• unlimited access with print and download
    $ 37 00
  • read full document, no print or download, expires after 72 hours
    $ 4 99
More info
Unlimited access including download and printing, plus availability for reading and annotating in your in your Udini library.
  • Access to this article in your Udini library for 72 hours from purchase.
  • The article will not be available for download or print.
  • Upgrade to the full version of this document at a reduced price.
  • Your trial access payment is credited when purchasing the full version.
Buy
Continue searching

Investigation of gallium nitride transistor reliability through accelerated life testing and modeling

ProQuest Dissertations and Theses, 2011
Dissertation
Author: Bradley D Christiansen
Abstract:
Gallium nitride (GaN) high electron mobility transistors (HEMT) are attractive to the United States Department of Defense for their ability to operate at high frequencies, voltages, temperatures, and power. Yet, there are concerns about the reliability, or short lifetimes, of these devices. Various degradation mechanisms and their causes are proposed in the literature. A variety of reliability tests were conducted to understand these mechanisms and causes. A multi-stressor experiment was performed on AlGaN/GaN HEMTs with high voltage and high power as stressors. The devices tested under high power generally degraded more than those tested under high voltage. In particular, the devices tested at high voltage in the OFF state did not degrade significantly as suggested by some papers in the literature. The same papers in the literature also suggest that high voltages cause cracks and pits in the AlGaN barrier layer. However, the high-voltage-tested devices in this study do not exhibit cracks or pits in transmission electron microscope images, while the high-power-tested devices do exhibit pits. The validity of Arrhenius accelerated-life testing when applied to GaN HEMT lifetime assessments was investigated. Temperature alone could not explain the differences in observed degradation. GaN HEMT reliability evaluations will benefit if other accelerants, such as voltage, are used. Such evaluations will consider failure mechanisms that are not primarily thermally accelerated in the complex electrothermomechanical system that is GaN. Reports to date of GaN HEMTs subjected to forward gate bias stress include varied extents of degradation. Reported herein is an extremely robust GaN HEMT technology that survived high forward gate bias (+6 V) and current (>1.8 A/mm) for >17.5 hours, exhibiting only a slight change in gate diode characteristic, little decrease in maximum drain current, with only a 0.1-V positive threshold voltage shift, and, remarkably, a persisting breakdown voltage exceeding 200 V. Several experiments to examine the time-dependence of GaN HEMT degradation were performed. The data fit best to an exponential model, unlike other reports. Also discovered was that the characterization temperature affects the level of degradation observed. Results of device testing under continuous- and pulsed-direct current (DC) stressing were compared. The comparison indicates that a pulse width of sufficient brevity is less stressful than continuous DC, possibly due to the device not reaching a higher steady-state channel temperature within the pulse ON time. For longer pulse widths that may attain the higher steady-state channel temperature, thermal cycling between the extremes of the temperature range may induce more degradation than continuous DC.

ix Table of Contents Page Abstract .............................................................................................................................. iv   Table of Contents ............................................................................................................... ix   List of Figures .................................................................................................................. xiv   List of Tables ................................................................................................................... xxi   List of Abbreviations ..................................................................................................... xxiv   I. Introduction .....................................................................................................................1   1.1.   Motivation ........................................................................................................1   1.1.1.   Desirable Performance Attributes ........................................................... 2   1.1.2.   Circuits and Applications ........................................................................ 3   1.2.   Accelerated Testing Research ..........................................................................3   1.2.1.   Problem Statement .................................................................................. 4   1.2.2.   Thesis Statement ...................................................................................... 4   1.2.3.   Contributions ........................................................................................... 4   1.3.   Modeling Research ..........................................................................................5   1.3.1.   Problem Statement .................................................................................. 5   1.3.2.   Thesis Statement ...................................................................................... 5   1.3.3.   Contribution ............................................................................................. 6   1.4.   Publications ......................................................................................................6   1.5.   Purpose .............................................................................................................7   1.6.   Document Overview ........................................................................................7   II. Background ....................................................................................................................9   2.1.   Brief History ....................................................................................................9   2.2.   Definitions......................................................................................................10  

