U NIVERSIT A` DEGLI S TUDI DI PADOVA D IPARTIMENTO DI I NGEGNERIA DELL’I NFORMAZIONE
D OTTORATO DI R ICERCA IN I NGEGNERIA E LETTRONICA E DELLE T ELECOMUNICAZIONI XVIII C ICLO
Reliability of AlGaN/GaN HEMTs for RF and microwave applications
D OTTORANDO :
S UPERVISORE :
Alberto Sozza
Prof. Enrico Zanoni Prof. Nathalie Labat
C OORDINATORE : Ch.mo Prof. Silvano Pupolin
Padova, 31 Dicembre 2005
Posta elettronica :
[email protected]
Numero telefonico : +39 049 827 7786
Indirizzo : Dipartimento di Ingegneria dell’Informazione Universit`a degli Studi di Padova via Gradenigo 6/B 35131 Padova Italia
to my family
Abstract Gallium Nitride (GaN) Heterostructure Field Effect Transistors (HFETs or HEMTs) have demonstrated a great potential as possible candidates for the next generation of RF and Microwave power amplifiers. The material properties of GaN such as wide bandgap, high current densities, high breakdown field are the key parameters of its superiority over semiconductors such as Si or GaAs. The present work recaps general notions on GaN HEMTs reliability and describes accelerated life test and complementary measurements carried out at the University of Padova (Italy), at the University of Bordeaux (France), and in the R&D laboratory of THALES, namely the III-V lab (France). Presentations and discussion are specially focussed on the thermal storage and DC life test. The results of these tests bring information on the reliability of various metallization schemes under thermal stress and highlights the role of hot electrons in the degradation process of transistors in operation.
Sommario I transistori ad effetto di campo a eterostruttura (HFET o HEMT) in nitruro di gallio (GaN) hanno dimostrato forti potenzialit´a come possibili candidati per la prossima generazione di amplificatori di potenza a radiofrequenza e a microonde. Le propriet´a fisiche e elettroniche del nitruro di gallio quali l’ampio bandgap, le alte densit´adi corrente e gli alti valori del campo critico di breakdown sono i parametri chiave che ne determinano la superiorit´a su altri semiconduttori quali Si o GaAs. Il presente lavoro ricapitola alcune nozioni generali sull’affidabilit´a degli HEMTs in nitruro di gallio e descrive le prove accelerate di vita e le misure complementari effettuate presso l’universit´a di Padova (Italia), l’universit´a di Bordeaux (Francia) e i laboratorio di recerca e sviluppo THALES, pi´u precisamente l’unit´a di ricerca III-V lab (Francia). La presentazione e la discussione dei risultati sono specialmente incentrate sulla prova di storage termico e sugli stress in polarizzazione in continua. I risultati di queste prove danno informazioni sull’affidabilit´a dei vari schemi di metallizzazione sottoposti a storage termico e sottolineano il ruolo degli elettroni caldi nel processo di degradazione dei transistori in funzionamento.
Resume´ Les transistors a` effet de champ a` heterostructure (HFET or HEMT) en nitrure de gallium se pr´esentent comme des candidats a` fort potentiel pour la prochaine g´en´eration d’amplificateur de puissance dans les domaines des radio fr´equences et des ondes millim´etriques. Ce sont ses propri´et´es physique et e´ lectronique telle que une grande largeur de bande interdite, un potentiel pour de forte densit´es de courant et un tr´es fort champ de claquage qui le rendent sup´erieur aux semiconducteurs tels que le silicium et l’ars´eniure de gallium. Ce manuscrit reprend des notions g´en´erales sur la fiabilit´e des HEMTs en GaN et d´ecrit certains tests de vieillissement acc´el´er´e et certaines mesures compl´ementaires r´ealis´es en collaboration avec l’Universit´e de Padoue (Italie), l’Universit´e de Bordeaux (France) et le laboratoire de Recherche et D´eveloppement de THALES, plus pr´ecis´ement dans l’unit´e III-V lab (France). Les pr´esentations s’int´eressent plus particuli´erement aux essais de stockage en temp´erature ainsi qu’aux essais sous polarisation statique continue en configuration de d´ebit. Les r´esultats des essais permettent de pr´eciser la fiabilit´e de differents contacts et empilements m´etalliques, tandis que les essais en d´ebit mettent en e´ vidence l’effet des e´ lectrons chauds sur le processus de d´egradation des transistors en fonctionement.
