27 to 1.78 on increasing annealing temperature from 300K to 673K, which was attributed to the degradation of metal semiconductor co ntact [43] Similar aging tests by annealing AlGaN/GaN HEMT samples showed inter diffusion of metals resulting in degradation of gate characteristics. It was obs erved that

PAGE 32

32 annealing a AlGaN/GaN sample, with Ni/Au as gate contact and Ti/Al/X/Ti/Au (X being Ir or TiBr 2 or Ni) as Ohmic contact, at 350 o C for 25 days showed an encroachment of Ohmic metal into gate metal for the case of Ni in O hm ic contact [44] This degradation was characterized by increase in gate current for the sample. In another report, AlGaN/GaN HEMT with Ti/Al/Pt/Au Ohmic contact and Pt/Au Schottky contact, when step stressed up to 400 o C, show ed marginal degradation up to 300 o C and significant degradation beyond [45] Surfa ce morphology of the metal contact showed good stability indicating that material defects are responsible for the high temperature instability. In a r ecent study Kuball et.al. [46] observed that an increase in trap density in electrically degraded Ni gated AlGaN/GaN HEMT followed a one dimensional diffusion model of the form given by a p seudo infinite source as [47] (2 2 ) Here N diff is the d iffusant concentration X is the d iffusion distance D is the d iffusion coefficient and t is the diff usi on time. It was inferred that pit like defects observed in the degraded devices formed possibly due to gate metal diffusion down pre existing threading dislocations near the gate edge. This could be i nitiated by electric field driven chemical reaction a t the Schottky interface Similar observations were confirmed by Lo et.al. in their study o f electrical degradation of Ni gated and Pt gated AlGaN/GaN HEMT s [48] It was found that devices with Ni gate appear t o degrade at a much faster rate and at a lower critical voltage than Pt gated devices, suggesting that degradation kinetics are different for different gate metals as shown in Figure 2 3 Figure 2 4 shows a TEM image of an electrically degraded Ni gated AlGaN/GaN HEMT illustrating diffusion of gate metal into the AlGaN barrier layer [49]

PAGE 33

33 2.5 Summary AlGaN/GaN HEMTs have been observed to degrade at high gate to source and gate to drain voltages. When the gate voltage is stepped, ultimately these dev ices undergo a catastrophic failure Several physical models have been published th at explain different mechanisms leading to device degradation. This chapter reviews in detail some of the widely accepted theories of degradation mechanisms in AlGaN/GaN HEM Ts. Specifically, degradation due to hot electron effects, degradation as a result of excessive strain due to inverse piezoelectric effect during high field operation and degradation due to field driven chemical reaction or diffusion at metal/semiconducto r Schottky interface are discussed in this chapter.

PAGE 34

34 Figure 2 1 EL intensity with respect to changing V G and V DS indicating that hot electron effects are maximized in the "semi on" mode of device operation [20] Figure 2 2 Degradation in various electrical quantities a bove critical voltage (Vcrit = 26V) in a step stress experiment at V DS = 0 V state in AlGaN/GaN HEMT [30]

PAGE 35

35 Figure 2 3 Improved electrical stability of Pt gated AlGaN/GaN HEMT over Ni gated AlGaN/GaN HEMT [48] Figure 2 4 TEM images of cross section of Ni gate/AlGaN layer interface in AlGaN/GaN HEMT (A) co ntrol sample and (B) after electrical stress. T wo white arrows highlight reg ions with Ni and oxygen diffusion in degraded sample [49]

PAGE 36

36 CHAPTER 3 EFFECT OF MECHANICAL STRESS ON CHANNEL RESISTANCE AND GATE LEAKAGE CURRENT IN ALGAN/GAN HEMT 3.1 M echanical Wafer Bending Wafer bending has been extensively used to study the effect of mechanical stress on semicon ductor performance [50], [51] Mechanical wafer bending offers a simple and cost effective method to study underly ing physics of strained semiconductors. Modifying fabrication steps to impart process induced stress in the device can be expensive and potentially alter other device characteristics. Also it can be difficult to quantify the internal stress imparted to the device accurately. Hence wafer bending can be useful to vary externally applied mechanical stress in a controlled manner in experiments. Several techniques have been employed to apply mechanical stress to wafers externally Smith [52] in his piezoresistance measurement, used hanging weights to apply uniaxial tensile stress to semiconductor slabs Cantilevers are beams anchored at one end and can be bent from other end to apply mechanical stress Three point ben ding fixtures may also be used to apply mechanical stress externally. However in all these methods accurate stress calibration is difficult because of non uniform stress profile along the bent structure. In this work, we employ four point based wafer bend ing fixtures to apply uniaxial tensile and compressive stress on AlGaN/GaN HEMT wafer samples, capable of applying high uniaxial stress with improved uniformity of stress profile along the bent structures between the load points [53] 3.1.1 Fo ur Point Bending S etup In a four point bending set up a beam held between four points, two anchoring points at the bottom and two load points on the top as shown in Figure 3 1. The sample

PAGE 37

37 is bent with a constant radius of curvature resulting in a uniform s tress profile between the inner rods, unlike in cantilevers and three point bending fixtures. The magnitude of uniaxial stress on the top surface of sample (homogeneous material) between the inner rods is given by following the analysis from Timoshenko [54] (3 1) H of the material and t is the sample thickness. L and a are spacing betwe en the rods as shown in Figure 3 1. For calibration, the magnitude of applied stress estimated from above analysis was compar ed with the readings from a noncontact fiber optical displacement system and strain gauge measurements [55] 3.1.2 Procedure for Stress Measurements in Small S amples Wafer bending methods are potentially destructive and hence AlGaN/GaN HEMT wafer pieces are diced into smaller pieces (~1cm 2 ). This maximizes the number of measurable samples and provide s a cost effective method to measure mechanical stress effects. Since these wafer pieces are too small to be directly bent in wafer bending fixture, they are attached to a heat treated high carbon stainless steel plate to prepare sizeable bendable samples using these steps. The steel pla te is sanded off and cleaned using alcohol, ethanol and DI water treatment to remove any dirt or grease. A thin layer of thermally conductive epoxy (EPOTEK H74) is applied in the midd le of steel plate, wafer piece is placed on it and gently placed down and excess epoxy is wiped around the sides. To remove an y air pocket s and ensure secure bond, a metal washer is then placed on top of wafer piece and is carefully clasped using binder clip and entire assembly is heat treated in oven for 5 min at 100 o C The washer and metal binder are removed and sample is further heat treated on a hot plate at 150 o C for 5 min to

PAGE 38

38 completely cure the epoxy. The wafer p iece attached on steel plate can now be loaded into bending fixture to apply mec hanical stress. A str ain gauge is mounted on top of wafer piece to calibrat e the stress transferred from steel plate to wafer piece. Stress is applied and released to the sample to ensure that stress applied is elastic, as the strain gauge reading returned to the starting value as shown in Figure 3 3 Also, the stainless steel plate does not undergo any permanent deformation for the amount of mechanical stress employed in this study. Effects of mechanical stress on AlGaN/GaN HEMTs are studied by doing elect rical measurements while simultaneously varying the applied stress. For this purpose, devices need to be wire bonded and attached with the probe tips using 1mil gold wire. Other than standard ball and wedge bonder s that may cause delamination of device ter minal pads a novel technique is employed to attach wires to the m. Using the ball bonder, a ball of size suitable for a respective bond pad is formed at the end of gold wire. Wire is then cut to about 1cm length and attached to the end of probe tip using e lectric ally conductive epoxy (EPOTEK EE1 29 4) and cured at room temperature for 24 hours. The probe tip is then mounted on the micro positioner and using the micro position er, the ball end of the wire is dipped in the electrically conductive epoxy and care fully placed on the device terminal pad and cured at room temperature for 24 hours After the epoxy is cured, the probe tip is slight ly lowered to allow slack on the gold wire for vertical displacement of the sample d uring application of mechanical stress. 3.2 Effect o f Mechanical Str ess on Channel Resistance of AlGaN/GaN HEMT The normalized change character ized as the Gauge Factor (GF), i.e.

PAGE 39

39 ( 3 2) The mechanical strain can be related to the applied stress via Yo of the material. A wide range of gauge factor values for AlGaN/GaN HEMTs have been reported in literature ranging from 4 to 40,000. This large discrepancy in the estimation of gauge factor could arise from the inaccurate estimation of stre ss applied using various methods and also due to changes in the trapped charge density during elapsed time of resistance measurement, as AlGaN/GaN HEMTs are known to contain high density of traps in form of point defects, dislocations and surface traps. Me thods used in previous studies on AlGaN/GaN HEMTs gauge factor involved cantilevers, circular membranes, complex mass system and three point bending set up which, as stated before, result in inaccurate stress calibration. In this work a four point bending fixture and a technique to mitigate effect of charge traps are employed to characterize stress sensitivity of channel resistance in AlGaN/GaN HEMTs by accurately meas uring the gauge factor value. 3.2.1 Effect of Trapped Charge s AlGaN/GaN HEMTs have a hig h density of traps that are created during f abrication [32] or may also be generated during typical de vice operation [37] [49] These traps may be present as bulk traps in AlGaN barrier la yer or in GaN layer or as surface traps at the AlGaN/GaN interface or on the AlGaN surface and can result in significant changes in device electrical characteristics such as current collapse [56] gate lag [26] drain lag [26] increase in gate leakage [30] and shift in threshold voltage [37]

PAGE 40

40 AlGaN/GaN HEMTs utilized in this study exhibit charge trapping effect through changes in drain c urrent and threshold voltage shift. The instability during device measurement occurs due to changes in density of trapped charge. Applying a V G pulse of 10V for 1 minute, fills up the trap states with electrons and shifts the threshold voltage less negati ve. Under illumination from incandescent microscope light on a sample with no field plate the trapped charge s are photoionized and hence threshold voltage shifts to more negative value Illuminating microscopic light during measurement of channel resistanc e results in 15% reduction in channel resistance during elapsed period of the measurement as shown in Figure 3 4 This is approximately similar to the change in channel resistance measured during gauge factor estimation of 40,000 in [57] T his high value of measured gauge factor could result from changes in the t rapped charge density during measurement. In this work, effect of parasitic charge trapping is eliminated by employing a continuous optica l illumination of sub bandgap energies during the measurement and photoionizing the trapped charges. Thus a constant trapped charge density is maintained and an accurate estimation of stress sensitivity of channel resistance is achieved. 3.2.2 Experimental Procedure to Measure Gauge Factor A schematic of the experimental set up is represented in Figure 3 5 The sample is prepared by attaching the wafer piece to high carbon steel plate using thermally conductive epoxy following the steps described before. Th e sample is loaded into flexure based four point wafer bending set up to apply uniaxial compressive and tensile stress. The stress de pendence of channel resistance, R ch is measured at V GS = 1V and V DS =0.1V using Keithley 4200 semiconductor characterization system. The measured resistance, R meas does not directly give the channel resistance but also includes source

PAGE 41

41 and drain contact resistance s R S & R D and external parasitic resistances, R ext as follows (3 3) The source and drain contact resistances, measured using TLM structures, are found to be R S = R D subtracted from R meas Parasitic ext ernal resist ances are excluded by using a four point Kelvin method to measure the resistance. Herein two sets of contacts with source and drain pads are made with wire bonds. One pair of source and drain contacts is used to sense the voltage drop between source and dr ain terminals and other pair measures the dc current flowing through the pads. Parasitic trapping/detrapping effect is eliminated b y use of a mercury arc ultraviolet (U V) spotlight in combination with a 380nm band pass filter to photoionize trapped electro ns affect ing the resistance measurement. UV spotlight has peak wavelength of 377.7 nm (3.284 eV), measured using a spectrometer, and hence has significant portion of photon energies above the band gap of GaN (3.4eV). A b and pass filter (CML=380nm and FWHM= 10nm), filters out the wavelengths below 365nm (or photon energies above 3.4 eV) and prevents band to band photogeneration of electron hole pairs by UV spotlight if it were used alone for optical illumination. The band pass filter is mounted on a 4 inch th ick polystyrene heal shield and placed before the UV spotlight, as shown in Figure 3 5 to block ambient heat from the mercury arc spotlight. The measured resistance is allowed to stabilize before mechanical stress is applied on the sample to eliminated in fluence of charge trapping during stress sensitivity measurement Using the above set up, the resistance is measured for 1500 seconds

PAGE 42

42 and stabilized un til the change in measured resist ance is about 0.02% as shown in Figure 3 6 Mechanical s tress is then in crementally applied to the wafer sample in the direction longitu dinal to the channel direction and in increments of 60 MPa, each increment held for 100 seconds Stress is applied to a maximum of 360MPa and then gradually released in similar steps to z ero, as shown by the dotted line in Figure 3 7 The channel resistance, R ch is measured simultaneously as the stress is being applied indicated by solid lines in Figure 3 7 3.2.3 Results on Gauge Factor Measurement R ch increases with increase in applied compr essive stress while R ch decreases with increase in applied tensile stress, rendering a negative a value for the gauge factor. The n ormalized change in R ch at the maximum stress of 360 MPa is ~0.83% / 100 MPa The c hange in measured R ch is identical for both applied compressive and tensile stress. The measured resistance returns to its initial unstressed value (at 0 MPa) after t he uniaxial stress is released, indicating th at the observed change in channel resistance is entirely due to applied mechanical stres s, is reversible and is not i nfluenced by parasitic effects such as charge trapping/detrapping. The measured change in normalized R ch with stress is much smaller than the one in (001)/<110> Si n M OSFET which is ~3.2% / 100 MPa. Gauge factor is calculated as described before at each stress increment by averaging out the measured R ch at that stress increment and a plot of averaged normalized change in R ch versus mechanical stress is obtained. Uncertainty in R ch measurement is estimated using 99% confidence interval (three times the standard deviation) and plotted as error bars in Figure 3 8 Total least square analysis is used to obtain a linear fit of change in R ch versus stress, which inc orporates the uncertainty in

PAGE 43

43 the measurement. The g a uge factor is det ermined from the slope of the linear fit and is found to be, G.F.= 2. 8 0.4. The gauge factor estimated accurately in this work is small relative to the range of values published in literature ( 4 to 40,000). The large variability in published gauge facto r values is primarily due to the trapping/de trapping phenomena occurring simultaneously during measurement which can potential ly influence the measured stress sensitivity of chann el resistance. Also inefficient techniques to apply stress such as hanging weights, cantilevers and t hree point bending set up, result in non uniform external mechanical stress on the sample and hence can potentially render an inaccurate estimation of gauge factor value. The channel resistance R ch is dependent on 2DEG sheet car s e [58] estimated a g auge factor value for AlGaN/GaN HEMT by simulating the stress sensitivity of R ch due to the stress induced changes in n s and e Stress induced changes in n s may arise from the additional polarization induced by applied stress since AlGaN and GaN are piezo electric in nature, while the stress induced changes in e in GaN may arise due to changes in effective mass through band warping. A simulated gauge factor value of 7.95.2 was obtained [58] The best fit set of elastic and piezoelectric coefficients were obtained by comparing the simulated ch with the experimental results and are listed in Table 3 1 [58] 3.3 Effect o f Mechanical Stress on Gate Leakage Current in AlGaN/GaN HEMT Defect formation during typical device op eration in AlGaN/GaN HEMT s at high bias increases the gate leakage current and reduces the output power [59] This has been attributed to crystallographic degradation of the AlGaN barrier layer due to excessive strain generated at regions of high electric field in the devi ce via the inverse piezoelectric effect [30] Such defect formation creates a low resistance path for gate