x 2.3.   Fabrication Processes .....................................................................................14   2.4.   GaN HEMT Structure ....................................................................................17   2.5.   Physics ...........................................................................................................19   2.5.1.   Basic Operation ..................................................................................... 19   2.5.2.   DC Performance .................................................................................... 19   2.5.3.   AC Performance .................................................................................... 23   2.6.   Accelerated Life Testing ................................................................................25   2.6.1.   Types of Stress Testing ......................................................................... 25   2.6.2.   Causes of GaN HEMT Failure in the Literature ................................... 26   2.6.3.   Arrhenius Relationship .......................................................................... 32   2.6.4.   Eyring Model ......................................................................................... 34   2.6.5.   Parameter Definitions and Failure Criteria ........................................... 35   2.7.   Modeling ........................................................................................................36   2.7.1.   Recent Examples ................................................................................... 36   2.7.2.   Current Industry Software – Synopsys TCAD ...................................... 37   2.8.   Chapter Summary ..........................................................................................38   III. Modeling Research .....................................................................................................39   3.1.   Introduction ....................................................................................................39   3.2.   General Motivation for Modeling Microelectronic Devices .........................39   3.3.   Approach ........................................................................................................41   3.4.   Modeling at AFIT ..........................................................................................43   3.5.   Future Work ...................................................................................................47   3.6.   Chapter Summary ..........................................................................................48  

xi IV. Experimental Procedures ............................................................................................49   4.1.   Introduction ....................................................................................................49   4.2.   Voltage Step-Stress ........................................................................................49   4.3.   300-hour Test .................................................................................................51   4.4.   1000-hour Test ...............................................................................................54   4.5.   600-hour Test .................................................................................................54   4.6.   Gate Bias Test ................................................................................................55   4.7.   Chapter Summary ..........................................................................................57   V. Reliability Testing of AlGaN/GaN HEMTs under Multiple Stressors ........................58   5.1.   Introduction ....................................................................................................58   5.2.   Experiment Description .................................................................................58   5.3.   Results and Discussion ..................................................................................59   5.4.   Conclusion .....................................................................................................66   VI. Benefits of Considering More Than Temperature Acceleration for GaN HEMT Life Testing................................................................................................................................67   6.1.   Introduction ....................................................................................................67   6.2.   Experiment Description .................................................................................70   6.3.   Results and Discussion ..................................................................................70   6.3.1.   300-hour Test ........................................................................................ 70   6.3.2.   600-hour High-Power Test .................................................................... 83   6.3.3.   Discussion ............................................................................................. 87   6.4.   Conclusion .....................................................................................................89   VII. A Very Robust AlGaN/GaN HEMT Technology to High Forward Gate Bias and Current ...............................................................................................................................91  

xii 7.1.   Introduction ....................................................................................................91   7.2.   Experiment Description .................................................................................92   7.3.   Results ............................................................................................................92   7.4.   Conclusion .....................................................................................................95   VIII. Time-dependent Electrical Degradation of AlGaN/GaN HEMTs Subjected to High DC Power and High Drain Bias .........................................................................................97   8.1.   Introduction ....................................................................................................97   8.2.   Experiment Description .................................................................................97   8.3.   Results and Discussion ..................................................................................98   8.3.1.   Voltage Step-Stress Test ....................................................................... 98   8.3.2.   1000-hour Test .................................................................................... 101   8.3.3.   600-hour Test ...................................................................................... 108   8.3.4.   Discussion ........................................................................................... 112   8.4.   Conclusion ...................................................................................................112   IX. Comparison of Pulsed- and Continuous-DC Stressing of AlGaN/GaN HEMTs .....114   9.1.   Introduction ..................................................................................................114   9.2.   Experiment Description ...............................................................................114   9.3.   Results and Discussion ................................................................................115   9.4.   Conclusion ...................................................................................................116   X. Conclusions ................................................................................................................118   10.1.   Overall Summary .........................................................................................118   10.2.   Contributions................................................................................................118   10.3.   Ideas for Future Research ............................................................................120   Appendix A. Pictorial Presentation of Test Preparation and Setup ................................124  