Introduction Wide bandgap semiconductors, such as SiC and GaN, are becoming extremely attractive for a lot of power applications from power switches to RF/microwave amplifiers. High current densities and high voltage handling are key-factors that can be achieved simultaneously in AlGaN/GaN High Electron Mobility Transistors (HEMTs). Since their introduction in 1994, power RF/Microwave transistors based on gallium nitride (GaN) are now trying to emerge as commercial products. Impressive power emission results have already been published both for single transistor, with power densities up 32W/mm at 4 GHz [1], as for complete amplifiers with a 230W CW output power at 2 GHz [2]. Initial theoretical predictions about the higher voltage handling and current densities level have now been experimentally confirmed with transistors that can achieve gate-to-drain breakdown voltages around 100V/µm and maximum current densities exceeding 1A/mm. These characteristics lead to power densities several times higher than commercially available devices. The applications of GaN-based transistors range from military to commercial. Military applications includes shipboard, airborne and ground radars in X- and S-band with transmit/receive module containing both power-amplifier and low-noise MMICs, and wideband high-power amplifier modules operating at frequencies of 2-20 GHz for applications such as jamming and electronic attack. The commercial applications include base-station transmitters, very small aperture terminal and broad-band satellites, local multipoint distribution systems and digital radio. Particularly, GaN-based transistors can be of immediate interest for 3rd generation wireless cellular networks using complex modulation scheme such as W-CDMA. At the moment, LDMOS dominate the market thanks to their excellent price-to-performance ratio compared with other commercially available technologies such as GaAs HEMTs and Si BJT. AlGaN/GaN HEMTs are just beginning now to emerge as contenders for high-power transistors. As GaN HEMTs find application in systems used in critical environments in which the maintenance is difficult or impossible (e.g. military systems, satellite systems, broadcasting systems), the devices reliability verification becomes a major issue. Only recently the importance of the reliability assessment has been understood and the reliability evaluation has become the number-one challenge. This fact is proved by high funding devoted by the American and European defence agencies to the improvement of the GaN microelectron-
xii
ics reliability research [3]. The aim is a great improvement in understanding the physical mechanisms behind device failures and the development of physical models to predict performance. As stated by various laboratories, GaN HEMTs degrade under stressful conditions such as high-temperature operation or RF bias. The main effect identified is a degradation of the drain current with time. A complete understanding of the the involved degradation mechanisms is still lacking. In this thesis, we investigate some reliability aspects of GaN HEMTs fabricated by various industrial and academic organisations and tested at the Alcatel/Thales III-V Lab (Paris, France) in collaboration with the university of Padova (Italy) and Bordeaux(France). The studies were based on two types of experiments: (1) storage tests on test structures on wafer (2) life test under biasing of packaged-transistors. The evolution of the electrical parameters of the transistors were monitored during the test. Complementary electrical characterization and chemical-physical analysis were conducted before and after the tests to understand the origin of the degradations. This PhD thesis was realized in collaboration with many partners, particularly FBH (Ferdinand-Braun-Institute f¨ur H¨oechstfrequenztechnik, Germany) and Daimler-Chrysler (Germany). This thesis is organized as follows: In Chapter 1, the use of GaN in the electronic industry is introduced. Its chemical and physical properties are presented for underlining its superior qualities for high power, high frequencies applications in comparison to other semiconductors. Then, the basic equations for DC and RF operation of HEMTs are recalled and a review of the GaN HEMT state of the art is presented. Finally, the principal steps for the fabrication of the HEMT are described and a rapid review of DC-to-RF dispersion problem is presented. Chapter 2 is devoted to the description of the adopted methodology for the study of GaN HEMTs reliability. After a brief overview of the basic mathematical tools, we describe the principal degradation modes and mechanisms that have been observed in similar devices for high-power, high-frequencies applications, for which an extensive treatment can be found in reliability literature (e.g. GaAs HEMTs). We also describe a secured methodology to obtain reliable data, starting from the sample selection up to the life test execution and the final data analysis. Chapter 3 deals with the description of the storage tests. After describing the analyzed devices, the performed measurements, and the measurement setup, the experimental results are reported. The chapter is mainly related to the study of the metallization stability. The chapter is divided in two sections to treat separately Schottky and ohmic metallization
xiii
schemes. Chapter 4 describes the tests bench and the realization of the DC-life test and of some RF test. The observed degradation are presented and a possible interpretation of the degradation mechanisms is proposed, particularly underlining the role of hot electrons. Finally, some conclusions are drawn on the carried-out tests, and perspective for future works are presented.