PAGE 44

44 leakage current resulting in an irreversible increase in the leakage current. Since strain is inhere ntly present in AlGaN/GaN HEMTs through lattice mismatch and due to the inverse piezoelectric effect from applied biases in typical device operation, it is essential to investigate the effect of mechanical stress on the gate current density (J G ) in these devic es to improve reliability. In this study, the stress sensitivity of J G is measured and a dominant transport mechanism for gate leakage is discussed. 3.3.1 Experimental Procedure to Measure Gate Current Stress Sensitivity A AlGaN/GaN HEMT wafer piece is att ached to a heat treated high carbon steel plate using thermally conductive epoxy as described earlier in this chapter. The sample is loaded into the flexure based four point wafer bending set up to apply external mechanical stress as shown in Figure 3 2 (B ). The stress sensitivity of J G is measured at different reverse gate biases, V GS = 0,25V, 0.5V, 1V, 2V, 4V and at V DS =0V state to isolate the effect of vertical electric field induced by applied gate bias At each applied bias, J G is measured using a Keithley 4200 semiconductor analyzer and is first allowed to stabilize as shown in Figure 3 9 At lower reverse gate bias, J G take s a lo nger time to stabilize than at higher reverse gate bias. After J G is stabilized, longitudinal mechanical stress is app lied in increments of 60 MPa up to a maximum of 360 MPa. Each increment is held for approximately 100sec and then stress is released in same steps ; while J G is simultaneously measured Both compressive and tensile stresses are applied. 3.3.2 Results on Gat e Current Stress Sensitivity Measurement The g ate current density J G increases with applied tensile stress and decrease s with compressive stress and returns to its initial unstressed value when the stress is released for all applied reverse gate bias es T his indicate s an irreversible change in J G

PAGE 45

45 observed entirely due to the applied mechanical stress and not due to any other transient effect s such as charge trapping. Normalized change in J G G /J G is plotted (solid line ) along with applied mechanical str ess (dashed line) for V GS = 0.25V and V GS = 4V in Figure 3 1 0 (A, B) respectively. J G is averaged at each stress incremental step G /J G is plotted versus applied mechanical stress for eac h gate bias. Figure 3 1 1 (A, B) r epresents such a plot for V GS = 0.25V and V GS = 4V. 99% confidence index is used to estimate the uncertainty in J G measurement and is indicated as errors bars. A linear fit to the plot is obtained by weighted total least s quare analysis. The stress sensitivity of J G is given by the slope of this linear fit and is found to decrease from 1.70.3% /100 MPa at V GS = 0.25V to 0.6 0.1% /100 MPa at V GS = 4V. J G stress sensitivity is estimated and is plotted as a function of each a pplied reverse gate bias represented (as data points) in Figure 3 1 2 It can be inferred that stress sensitivity of J G decreases with increase in applied reverse gate bias. 3.3.3 Gate Current Measurement as a Function of Temperature In order to analyze dom inant gate leakage transport mechanisms in AlGaN/GaN HEMT DC characterization of gate current as a function of temperature is performed. J G is measured by sweeping V G from 5V to 0V at V DS =0V at different chuck temperatures, varying from 300K to 400K in s teps of 25K inside Lakeshore cryo stat Since PF emission is a thermal assisted phenomenon whose efficiency increases with increase in temperature, J G measurement was done at temperatures higher than room temperature. J G is observed to increase with increa se in chuck temperature as shown in Figure 3 13

PAGE 46

46 3.3.4 Discussion on Gate Leakage Mechanism The g ate leakage mechanism in AlGaN/GaN HEMT is widely discussed in literature and s everal transport mechanism s have been proposed, for e g. D irect or Fowler Nordhe im (F N) tunneling [60 62] thermionic emission [15 16] trap assisted tunneling [62], [64] mult i step trap assisted tunneli ng [65] Poole Frenkel (PF) emission [66 68] and tunneling t hrough a thin surface barrier [69] The stre ss sensitivity of J G estimated in the previous section can be understood by analyzing the dominant gate leakage mechanism. The t emperature depen dence of gate leakage ( Figure 3 13 ) helps to get an insight into probable dominant leakage mechanism by compar ing measurement with gate leakage transport models. Simulations indicated that F N tunneling and thermionic emission models underestimate experimentally measured J G while PF emission model closely matches with it for the gate bias |V G | < |V TH | [38] Hence Poole Frenkel (PF) emission model is analyzed in detail to understand gate leakage transport in AlGaN/GaN HEMTs for |V G | < |V TH | and stress sensitivity of gate leakage is explained Poole Frenkel Emission Poole Frenkel emission is the electric field assisted emission of trapped charges into conduction band. This mechanism is manifested when the Coulombic potential barrier for the trapped charge is lowered due to applied field and trapped charge can easily be emitted into the conduction band The general expression for PF emission current is given by [24, 25] (3 4 ) (3 5 )

PAGE 47

47 Here C is proportionality constant based on mobility and density of T is the trap activation energy, is the vacuum permittivity, is the relative permittivity of the AlGaN, E AlGaN is the electric field in AlGaN layer compensation parameter with 1 < r < 2, depen ding upon the density of deep acceptor traps in the host material The acceptor traps compensate the donor traps and effectively lower the Fermi level between donor trap level and conduction band, thus reducing free carrier concentration contributing in PF emission. For limiting case s assumes a value of 1 if acceptor density is high enough to compensate approximately all donor traps and lowering the Fermi level t o donor trap level there are no acceptor traps and all donor t raps participate in PF emission [14 15] Traps are distributed both in energy and space in AlGaN Traps participating in PF emission are assumed to reside close to the AlGaN surface near the metal Fermi level and hence are filled otherwise thermal emission of carriers fr om metal into traps would have to occur Thus leakage current is limited by the emission into conduction band T can be estimated by taki ng natural log of E quation 3 4 which gives the linear form of PF model as follows (3 6 ) (3 7 ) (3 8 ) Temperature dependence of gate leakage ( Figure 3 13 ) is re plotted using the above analysis as the natural logarithm of J G divided by electric field versus square root of

PAGE 48

48 electric field for |V G | < |V TH | and a linear fit i s obtained as shown in Figure 3 14 The linearity of this plot signifies that PF emission is the dominant mechanism fo r gate current transport for the specified voltage range. Experimentally determined 1D E AlGaN values from [74] AlGaN = 5.1 [75] are used for the analysis From the slope of linear fit m(T) vs 1/T plot, the value of compensation using E qua tion 3 7 is evaluated to be 1.07 0.02 which signifies high acceptor compensation The y intercept b(T) of linear fit vs 1/T plot gives trap acti vation energy using E quation 3 8 T = 0.40 0.05 eV. This is effective activation energy of traps participating in PF emissio n. T the uncertainties in slope and y intercept of linear fit obtained. Since traps are assumed to exist near metal Fermi level and the Ni/AlGaN Schottky barrier is ~1.4eV, a trap level of ~0.4eV suggests th at emission of trapped charges does not occur into the conduction band of AlGaN. Instead PF emission occurs from trap state into a dislocation band which exist 0.4e V above the trap state and provides a conductive path for the leakage current from metal to the channel as shown in Figure 3 1 5 AlGaN/GaN HEMT have a high density of threading d islocations which could possibly provide conductive dislocation bands for PF emission. Stress sensitivity of J G can be understood by the following analysis of PF emission dominant gate current (3 9 ) (3 10 )

PAGE 49

49 H ere J PF and J PF (0) is the current under no stress. T related changes in trap level and compensation parameter respectively. Increase i n J G w ith applied tensile stress as shown in Figure 3 11 can be explained by resultant de crease in T and trap level moves closer to conductive dislocation band This potentially de creases acceptor compensation and thus increasing Similarl y with applied compressive T in creases de creases resulting in decrease in J G By s olving E quation 3 1 0 simultaneously at biases V GS = 0.5V and 1V, changes in T to be T = r = 0.00170.00 01/100 MPa. As indicated, tensile (compressive) stress decreases (increases) T The bias points selected are in the range where gate leakage is PF emission dominant. PF PF (0) using E quation 3 1 0 offers a reaso nable fit to the experimentally measured J G stress s ensitivity as shown in Figure 3 1 2 (solid line) Small observed discrepancy in the fit at low V G can be due to the presence of reverse current prominent at lower gate biases. There is relatively high E fi eld in AlGaN layer (0.75 MV/cm) at zero gate bias due to the presence of polarization which can potentially produce forward gate current at zero gate bias. Yan et. al. [68] proposed existence of a reverse current J R J R exp( 3/2 / E AlGaN which is of equal magnitude to balance the gate current at equilibrium or zero gate bias. This reverse gate biases dec reases with higher gate biases. For |V G | > |V TH |, E AlGaN a t the gate center begins to saturate while continues to increase at the edges and thickness of the AlGaN potential barrier reduces with increase in revere gate bias [74] As a result, Fowler Nordheim (F N) tunnelin g wi ll

PAGE 50

50 occur at the gate edges while PF emission will dominate in the middle of gate. F N tunneling current becomes significant for E AlGaN > 3MV/cm [76] .The stress dependence of FN tu nneling is prominently due to out of plane effective mass, which is relatively independent of stress [77] This potentially explains the decrease in stress sensitivity of J G with increasing rever se gate bias. At extremely high E fields at which device degradation occurs the gate leakage will likely be dominated by FN tunneling 3.4 Summary This chapter investigates the effect of longitudinal uniaxial externally applied mechanical stress on electr ical parameters, such as channel resistance and gate leakage current, of AlGaN/GaN HEMT by eliminating the charge trapping effects and utilizing a four point wafer bending set up. Wires are bonded to the device pads to simultaneously perform electrical mea surements while mechanical stress is being applied incrementally. An accurate estimation of gauge factor for AlGaN/GaN is found to be G.F. = 2.8 0.4 by illuminating the device with a filtered UV light (phonon energies within GaN bandgap). Negative value of G.F. indicates that tensile (compressive) stress decreases (increases) the channel resistance. A small value of G.F. can be attributed to small changes in carrier concentration and carrier mobility due to applied mech anical stress. Mechanical stress se nsitivity of ga te current decreases with increasing reverse gate bias. Tensile (compressive) stress increases (decreases) gate current for all applied gate biases ( V GS = 0.25V to 4 V). PF emission dominates gate leakage transport for |V G |<|V TH | and stres s sensitivity of gate current is attributed to variation in trap activation energy and compensation factor under the applied mechanical stress.

PAGE 51

51 For |V G |>|V TH |, E field at the gate edges increases due to 2D effects and FN tunneling is l ikely to dominate gat e leakage.

PAGE 52

52 Table 3 1 Best fit material coefficient values for AlN and GaN [58] Material Coefficients GaN Piezoelectric: [78] e 14 = 0.375 C/m 2 e 15 = e 31 = (1/ 3)e 14 = 0.217 C/m 2 e 33 = 2e 31 = 0.433 C/m 2 Elastic: [79] c 1 1 = 365 GPa c 1 2 = 135 GPa c 1 3 = 114 GPa c 33 = 381 GPa c 44 = 109 GPa c 66 = 115 GPa AlN Piezoelectric: [80] e 15 = e 24 = 0.4 8 C/m 2 e 31 = e 3 2 = 0.58 C/m 2 e 33 = 1.55 Cm 2 Elastic: [81] c 1 1 = 410 GPa c 1 2 = 140 GPa c 1 3 = 100 GPa c 33 = 390 GPa c 44 = 120 GPa

PAGE 53

53 Figure 3 1 Schematic of four point mechanical wafer bending set up As indicated, top surface is under tensile stress while bottom surface is under compres sive stress. A neutral axis exists at the center of the bent sample which is a zero mechanical s tress axis Figure 3 2 Four point mechanical wafer bending fixture. A) Top plate and bottom plate based bending set up B) Flexure based bending set up

PAGE 54

54 Figure 3 3 Calibration of stress applied by four point flexure based bending set up using a strain gauge mounted on top of wafer, under stress. As shown stress applied by bending set up is elasti c. Figure 3 4 A 15% reduction in channel resistance i s observed during 1200 seconds of measurement under optical illumination with incandescent microscope light

PAGE 55

55 Figure 3 5 Experimental setup to measure stress sensitivity of channel resistance (R ch ), i.e. gauge factor, while illuminating with sub bandg ap photon energies using a filtered UV light to eliminate influence of charge trapping without photogenerating band to band electron hole pair s. Figure 3 6 Measured channel resistance ( R ch ) stabilized to less than 0.02% variation over 1500 seconds of m easurement under illumination with filtered UV light

PAGE 56

56 Figure 3 7 Change in normalized R ch measured while incrementally applying longitudinal uniaxial stress. [ Reprinted with permission, from A.D. Koehler, et al IEEE Elec. Dev. Lett., Vol. 31, pp. 665 667, Figure 2, July 2010] Figure 3 8 ch measurements averaged at eac h increment of applied stress. Error bars represent 99% confidence interval of uncertainty in the measurement

PAGE 57

57 Figure 3 9 Measured J G stabilized before appl ying external mechanical stress

PAGE 58

58 Figure 3 1 0 Normalized change in J G for incrementally increasing and decreasing longitudinal uniaxial mechanical stress for gate biases A) V G = 0.25V and B) V G = 4 V

PAGE 59

59 Figure 3 1 1 Normalized change in J G averaged a t each increment of applied stress, for gate biases A) V G = 0.25V and B) V G = 4 V. Error bars represent 99% confidence interval of uncertainty in the measurement

PAGE 60

60 Figure 3 12 Experimental and simulated mechanical s tress sensitivity of J G (per 100 M Pa of stress), including T and r variation with mechanical stress, as a function of reverse gate bias.

PAGE 61

61 Figure 3 13 Measured J G vs V G for different chuck temperatures from T C = 300K to 4 00K Figure 3 14 PF plot show ing linear fit to the measured J G vs V G data for T C = 300 K to 400 K

PAGE 62

62 Figure 3 1 5 Schematic of PF emission through a dislocation band and conceptual reverse current in AlGaN/GaN HEMTs

PAGE 63

63 CHAPTER 4 TIME DEPENDENT ELECTRICAL DEGRADATION OF ALGAN/GAN HEMT 4.1 Background 4.1.1 Liter ature Review on Device Degradation by Electrical Step Stress Several device electrical parameters can be measured as signature s for device degradation. Electrical degradation in AlGaN/GaN HEMT due to DC stressing has been characterized by changes measured in different device parameters, such as increase in source and drain resistance, decrease in saturation drain current, decrease in G m V T shift and sub threshold chang e [37] [56] [81 82] An incremental technique to study degradation in these devices is termed electrical step stress test whereby voltage on the device is incremented until a critica l voltage is determined [30] J. Joh et al performed a step stress experiment on GaN HEMT with L G from V GS = 10V to 50V while keeping V DS =0V. At this condition no channel current flows in the device and at a critical voltage, degradation in several device characteristics start s sharply and increases as the stress p roceeds, as shown in Figure 4 1 [30] A RF stress experiment in these devices can also cause degradation in I D MAX and output power. A GaN HEMT biased under class AB condition at V DS =30V or 40V, was stressed at input power for which gain compression was 3dB. The degradation of P out a fter initial burn in is shown in Figure 4 2 [84] In another study [85] a n AlGaN/GaN HEMT sample was electrically stressed by applying current s tress pulses on drain terminal while under floating g ate and grounded gate condition. C atastrophic device degradation is typically accompanied by significant increase in gat e current and decrease in R S and R D which is indicative of degradation of both Schottky and ohmic contacts. This has been