xiii Appendix B. Scripts, Command Files, and Macros ........................................................127   B.1.   Modeling ......................................................................................................127   B.1.1.   Unix script .......................................................................................... 127   B.1.2.   Sentaurus Structure Editor command files ......................................... 127   B.1.3.   Sentaurus Device command file ......................................................... 138   B.1.4.   Sentaurus Device parameter file ......................................................... 139   B.1.5.   Tecplot macro ..................................................................................... 140   B.1.6.   MATLAB files.................................................................................... 140   B.2.   Data Processing ............................................................................................142   B.2.1.   Unix scripts ......................................................................................... 142   B.2.2.   MATLAB files.................................................................................... 143   B.3.   Voltage step-stress Excel macro ..................................................................146   B.4.   Accel-RF test sequence ................................................................................155   Appendix C. Data Tables ................................................................................................160   Bibliography ....................................................................................................................172   Vita ..................................................................................................................................181  

xiv List of Figures Page Figure 1. Relative size comparison of a transistor (left) and a vacuum tube (Nobelprize.org, 2010) ................................................................................................. 3   Figure 2. Cell phone base station (Statemaster.com, 2010) ............................................... 4   Figure 3. National Missile Defense X-band radar (U.S. DoD, 2010) ................................ 4   Figure 4. AlGaN/GaN HEMT (after (Liddle, 2008)) ...................................................... 11   Figure 5. Energy band diagram of AlGaN/GaN hetero-junction with 2DEG area identified (after (McClory, 2008)) ............................................................................. 11   Figure 6. Illustrations of (a) lattice-matched, (b) strained, and (c) unstrained (relaxed) heteroepitxial structures (May, 2004:153) ................................................................. 12   Figure 7. CdS wurtzite structure with lattice constants a and c (Ullrich, 2010) .............. 13   Figure 8. Calculated (solid line) lattice constant (a in (a) and c in (b)) dependence on aluminum composition (x) in Al x Ga 1-x N alloys (Dridi, 2010) ................................... 14   Figure 9. MOCVD reactor (EENG 596, 2008:slide 9) .................................................... 15   Figure 10. MOCVD multiwafer processing (EENG 596, 2008:slide 13) ........................ 15   Figure 11. MBE system (May, 2004:149) ....................................................................... 15   Figure 12. Typical HEMT creation steps (Ali, 1991:85) ................................................. 16   Figure 13. (a) Cross-sectional view of an undercut T-shaped resist cavity with a 0.15-μm bottom opening. Three layers of electron-beam resist are used to form the cavity. (b) Submicron T-gate on the channel of a HEMT after removing the trilayer resist structure shown in (a). (Ali, 1991:88) ....................................................................... 17   Figure 14. AlGaN/GaN HEMT with gate-integrated and source-connected field plates 18  

xv Figure 15. Drain current versus drain voltage at multiple values of gate voltage for an NMOSFET (Circuits Today, 2010) ............................................................................ 21   Figure 16. Drain current versus drain voltage at multiple values of gate voltage for a HEMT used in this research ....................................................................................... 22   Figure 17. Transfer curve for an NMOSFET (after (Circuits Today, 2010)) .................. 23   Figure 18. Transfer curve for a HEMT used in this research .......................................... 23   Figure 19. Transfer and transconductance (g m ) curves for a HEMT used in this research .................................................................................................................................... 24   Figure 20. Pictorial representation of lattice disruption creation by an etching process as proposed by Smith et al. (Smith, 2009). .................................................................... 27   Figure 21. (a) Change in normalized I Dmax in step-stress experiments for three different stress conditions. Dashed line represents the estimated change in I Dmax in the high- power state removing the effect of V T change. (b) Change in the gate leakage current I Goff (gate current at V DS = 0.1 V and V GS = −5 V) in the same experiment (Joh, 2008). .................................................................................................................................... 28   Figure 22. (a) and (b) Cross-sectional HREM and (c) Z-contrast images of three stressed devices. Material below the horizontal interface is semiconductor; the trapezoidal shape defines the gate metal. Right side is toward the drain, and left side is toward the source in all three images. (a) shows the formation of pits on both the source- and drain-side edges of the gate, (b) shows the formation of a crack, and (c) shows a severe case of degradation where the gate metal (Pt) has diffused into the crack formed (Chowdhury, 2008). (b) and (c) have roughly the same scale. ..................... 29  