Acknowledgments I would like to thank Prof. Enrico Zanoni , who allowed me to pursue my PhD studies, and Nathalie Labat, who accepted with enthusiasm the co-tutorship of the thesis, for their support and guidance. I appreciate Thales Research & Technology (TRT), especially Dr. Sylvain Delage for the opportunity of spending a period of study in the laboratories. I would like to thank the other PhD students Alex, Raphael, Nicolas; Dr Erwan Morvan and Dr Christian Brylisnki for their constant scientific support; the technicians Marcel and Marceline and all the coworkers at the laboratories. I am specially grateful to Mr. Christian Dua, my advisor during my stay at TRT. Not only he learned me how to work in a big research environment, but also he showed me that to work with a smile is better. I thank Prof. Jean-Claude De Jeager and all his team at IEMN (Lille, France) for supplying me devices in the frame of the common laboratory TIGER (THALES-IEMN). I am also grateful to Prof. Hans-Joachim W¨urfl, Dr. Richard Lossy and Dr. Reza Behtas of FBH (Berlin, Germany) for supplying me devices and for the helpful discussions. I thanks all members of the Microelectronics lab (”casetta”) and Zanoni’s group at the Department of Information Engineering of Padova provided a supportive, enjoyable environment to study in and I am grateful to all of them: Andrea B., Fabiana, Francesca, Simone L., Salvatore, Daniele, Leonardo, Giorgio, Luisa, Andrea M.; prof. Gaudenzio Meneghesso for his help in preparing papers; prof. Andrea Neviani for his help in administrative problems. I am very grateful to Dr Arnaud Curutchet for low frequency noise measurements and his help during my stay in Bordeaux. My deep gratitude is due to my parents for their continuous guidance, encouragement, and support during my life. I would like also to extend my gratitude to my brothers for their help and cooperation. Many thanks are due to Francesca for her patience and encouragement during the exacting moments of the thesis.
Contents
Abstract
1
Sommario
vii
Resum´e
ix
Introduction
xi
Acknowledgments
xv
GaN-based High Electron Mobility Transistors
1
1.1
Wide BandGap Semiconductors . . . . . . . . . . . . . . . . . . . . . .
1
1.2
GaN Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
1.2.1
Physical Properties . . . . . . . . . . . . . . . . . . . . . . . . .
4
1.2.2
Two Dimensional Electron Gas (2DEG) . . . . . . . . . . . . . .
7
1.3
2
v
Basic principles of HEMTs
. . . . . . . . . . . . . . . . . . . . . . . .
8
. . . . . . . . . . . . . . . . . . . . . . . . . .
10
1.3.1
Basic equations
1.3.2
Small-signal model
. . . . . . . . . . . . . . . . . . . . . . . .
15
1.3.3
Large-signal model
. . . . . . . . . . . . . . . . . . . . . . . .
16
1.4
GaN HEMT Performance Review . . . . . . . . . . . . . . . . . . . . .
18
1.5
GaN HEMT process . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
1.5.1
GaN Epitaxy . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
1.5.2
GaN HEMT Technological Process . . . . . . . . . . . . . . . .
22
1.6
DC-to-RF Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
1.7
Thermal Resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
34
Methodology for Reliability Evaluation
37
2.1
A Mathematical Overview . . . . . . . . . . . . . . . . . . . . . . . . .
37
2.1.1
Probability and Reliability . . . . . . . . . . . . . . . . . . . . .
37
2.1.2
Mean-Time-To-Failure and Median Lifetime . . . . . . . . . . .
38
2.1.3
Statistical Distributions
39
. . . . . . . . . . . . . . . . . . . . . .
xviii
Contents
2.1.4
Arrhenius Law . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
2.2
Accelerated Ageing Tests
. . . . . . . . . . . . . . . . . . . . . . . . .