PAGE 64

64 hypothesized to occur due to critical temperature increase caused by power dissipation under high current levels used for stressing. 4 .1.2 Models for AlGaN/GaN HEMT Device Degradation Electrical degradation in AlGaN/GaN HEMT due to DC stressing has been characterized by changes measured in different device parameters, such as increase in source and drain resistance, decrease in saturatio n drain current, decrease in G m V T shift, sub threshold change and increase in gate leakage [30] [37] [86] T wo theories, expla ining the cause of electrical degradation in GaN HEMT, have been published in the literature. According to one popular theory, such degradation is widely accepted to be electric field driven and is attributed to the stress result ing from the inverse piezoe lectric effect due to the applied field [30], [38], [88 90] According to this theory, electric field due to an applied voltage generates a tensile stress in the AlGaN layer which along with intrinsic stress in the device due to lattice mismatch, inc reases the elastic energy of AlGaN layer. When total elastic energy of AlGaN layer reaches a critical value, it undergoes strain relaxation through crystallographic defect formation under the gate edge where peak electric field occurs [37] A nother theory associates electrical degradation of GaN HEMTs to thermal or chemical reaction occurring at the metal semiconductor interface resulting in diffusion of gate metal into the semiconductor layer [87], [91] or consumption of interfacial oxide present at the metal semiconductor interface [92] Hence the interfacial chemical reaction occurring in this case is a field driven mechanism. Gate sinking or gate metal in diffusion into the semiconductor bulk is a well known effect in III V HEMTs causing degradation [39], [40] and resultant variation in the Schot tky interface characteristics [93] The analysis of device degradation upon electrical stressing is severely complicated due to trappin g phenom ena occurring during

PAGE 65

65 device operation. One of the widely accepted model s i s [19] According to this model, surface states between the drain and gate may capture electrons during high drain field stressing. These trapped charges act as a virtual gate causing localized depletion of 2DEG and resulting in an increase in the drain resistance and decrease in the drain current. 4.1.3 Role of Threading Dislocations in AlGaN/GaN HEMT Dislocations are chains of discontinuities in a semiconductor material crystal, which provide preferential path s for either material diffusion or charge migration [32] and hence can influence the device electrical properties. Epitaxially grown GaN films contain a high density of threading dislocations, about 10 5 10 6 cm 2 [32] Types of threading dislocations associated with hexagonal shaped crystals such as GaN a re edge type, screw type and mixed type. The threading dislocations are normally oriented along the c axis of the material. They are believed to affect device performance in GaN HEMTs by reducing carrier mobility [94], [95] and by increasing leakage current [96], [97] A dislo cation core can be filled up with excess Ga atoms or impurities, thus modifying the core structure and creating new energy states in the material bandgap [98] The dangling bonds along an edge dislocation line can act as electron acceptor sites which can be filled by electrons from the 2DEG. These negatively charged sites induce electron scattering and reduce carrier mobility in the channel. Resultantly, the occurrence of trap sites reduces the density of free charge carriers in the channel. The gate l eakage current in GaN HEMTs is directed vertically across the AlGaN layer and is enhanced by the presence of dislocations. Dislocations in AlGaN are presumed to be similar to or follow most of those from GaN layers. The strain in the AlGaN layer generated due to lattice mismatch are believed to bend threading dislocations in th e

PAGE 66

66 layer slightly along the growth direction but do not obstruct the propagation path to the device surface [98] Several studies have reported electrical conduction through such threading dislocations [99 101] proving them to be highly charged. A study by Y. Kamimura [102] shows a large amount of conduction through threading dislocations in a GaN single crystal, grown using hydride vapor phase epitax y followed by the epitaxial lateral overgrowth technique. A sample of this cry stal was deformed by 5% under a uniaxial compression. The density of dislocations or spots increased to 10 9 10 10 cm 2 The electrical conduction along these dislocations studie d using scanning spread ing resistance microscopy (SSRM) indicate current values at these spots 1000 to 100 times higher than the s urrounding area as shown in Figure 4 3 [102] Hence due to such a high current flow these dislocations may act as hot spots leading to localized heating. Hsu et. al. [103] have also reported similar high current density throu gh screw dislocations in GaN, several orders of magnitude higher in samples grown under Ga rich condition s than under N rich conditions. Device degradation observed in GaN HEMTs under electrical stressing can be attributed to an increase in threading dislo cation density when the electric field in the device reaches a critical value or the high current propagating through these dislocations induc es an electro chemical reaction at the gate semiconductor interface. Increased density of dislocations provides ad ditional propagation paths for reverse bias gate current, hence increasing the gate leakage. Occurrence of additional trapping sites due to increase in threading dislocations induces additional electron scattering and reduction of free charge concentration in the channel, hence causing decrease in drain

PAGE 67

67 current and increase in R on A close link between the surface pits acting as leakage paths and physical device degradation near the gate edge was recently observed by Bajo et. al. [104] in a GaN HEMT sample with L G V DS = 40V and V GS = 15V. 4.2 Electrical Step Stress with Varying Time Step As discussed in Section 4.1, the onset of electrical degradation in AlGaN/GaN HEMTs has been studied by conducting an electrical step stress ex periment [30] [59] [48] ; where the electrical bias is stepped incrementally while simultaneously monitoring an electrical parameter. The voltage at which the device undergoes a sudden increase in the degradation of the electrical parameter is termed the critical voltage (V crit ) However, it has not b een proven that the so called critical voltage is actually a threshold, independent of the duration of the voltage step. A modified electrical step stress experiment is conducted in this work where a set of degradation tests are performed with varying time step s (or length of time for which each voltage step is applied) and the critical voltage is determined for each test The main objective is to investigate the effect of varying the time step s on the critical voltage at w hich device degradation occurs If the so called critical voltage varies as a function of the time step, then the picture of degradation cannot be as simple as a single critical voltage. Understanding the time dependence may provide insight into the complex degradation mechanism that d etermine the reliability of AlGaN/GaN HEMTs. 4.2.1 Experimental Procedure A finger Al 0.2 8 Ga 0.7 2 N/GaN HEMT grown on SiC substrate, with gate length L G X passivation is used for this experiment. Electrical step stress is performed by stepping the reverse gate bias in steps of

PAGE 68

68 1V/step, beginning with V GS = 10V at V DS =0V and at room temperature. The l ength of time for each voltage step is varied as t=10 sec, 100sec, 500sec and 1000sec for four different tests performed separately on f ou r identical devices for comparison. Devices are assumed to be id entical if they demonstrate similar electrical characteristics before the electrical stress, such as saturation drain current, threshold voltage and gate leakage current. The g ate current I G is simultaneously monitored while the voltage step stress is app lied using a Keithley 4200 semiconductor characterization system with the I G to prevent internal heating of the device A user defined module is coded , to apply any desired voltage s tep stress 4.2.2 Results and Discussion All four step stress tests depict I G trapping transients at each voltage step prior to breakdown When a critical voltage is reached, a sudden increase in I G is observed indicating onset of irreversible degradation. This is followed by multiple smaller jumps in I G and I G continues to increase beyond the critical voltage un ti l it hits the set current compliance limit This indicates that once initiated, device degradation continues to proceed un til compliance is rea ched which is preset on the Keithley 4200 to protect the instrument Figure 4 4 depicts I G degradation due to electrical step stress with 100sec/step time step. It is observed that the critical voltage at which I G begins to degrade depends on the time step (or length of time for each applied voltage step) and it decreases with increase in time steps as shown in Figure 4 5 Devices that are electrically step stress ed with longer time steps begin to degrade sooner (at smaller critical voltage) than the device s that are step st ressed with smaller time steps. Figure 4 6 shows a plot indicating linear dependence between critical voltage or breakdown

PAGE 69

69 voltage (from each electrical step stress test) versus square root of corresponding time step. This implies that a diffusion phenomenon may be playing a role in determining the critical voltage for a device because on shows a as described in detail in Chapter 5 There exists a thin oxide layer at the gate metal/semiconductor Schottky interface. It can be hypothesized that this oxide layer undergo es a diffusion due to excessive strain under the gat e from the high electrical field ( from the applied voltage step ) or due to the presence of threading dislocations that potentially act as hot spots. A longer time step (at each voltage step ) allows more time for slow diffusion of the oxide layer at each vo ltage step, until the oxide layer weakens, loses its integrity and ruptures potentially around the threading dislocations This is observed as a sharp increase in I G and marked by a critical vol t age Hence the electrical step stress test with longer time step gives a smaller critical voltage for the device. The inset of Figure 4 4 shows the time evolution of I G during the critical voltage step for electrical step stress test with 100 sec/step time step It is observed that the gate current undergoes sign ificant enhancement not immediately but after some elapsed time of the application of the critical voltage indicating that degradation in AlGaN/GaN HEMTS is time dependent. The diffusion of oxide layer is potentially a time dependent phenomenon. 4.3 Summa ry A detailed literature review of existing models for AlGaN/GaN HEMT degradation is presented in this chapter. AlGaN/GaN HEMTs are known to have a high density of threading dislocations, generated during epitaxial growth. The r ole of threading dislocation s is discussed in this chapter They are believed to affect device performance

PAGE 70

70 in GaN HEMTs by reducing carrier mobility and by provid ing leakage paths for high gate current which may act as hot spots leading to localized degradation within the device In addition a modified electrical step stress experiment is presented to investigate the effect of time step on the at which the gate current increases abruptly It is inferred that the critical voltage decreases with increase in length of time steps indicating a time dependent electrical degradation in the AlGaN/GaN HEMT A linear dependence of the critical voltage on the square root of time steps may imply that a diffusion phenomenon is playing a role in determining the critical voltage f or the device

PAGE 71

71 Figure 4 1 Change s in normalized I DMAX R D R S I Gstress and I GOFF versus voltage st ep at V DS =0V state in a step stre ss experiment [30] Figure 4 2 O utput power degradation during RF stress of AlGaN/ GaN HEMT with field plate d gate [84]

PAGE 72

72 Figure 4 3 SSRM images of (A) unde for m ed and (B) 5% deformed samples respectively [105] Figure 4 4 I G degradation of AlGaN/GaN HEMT at V DS =0V during electrical step stress test with 100sec/step time step. Inset shows I G progression indicating time dependent degradation of I G

PAGE 73

73 Figure 4 5 D of I G breakdown) on time step (length of time for each applied voltage s tep ) in electrical step stress test Figure 4 6 square root of the time step s ind icates potential role of diffusion phenomenon in determining critical voltage.

PAGE 74

74 CHAPTER 5 STRAIN RELAXATION IN AlGa N/G a N HEMT 5.1 Review of Strain Relaxation in Heterostructures Heterostructures, such as Si/SiGe superlattices, or compositionally modulated films under excessive strain undergo relaxation to lower the system energy either though attaining homogeneity of the entire structure [106] or generation of misfit dislocations at the interface [107] Both of these mechanisms tend to compete and the one with more favorable kine tic conditions takes place in the heterostructures. Strained structures relax to lower energy or less strained state when an external excitation is applied such as temperature, electrical bias, and mechanical stress. The strain in such structures is genera ted due to varying factors, such as lattice mismatch between epitaxially grown layers [108] compositional gradient [109] thermal mismatch during annealing or self heating [110] or in some cases inverse piezoelectric effect such as in GaN HEMTs [30] 5.2 Strain Relaxation through Trap Generation In as grown structures, stress arises mainly due to lattice mismatch or compositional gradient (Si Si x Ge 1 x ) and is dependent on growth conditions, composition and layer thickness and leads to the formation of misfit dislocations at the hetero e pitaxial interfaces [111 114] In cases where stress is accommodated by an array of dislocations near the edge in the as grown state, there is tendency to relax by generating additional misfit dislocations which depends on the initial dislocation density in the structure. D i sloc ations present at the interface play a catalytic role in the generation of additional dislocations. It has been proposed by H agen and Strunk [115] and Vdovin el al. [116] that when an orthogonal network of dislocations is present, the

PAGE 75

75 intersection between the d islocations with identical Burger vectors can be a source for dislocation multiplication. Misfit dislocations when present in the active areas of a device can be electrically active and may act as generation recombination centers and facilitate dislocation multiplication to achieve strain relaxation. The kinetics of trap generation can be expressed through trap generation and recombination rate equations and hence gives the transient nature of strain relaxation. The strain relaxation through dislocation mul tiplication in III V material system can occur through c ondensation of point defects, dislocation half loop glide and e dge dislocation tilt [117] 5.2.1 Rate of Dislocation multiplication As mentioned before, a strained layer structure may be subjected to loading conditions of biaxial shear strain through internal lattice misfit, thermal strain, external mechanical strain or strain due to inverse piezoelectric nature. Strain relaxation occurs through introductio n of a network of additional misfit dislocations at the overlay interface, if the kinetic barrier of such an event is lowered [118 121] The energy barrier for strain relaxation can be low ered by external excitation such as applied temperature, electrical bias or externally applied mechanical stress. Such strain relief is proportional to the dislocation density and i s analytically given via the following kinetic model [11 8] (5 1) the degree of strain relaxation, m is the density of misfit dislocation s and b is the magnitude of the Burger vector The strain relaxation via dislocation m ultiplication can be understood by the phenomenological model developed by Alexander and Haasen [119] for dislocation dynamics. According to this model, the

PAGE 76

76 effective stress in a strained layer driving dislocation motion induces a dislocation glide the local stress as [118] (5 2 ) Here E A is the activation energy of the dislocation glide velocity eff is the effective stress in the epitaxial layer. In AlGaN/GaN HEMTs, additional stress in the layer can result from applied electrical bias due to the inverse piezoelectric effect, thermal mismatch during device operation or externally applied mechanical stress. The strain relaxation occurs through multiplication of moving dislocations that occur s in proportion to their moving length and the distance travelled. Under thermally activated glide velocity and local strain field, the rate of dislocation multiplication is given as [119] (5 3) 5.2.2 Di slocation Motion and Dislocation Glide Velocity in Heterostructure s The activation energy E A of the dislocation glide velocity is expected to be dependent on the bond energy because the dislocation glide motion occurs via a bond breaking process [122] Hence E A depends on the lattice energy or the cumulative bond energy of the crystal. Figure 5 1 [122] represents E A for differe nt types of dislocations in III V binary compounds which dislocations have the largest mobility and hence are expected to dominate the degradation rate [122] GaN based materials have a high density of threading dislocations parallel to the c axis of the GaN wur t zite crystal. These dominant threading dislocations have parts on the (0001) basal plane and are highly mobile and hence may dominate the degradation of AlGaN/GaN HEMT under the influence of enhanced stress in the epitaxial layer. As

PAGE 77

77 mentioned before, the enhanced stress in AlGaN/GaN HEMT may occur due to applied bias and thermal mismatch during device operation or externall y applied mechanical stress. E A for GaN semiconductor is found to be ~ 2.1eV from the plot of E A vs lattice energy in Figure 5 1. 5.3 Strain Relaxation through Diffusion 5.3.1 The one s of diffusion and can be formulated in terms of a rate of increase in concentration (of diffusant) with time as follows [123] (5 4) Here D is the diffusion coefficient or diffusivity which is temperature dependent and is assumed to be independent of position. N is the concentration of the impurity specie s In semiconductors, the solution to the partial differential equation in Eq uation 5 4 is obtained using two specific boundary conditions. The first boundary condition is a constant source that corresponds to a constant concentration of impurity (o r diffusant) at the surface of the semiconductor epilayer. For a semi infinite wafer with N 0 impurity concentration at the surface, the solution to Eq uation 5 4 is given by a complementary error function as [124] (5 5) The diffusion front (x) continues to proceed into the semiconductor bulk with time (t) while the surface concentration is assumed to remain constant and supplies a continual flow of diffusant at oms into the bulk. Another boundary condition relates to a limited

PAGE 78

78 source which implies that the impurity concentration at the surface (supplying diffusant atoms) is limited. It can be mathematically modeled as an impulse function and is referred to as a t The solution to the basic diffusion equation in Eq uation 5 4 in this case is given by a Gaussian distribution as [124] (5 6) As the diffusion front (x) moves with time (t) the surface concentration decreases while the total dose of impurity diffusing into the semiconductor bulk remains constant. 5.3.2 Self D iffusion and Inter D iffusion in Heterostructures When an energy barrier to trap gen eration exists, strain relaxation may occur through strain enhanced diffusion in the structures. As in Si Ge superlattices [125] such tetragonal strain is relieved by causing Ge to out diffuse into the substrate reducing strain and resulting in homogenization at the interface Strain enhanced diffusion is a well studied phenom enon in metallic systems. Cahn [126] proposed that an increase in elastic energy in such systems due to excess strain is seen to act as the driving force for the diffus ion to re lieve strain. Spaepen [127] discussed that such strain enhanced diffusivity is proportion non linear and transient nature of the strain enhanced diffusion phenomenon arises from the dependence of diffusion on the concentration of diffusing material that changes with time as the diffu sion proceeds. Also, d iffusion is influenced by the density of misfit dislocations that may be present in the structures or may be generated during strain relaxation.