xvi Figure 23. Average percent decrease of the maximum transconductance measured at V DS

= 10 V during 10-hour ON-state tests (V DS = 20 V, V GS = 0 V; diamonds), OFF-state tests (V DS = 20 V, V GS = −7.7 V; squares), and semi-ON-state tests (V DS = 20 V, V GS = −5.5 V; triangles) (Meneghesso, 2008) ...................................................................... 30   Figure 24. Atomic configurations of triply hydrogenated (a) gallium vacancy,(b) nitrogen antisite, and (c) divacancy. (Puzyrev, 2011) ............................................... 31   Figure 25. Electron tunneling leakage from the gate electrode and possible current paths (Trew, 2009) ............................................................................................................... 32   Figure 26. Flow chart for automated device modification. (Coutu, 2011) ..................... 42   Figure 27. Manual electromigration example. Current densities range from 1 kA/cm 2

(and less) in blue to 100 kA/cm 2 (and greater) in red. The last frame (not shown) is a complete void. ............................................................................................................ 44   Figure 28. Automated electromigration example. Current densities range from 1 kA/cm 2

(and less) in blue to 100 kA/cm 2 (and greater) in red. ............................................... 46   Figure 29. Transfer and transconductance curves at 0 and 300 hours of typical high- voltage-tested device. Device 7579 was tested at Condition 6. (Christiansen, 2011b) .................................................................................................................................... 61   Figure 30. Transfer and transconductance curves at 0, 300, and 1016 hours of exceptional high-voltage-tested device. Device 001 was tested at Condition 6. (Christiansen, 2011b) ................................................................................................. 61   Figure 31. Representative transfer and transconductance curves at 0, 300, and 343 hours of high-power-tested device. Device 007 was tested at Condition 1. (Christiansen, 2011b) ........................................................................................................................ 62  

xvii Figure 32. Device 001 (high-voltage-tested) at a baseplate of 85 °C. The upper middle spot was targeted for TEM imaging. (a) IR (radiance) image at 15X magnification. V DS = 40 V, I D = 10 mA, V GS = −2.42 V, I G = −5 μA. (b) PE image at 20X magnification. V DS = 100 V, I D = 11 μA, V GS = −10 V, I G = −12 μA. (Christiansen, 2011b) ........................................................................................................................ 64   Figure 33. Device 007 (high-power-tested) at a baseplate of 85 °C. The lower left spot was targeted for TEM imaging. (a) IR (radiance) image at 15X magnification. V DS = 28 V, I D = 10 mA, V GS = −1.69 V, I G = −3.3 μA. (b) PE image at 50X magnification. V DS = 10 V, I D = 3.2 mA, V GS = −1 V, I G in nA range. (Christiansen, 2011b) .......... 65   Figure 34. TEM images of (a) Device 001 (high-voltage-tested) and (b) Device 007 (high-power-tested). Notice absence of a pit or crack in Device 001 and the presence of a pit in Device 007. (Christiansen, 2011b) ........................................................... 65   Figure 35. Normalized pre- and post-stress values of (a) I Dmax and (b) I DSS for high-power conditions. The top three lines (red) of the legend are Condition 1, the middle three (blue) Condition 2, and the bottom three (green) Condition 3. (Christiansen, 2011a) .................................................................................................................................... 74   Figure 36. Normalized pre- and post-stress values of (a) I Dmax and (b) I DSS for high- voltage conditions. The top two lines (red) of the legend are Condition 4 and the bottom two (blue) Condition 6. (Christiansen, 2011a) ............................................. 76   Figure 37. Comparing acceleration factors based on initial T ch estimates and observed degradation. The reference slope line assumes an activation energy of 2.09 eV. (Christiansen, 2011a) ................................................................................................. 78  