42
2.3
Failures Modes and Mechanisms . . . . . . . . . . . . . . . . . . . . . .
44
2.4
Life Test Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
2.4.1
Investigated Devices . . . . . . . . . . . . . . . . . . . . . . . .
48
2.4.2
Preliminary Considerations
. . . . . . . . . . . . . . . . . . . .
48
2.4.3
Monitoring and Read-Out Measurements . . . . . . . . . . . . .
49
Electrical Characterization Techniques . . . . . . . . . . . . . . . . . . .
50
2.5.1
TLM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
50
2.5.2
Low-Frequency Noise . . . . . . . . . . . . . . . . . . . . . . .
50
2.5.3
Transconductance Frequency Dispersion . . . . . . . . . . . . .
54
2.5.4
Gate-Lag Measurements . . . . . . . . . . . . . . . . . . . . . .
56
2.5.5
Light Emission Measurements . . . . . . . . . . . . . . . . . . .
60
Physical-chemical Analysis Techniques . . . . . . . . . . . . . . . . . .
62
2.6.1
Auger Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
2.6.2
X-ray Diffraction Analysis . . . . . . . . . . . . . . . . . . . . .
64
2.6.3
Microsectioning . . . . . . . . . . . . . . . . . . . . . . . . . .
65
2.5
2.6
3 Reliability of AlGaN/GaN HEMT: Storage Test
67
3.1
Investigated Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
3.2
Performed Measurements and Measurement Setup . . . . . . . . . . . .
68
3.3
Ohmic Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
69
3.3.1
Ti/Al/Ni/Au (A-type) . . . . . . . . . . . . . . . . . . . . . . . .
69
3.3.2
Ti/Al/Ni/Au (B-type) . . . . . . . . . . . . . . . . . . . . . . . .
71
3.3.3
Ti/Al/Ti/Au/WSiN . . . . . . . . . . . . . . . . . . . . . . . . .
72
Schottky Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
74
3.4.1
Mo/Au . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
75
3.4.2
Ni/Au . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79
3.4.3
Pt/Ti/Au . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79
Conclusions on Storage Test . . . . . . . . . . . . . . . . . . . . . . . .
81
3.4
3.5
4 Reliability of AlGaN/GaN HEMT: DC Life Test
83
4.1
84
Investigated Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contents
xix
4.2
Measurement Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
87
4.3 4.4
Performed Tests . . . . . Preliminary RF Life Test 4.4.1 Sample . . . . . 4.4.2 Test Conditions .
. . . .
90 93 93 94
4.4.3 Result of the RF Test . . . . . . . . . . . . . . . . . . . . . . . . Experiments Discussion: A-Type HEMT . . . . . . . . . . . . . . . . . . 4.5.1 HFGC Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
95 96 96
4.5
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
4.6
4.5.2 HTRB Test . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5.3 HTO Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5.4 Complementary Analysis and Conclusions on A-type HEMTs Experiments Discussion: B-Type HEMT . . . . . . . . . . . . . . . .
. . . .
. 97 . 98 . 100 . 109
4.7
4.6.1 HTRB Test . . . . . . . . . . . 4.6.2 HTO Test . . . . . . . . . . . . 4.6.3 Conclusions on B-type HEMTs Experiments Discussion: C-Type HEMT
. . . .
. . . .
4.7.1 4.7.2 4.7.3 4.8 5
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
HTRB Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 HTO Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 Conclusions on C-type HEMTs . . . . . . . . . . . . . . . . . . 122
Conclusions on DC Life Test . . . . . . . . . . . . . . . . . . . . . . . . 123
Conclusions and Future Works
125
A Oscillation problems A.1 A.2 A.3 A.4
109 111 113 114
Introduction . . . . . . . . DC test and measurements Oscillations problems . . A case of study . . . . . .
127 . . . .
. . . .