PAGE 79

79 5.3.3 Effect of T raps on Diffusion Rate Cowern et.al [128] quantitatively studied the defect mediated diffusion in Si/SiGe heterostructures via vaca ncy and interstitials injection In a strained layer, the diffusivity generalizes to [128] (5 7) H ere D (i) and D (v) are diffusion coefficients related to interstitial and v acancy traps respectively under strain free conditions and E (i) and E (v) are the corresponding activation energies of diffusion for interstitial and vacancy mediated diffusion respectively. It was observed that in a Si Ge superlattice there is an incr ease in diffusivity of As as the lattice is compressed [129] and a corresponding dec rease in diffusivity of B [130] Int erstitial defects cause an outward relaxation of the surrounding lattice, similar to in a Si lattice around migrating B interstitials, resulting in an increase in defect formation energy in a compressively strained lattice. This explains the decrease in B diffusion as the Si lattice is compressed. On th e other hand substantial inward relaxation due to bulk vacancy d efects causes a decrease in defect formation energy and corresponding change in activation energy of defect mediated diffusion. Hence there is a n increase in diffusivity of As in com pressively strained Si lattice. Strain is inherent in the AlGaN/GaN HEMT due to lattice mismatch. An excessive strain can also be generated by the inverse piezoelectric effect during typical device operation which may lower the energy barrier for a diffusion process to occur in these devices. There exists an oxide layer at the metal/semiconductor junction in AlGaN/GaN HEMT. Activation energy o f oxygen diffusion into the GaN epitaxial layer is reported to be 0.23 eV [131] while activation energy of dislocation motion in Ga N is estimated to be

PAGE 80

80 2.1 eV [122] Hence an oxygen diffusion process is more likely to occur as a result of strain relaxation in AlGaN/GaN HEMT and may affect device reliabil ity A high density of threading dislocations can provide a low resistance path for the oxygen mediated diffusion process and an elevated temperature can accelerate the diffusion process by providing the required energy for the diffusion kinetics. 5.4 Summ ary This chapter discusses the strain relaxation phenomenon widely observed in heterostructures, such as Si/SiGe and III V heterojunctions. Heterostructures undergo relaxation to release excessive strain and lower system energy, either by trap generation o r diffusio n. These phenomena are aided by high density of already existing point defects and dislocations, temperature as well as excess ive strain. T hese mechanisms tend to compete and the one with more favorable kinetic conditions dominate s This chapte r reviews the m odels for strain relaxation through trap generation and strain relaxation through diffusion published in literature O ne dimensional diffusion semiconductors are also briefly described in this chapter Strain relaxation due to a d iffusion process is more likely to occur in AlGaN/GaN HEMT because of the lower activation energy of oxygen diffusion than the activation ener gy of dislocation motion in GaN and hence can potentially affect the device reliability.

PAGE 81

81 Figure 5 1 P lot of activation energy of dislocation motion (E A ) versus lattice energy in III V materials [122]

PAGE 82

82 CHAPTER 6 EFFECT OF ELECTRICAL BIAS, TEMPERATURE, MEC HANICAL STRESS AND AMBIENT ON AlGaN/Ga N HEMT TIME DEPENDENT DEGRADATION 6.1 Introduction and Research Objective To realize the potential of AlGaN/GaN HEMTS including its built in polari zation and wide band gap for high pow er and high frequency operation, the reliability of AlGaN/GaN HEMTS needs to be improve d. Irreversible physical damages have been observed in these devices during typical operation causing permanent degradation of elect rical parameters and compromising device reliability. Electrical degradation in AlGaN/GaN HEMTs has been studied using voltage step stress measurements Under such measurements, a critical voltage is observed at which a sharp increase in gate current oc curs. Some studies, as discussed in previous chapters associate such degradation to high electric fields generated at the critical voltage that creates a mechanical strain via inverse piezoelectric effect in addition to the inherent strain from lattice mi smatch and causes crystallographic defect formatio n [30] [37] Other grou ps explain this type of electrical degradation by an electric field driven chemical reaction or thermal reaction occurring at the metal semiconductor interface and potentially causing diffusion of gate metal or other chemical specie s into the semiconductor epi taxial layer [87], [91] Thermal reactions based study of these devices has mostly focused on the thermal stability of the Schottky and Ohmic contacts during the annealing process [43 ], [44] However temperature has also been shown to accelerate electrical degradation of AlGaN/GaN HEMTs by trap generation via thermal mismatch and hot electron effects [33] In addition, F.Gao et. al. [132], [133] have observed that variation s in the ambient condition significantly influences AlGaN/GaN HEMT degradation. The effect of ambient has been postulated as follows: the p res ence of oxygen causes the oxidation of the

PAGE 83

83 device surface while moisture can form a thin water laye r on hydrophilic surfaces (such as AlGaN) and both of these processes are hypothesized to create new trapping centers at the interface [132], [133] It can be seen that there exists a wide disagreement on a variety of f undamental degradation mechanism models. Hence there is a need for a more comprehensive study on AlGaN/GaN HEMT degradation occurring during device operation. The main objective of th e work in this chapter is to conduct a systematic study of time dependent degradation observed in AlGaN/GaN HEMTs by varying electrical bias, temperature, mechanical stress and ambient conditions. As stated in previous chapters, electrical degradation can occur in AlGaN/GaN HEMTs even at voltages below the critical voltage but it takes a longer time for the device to degrade at lower voltages than at higher voltages. This time dependence in device failure is key in understanding the degradation mechanisms The s train relaxation phenomenon, discussed in detail in chapter 5, is t ypical to heterostructures Ex cessive strain generated in epitaxial layers due to lattic e mismatch, thermal mismatch, compositional variation or by inverse piezoelectricity causes the heterostructure to undergo relaxation through trap generation or diffus ion process es p otentially degrading it s electrical perform an ce The rate of t rap generation and diffusion depend on electrical bias, temperature, mechanical stress, and ambient and can give insight into the time dependent degradation observed in AlGaN/GaN HEMTs. These processes are kinetically triggered by application of electrical bias and temperature S ince strain is inher ent to these processes, externally applied mechanical stress will also play an effective role in investigating these processes Hence t his work endeavors to gain an understanding of fundamental physics

PAGE 84

84 behind degradation of AlGaN/GaN HEMTs through detailed experimentation using applied bias, temperatu re, mechanical stress and ambient and theoretical analysis. 6.1.1 AlGaN/GaN HEMT Sample Description The AlGaN/GaN HEMT structure used in this study is sho w n in Figure 6 1. The heterostructure consists of a 15nm thick Al 0.28 Ga 0.72 m thick Fe doped GaN buffer layer grown epitaxially on a substrate with ICP (inductively coupled plasma) etched 100nm mesas Two types of samples are used, one grown on semi insulating 6H SiC substrate and the other grown on Si substrate There exists an unintentionally doped 0.3nm thick GaN cap layer on top of the AlGaN laye r. Also, a n AlN nucleation layer is present between the GaN layer and substrate to improve lattice m ismatch Source and drain o hmic contacts are made of T i/Al/Ni/Au stack while Ni/Au f orm the Schottky gate contact s. The AlGaN/GaN HEMT test wafer has a reticle layout, and each reticle is comprised of four sub reticles, A to D. Each sub reticle is populated long centered gate HEMT structure, one sub micron off centered gate HEMT, one sub micron centered gate HEMT, one ungated HEMT site, iso lation structures, van der Pauw test structures and Transfer Length Module (TLM) structures. Among the subreticles, the sub micron gate lengths vary as (A) 100nm, (B) 125nm, (C) 140nm and (D) 170nm. All gated HEMT structures have finger gates. All structures have a wide band gap SiN x passivation layer but no field plate. The s centered 6.1.2 Sample Requirement for Experimental Study The AlGaN/GaN wafer sample was diced into smaller pieces for cost effective utilization of the sample for the experimental study. The effect of mechanical stress was

PAGE 85

85 investigated by attaching a small wafer piece on a high carbon steel plate with a thermally conductive epoxy u sing the procedure described in chapt er 3. The prepared sample is then inserted into a four point wafer bending jig to apply mechanical stress. Since the wafer sample is ~ 4 0 0 = 460 GPa [134] ) is a much stiffer material <110>(100) Si = 169 GPa [135] ), it is difficult to transfer mechanical stress to the samples fabricated on SiC substrate using a st eel plate in a wafer bending jig Hence only t he samples fabricated on Si substrate are chosen to study the effect of mechanical stress. All othe r experimental studies on time dependent degradation can be performed on samples fabricated on both Si and SiC substrate. For all time dependent degradation experiments, AlGaN/GaN HEMTs micron devices have 2D edge effects in the E field profiles under the gate and other short channel effects which can potentially complicate the study of the fundamental physics behind time dependent device degradation However in some cases where no measurable effect of external conditions (such as ambient) on device degradation is observed in sub micron devices are utilized to gain further insight into possible failure mechanism s. In order to compare the effect of varia tions in an external parameter (electrical bias, temperature, mechanical stress or ambient ) on time dependent degradation of AlGaN/GaN HEMT, identical devices need to be selected for each experimental study. This is accomplished by doing pre degradation DC characterization of the devices. First DC characterization is conducted by measur ing I G while sweeping V GS from 5V to 1V at V DS =0V and by measuring I D while sweeping V GS from 5V to 0V at V DS =0.1V The

PAGE 86

86 transconductance (G m ), th reshold voltage (V TH ), saturation drain current (I DSAT ) from I D V GS measurement and forward and reverse gate current from I G V GS m easurement are compared and devices with similar electrical characteristics are chosen for the degradation study. It is assume d that devices with similar G m V TH I DSAT and g ate leakage current will exhibit similar degradation under identical exte rnal conditions or excitations. 6.2 Effect of Electrical Bias on AlGaN/GaN HEMT Degradation 6.2.1 Experimental Details The e ffect of el ectrical bias on AlGaN/GaN HEMT degradation is studied via constant voltage stressing (CVS). CVS offers a simpler way to study time dependent degradation over conventional step stress measurements in AlGaN/GaN HEMTs by monitoring the time evol ution of an e lectrical quantity under constant applied bias A rever s e gate bias is applied at V DS =0V state grown on SiC substrate and gate leakage current is simultaneously monitored over time at room temperature The c ritical voltage (V crit ) determined from electrical step stress of these devices helps to a s certain the gate b iases needed to capture the time dependent degradation within a reasonable time. These AlGaN/GaN HEMT s undergo degradation at V crit ~ 30 V when the gate voltage is stepped from V GS = 10V at V DS =0V w ith each voltage step held for 1 00 sec Thus three gate bia ses are used for this e xperiment, V GS = 25V, 30V and 35V which are applied separately to three identical devices, selected from pre degradation DC characterization The V DS =0V state allows the effect of vertical field under the gate to be isolated and sim plif ies the theoretical analysis of failure mechanisms and also eliminate s any possible device self heating which would occur if high drain bias were applied.

PAGE 87

87 Electrical characterization is performed by the Keithley 4200 semiconductor characterization syst em to measure (or sample) I G under the applied bias condition during degradation experiment. However the inbui lt interactive module (ITM) in the K eithley can only capture 4096 samples over the time during an electrical measurement in the sampling mode whic h is insufficient to observe time dependent degradation especially at smaller biases Hence a compatible user defined module was coded to run wi collect a large number (>16000) of samples over the entire span of the time d ep endent degradation test 6.2.2 Results and Discussion A sudden increase in I G is observed after an elapsed period of time indicating the onset of device degradation under the applied high reverse gate bias. I G continues to increase further un til it reaches a set current compliance. The I G compliance is set at It is observed that device s biased at higher reverse gate bias es (such as V GS = 35V) begin to degrade sooner than device s biased at smaller gate bias es (such as V G S = 25V), as shown in Figure 6 2 A h igher gate bias results in a higher E field under the gate, especially at the gate edges Hence it can be inferred that high E field accelerates device degradation, possibly by adding to the existing stress through the i nverse piezoelectric effect and enhancing the strain relaxation process under the gate either through gene ration of new trapping centers or field driven chemical reaction resulting in diffusion and altering of the metal semiconductor interface This can al ter the Schottky barrier height and also provide additional low resistance path s for the leakage current and hence may be manifested as an increase in the gate current, a decrease in the saturation drain current and a shift in V TH in the degraded samples.

PAGE 88

88 In order to analyze the failure mechanism causing I G degradation in AlGaN/GaN gate failure I Gfailure_a vg ) is defined as follows (6 1) Here t 1 is the time at which sudden sharp increase in I G is observed which is believed to be start of permanent degradation in the device. Time t 2 is t he time at which I G reaches the set current compliance and t deg is the total time during which permanent degradation in device occurs ( manifested by the increas e in gate current till I G _max=I G (t 2 ) as defined ) and is given by (6 2) The quantity I Gfailure_ avg i s an estimate of the average current that flows through t he device during degradation, calculated over the critical failure time interval of device degradation t deg =t 2 t 1 and hence can po tentially give insight into the time dependent device degradation. The validity of I Gfailure_avg as a metric to analyze failure mechanism in AlGaN/GaN HEMT is explained in Section 7.2. It is expected that I Gfailure_avg would be larger for higher applied vo ltages due to the larger gate current and the shorter critical failure time interval of device degradation. I Gfailure_avg is estimated from the time dependent I G degradation for each applied bias and is plotted as a function of the applied reverse gate bia s in Figure 6 3. It can be directly inferred from the plot that the average gate failure current increases with increase in applied reverse gate bias indicating that higher V GS accelerates the degradation process in these devices. I Gfailure_avg for each ap plied reverse gate bias is also plotted versus the square root of the time to degrade deg ) at that gate bias as shown in Figure 6 3. The plot shows a strong

PAGE 89

89 linear relationship between I Gfailure_avg and t deg Hence it can be hypothesized from this plot that the diffusion process may potentially be influencing the device degradation in dependence, as described in detail in Chapter 5. 6.3 Effect of Electrical Bias and Temperature on A lGaN/GaN HEMT Degradation 6.3.1 Experimental Details The e ffect of electrical bias and temperature is studied at a con stant high reverse gate bias at V DS =0V with varying chuck temperature. To alter the chuck temperature, a th ermal stress set up is employed as shown in Figure 6 5 that consists of a P eltier heater attached to the back of an aluminum block which acts as a heat sink for the heater as well as a chuck to place the sample being measured. A thermistor is attached to the setup to monitor the chuck temperature. A PID controller controls the chuck temperature via a PC (personal computer) The c huck temperature (T C ) is varied as 375K 400K and 450K for three different degradation tests in this experimental study. 1um gate lengt h AlGaN/GaN HEMTs fabric ated on SiC substrate are used for this study. Three id entical devices are employed to compare the time dependent degradation at three different chuck temperatures listed above. Devices are biased at V GS = 25V, V DS =0V state and I G is simult aneously measure d (or sampled) over time. The d evice temperature is monit ored using a RTD sensor, and Kei t hley 4200 is employed for electrical measurements similar to the previous experimental study. 6.3.2 Results and Discussion Similar to the previous test in S ection 6.2 a n onset of device degradation is marked by a sudden increase in I G observed after an elapsed period of time under the applied high reverse gate bias and elevated chuck temperature I G continues to increase