xviii Figure 38. Comparing Agilent power supply measurement error and initial T ch estimates in 300-hour test. (Christiansen, 2011a) ..................................................................... 80   Figure 39. Magnified portion of Figure 38 but on different scales. (Christiansen, 2011a) .................................................................................................................................... 81   Figure 40. Normalized values (to the 1-hour, 245-°C measurements) of I Dmax over time during the 600-hour test. (Christiansen, 2011a) ........................................................ 84   Figure 41. Normalized values (to the 0-hour, 70-°C measurements) of I Dmax over time during the 300-hour, Condition-1 (lines 2-4 in the legend, in red) and 600-hour, Condition-7 (lines 5-7 in the legend, in green) tests. (Christiansen, 2011a) ............. 85   Figure 42. Comparing Agilent power supply measurement error and initial T ch estimates in Conditions 1 and 7. (Christiansen, 2011a) ............................................................ 87   Figure 43. Transfer curves (a), and associated gate current in absolute value (b) and transconductance (c) of the device as measured during characterizations between gate stressing events. Insets show detail at regions of interest in the same data sets. Extra gate current is seen in (b) above V G ≈ −3.5 V after 210 minutes stress (top curve) that is not seen after longer stress time (second-to-top curve). It is not known if there was a temporary test issue or if that is indeed real. ........................................................... 94   Figure 44. Gate diode curves during the stressing. Insets show additional detail for regions of interest of the same curves as the main plot and share the same units (i.e., mA/mm and V) as the main figure. The data was collected at stress times represented in Figure 43. Black curves represent the initial Vg = +2.5 V gate stresses. Red curves represent the gate voltage stress ramps of increasing magnitude collected just prior to the red curves of Figure 43. Green curves are gate voltage

xix stress ramps collected just after the total stress times represented by the green curves of Figure 43. (Christiansen, 2011c) ........................................................................... 95   Figure 45. Drain and gate current of Device H11V05R during voltage step-stress test. . 98   Figure 46. Evolution of I Dmax , I DSS , and I G of Device H11V05R during voltage step-stress test .............................................................................................................................. 99   Figure 47. Evolution of I Dmax and |I Goff | for Device R10C2 during del Alamo’s “V DS = 0 step-stress experiment” ............................................................................................ 100   Figure 48. Evolution of I Dmax and |I Goff | for Device R9C2 during V DG step-stress experiment (a modified del Alamo test) ................................................................... 101   Figure 49. Normalized values (to the 1-hour measurements) of I Dmax over time during the 1000-hour test. The top three lines (red) of the legend are Condition 4, the next line (blue) is Condition 5, the next two (green) are Condition 6, and the bottom line is the model with the means of curve-fitting parameters. .................................................. 103   Figure 50. Fits of Channel 15 normalized I Dmax data to Equation (19) in blue (fit 166) and to the stretched exponential in red (fit 165). ............................................................ 105   Figure 51. Last hours of 1000-hour test showing the similarity of devices’ responses to OFF-state stress independent of applied drain voltage. ........................................... 108   Figure 52. Normalized values (to the 1-hour measurements) of I Dmax over time during the 600-hour test. ............................................................................................................ 109   Figure 53. Fits of Channel 26 normalized I Dmax data to Equation (19) in blue (fit 188) and the stretched exponential in red (fit 187). ................................................................ 110   Figure 54. Packaged devices as sent from vendor ......................................................... 124   Figure 55. Packaged devices with lead frames removed ............................................... 124  

xx Figure 56. Packaged device attached to test module ..................................................... 125   Figure 57. Heater bar for three test modules .................................................................. 125   Figure 58. Clamps attached above devices and test modules attached to heater bar ..... 125   Figure 59. Test module covers and nitrogen tubes attached .......................................... 126   Figure 60. DC test station .............................................................................................. 126   Figure 61. Thermal and photoemission imaging system ............................................... 126  