B PCM (Process Control Monitoring) Bibliography
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
127 127 129 132 139 151
1 GaN-based High Electron Mobility Transistors 1.1 Wide BandGap Semiconductors Wide Bandgap Semiconductors (WBS) find applications in a lot of domains. They represent the right choice for high-temperature application because they present low concentration of intrinsic carriers up to 1000 ◦ C and moreover because they can support high temperature thanks to their remarkable physical properties. Various technologies of devices are implemented with this type of semiconductor (LED, HEMT, MESFET,...). One of the first great domain of application of WBS has been the optolelectronics one. Optoelectronic devices such as LEDs (Light Emitting Diodes) and LDs (Laser Diodes) based on WBS not only are capable of making full color display, high density optical storage technologies (>20 Gb), but also can reduce the consumption by about 10-20 of power energy. Notably, although red (650 nm) and yellow emitters based on AlGaInP material had been demonstrated successfully , and 630/650-nm LDs have been commercialized for digital versatile disk (DVD) systems, the shortest wavelength available from this material system is restricted to about 560 nm. In fact, the density of storage on optic disk is limited by the wavelength of the radiance given out by the laser diodes used to read information, and this density varies as λ−2 . In GaAs technology the used lasers give out an infrared radiance of around 0.8µm lenght. The utilization of blue or ultraviolet lasers based on III-N semiconductors allows to increase the capacity of storage considerably (by a factor 4). Besides, these wide gap lasers can work to relatively elevated temperatures without their performances being affected in a significative way. With regard to detectors, WBS large energy gap makes them at the same time effective in the ultraviolet spectrum, and blind to the visible radiance. It allows to make the UV
2
1. GaN-based High Electron Mobility Transistors
imagery in presence of the sun while avoiding all phenomena of blinding. One of the emerging domain is related to power devices thanks to their high thermal conductivity, which allows efficient heat dissipation, and their high breakdown field, which ensures high working voltage in off-state condition. Particularly, the high frequency, high power domain is nowadays considered strategic for military applications (e.g. phased-array radar) and very attractive for civil and space applications (e.g. C-band Satcom, Ku-K small aperture terminal, digital radio). In this case, WBSs allow size and weight reduction of the devices, since they allow simpler matching circuits. These two parameters, size and weight, are important for all devices and systems which are loaded into vehicles and especially for spatial applications. Fig. 1.1 shows a diagram indicating the RF application for microwave devices [4] and fig. 1.2 focus on various applications of GaN HEMTs [5].
Figure 1.1: RF output-power vs frequency for various electronic devices
Among the WBS, the most promising are the gallium nitride, the silicon carbide, the diamond and the compound semiconductor III-V (BN, AIN, InN). Certain semiconductors II-VI, as ZnS, ZnSe and their alloys ZnSSe, are WBS, but they present too weak bond energy to withstand high temperatures without damages. A remarkable property of the III-N compounds (GaN, GaInN, AlGaN) is possibility of
1.1. Wide BandGap Semiconductors
3
Figure 1.2: Power vs Frequency potential domains of application of GaN HEMTs
heterostructures. In fact, the bands discontinuities can be large, which allows to reach high carriers density (> 1013 cm−2 ) for the bidimensional electronic gas (AlGaN/GaN structures essentially). As a consequence, the FETs realized with III-N compounds present high drain current.
The impressive physical-chemical properties of WBS are not univocally favorable. In fact, the high values of the bond energy, which allow chemical stability at high temperature, are on the contrary sources of difficulties in the device making because they complicate all the fabrication process: mechanical polishing, chemical etching, doping, realization of Schottky contacts, etc. This induces some constrains which need new solutions different from those with are used for usual semiconductors.
4
1. GaN-based High Electron Mobility Transistors
1.2 GaN Properties 1.2.1 Physical Properties The sable polytype of GaN crystal is hexagonal 2H (Wurtzite). Thin layer of cubic polytype 3C (Zincblend) can be grown on cubic substrate. GaN hexagonal (Wurtzite,Wz) or cubic (Zincblend, Zb) crystal forms are reported in fig. 1.3 [6]. Despite the predictable merits for the zincblend crystals, such as reduced number of surface dangling bonds, and smaller effective mass these crystals due to their instability cannot be used yet for devices. Hexagonal crystal layers grown on sapphire or 6H/4H SiC, constitute almost all the GaN substrates exploited for the research and development of GaN-based electronic devices. More recently, many groups have tried to exploit silicon substrate to grow GaN films to benefit from both the advancements already made on the large wafer silicon technology and the opportunity of the chip-level integration with silicon-based circuitry.
Figure 1.3: Unit cells of GaN. Zincblend structure (left); Wurtzite structure (right).