PAGE 90

90 further un til it reaches a set current complian ce of The i ncrease in chuck temperature is observed to enhance device degradation in AlGaN/GaN HEMT, as shown in Figure 6 6 The d evice stressed under high reverse gate bias at T c = 450K shows a quicker and more enhanced degradation than the device st ressed at T c = 375K This possibly indicates that elevated temperature accelerates device degradation through a thermally activated process such as trap generation or chemical reaction resulting in diffusion and modifying the metal semiconductor Schottky int erface. This may be manifested as an increase in gate leakage current, decrease in saturation drain current and shift in V TH in the degraded devices Since all the devices are stressed at V DS = 0V state, no or little self heating occurred and the effect of externally applied electrical bias and elevated temperature can be independently studied by externally controlling the device temperature. T o get insight into the possible mechanism causing device degradation, the average gate failure current I Gfailure_avg and time to degrade t deg are estimated from these tests as described in S ection 6.2.2. t deg is the time during which I G is observed to undergo degradation, hence 1 / t deg correspond s to the rate of I G degradation and reflects the failure mechanism causing d evice degradation. A plot of Ln (1/ t deg ) v ersu s 1/T, where T is the chuck temperature, is shown in Figure 6 7 At each temperature a set of three devices is stressed under constant voltage stressing. Uncertainty in Ln (1/ t deg ) at each temperature is estima ted using 99% confidence interval and plotted as error bars in Figure 6 7. A thermal activation energy E A =0.2 80.06 eV is computed from the slope of the plot of Ln (1/ t deg ) vs 1/T, which is consistent with the activation energy published in literature for oxygen di ffusion along dislocations in III n itride layers [46] [132 133]

PAGE 91

91 Pearton et. a l. [131] estimate an activation energy of 0.23eV for diffusi on of oxygen into GaN epi taxial layer. The devices used in this study of time dependent degradation potentially have a thin oxide layer between the Ni gate metal and AlGaN layer. Under appropriate condition s of applied bias and temperature, this oxide la yer may weaken and eventually rupture which may cause the onset of permanent device degradation marked by sudden increa se in gate current. Further, oxygen from the oxide layer diffuse s into the AlGaN layer which may be accompanied by the d iffusion of gate metal into the epitaxial l ayer, as observed in TEM images of degraded AlGaN/GaN HEMT samples in [49] F igure 6 8 shows a plot of average gate failure current I Gfailure_avg (computed for each test done in this study) versus the square root of time to degrade deg ) We observe a linear relationship between I Gfailure_avg and t deg which may be indicative of possible diffusion causing device degradation, since However we find a weak linear dependence of I Gfailure_avg with t deg in this study compared to the time dependent degradation study under varying gate biases. This could be due to the fact that under elevated temperature s the increase in gate current may also be caused by temperature dependent ther mal emission along with the possible device degradation. 6.4 Effect of Electrical Bias and Mechanical Stress on AlGaN/GaN HEMT Degradation 6.4.1 Experiment Details The effect of mechanical stress on I G degradatio n in AlGaN/GaN HEMTs is studied on samples fabricated on Si substrate, by applying externa l l y uniaxial mechanical stress using a 4 point wafer bending jig as shown in Figure 6 9 A sample is

PAGE 92

92 prepared by attaching a small wafer piece on a high carbon steel metal plate using a thermally conductive epoxy and the sample is then inserted into the jig to apply mechanical stress The device is wire bonded, using the procedure described in Section 3.1.2, so that uniaxial mechanical stress can be applied simultaneo usly while it is electrically stressed. Electrical step stress test is first performed at V DS =0V and varying reached. Reverse gate bias for constant voltage stressing (CVS) is chosen close to the critical voltage of device As I G begins to undergo degradation during CVS, mechanical stress is applied on the sample and varied between two values in a cyclic manner as shown in Figure 6 10 This procedure allows the study o f the effect of varying mechanical stress on the rate of time dependent I G degradation on the same device just as it begins to exponentially degrade thereby eliminating the influence of device to device variation during such tests A uniaxial tensile stre ss, longitudinal to channel direction, is applied using the bending jig on an AlGaN/GaN HEMT sample which is electrically stressed at constant voltage at V DS =0V a nd a high reverse gate bias of V GS = 65V at room temperature. Tensile stress is varied between 0 MPa and 105 MPa in a cyclic manner simultaneously while I G is being monitored simultaneously. Similarly o n another AlGaN/GaN HEMT sample, a uniaxial compressive stress, longitudinal to channel direction, is applied using the bending jig while it is elec trically stressed at constant voltage at V DS =0V and V GS = 35V at room temperature (Note the device to device variation as this device has a different critical voltage from the previous device). Compressive stress is varied between 0 MPa and 50 MPa i n a cyc lic manner while I G is being monitored simultaneously.

PAGE 93

93 6.4.2 Results and Discussion It is observed, as shown in Figure 6 11 that the rate of I G degradation decreases under externally applied uniaxial tensile stress while it begins to increase after the te nsile stress is released from the sample. It is observed in Figure 6 13 that the rate of I G degradation increases under externally applied uniaxial compressive stress and it begins to decrease after the compressive stress is released from the sample. Figur e 6 12 and Figure 6 14 represent the % change in I G during the time dependent degradation of AlGaN/GaN HEMT while being electrically stressed at constant voltage at V DS =0V and high reverse gate bias, during two cycles of varying external uniaxial tensile a nd compressive stress respectively. A moving average of data po ints is also plotted alongside ( with subset of 100 points ) to smooth out local fluctuations in each gate current measurement and highlight the overall trend of rate of I G degradation during the course of application of external mechanical stress A slope is also estimated using linear regression for the plots of % change in I G during AlGaN/GaN time dependent degradation under the externally applied mechanical stress and a high reverse gate bias The moving average plots as well as the corresponding slopes, as shown in Figure 6 12 and Figure 6 14, s uggest that the rate of I G degradation decreases under the application of uniaxial tensile stress while it increases under the application of uniaxial compression stress Uniaxi al tensile (compressive) stress is applied in plane to AlGaN layer and longitudinal to channel direction. Resultantly, this generates compressive (tensile) stress transverse to the channel direction. This resultant compressive (te nsile) stress can be estimated using the strain). Externally applied l ongitudinal tensile (compressive) stress together with resultant tr ansverse compressive (tensile) stress can potentially decelerate (accelerate)

PAGE 94

94 the gate current degradation. One possible explanation is the combination of electric field and mechanical transverse compressive (tensile) stress decreases (enhance s ) diffusion of both oxygen and gate metal at the gate metal/AlGaN layer i nterface that causes device degradation at V DS =0V and high reverse gate bias although proof would require material characterization studies. 6.5 Effect of Electrical Bias and Ambient on AlGaN/GaN HEMT Degradation 6.5.1 Experimental Details A state of the a rt cryogenic probe statio n chamber from Lakeshore Cryotronics Inc. shown in Figure 6 1 5 is employed to study the effect of varying ambient on I G degradation at V DS =0V state in AlGaN/GaN HEMT samples fabricated on SiC substrate The a mbient is varied by p urging the cryostat chamber using the procedure described below. A device is electrically stressed at constant voltage at V DS =0V and a high reverse gate bias while monitoring I G simultaneously. The r everse gate bias for constant voltage stressing (CVS) is chosen close to the critical voltage of the device, si milar to as described in S ection 6.4.1. As I G begins to undergo degradation during CVS, the ambient is varied between two gases used for the study, in a cyclic manner as shown in Figure 6 1 6 similar t o as described in S ection 6.4.1 This procedure allows the study of the effect of ambient variation on the rate of time dependent I G degradation on the same device and eliminates the potential influence of device to device variation on the outcome. The stu dy of ambient variation is done at elevated temperature s to enhance potential d iffusion of ambient gas through the thick SiN X passivation on the device. This probe station consists of a three stage heating set up to control the temperature of inside of the chamber as well as of the stage where the device is placed. The

PAGE 95

95 temperature range su pported by this set up is 77K 45 0K. A RTD temperature sensor is used to monitor the temp erature of the device being tested. Two tests are performed on two separate devices for this study. In the first test, the device is electrically stressed at V GS = 25V and V DS =0V while I G is being monitored. The stage temperature is set at 400K and ambient is varied between Nitrogen gas and Oxygen gas in a cyclic manner for the test. In a second test, the device is electrically stressed at V GS = 25V and V DS =0V while I G is being monitored, and the stage temperature is set at 450K and the ambient is varied between Nitrogen gas and Forming gas (4% Hydrogen +96% Nitrogen ) in a cyclic manner. P urging Procedure. Lakeshore probing chamber consists of a p urge valve on one side (Figure 6 1 5 (B )) and a 1/3 psi relief va lve on the other side (Figure 6 1 5 (C)). The 1/3 psi means that the relief valve will begin to open when the pressure inside reaches 1/3 psi above atmospheric pressure and maintains 1/3 psi inside the chamber. Purging Lakeshore chamber with a gas requires displacing the already present atmospheric air with a const ant flow of the gas through purg e valve. A g as line is attached to the pur ge valve. The purge valve is slowly opened and a regulated flow of gas is maintained so that the pressure inside the chamber does not exceed 1/3 psi. The relief valve functions on a spring system. As the pressure inside the chamber exceeds beyond the marke d limit, it stretches the spring associated with relief valve and opens the valve to release the excess pressure. If the chamber is over pressurized beyond the safe limit, it causes the spring to stretch beyond its elastic limit and potentially break it an d resultantly the relief valve will not reseal. Hence the gas flow through the purge valve must be limited and regulated by using a precision pressure regulator attached to

PAGE 96

96 the gas cylinder. The lids of the chamber are left unscrewed/loosened to act as sec ondary relief valve and prevent over pressuri zi ng inside the chamber. At 1/3 psi, the gas flow rate is approximately 5 liters per minute (LPM) while the volume of cryostat chamber is about 35 liters as shown in Figure 6 15 Hence the gas being purged will ideally take approximately 5 minut es to completely fill the chamber It is essential to completely equilibrate the device in the desired ambient gas in side the chamber. Hence to successfully purge the cryostat chamber completely with desired ambient the a mbient gas must completely displace all atmospheric air and moisture from the chamber. Displacing moisture out of chamber is the hardest to achieve. Hence a pre purging procedure is employed to facilitate displacement of moisture from the Lakeshore chamber T he chamber is purged with the gas for a s hort period of time (5 10 min), followed by creating vacuum by running a Agilent DS 101 roughing pump. This process is repeated about 4 5 times, thereby displacing moisture from inside the chamber. Moisture can b e more easily displaced by vacuum pump than the exhaust gas. The purging procedure is summarized in a flow chart in F ig ure 6 17 6.5.2 Results and Discussion Figure 6 1 8 shows the effect of varying ambient between O xygen and N itrogen in a cyclic manner at 400K on I G degradation while the device is electrically stressed at constant voltage at V GS = 25V and V DS =0V Figure 6 1 9 shows the effect of varying ambient between Forming gas and Nitrogen in a cyclic manner at 450K on I G degradation, while the device i s electrically stressed at constant voltage at V GS = 25V and V DS =0V. In both of the tests, no obser vable effect of ambient variation on rate of I G degradation is captured during the measurement This could be potentially due to the presence of a thick pass ivation layer on AlGaN/GaN HEMT sample s used in this study.

PAGE 97

97 Also the cycle time used in thes e tests to allow the device to equilibrate in a given ambient gas at a time is considerably short. Thus, the thick passivation layer and the short cycle time, would prevent the diffusion of ambient gas molecules through the passivation into the device to influence any device performance. However the device s continue to degrade under constant voltage stressing at a certain rate un til the gate current reaches the set c urrent compliance. 6.6 Summary This chapter focuses on a detailed experimental study to investigate the effect of varying external conditions such as bias, temperature, mechanical stress and ambient on the observed time dependent degradation in AlGaN/GaN HEMT Device degradation is studied through increase in gate current at high reverse gate bias at V DS =0V state. It is observed that I G degradation increases with increase in applied gate bias and also with applied temperature. A linear dependence with the square root of the time to degrade t deg ) indicates that observed time dependent degradation is potentially via diffusion mechanism occurring within the device. An activation energy of E A =0.2 80.06 eV is obtained from I G degradation under varying temperature which may correspond to the dif fusion of oxygen from the oxide present at the interface into the GaN epitaxial layer The e ffect of externally applied mechanical stress on time dependent degradation is studied by varying mechanical stress in a cyclic manner on the same device. It is obs erved that the rate of I G degradation decreases (increases) under external uniaxial tensile (compressive) stress applied longitudinal to channel direction. The e ffect of ambient variation is also studied by varying ambient gases in a cyclic

PAGE 98

98 manner inside a cryostat chamber. No observable effect of ambient variation is measured on the rate of I G degradation.

PAGE 99

99 Figure 6 1 Schematic of the cross section of AlGaN/GaNHEMT hterostructure used for time dependent degradation study Figure 6 2 Time dependent I G degradation of AlGaN/GaN HEMT on SiC substrate by varying reverse gate bias (V GS = 25V, 30V, 35V) at V DS =0V state and at room temperature

PAGE 100

100 Figure 6 3 I Gfailure_a vg estimated over time of degradation increases with increase in applied reverse gat e bias indicating that degradation is accelerated with increase in applied bias in AlGaN/GaN HEMT F igure 6 4 Strong linear dependence between I Gfailure_a vg (estimated over time of degradation) and square root of time to degrad e for the time dependent degradation test at varying reverse gate biases. It indicates a potential role of diffusion process on device degradation in AlGaN/GaN HEMT

PAGE 101

101 Figure 6 5 Peltier heater set up to vary chuck temperature Figure 6 6 Time dependent I G degradation of A lGaN/GaN HEMT by varying the chuck temperature (T= 375K, 400K, 450K ) at V GS = 25V and V DS =0V

PAGE 102

102 Figure 6 7 Plot of log of time to degrade versus chuck temperatu re. Activation energy of E A =0.28 0.06 eV is obtained Figure 6 8 Linear dependence betwee n I Gfailure_avg and square root of time to degrade for the time dependent degradation test at varying temperatures. It indicates a potential role of diffusion process on device degradation in AlGaN/GaN HEMT.

PAGE 103

103 Figure 6 9 Ex perimental set up to investig ate effect of mechanical stress on rate of I G degradation in AlGaN/GaN HEMT Figure 6 10 Varying two different mechanical stresses alternatively to study effect of mechanical stress on rate of AlGaN/GaN HEMT d egradation

PAGE 104

104 Figure 6 11 Effect of uni axial tensile stress on rate of I G degradation in AlGaN/GaN HEMT on Si substrate which is electrically stressed at V DS =0V, V GS = 65V at room temperature Figure 6 12 % Change in I G during time dependent degradation in AlGaN/GaN HEMT at V DS =0V, V GS = 65V at room temperature during two cycles of varying external uniaxial tensile stress (as in Figure 6 11 )

PAGE 105

105 Figure 6 1 3 Effect of uniaxial compressive stress on rate of I G degradation in AlGaN/GaN HEMT on Si substrate which is electrically stressed at V D S =0V, V GS = 3 5V at room temperature Figure 6 14 % Change in I G during time dependent degradation in AlGaN/GaN HEMT at V DS =0V, V GS = 3 5V at room temperature during two cycle s of varying external uniaxial compressive stress (as in Figure 6 13).