xxi List of Tables

Page Table 1. Definitions of Experimentally Measured or Derived Parameters (Via, 2010) .. 35   Table 2. Failure Criteria ................................................................................................... 35   Table 3. Test Conditions for 300-hour, 1000-hour, and 600-hour Tests (Christiansen, 2011a) ......................................................................................................................... 52   Table 4. Average Absolute Percentage Changes in Parameters after 300 Hours (Christiansen, 2011b) ................................................................................................. 60   Table 5. Initial Variation of Parts Tested for 300 Hours (Christiansen, 2011a) .............. 71   Table 6. Parameter Changes by Device and Condition in 300-hour Test (Christiansen, 2011a) ......................................................................................................................... 72   Table 7. Arrhenius Acceleration Factors Between Test Conditions (Christiansen, 2011a) .................................................................................................................................... 72   Table 8. Slopes of Normalized I Dmax and I DSS Lines for High-power Conditions (Christiansen, 2011a) ................................................................................................. 75   Table 9. Slopes of Normalized I Dmax and I DSS Lines for High-voltage Conditions (Christiansen, 2011a) ................................................................................................. 76   Table 10. Initial Measured Parameter Values in 300-hour Test (Christiansen, 2011a) ... 77   Table 11. Sensitivity Analysis of Thermal Model R th for Conditions 1, 2, and 3 (Christiansen, 2011a) ................................................................................................. 82   Table 12. Initial Variation of Parts Tested for 600 Hours (Christiansen, 2011a) ............ 83   Table 13. Parameter Changes by Device and Stress Time Y in 600-hour Test (Christiansen, 2011a) ................................................................................................. 83  

xxii Table 14. Initial Measured Parameter Values in 600-hour Test (Christiansen, 2011a) ... 85   Table 15. Parameter Percentage Changes by Channel and Condition in 1000-hour Test (T bp = 245 °C; t = 1 hour vs. t = 1000 hours) ........................................................... 102   Table 16. Model Parameter Solutions and Statistics for Fitting 1000-hour Normalized I Dmax Data to Equation (19) ...................................................................................... 106   Table 17. Parameter Percentage Changes by Channel and Condition in 1000-hour Test (T bp = 245 °C; t = 0 hours vs. t = 1000 hours) .......................................................... 106   Table 18. Parameter Percentage Changes by Channel and Condition in 1000-hour Test (T bp = 70 °C; t = 0 hours vs. t = 1000 hours) ............................................................ 107   Table 19. Parameter Percentage Changes by Channel and Stress Time in 600-hour Test (T bp = 245 °C; t = 1 hour vs. t = Y hours) ................................................................ 109   Table 20. Model Parameter Solutions and Statistics for Fitting 600-hour Normalized I Dmax Data to Equation (19) ...................................................................................... 111   Table 21. Parameter Percentage Changes by Channel and Stress Time in 600-hour Test (T bp = 245 °C; t = 0 hours vs. t = Y hours) ............................................................... 111   Table 22. Parameter Percentage Changes by Channel and Stress Time in 600-hour Test (T bp = 70 °C; t = 0 hours vs. t = Y hours) ................................................................. 112   Table 23. Average I Dmax Degradation in Continuous- and Pulsed-DC, Same-Bias Tests .................................................................................................................................. 116   Table 24. Absolute Percentage Changes in Parameters after 300 Hours (see Table 4) . 161   Table 25. Parameter Changes by Device and Condition in 300-hour Test (see Table 6) .................................................................................................................................. 162  