As grown, GaN epilayers were always n-type with carrier concentration of about 1016 cm−3 . Two hypothesis in conflict were made to explain this natural doping. The first hypothesis based on calculations of Perlin et al. [7] showed that the nitrogen vacancies could act as shallow donors and they could give rise to n-type conductivity in nominally undoped GaN epilayers. The energy of formation of nitrogen vacancy was lower than that of Ga interstitials. On the other hand, Neugebauer and Van de Walle [8] showed that the
1.2. GaN Properties
5
necessary energy to create a nitrogen vacancy was 4eV. Therefore, the vacancy at room temperature will not be present in high concentrations. Experiments also showed that oxygen atoms act as donors in GaN and most of the interstitial community now considers that the residual impurities (Si, and O) are responsible for n-type conductivity in nominally undoped GaN. Si substitutes Ga sites in GaN, and O substitutes N site both forming shallow donors. Activation energy of Si in GaN, determined by Tanaka et al. [9], using Hall effect measurements is 30meV. The interest of GaN for high-power microwave application stems from its intrinsic properties thanks to its wide bandgap nature. As previously discussed, materials such as GaN and SiC, are of great interest for high temperature applications because their intrinsic level is much lower as compared to materials like Ge, Si and GaAs. Due to their heterojunction capability and cost advantages, GaN family of semiconductors is the best choice among other wide-bandgap families. Although, the maximum electron drift mobility of two dimensional electron gas in GaN is predicted to be much lower than that of GaAs, other characteristics of this material have provided it with a great superiority over GaAs, for power applications. Using these wide bandgap materials not only extends the output power to values beyond that of small-bandgap based devices, but also increases the radiation hardness and temperature tolerance of the device. Another attractive property of GaN and III-N is their high breakdown fields. This breakdown field scales roughly with the square of the energy band gap, and is estimated to be in the range 2-4 MV/cm for GaN [10], as compared to 0.2 and 0.4 MV/cm for Si and GaAs, respectively. GaN has also excellent electron transport properties, including good Properties
Si
GaAs
4H-SiC
GaN
Energy Bandgap Eg (eV )
1.12
1.42
3.25
3.40
Relative dielectric constant ε0
11.8
12.8
9.7
9.0
Breakdown field EB (MV /cm)
0.25
0.4
3.0
2.5
µ(cm2 /V s)
1350
6000
800
1300
Saturation velocity vs (107 cm/s)
1.0
2.0
2.0
2.2
Thermal conductivity χ(W /cmK)
1.5
0.5
4.9
1.3-2
1
7
229
215
Electron mobility
JFM
Table 1.1: Material properties for several semiconductors.
mobility, and high saturated drift velocity as shown in Fig. 1.4 [11], thus making this
6
1. GaN-based High Electron Mobility Transistors
material suitable for general electronics, and promising for microwave amplifiers. The
Figure 1.4: Electron drift velocity at 300K in GaN, SiC,Si and GaAs computed using the
Monte Carlo technique
typical values for the material properties associated with high temperature, high power, and high frequency application of GaN and other more conventional semiconductors are summarized in table 1.1. In order to better understand the interest of GaN-based devices for high power microwave devices, some figure of merits were introduced (Johnson’s figure of merit, Baliga’s figure of merit, Keyes’ figure of merit). They allow us to point out the power-frequency limit exclusively starting from the material properties. A Combined Figure-Of-Merit (CFOM) can be expressed by : CFOM =
χε0 µvs EB2 (χε0µvs E B 2 )Si
(1.1)
By comparing the parameters listed in table 1.1 it can be seen how wide band gap materials, such as GaN and SiC, take tremendous advantage from their intrinsic properties, reaching a CFOM several time higher than other conventional semiconductors. Even if both GaN and SiC show high value of the figure of merit, GaN-based devices are more attractive than SiC-based ones, mainly thanks to availability of the heterostructure junction formed by using gallium nitride in combination with the aluminum gallium nitride or indium gallium nitride alloys.