PAGE 106

106 Figu re 6 15. Lake shore cryogenic probe station chamber u sed for the study of effect of ambient variation on time dependent degradation in AlGaN/GaN HEMT (B) Purge Valve (C) Relief Valve Figure 6 1 6. Varying two different ambients alternatively in cyclic m anner while device is being electrically stressed under constant voltage to study the effect of ambient variation on time dependent degradation in AlGaN/GaN HEMT

PAGE 107

107 Figure 6 1 7. Flow chart of the steps to purge cryostat chamber with desired ambient gas completely

PAGE 108

108 Figure 6 1 8. Effect of varying ambient between Oxygen and Nitrogen at 400K on I G degradation while device is electrically stressed at V GS = 25V and V DS =0V Figure 6 1 9. Effect of varying ambient between Forming gas and Nitrogen at 45 0K o n I G degradation while device is electrically stressed at V GS = 25V and V DS =0V

PAGE 109

109 CHAPTER 7 MODELLING AND ANALYSIS OF EFFECT OF ELECTRICAL BIAS, TEMPERATURE, MEC HANICAL STRESS AND AMBIENT ON AlGaN/Ga N HEMT TIME DEPENDENT DEGRADATION 7.1 Resultant Mechanica l Stress in AlGaN layer due to Applied Bias One of the widely accepted theories for AlGaN/GaN HEMT degradation is the creation of additional mechanical stress at h igh operating voltages due to inverse piezoelectric effect [30] in addition to in built strain in AlGaN layer from latt ice mismatch The added mechanical stress increases the and w hen a critical limit is reached, the epitaxial layer undergoes relaxation through crystallographic defect crea tion in form of cracks and pits. These defects may act as t rapping centers for electrons or provide additional transport path for gate leakage current; hence degrading device performance and reliability. This chapter identifies the effect of electrical bias, temperature, and mechanical stress on the time dependent degradation of AlGaN/GaN HEMTs and carefully explains possible degradation mechanism through diffusion kinetics as hypothesized in Chapter 6 causing irreversible device failure. In order to perform a quantitative analysis of mechanical stress generated i n AlGaN layer due to inverse piezoelectric effect as a result of applied bias, following c onstitutive r elation s for a piezoelectric m aterial need to be considered [137] (7 1) (7 2) Here, T = Stress matrix (6X1) S = Strain matrix (6X1) E = Electric Field (3X1)

PAGE 110

110 For AlN and GaN wurzite crystal s the stiffness and piezoelectric coefficients matrix is given by [137] Assuming free standing and clamped model for thin AlGaN layer grown (in x y plane) pseudomorphically on relatively relaxed GaN layer, the me chanical stress in zz, xz and yz directions can be neglected (T zz ,T xz ,T yz =0) Also the planar strain in the xx and yy directions ( S xx S yy ) can be assumed to be equal when mechani cal stress is generated due to a transverse a pplied field in the zz direction Hence using the piezoelectric constitutive relation s (E quations 7 1 and 7 2) and following these assumptions, the mechanical stress due to the vertical electric field E zz is given by ( 7 3) Stiffness coefficien t C E = Piezoelectric coefficient e T = c E = stiffness matrix (6X6) s E = compliance matrix (6X6) d t ,e t = piezoelectric matrix (3X6)

PAGE 111

111 The first term in E quation 7 3 corresponds to the stress generated in the AlGaN layer due to lattice mismatch. Lattice mismatch between AlGaN and GaN layers is associated with the in plane (x y plane) strain (S xx and S yy ) in the AlGaN layer The s econd term in the equation gives the resultant stress due to the applied bias via the inverse piezoelectric effect. The piezoelectric constants and stiffness coefficients for GaN and AlN are listed in Table 3 1 [38] The field values are estimated by performing 2D simulations using FLOODS for the applied reverse gate bias with the source and drain grounded. Only the vertical field values E zz are employed for the calculation of stress in the above eq uation. The simulated contours of the total electric field values under the gate edges for varying reverse gate bias and at V DS =0V is shown in Figure 7 1. Figure 7 2 indicates the simulated vertical E field values used to estimate the resultant stress due to the inverse piezoelectric effect for the applied reverse gate bias at V DS =0V, with E field values extrapolated to larger reverse gate biases The in plane (x y plane) resultant stress T xx generated in the AlGaN layer due to the applied reverse gate bias is plotted as a function of electric field in Figure 7 3 and is found to be ~ 10 MPa for 1V of applied gate bias As mentioned above this model of device degradation is based on the elastic energy stored in the AlGaN layer and its modification due to the electric field under the gate through the inverse piezoelectric effect. The e lastic energy stored in AlGaN layer corresponding to resultant strain can be calculated using the following e quation [138] (7 4) Here, W elastic is the elastic energy stored in AlGaN layer, Y AlGaN is the Y modulus of AlGaN, t AlGaN is the thickness of AlGaN layer and S zz is the strain in the

PAGE 112

112 vertical direction in the AlGaN layer. The t hickness of the AlGaN layer in AlGaN/GaN HEMTs under study is 15nm, as mentioned in C hapter 6. 7. 2 Metal to S emiconductor Current T ransport Time dependent I G degradation in AlGaN/GaN HEMT is further investigated in this section by analyzing a model for the tunneling current through a reverse biased Schottky interface and u tilizing results from the set of experimental results from Chapter 6 to gain further insight into possible degradation mechanism s in AlGaN/GaN HEMTs The g ate in AlGaN/GaN HEMTs forms a Schottky interface with the underlying semiconductor epilayer, in this case AlGaN layer. Several theories have been proposed in the literature explaining the degradation of the AlGaN/GaN HEMT during t ypical device operation; namely inverse piezoelectric effec t [37] h ot electron effect [31] diffusion at gate/AlGaN Schottky interface [49] and virtual gate model [19] among others. This section identif ies the effect of electrical bias, temperature and mechanical stress on the time dependent degradation of AlGaN/GaN HEMTs and carefully explain s probable degradation mechanisms through diffusion kinetics as hypothesized in Chapter 6 causing irreversible device failure The analytical model formulated in this chapter is based on the analysis of the tunneling current in metal GaN Schottky diode s [60] [139] The tunneling current model for Schottky diodes applied to a metal Al GaN Scho ttky diode involves the gate current density J due to electrons tunneling from metal into semiconductor under reverse biased gate in AlGaN/GaN HEMTs. The d egradation in AlGaN/GaN HEMTs is believed to occur under the gate near the gate edges where the electric field is the highest in the device [37] [49] Hence any catastrophic failure of the gate semiconductor junction will be reflected in the degradation of I G which is a useful metric in thi s study to analyze

PAGE 113

113 device degradation. Current flow due to electron tunneling from the metal to semiconductor is given by the basic equation as follows [60] ( 7 5 ) Here A* is the effec the Fermi Dirac distribution function s in me tal and semiconductor, k is Boltzmann constant, T is t emperature in Ke lvin (K) and the range of energies of electrons that will participate i n tunneling from metal into the semiconductor conduction band Figure 7 4 shows a schematic of the energy band diagram at V DS =0V state and defines relevant parameters, including energies qV b which are used in this derivation. Since only a few electrons will tunnel from the metal into semiconductor at the base of triangular barrier, most states in the semiconductor c onduction band participating in this derivation can be as sumed to be unoccupied, giving (1 F s ) ~ 1 In case of metal, since the Fermi level is within the conduction band, all the participating states can be assumed to be occupied yielding F m ~ 1. The quant the quantum transmission coefficient of the electrons tunneling through an approximate ly triangular Schottky barrier and is expressed as (7 6) in the above equation is the width of the tunneling barrier Electrons in the metal may tunnel through the barrier throughout the range of ene rgies used in the integral in Eq uation 7 5. Hence the he energy dependence as fo llows

PAGE 114

114 (7 7) Here the is the physical distance that electrons must travel at the base of the triangular barrier and corresponds to the thickness of the AlGaN barrier layer (d AlG aN ) in devices use is very important in this study as it can be modified due to potential gate metal diffusion into the AlGaN epilayer as a result of degradation of the metal semiconductor junction If gate metal diffusio n into the AlGaN barrier layer occurs at high gate E fields, it would reduce the AlGaN barrier physical thickness and resultantly narrow the t unneling barrier for electrons increasing G and potentially leading to device failure Two useful assumptions can be made in this analysis : 1) limited source diffusion model can be assumed for the potential diffusion occurring at the gate/AlGaN junction, described in detail in Chapter 5 and is given by [124] (7 8) Here Q is the total dose of diffusant, D is the diffusion coefficient that depends on temperature and x is the distance by which diffusion front moves in time t 2) During device degradation due to metal diffusion under the gate, the gate current I G can be assumed to be concentrated near localized areas of gate metal diffusion as shown in Figure 7 5. Using the assumption (1) stated above, the distance, d diff to which metal diffusion occurs into the AlGaN layer can be estimated by using a condition where diffusant concentration is negligible (N(x,t) ~0 in Eq uation 7 8) Hence, for N(x,t) = 0.002 or 0.2%

PAGE 115

115 ( 7 9) Therefore, the effective distance at the base of tunneling barrier at time t during time dependent device degradati on due to gate metal diffusion will be given by (7 10) uation 7 6 to compute the transmission coefficient which is then plugged into the integral in Eq uation 7 5. The integral is computed numerically to estimate the gat e current density J TEM imaging by Ray [140] shows the regions of metal diffusion under the gate in a degraded AlGaN/GaN HEMT and the associate d cross section area is found to be ~10 1 5 m 2 Using assumption (2), the gate current can be estimated from the simulated J using the best fit for the area of the localized metal diffusion where the gate current is assumed to be concentrated. The above simulation steps to estimate I G are summarized in a flowchart in Figure 7 6. The g ate current I G as a function of time is simulated for varying applied gate bias es as well as varying temperature s and compared with the experimentally measured gate current during the time dependent degradation study described in Chapter 6 Figure 7 7 shows the simulated I G (dotted curve) for v arying reverse gate biases, V GS = 25V and V GS = 30V, at room temperature and compared with experimentally measured I G under similar conditions. Figure 7 8 shows the simulated I G (dotted curve) for varying temperatures, T=375K and T=400K, at V GS = 25V and c ompared with experimentally measured I G under similar conditions. It is observed that the simulated I G shows a close match with the measured I G for the range of time span deg the figure) during which irreversible time dependent device (or I G ) degradation is believed to occur. Hence it can be inferred that diffusion at the gate

PAGE 116

116 semiconductor Schottky interface due to high E field (due to applied bias) or elevated temperature may be playing a significant role in the observed time dependent deg radation in AlGaN/GaN HEMTs. The best fit values of the parameters employed in this simulation are listed in Table 7 1. To validate the significance of the ( I Gfailure_avg ) as a metric to analyze device degradation, t he gate current I G is simulated using the above model for device degradation assuming gate metal diffusion into the AlGaN layer for varying reverse gate biases. The s imulated I G for varying reverse gate biases, at V DS =0V and room temperature, is plotted in Figure 7 9. I Gfailure_avg is estimated at each reverse gate bias from the simulated I G using the pr ocedure described in Section 6.2 .2 and is found to have a strong linear dependence with the square root of the time to degrade t deg ) as shown in Figure 7 10. This implies that if device degradation occurs via gate metal diffusion into the AlGaN layer then I Gfailure_avg will have a linear dependence with t deg The plot between I Gfailure_avg versus t deg estimated from the measured I G ( Figure 6 4 and Figure 6 8 ) shows slight deviation from the linearity. This could be due to the fact that the measured gate current may also be influenced by trap assisted leakage mechanism s as well as thermionic emission at elevated temperature s 7. 3 Summary Fundamental mathem atical modelling has been presented for a quantitative analysis of the effect of varying bias, temperature and mechanical stress on time dependent degradation in AlGaN/GaN HEMTs. It is found that for 1V of applied gate bias, approximately 10 MPa of biaxi al tensile stress is generated in the AlGaN layer due

PAGE 117

117 to inverse piezoelectric effect To utilize the V DS =0V state, only the vertical electric field under the gate is employed for the computation of the resultant stress This chapter also investigates a me tal semiconductor tunneling transport model to get an insight into the physics of the failure mechanism causing device degradation in AlGaN/GaN HEMTs The t ransmission coefficient for electrons tunneling from metal into semiconductor depends on the tunneli ng barrier width. A potential metal diffusion of gate metal into the AlGaN epilayer (initiated by the oxide layer rupture followed by its diffusion into AlGaN epilayer) can alter the barrier width thus increasing the gate current and resulting in permanen t device degradation. Assuming a limited source diffusio n model and the gate current will mainly be concentrated at localized areas of metal diffusion, a good fit is obtained between the gate current measured during device degradation and the gate current sim ulated using this model.

PAGE 118

118 Table 7 1 Best fit of parameters used in I G simulation by metal semiconductor tunneling model Parameter Fitting Value Barrier Height ( b ) 1. 4 eV [74] Electron Effective Mass (m*) 0.25 m e 0 [141], [142] Area (A) (where I G is concentrated) 1 E 1 5 m 2 [140] Diffusion Coefficient (D) at 300K 2E 18 c m 2 s 1 [46]

PAGE 119

119 Figure 7 1 Simulated E field contours under the gate in AlGaN layer for varying reverse gate bias es at V DS =0V Figure 7 2 Simulated and extrapolated vertical E field values (E xx ) under the gate for varying reverse gate bias at V DS =0V

PAGE 120

120 Figure 7 3 Simulated in plane m echanical s tress generated in AlGaN layer due to a pplied gate b ias as a result of inverse piezoelectric ef fect Figure 7 4 Schematic of energy band diagram in AlGaN/GaN HEMT under reverse gate bias showing electron tunneling from metal into semiconductor.

PAGE 121

121 Figure 7 5 Schematic showing concentration of gate current near localized areas of gate metal di ffusion in AlGaN/GaN HEMT. Figure 7 6 Flow chart showing steps to simulate I G in AlGaN/GaN HEMT under reverse gate bias including diffusion under the gate at Schottky interface.

PAGE 122

122 Figure 7 7 Simulated and measured I G indicating time dependent degra dation in AlGaN/GaN HEMT at varying reverse gate bias es Figure 7 8 Simulated and measured I G indicating time dependent degradation in AlGaN/GaN HEMT at varying temperature s

PAGE 123

123 Figure 7 9 Simulated I G during AlGaN/GaN HEMT degradation at varying re verse gate biases due to gate metal diffusion into the AlGaN layer Figure 7 10. Linear dependence between I Gfailure_a vg and square root of time to degrade I Gfailure_a vg is estimated from the simulated I G during device degradation at varying reverse ga te biases due to gate metal diffusion into the AlGaN layer.

PAGE 124

124 CHAPTER 8 CONCLUSION AND FUTURE WORK 8.1 Conclusion The AlGaN/GaN high electron mobility transistor ( HEMT ) has emerged as the most attractive candidate for high power and high frequency applica tions over its other semiconductor counterparts, namely Si MOS FETs and AlGaAs/GaAs FETs. Existence of wider band gap, higher saturation velocity and higher breakdown field are some of the material parameters which render high performance capabilities to th is technology AlGaN/GaN HEMT s have demonstrated both reversible as well as irreversible electrical degradation during its device operations, marked by increase in source/drain resistance current collapse decrease in transconductance, threshold voltage s hift increase in gate leakage etc. A body of work has been published proposing several theories to explain the degradation mechanism s in AlGaN/GaN HEMTs. Hot electron effect, strain due to inverse piezoelectric effect, formation of virtual gate on AlGaN s urface between gate and drain contacts and occurrence of chemical reaction at the gate/AlGaN Schottky interface resu l ting in diffusion of gate metal or other chemical species are some of the widely accepted physical models for the degradation of GaN based HEMTs. Clearly there is a lack of agreement on the physics behind the observed device degradation and reliability of these devices still remains a major concern This work has endeavor ed to investigate the fundamental physics of AlGaN/GaN HEMT device fail ure through a series of systematic experimentation and to perform a qualitative analysis of existing physical model for gate metal diffusion into the semiconductor causing degradation in these devices

PAGE 125

125 T he effect of mechanical stress on device performance is studied through stress sensitivity of the channel resistance and gate current A four point wafer bending apparatus is described which allows application of external mechanical stress in a controlled manner. Also, a cost effective method is presented to apply mechanical stress externally to a small wafer piece by attaching it to a high carbon steel plate. The effect of mechanical stress on AlGaN/GaN HEMT channel resistance and gate leakage is studied eliminating the effect of charge trapping. An accurate value of gauge factor (G.F.= 2.80.4) is extracted indicating a small stress dependence on device resistivity. Also, from the stress sensitivity of gate current, it is proposed that PF emission dominates the gate leakage for low gate bias es above the thr eshold voltage while the gate leakage, at high depletion gate biases below the threshold voltage, could potentially be dominated by FN tunneling A thorough literature review on the existing models exp laining electrical degradation in AlGaN/GaN HEMT is pre sented. Physical models for strain relaxation phenomenon in heterostructure s are also discussed. Strain relaxation in heterostructures can occur either through trap generation or self or inter diffusion and the process with more favorable kinetic condit ions takes place. A modified electrical step stress experiment is designed to critical voltage (at which permanent electrical degradation begins) on varying time step s or the length of time for which each voltage step is app lied It is observed that degradation in AlGaN/GaN HEMTs is time dependent and can also take place at voltage s below the critical voltage.