xxiii Table 26. Slopes of Normalized I Dmax and I DSS Lines for High-power Conditions (see Table 8) .................................................................................................................... 163   Table 27. Slopes of Normalized I Dmax and I DSS Lines for High-voltage Conditions (see Table 9) .................................................................................................................... 163   Table 28. Parameter Changes by Device and Stress Time Y in 600-hour Test (see Table 13) ............................................................................................................................ 164   Table 29. Parameter Percentage Changes by Channel and Condition in 1000-hour Test (T bp = 245 °C; t = 1 hour vs. t = 1000 hours) (see Table 15) ................................... 165   Table 30. Parameter Percentage Changes by Channel and Condition in 1000-hour Test (T bp = 245 °C; t = 0 hours vs. t = 1000 hours) (see Table 17) .................................. 166   Table 31. Parameter Percentage Changes by Channel and Condition in 1000-hour Test (T bp = 70 °C; t = 0 hours vs. t = 1000 hours) (see Table 18) .................................... 167   Table 32. Parameter Percentage Changes by Channel and Stress Time in 600-hour Test (T bp = 245 °C; t = 1 hour vs. t = Y hours) (see Table 19) ........................................ 168   Table 33. Parameter Percentage Changes by Channel and Stress Time in 600-hour Test (T bp = 245 °C; t = 0 hours vs. t = Y hours) (see Table 21) ....................................... 169   Table 34. Parameter Percentage Changes by Channel and Stress Time in 600-hour Test (T bp = 70 °C; t = 0 hours vs. t = Y hours) (see Table 22) ......................................... 170   Table 35. I Dmax Degradation in Continuous- and Pulsed-DC, Same-Bias Tests (see Table 23) ............................................................................................................................ 171  

xxiv List of Abbreviations

2DEG Two-Dimensional Electron Gas AFRL Air Force Research Laboratory Al Aluminum AlGaAs Aluminum Gallium Arsenide AlGaN Aluminum Gallium Nitride AlN Aluminum Nitride Au Gold C Carbon Cd Cadmium CdS Cadmium Sulfide CMOS Complementary Metal Oxide Semiconductor DC Direct Current DoD Department of Defense EBL Electron-Beam Lithography EDS Energy-Dispersive X-Ray Spectroscopy EELS Electron Energy Loss Spectroscopy FET Field-Effect Transistor Ga Gallium GaAs Gallium Arsenide GaN Gallium Nitride HEMT High Electron Mobility Transistor IR Infrared

xxv MBE Molecular Beam Epitaxy MOCVD Metal-Organic Chemical Vapor Deposition MURI Multidisciplinary University Research Initiative N Nitrogen NAVSEA Naval Surface Warfare Center Ni Nickel NMOSFET N-channel Metal-Oxide-Semiconductor FET NPS Naval Postgraduate School O Oxygen ONR Office of Naval Research Pd Palladium PCA Pulsed Condition A PCB Pulsed Condition B PCC Pulsed Condition C PE Photoemission P-HEMT Pseudomorphic HEMT RF Radio Frequency RYD Aerospace Components and Subsystems Division S Sulfur Si Silicon SiC Silicon Carbide SiN Silicon Nitride SEM Scanning Electron Microscope

xxvi SMU Source/Monitor Unit TCAD Technology Computer-Aided Design TEM Transmission Electron Microscope UV Ultraviolet

1 INVESTIGATION OF GALLIUM NITRIDE TRANSISTOR RELIABILITY THROUGH ACCELERATED LIFE TESTING AND MODELING

I. Introduction The material properties of gallium nitride (GaN) enable the production of high electron mobility transistors (HEMT) with characteristics attractive to the United States Department of Defense (DoD) for application in communications and sensing systems. Interest in this technology is demonstrated by the Defense Advanced Research Projects Agency’s Wide Bandgap Semiconductor initiative and by the Multidisciplinary University Research Initiatives (MURI) funded by the Office of Naval Research (ONR) and Air Force Office of Scientific Research. Despite the advantages of GaN HEMTs, there is concern that their reliability is low or, in other words, they do not have sufficiently long lifetimes for military systems. This concern has hampered their widespread acceptance and use (Christiansen, 2011b). Determining GaN HEMT lifetimes is usually accomplished with life testing that is accelerated with temperature, although there may be other stressors that hasten transistor failure. The lifetimes calculated at operating temperatures are estimates since they are extrapolated from the lifetimes at high temperatures. Modeling and simulation of GaN HEMTs is an alternative to life testing for estimating lifetimes. 1.1. Motivation This section presents several reasons why the knowledge and research of GaN HEMTs is important to the DoD.