PAGE 126

126 Following this inference, a series of systemic experiments are performed to investigate the effect s of electrical bia s, temperature, mechanical stress and ambient (N 2 O 2 f orming gas ) on the time dependent device degradation by monitoring the increase in gate leakage (or I G degradation ) at V DS =0V state in AlGaN/GaN HEMTs It is observed that I G degradation is enhanced by an increase in reverse gate bias as well as chuck temperature. The a verage failure gate current (I Gfailure_avg ) is estimated from the measured I G degradation and is shown to have a linear dependence with the square root of time for the observed irrevers ible degradation deg ) This implies that the observed degradation can be due to a potential diffusion phenomenon. An activation energy E A =0.2 80.06 eV is extracted from the I G degradation observed under varying temperatures, which is consistent with the activation energy of o xygen diffusion into the GaN epitaxial layer published in literature. The AlGaN/GaN HEMTs with Ni gate, used in this study have a thin oxide layer at gate/semiconductor junction due to oxidation of Ni during device fabrication It i s hypothesized that this oxide layer potentially undergoes slow degradation and diffusion u nder appropriate conditions of applied bias and temperature, until it weakens and e ventually rupture s most probably around threading dislocations. This is observed a s a sudden increase in gate current and causes the onset of permanent device degradation The diffusion process may be influenced by the high field driven excessive strain or elevated temperature. The diffusion of oxygen from the oxide layer may be ac compa nied by the d iffusion of gate metal into the AlGaN epita xial layer, providing a low resistance path for enhanced gate leakage current. The e ffect of mechanical stress on device degradation is studied using a four point wafer bending jig and it is observed that the rate of I G degradation

PAGE 127

127 decreases with applied uniaxial tensile stress longitudinal to channel direction. A state of the art cryogenic probe station chamber is utilized to study the effect of ambient variation on time dependent I G degradation A m e tal semiconductor tunneling model through a triangular barrier to simulate gate current across Ni AlGaN Schottky interface is studied by varying the tunneling distance. The t unneling distance is varied due to gate sou rce diffusion model. A close fit of simulated current, under varying biases and temperatures with experimentally measured I G degradation provides further evidence that a metal diffusion mechanism may potentially be causing the observed time dependent I G d egradation in AlGaN/GaN HEMTs. Degradation in AlGaN/GaN HEMTs due to strain relaxation via dislocation motion is improbable d ue to high activation energy of such mechanism. 8.2 Recommendations for Future Work One of the major challe nges in this work is th e device to device variation across the wafer. Device to device variation makes it challenging to compare the effects of an external condition on device degradation. In order to investigate the effect of variation in external conditions (temperature, mecha nical stress, ambient and o t hers) on device degradation reliably, measurement techniques must be developed to apply such variations on the same device being tested. AlGaN/GaN HEMTs have a high density of point defects and threading dislocations. Constant t rapping/de trapping interfere s with the reliable electrical measurement to investigate degradation p rocess Hence methods to eliminate trapping/de trapping phenomenon during an electrical measurement need to be employed while investigating device degradat ion One possible way to accomplish this is to employ a filtered UV light source to de trap any trapped charges without causing photo generation of carriers during the course of electrical measurement for

PAGE 128

128 device degradation. To gain a better insight into p hysics of mechanisms resulting in device degradation, new measurement techniques need to be employed to measure temporal variation in other electrical metrics such are source/drain resistance, transconductance, threshold voltage, position of Fermi level an d others during time dependent device degradation. Finally, a study of AlGaN/GaN HEMT device degradation through electrical measurements is recommended on samples with different gate metals, passivation layers and substrates and varying Al mole fraction i n AlGaN epilayer to gain a better understanding of failure mechanisms in AlGaN/GaN HEMTs and to develop robust mathematical models to explain observed device degradation

PAGE 129

129 LIST OF REFERENCES [1] S. Singhal, T Li, A. Chaudhari, A. W. Hanson, R. Therrien, J. W. Johnson, W. large periphery GaN on ROCS Workshop, 2005. [Reliability of Compound Semiconductors] pp. 135 149 2005. [2] A. M. Conway, M. Chen, P. Hashimoto, P. J. Willadsen, and M. Micovic, Reliability physics symposium, 2007. proceedings. 45th annual. ieee international pp. 472 475, 2007. [3] [4] Zhurnal Tekhnicheskoi Fiziki vol. 21, no. 2, p. 239, 1951. [5] IBM Journal of Resear ch and Development vol. 4, no. 3, pp. 283 287, 1960. [6] Solid State Electronics vol. 22, no. 9, pp. 783 791, Sep. 1979. [7] insic heterojunctions in thermal Solid State Electronics vol. 23, no. 10, pp. 1069 1075, Oct. 1980. [8] Electron Devices, IEEE Transactions on vol. 33, no. 5. pp. 616 619, 1986. [9] x)AlxAs/Ga(1 y)AlyAs double heterostructure room temperature lasers grown by metalorganic chemical vapor Applied Physics Letters vol. 31, no. 12, pp. 839 841, Dec. 1977 [10] modulation Applied Physics Letters vol. 33, no. 7, pp. 665 667, Oct. 1978. [11] T. Mimura, S. Hiyamizu, T. Fujii, a Effect Transistor with Selectively Doped GaAs/n Al x Ga 1 Japanese Journal of Applied Physics vol. 19, no. 5, pp. L225 L227, May 1980. [12] ectron mobility transistor based on a GaN Al[sub x]Ga[sub 1 Applied Physics Letters vol. 63, no. 9, pp. 1214 1215, Aug. 1993.

PAGE 130

130 [13] O. Ambacher, J. Smart, J. R. Shealy, N. G. Weimann, K. Chu, M. Murphy, W. J. Schaff, L. F. Eastman, R Dimitrov, L. Wittmer, M. Stutzmann, W. Rieger, and J. dimensional electron gases induced by spontaneous and piezoelectric polarization charges in N and Ga face AlGaN/GaN Journal of Applied Physics vol. 85, no. 6, pp. 3222 3233, Mar. 1999. [14] J. P. Ibbetson, P. T. Fini, K. D. Ness, S. P. DenBaars, J. S. Speck, and U. K. Applied Physics Le tters vol. 77, no. 2, pp. 250 252, Jul. 2000. [15] [16] U. K. Mishra, Y. F. Wu, B. P. Keller, S. Keller, a Millimeter Waves, 1997 Topical Symposium on pp. 35 39, 1997. [17] G. J. Sullivan, M. Y. Chen, J. A. Higgins, J. W. Yang, Q. Chen, R. L. Pierson, and power 10 GHz operation of AlGaN HFE Electron Device Letters, IEEE vol. 19, no. 6. pp. 198 200, 1998. [18] Y. F. Wu, A. Saxler, M. Moore, R. P. Smith, S. Sheppard, P. M. Chavarkar, T. W/mm GaN HEMTs by field plate optimizatio Electron Device Letters, IEEE vol. 25, no. 3. pp. 117 119, 2004. [19] IEEE Transactions on Electron Devices vol. 48, no. 3, pp. 560 566, Mar. 2001. [20] G. Meneghesso, G. Verzellesi, F. Danesin, F. Rampazzo, F. Zanon, A. Tazzoli, Electron Mobility Device and Materia ls Reliability, IEEE Transactions on vol. 8, no. 2. pp. 332 343, 2008. [21] A. Sozza, A. Curutchet, C. Dua, N. Malbert, N. Labat, and A. Touboul, Microelectronics Reliab ility vol. 46, no. 9 11, pp. 1725 1730, Sep. 2006. [22] N. Sghaier, M. Trabelsi, N. Yacoubi, J. M. Bluet, A. Souifi, G. Guillot, C. Gaquire, Microelectronics Journal vol. 37, no. 4, pp. 363 370, Apr. 2006.

PAGE 131

131 [23] frequency dispersion Electronics Letters vol. 31, no. 22. pp. 1951 19 52, 1995. [24] characteristic collapse in AlGaN/GaN heterostructure insulated gate field effect Electronics Letters vol. 30, no. 25. pp. 2175 2176, 199 4. [25] GaN High Electron Device Letters, IEEE vol. 29, no. 7, pp. 2008 2010, 2008. [26] S. C. Binari, K. Ikossi, J. A. Roussos, W. Kruppa, D. Park H. B. Dietrich, D. D. Electron Devices, IEEE Transactions on vol. 48, no. 3. pp. 465 471, 2001. [27] S. Trassaert, B. Boudart, C. Gaquiere, D. Theron, Y. Crosnier, F. Huet, and M. A. Electronics Letters vol. 35, no. 16. pp. 1386 1388, 1999. [28] H. Kim, J. Lee, D. Liu, and W. Lu, mechanism in unpassivated AlGaN/GaN high electron mobility transistors by post Applied Physics Letters vol. 86, no. 14, pp. 143503 143505, Apr. 2005. [29] G. Meneghesso, A. Chini, E. Zanoni, M. Manfred i, M. Pavesi, B. Boudart, and C. Electron Devices Meeting, 2000. IEDM pp. 389 392, 2000. [30] gradation of GaN High IEEE Electron Device Letters vol. 29, no. 4, pp. 287 289, Apr. 2008. [31] G. Meneghesso, R. Pierobon, F. Rampazzo, G. Tamiazzo, E. Zanoni, J. Bernat, P. Kordos, A. F. Basile, A. Chim, and G. Verzellesi electron stress Reliability Physics Symposium, 2005. Proceedings 43rd Annual IEEE International 2005, pp. 415 422. [32] gh dislocation densities in high efficiency GaN based light Applied Physics Letters vol. 66, no. 10, p. 1249, 1995.

PAGE 132

132 [33] G. Meneghesso, A. Mion, Y. Haddab, M. Pavesi, M. Manfredi, C. Canali, and E. InGaAs high electron mobility transistors: Journal of Applied Physics vol. 82, no. 11, pp. 5547 5554, Dec. 1997. [34] H. Kim, A. Vertiatchikh, R. M. Thompson, V. Tilak, T. R. Prunty, J. R. Shealy, and L. Microelectronics Reliability vol. 43, no. 6, pp. 823 827, Jun. 2003. [35] C. Canali, A. Neviani, C. Tedesco, E. Zanoni, A. Centronio, and C. Lanzieri, Electron Devices, IEEE Transactions on vol. 40, no. 3. pp. 498 501, 1993. [36] A. P. Edwards, J. A. Mittereder, S. C. Binari, D. S. Katzer, D. F. Storm, and J. A. GaN HEMTs using an NH3 plasma Electron Device Letters, IEEE vol. 26, no. 4. pp. 225 227, 2005. [37] Electron Devices Meeti ng, 2006. IEDM International pp. 1 4, 2006. [38] M. Chu, Characterization and modeling of strained Si FET and GaN HEMT devices University of Florida, 2011. [39] nanometer scale di ffusion depth of Pt/Ti/Pt/Au gate metallization on InAlAs/InGaAs high electron Applied Physics Letters vol. 87, no. 23, pp. 232102 232103, Dec. 2005. [40] e Electron Device Letters, IEEE vol. 7, no. 3. pp. 185 187, 1986. [41] barrier height variati ons in Au -W(Ti)/n Journal of Vacuum Science and Technology vol. 14, no. 4, pp. 939 942, Jul. 1977. [42] J. Wrfl, V. Abrosimova, J. Hilsenbeck, E. Nebauer, W. Rieger, and G. Trnkle, nitride microelectro Microelectronics Reliability vol. 39, no. 12, pp. 1737 1757, Dec. 1999. [43] type AlGaAs grown by molecular beam Materials Science and Engineering: B vol. 60, no. 3, pp. 189 193, Jun. 1999.

PAGE 133

133 [44] R Khanna, L. Stafford, L. F. Voss, S. J. Pearton, H. T. Wang, T. Anderson, S. C. and Light Emitting Diodes With TiB2 and Ir Device and Materials Reliability IEEE Transactions on vol. 8, no. 2. pp. 272 276, 2008. [45] Y. C. Chou, D. Leung, I. Smorchkova, M. Wojtowicz, R. Grundbacher, L. Callejo, under elevated temperature lifetest Microelectronics Reliability vol. 44, no. 7, pp. 1033 1038, Jul. 2004. [46] HEMT device reliability and degradation evolution: Importance of diffusion Microelectronics Reliability vol. 51, no. 2, pp. 195 200, Feb. 2011. [47] M. E. Glick sman, state principles, and applications New York: Wiley, 2000. [48] C. State Stress Critical Voltage by Using Pt Gated Electrochemical and solid state letters vol. 14, no. 7, p. H264, 2011. [49] L. Liu, T. S. Kang, D. a. Cullen, L. Zhou, J. Kim, C. Y. Chang, E. a. Douglas, S. plate on the characteristics of off state, step stressed AlGaN/GaN high electron Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures vol. 29, no. 3, p. 032204, 2011. [50] oxide semiconducto Journal of Applied Physics vol. 44, no. 10, pp. 4607 4615, Oct. 1973. [51] Solid State Electronics vol. 16, no. 2, pp. 253 263, Feb. 1973. [52] Physical Review vol. 94, no. 1, pp. 42 49, Apr. 1954. [53] Coefficients of (100) Silicon nMOSFETs Measured at Low and High (~1.5 GPa) Chann Electron Device Letters, IEEE vol. 28, no. 1. pp. 58 61, 2007. [54] S. Timoshenko, Strength of materials. New York: Van Nostrand, 1955. [55] K. Wu, Strain effects on the valence band of silicon piezoresistance in p type silicon and mobility en hancement in strained silicon pMOSFET [Gainesville, Fla: University of Florida, 2005.

PAGE 134

134 [56] J. A. Mittereder, S. C. Binari, P. B. Klein, J. A. Roussos, D. S. Katzer, D. F. Storm, AlGaN/GaN high electron Applied Physics Letters vol. 83, no. 8, p. 1650, 2003. [57] B. S. Kang, S. Kim, F. Ren, J. W. Johnson, R. J. Therrien, P. Rajagopal, J. C. Roberts, E. L. Piner, K. J. Linthicum, S. N. G. Chu, K. Baik, B. P. Gila, C. R. Induced Changes in the C onductivity of Electron Mobility Applied Physics Letters vol. 85, no. 14, pp. 2962 2964, Oct. 2004. [58] AlGaN/GaN high electron mobility transistor gauge factor based on two Journal of Applied Physics vol. 108, no. 10, pp. 104502 104506, Nov. 2010. [59] degradation of GaN HEMTs and its correlation with DC stress and I DLTS Microwave Integrated Circuits Conference, 2009. EuMIC 2009. European pp. 132 135, 2009. [60] bias leakag e current mechanisms in GaN grown by molecular Applied Physics Letters vol. 84, no. 4, pp. 535 537, Jan. 2004. [61] Electron Devices, IEEE Transactions on vol. 54, no. 10. pp. 2614 2622, 2007. [62] S. H. and B. S. and F. J. X. and F. L. and Z. L. M. and J. S. and L. L. and L. B. C. and L. W. S. and Z. X. Q. and Z. J. Y. and G. current mechanism in Schottky contacts to Al 0.25 Ga 0.75 N/GaN Semiconductor Science and Technology vol. 24, no. 5, p. 55005, 2009. [63] of gate leakage current in AlGaN/GaN high electron Applied Physics Letters vol. 82, no. 18, p. 3110, 2003. [64] mechanism in unpassivated AlGaN/GaN high electron mobility transistors by post App lied Physics Letters vol. 86, no. 14, pp. 143503 143505, Apr. 2005.