2 1.1.1. Desirable Performance Attributes The physical attributes of HEMTs result in desirable performance in high-speed, high-temperature, high-voltage, and high-power applications. High speed, usually expressed by a frequency, results from the high electron mobility. The intended use of the first HEMT built (by Mimura et al.) was high-speed digital applications at low temperatures (Ali, 1991:91). HEMTs have been used in flip-flop circuits operated at 5.5 GHz and 300 K (Neaman, 2003:608). GaN HEMTs have demonstrated high- frequency operation: 190 GHz for unity current gain cut-off frequency (f T ) and 251 GHz for maximum frequency of oscillation (f max ) (Higashiwaki, 2008). High-temperature operation results from the wide bandgap semiconductor of HEMTs. This high-temperature operation reduces the cooling requirement (Mishra, 2002). Although gallium arsenide (GaAs) was the wide bandgap material used for the first HEMT, GaN is now being employed in HEMTs for power applications. For a given doping concentration, GaN has a breakdown voltage that is an order of magnitude greater than that of GaAs (Liddle, 2008). GaN also has high current capacity resulting from additional carriers created by spontaneous and piezoelectric polarization effects (Sze, 2007:409). The high breakdown voltage and high current enable high power operation. With these desirable attributes, HEMTs are beginning to compete with vacuum tubes that have dominated the areas of high RF power at high temperatures and frequencies (Trew, 2005). Compared to vacuum tubes, HEMTs also have a considerable size advantage as depicted in Figure 1.

Full document contains 210 pages
Abstract: Gallium nitride (GaN) high electron mobility transistors (HEMT) are attractive to the United States Department of Defense for their ability to operate at high frequencies, voltages, temperatures, and power. Yet, there are concerns about the reliability, or short lifetimes, of these devices. Various degradation mechanisms and their causes are proposed in the literature. A variety of reliability tests were conducted to understand these mechanisms and causes. A multi-stressor experiment was performed on AlGaN/GaN HEMTs with high voltage and high power as stressors. The devices tested under high power generally degraded more than those tested under high voltage. In particular, the devices tested at high voltage in the OFF state did not degrade significantly as suggested by some papers in the literature. The same papers in the literature also suggest that high voltages cause cracks and pits in the AlGaN barrier layer. However, the high-voltage-tested devices in this study do not exhibit cracks or pits in transmission electron microscope images, while the high-power-tested devices do exhibit pits. The validity of Arrhenius accelerated-life testing when applied to GaN HEMT lifetime assessments was investigated. Temperature alone could not explain the differences in observed degradation. GaN HEMT reliability evaluations will benefit if other accelerants, such as voltage, are used. Such evaluations will consider failure mechanisms that are not primarily thermally accelerated in the complex electrothermomechanical system that is GaN. Reports to date of GaN HEMTs subjected to forward gate bias stress include varied extents of degradation. Reported herein is an extremely robust GaN HEMT technology that survived high forward gate bias (+6 V) and current (>1.8 A/mm) for >17.5 hours, exhibiting only a slight change in gate diode characteristic, little decrease in maximum drain current, with only a 0.1-V positive threshold voltage shift, and, remarkably, a persisting breakdown voltage exceeding 200 V. Several experiments to examine the time-dependence of GaN HEMT degradation were performed. The data fit best to an exponential model, unlike other reports. Also discovered was that the characterization temperature affects the level of degradation observed. Results of device testing under continuous- and pulsed-direct current (DC) stressing were compared. The comparison indicates that a pulse width of sufficient brevity is less stressful than continuous DC, possibly due to the device not reaching a higher steady-state channel temperature within the pulse ON time. For longer pulse widths that may attain the higher steady-state channel temperature, thermal cycling between the extremes of the temperature range may induce more degradation than continuous DC.