PAGE 135

135 [65] J. C. Carrano, T. Li, P. A. Grudowski, C. J. Eiting, R. D. Dupuis, and J. C. based metal -semiconductor -Applied Physics Letters vol. 72, no. 5, pp. 542 544, Feb. 1998. [66] Frenkel electron emission from the traps in Journal of Applied Physics vol. 95, no. 11, pp. 6414 6419, Jun. 2004. [67] H. Zhang, E. beam Journal of Applied Physics vol. 99, no. 2, p. 023703, 2006. [68] D. Yan, H. Lu, D. Cao, D. Chen, R. Zh Applied Physics Letters vol. 97, no. 15, p. 153503, Oct. 2010. [69] AlGaN Sch Applied Physics Letters vol. 84, no. 24, pp. 4884 4886, Jun. 2004. [70] observation of Poole Thin Solid Films vol. 405, no. 1 2, pp. 205 217, Feb. 2002. [71] Frenkel Effect with Compensation Journal of Applied Physics vol. 39, no. 12, pp. 5600 5604, Nov. 1968. [72] A. G. Milnes, Deep impurities in semiconductors. New York: Wiley 1973. [73] The theory of electrical conduction and breakdown in solid dielectrics Oxford: Clarendon Press, 1973. [74] ity of Florida, United States -Florida, 2011. [75] M. E. Levinshtein, S. L. Rumyantsev, and M. S. Shur, Properties of advanced New York: Wiley, 2001. [76] K. Rim, S. Koester, M. Hargrove, J. Chu, P. M. Mooney, J. Ott, T. Kanarsky, P. Ronsheim, M. Ieong, A. Grill, and H. VLSI Technology, 2001. Digest of Technical Papers. 2001 Symposium on pp. 59 60, 2001.

PAGE 136

136 [77] S. Thompson, N. An and, M. Armstrong, C. Auth, B. Arcot, M. Alavi, P. Bai, J. Bielefeld, R. Bigwood, J. Brandenburg, M. Buehler, S. Cea, V. Chikarmane, C. Choi, R. Frankovic, T. Ghani, G. Glass, W. Han, T. Hoffmann, M. Hussein, P. Jacob, A. Jain, C. Jan, S. Joshi, C. Kenyon, J. Klaus, S. Klopcic, J. Luce, Z. Ma, B. McIntyre, K. Mistry, A. Murthy, P. Nguyen, H. Pearson, T. Sandford, R. Schweinfurth, R. Shaheed, S. Sivakumar, M. Taylor, B. Tufts, C. Wallace, P. nm strained silicon channel transistors, 7 layers of Cu interconnects, low k ILD, and 1 Electron Devices Meeting, 2002. IEDM pp. 61 64, 2002. [78] A. D. Bykhovski, V. V Kaminski, M. S. Shur, Q. C. Chen, and M. A. Khan, Applied Physics Letters vol. 68, no. 6, pp. 818 819, Feb. 1996. [79] M. Yamaguchi, T. Yagi, T. Sota, T. Deguchi, K. Shimada, and S. Nakamura, Journal of Applied Phy sics vol. 85, no. 12, pp. 8502 8504, Jun. 1999. [80] 1983 Ultrasonics Symposium pp. 299 310, 1983. [81] C. Deger, E. Born, H. Angerer, O. Ambacher, M. St utzmann, J. Hornsteiner, E. x]N thin films obtained by surface acoustic Applied Physics Letters vol. 72, no. 19, pp. 2400 2402, May 1998. [82] H. Kim, a. Vertiatchikh, V. Tilak, R. M. Thompson, T. Prunty, J. R. Shealy, and L. GaAs Reliability 2002 Workshop pp. 5 6, 2002. [83] E. Zanoni, F. Danesin, M. Meneghini, a. Cetronio, C. Lanzieri, M Peroni, and G. IEEE Electron Device Letters vol. 30, no. 5, pp. 427 429, May 2009. [84] IV characteristics of AlGaN/GaN HEMTs with field Electronics Letters vol. 40, no. 24, pp. 1547 1548, 2004. [85] Solid State Electronics vol. 48, no. 2, pp. 271 276, Feb. 2004. [86] S. Y. Park, C. Floresca, U. Chowdhury, J. L. Jimenez, C. Lee, E. Beam, P.

PAGE 137

137 Microelectronics Reliability vol. 49, no. 5, pp. 478 483, May 2009. [87] U. Chowdhury, J. L. Jimenez C. Lee, E. Beam, P. Saunier, T. Balistreri, J. Wang, and Pit Shaped IEEE Electron Device Letters vol. 29, no. 10, pp. 1098 1100, Oct. 2008. [88] M. F aqir, G. Verzellesi, G. Meneghesso, E. Zanoni, S. Member, and F. Fantini, Electric vol. 55, no. 7, pp. 1592 1602, 2008. [89] n of GaN High Stress: The International Journal on the Biology of Stress pp. 5 8. [90] A. Sarua, H. Ji, M. Kuball, M. J. Uren, T. Martin, K. J. Nash, K. P. Hilton, and R. S. structure field effect transistors Applied Physics Letters vol. 88, no. 10, pp. 103502 103503, Mar. 2006. [91] C. F. Lo, L. Liu, T. S. Kang, R. Davies, B. P. Gila, S. J. Pearton, I. I. Kravchenko, O. Laboutin, Y. Cao, W. J. Johnson, and F. Re State Stress Critical Voltage by Using Pt Gated AlGaN/GaN High Electron Mobility Electrochemical and Solid State Letters vol. 14, no. 7, p. H264, 2011. [92] E. a. Douglas, C. Y. Chang, B. P. Gila, M. R. Holzworth, K. S Jones, L. Liu, J. temperature during off state degradation of AlGaN/GaN High Electron Mobility Microelectronics Reliability vol. 52, no. 1, pp. 23 28, Jan. 2012. [93] barrier height variations in Au W(Ti)/n Journal of Vacuum Science and Technology vol. 14, no. 4, p. 939, Jul. 1977. [94] N. G. Weimann, L. F. Eastman, D. Doppalapudi, H. M. Ng, and T. D. Moustakas, Journal of Applied Physics vol. 83, no. 7, p. 3656, 1998. [95] J. H. You, J. dislocations in n Journal of Applied Physics vol. 99, no. 3, p. 033706, 2006. [96] J. W. P. Hsu, M. J. Manfra, D. V. Lang, S. Richter, S. N. G. Chu, a. M. Sergent, R.

PAGE 138

138 of reverse bias leakage Applied Physics Letters vol. 78, no. 12, p. 1685, 2001. [97] Schottky Barrier Diodes with Multi Journal of The Electrochemic al Society vol. 155, no. 10, p. H716, 2008. [98] Y. Y. Wong, E. Y. Chang, T. H. Yang, J. R. Chang, J. T. Ku, M. K. Hudait, W. C. Chou, M. Chen, and K. Properties of AlGaN/GaN Heterostructure Grown Journal of The Electrochemical Society vol. 157, no. 7, p. H746, 2010. [99] the electrostatic potential of charged dislocations in n GaN and n ZnO epilayers by tran Physical Review B vol. 73, no. 24, p. 245316, Jun. 2006. [100] related Journal of Applied Physics vol. 102 no. 2, p. 026103, 2007. [101] Physical Review Letters vol. 87, no. 20, p. 205504, Oct. 2001. [102] Y. Kamimura, T. Yokoyama, H. Oiwa, K. Edagawa, and I. Yonenaga, IOP Conference Series: Materials Science and Engineering vol. 3, p. 012010, Jul. 2009. [103] imaging of reverse bias leakage through pure screw dislocations in GaN films Applied Physics Letters vol. 81, no. 1, p. 79, 2002. [104] elec troluminescence, gate current leakage, and surface defects in AlGaN/GaN high electron mobility transistors upon off Applied Physics Letters vol. 101, no. 3, p. 033508, 2012. [105] Y. Kamimura, T. Yokoyama, H. Oiwa, K. Edagawa, and I. Yonena IOP Conference Series: Materials Science and Engineering vol. 3, p. 012010, Jul. 2009. [106] Ge superlattices Journal of Applied Physics vol. 65, no. 12, p. 4693, 1989.

PAGE 139

139 [107] strained layer GexSi1 Physical review. B, Condensed matter vol. 31, no. 6. pp. 4063 4065, 15 Mar 1985. [108] x]Al[x]Sb strained layer Applied Physics Letters vol. 60, no. 8, pp. 998 1000 Feb. 1992. [109] F. Romanato, E. Napolitani, A. Carnera, A. V Drigo, L. Lazzarini, G. Salviati, C. In[x]Ga[1 Journal of Applied Physics vol. 86, n o. 9, pp. 4748 4755, Nov. 1999. [110] H. Zogg, S. Blunier, A. Fach, C. Maissen, P. Mller, S. Teodoropol, V. Meyer, G. mismatch strain relaxation in epitaxial CaF2, BaF2/CaF2, and PbSe/BaF2/CaF2 layers on Si(1 11) after many Physical review. B, Condensed matter vol. 50, no. 15. pp. 10801 10810, 15 Oct 1994. [111] Journal of Applied Physics vol. 34, no. 1, p. 123, 1963 [112] Journal of Crystal Growth vol. 27, no. 0, pp. 118 125, Dec. 1974. [113] cal Journal of Applied Physics vol. 64, no. 10, p. 4968, 1988. [114] mismatch for Ge[x]Si[1 x]/Si strained Applied Physics Letters vol. 47, no. 3, pp. 322 324, Aug. 1985. [115] EPITAXIAL Zeitschrift fr Kristallographie 1977, vol. 146, no. 1 3, pp. 152 153. [116] V. I. Vdovin, L. a. Matveeva, G. N. Semen ova, M. Y. Skorohod, Y. a. Tkhorik, and Physica Status Solidi (a) vol. 92, no. 2. pp. 379 390, 16 Dec 1985. [117] M. E. Twigg, N. D. Bassim, M. a. Mastro, C. R. Eddy, R. L. Henry, J. C. and dislocation motion in III nitride films grown on stepped and step free 4H SiC Journal of Applied Physics vol. 101, no. 5, p. 053509, 2007.

PAGE 140

140 [118] layer semiconductor Applied Physics Letters vol. 51, no. 17, pp. 1325 1327, Oct. 1987. [119] c Flow in the Diamond Solid State Physics vol. Volume 22, D. T. and H. E. B. T. S. S. P. Frederick Seitz, Ed. Academic Press, 1969, pp. 27 158. [120] D. Maroudas, L. a. Zepeda relaxation through misfit dislocation formation in the growth of epitaxial films on Applied Physics Letters vol. 73, no. 6, p. 753, 1998. [121] semiconductor crystals grown f rom the melt: analysis of the Haasen model for Journal of Crystal Growth vol. 108, no. 1 2, pp. 399 415, Jan. 1991. [122] compound semiconductor light em Applied Physics Letters vol. 70, no. 10, p. 1317, 1997. [123] Philosophical Magazine Series 4 vol. 10, no. 63, pp. 30 39, Jul. 1855. [124] R. C. Jaeger, Introduction to Microelectronic Fabrication Boston, MA, USA: Addison Wesley Longman Publishing Co., Inc., 1987. [125] Ge Journal of Applied Physics vol. 65, no. 12, pp 4693 4698, Jun. 1989. [126] Transactions Of The Metallurgical Society Of AIME vol. 242, no. 2, p. 166 &, 1968. [127] ultilayer Applied Physics Letters vol. 47, no. 3, pp. 234 236, Aug. 1985. [128] N. E. B. Cowern, P. C. Zalm, P. van der Sluis, D. J. Gravesteijn, and W. B. de Physical Review Letters vol. 72, no. 16, pp. 2585 2588, Apr. 1994. [129] A. Pruijmboom, C. E. Timmering, J. M. L. van Rooij Mulder, D. J. Gravesteijn, W. B. de Boer, W. J. Kerste n, J. W. Slotboom, C. J. Vriezema, and R. de Kruif, Microelectronic Engineering vol. 19, no. 1 4, pp. 427 433, Sep. 1992.

PAGE 141

141 [130] N. Moriya, L. C. Feldman, H. S. Luftman, C. A. King, J. Bevk, and B. F reer, Physical Review Letters vol. 71, no. 6, pp. 883 886, Aug. 1993. [131] S. J. Pearton, H. Cho, J. R. LaRoche, F. Ren, R. G. Wilson, and J. W. Lee, capped Ga Applied Physics Letters vol. 75, no. 19, pp. 2939 2941, Nov. 1999. [132] F. Gao, D. Chen, B. Lu, H. L. Tuller, C. V Thompson, S. Keller, U. K. Mishra, and Co Electron Device Letters, IEEE vol. 33, no. 10. pp. 1378 1380, 2012. [133] of oxygen in the OFF state degradation of AlGaN/GaN high electron mobil ity Applied Physics Letters vol. 99, no. 22, pp. 223503 223506, Nov. 2011. [134] Deutsche Chemische Gesellschaft., R. J. Meyer, F. Peters, L. Gmelin, E. H. E. Pietsch, K. C. Buschbeck, and M. Becke Goehring, Gmelins Handbuch der anorganische n Chemie Weinheim/Bergstr.: Verl. Chemie, 1959. [135] Journal of Applied Physics vol. 36, no. 1, pp. 153 156, Jan. 1965. [136] K. Harafuji and Japanese Journal of Applied Physics vol. 44, no. 9A, p. 6495 LP 6504. [137] ANSI/IEEE Std 176 1987 pp. 0 66, 1988. [138] L. B. Freund and S. Suresh, Thin Film Materials Cambridge University Press, 2004. [139] the tunneling component in the current -voltage characteristics of metal GaN Journal of Applied Physics vol. 84, no. 4, pp. 2099 2104, Aug. 1998. [140] [141] S. Bloom, G. Harbeke, E. Mei physica status solidi (b) vol. 66, no. 1, pp. 161 168, 1974.

PAGE 142

142 [142] Y. Physical Review B vol. 48, no. 7, pp. 4335 4351, Aug. 1993.

PAGE 143

143 BIOGRAPHICAL SKETCH Amit Gupta was born in Luc k now, India, in the year 198 4 He received his Bachelor of Technology (B. Tech.) degree in electronics and communication e ngineering from the National Institute of Technology, Warangal, India in 2007. During the undergraduate studies, he interned at the Indian Institute of Technology (IIT), Kanpur, India during the summer of 2005 and at the Indian Institute of Science (IISc), Bangalore, India during the summer of 2006. He joined the Interdisciplinary Microsystems Group (IMG) at the University of Florida in August, 2008 and started working as a graduate research assistant for his advisor, Dr. Toshi Nishida, to pursue his doctoral research in semiconductor device phy sics. He earned his Mast er of Science (M.S.) degree in electrical and computer e ngineering with a minor in Physics from the University of Florida in 2010. He received his Doctor of Philosophy (PhD) in electrical and computer e ngineering from the University of Florida in 2013. His PhD research was focused on investigati ng the reliability of AlG a N/G a N HEMTs His technical expertise is in semiconductor device physics focusing on electrical characterization, modeling and reliability.