1.2. GaN Properties
7
1.2.2 Two Dimensional Electron Gas (2DEG) The possibility of heterostructure technology can be considered as the strongest feature of the III-V nitrides compared to SiC. Quantum well, modulation-doped on piezoelectric heterointerface, and heterojunction structure can all be made in this system, giving access to new spectral regions for optical devices and new operation regimes for electronic devices. From this point of view, III-V nitrides can be considered the wide band gap equivalent of the AlGaAs/InGaAs system which has set the modern benchmark for microwave device performance. In AlGaN/GaN HEMTs, the conductive channel is created at the heterointerface by a Two Dimensional Electron Gas (2DEG). The main advantage of having a 2DEG channel is the possibility of increasing the conductivity by increasing the carriers concentration without suffering the mobility degradation effects due to the impurity scattering. Experimentally, the existence of the 2DEG can be evaluated by measuring the temperature dependence of the carrier mobility and carrier concentrations using low temperature Hall measurement [12]. The main difference between the heterojunction in gallium arsenide and in gallium nitride is represented by the method of populating the 2DEG gas. In fact, in GaAs-based HEMT, it is usually achieved through the doping of the AlGaAs barrier layer close to the AlGaAs/GaAs heterojunction (modulation doping). In these devices the accumulation of electrons in a triangular shaped potential well in the GaAs layer close to the heterointerface is reached by depleting the region of the barrier close to the interface [13], reaching high mobility thanks to the distance between the electrons in the channel and the dopants in the barrier layer. On the contrary, in Alx Ga1−x N/GaN HEMTs, doping is not required to form a two dimensional electron gas. For this heterojunction, Ambacher et al. [14] demonstrated that the formation of 2DEG in even undoped AlGaN/GaN structures relies mainly on piezoelectric and spontaneous polarization induced effects. The engineering of 2DEG carrier concentration in AlGaN/GaN HEMT is achieved by changing the Al composition and thickness of the Alx Ga1−x N barrier [12](fig. 1.5), and the cap layer thickness [15](fig. 1.6). See section 1.5 for a typical epitaxial structure used for GaN HEMT. Although increasing the Al mole fraction improves sheet charge density and carrier confinement, it also reduces the carrier mobility due to different effects including, inter-
8
1. GaN-based High Electron Mobility Transistors
Figure 1.5: Dependence of 2DEG sheet density on Al mole fraction
subband scattering, increased alloy disordering in AlGaN, increased density of interface charges and larger potential fluctuations at the interface due to the surface roughness. Furthermore, theoretical analysis of the magneto-resistance data has shown that, while for 2DEG densities below 7x1012 cm−2 , the electron mobility is limited by the interface charges, AlGaN/GaN interface roughness plays a more pronounced role at higher sheet carrier densities where the electron wave function penetrates deeper into the AlGaN barrier [9-11]. This susceptibility of the electron mobility to the surface roughness necessitates a compromise between the desire for high sheet carrier concentration and the carriers mobility.
1.3 Basic principles of HEMTs The physical principle of the heterojunction beyond GaN devices based on 2DEG is the same as for GaAs based devices. There are many names in the literature for the same kind of devices. After the discovery of the 2DEG at the n-AlGaAs/GaAs interface, many research groups in France [16], Japan [17] and US [18] have successfully fabricated the conventional AlGaAs/GaAs heterostructure FET and they called this new device by different names based on the underlying physical mechanism or structural configuration, e.g. the
1.3. Basic principles of HEMTs
9
(a)
(b)
Figure 1.6: a) Influence of AlGaN thickness on sheet carrier density and Hall mobility, for
AlGaN/GaN single heterostructure.The green points are the measured Hall mobility and the black lines are simulation fits; the solid line assuming no point defects in the structure, the dashed line using a shallow acceptor concentration in the AlGaN layer as a fitting parameter; b)Effect of GaN cap layer thickness on sheet carrier density and Hall mobility, for a GaN/AlGaN/GaN heterostructure with a fixed AlGaN layer thickness of 20 nm. The black solid line is a fit to simulations, and the gray points are the Hall measurements.
10
1. GaN-based High Electron Mobility Transistors
High Electron Mobility Transistor (HEMT), the Selectively Doped Heterojunction Transistor (SDHT), the Modulation-Doped FET (MODFET), the Two-Dimensional Electron Gas FET (TEGFET), and Heterojunction FET (HFET). At the moment, the most reported names in scientific literature for heterojunction-based transistor on GaN are HFET and HEMT. In this thesis, we will always use HEMT to avoid any ambiguity. Regarding the applications, there are some important differences between HEMTs used in power and small-signal applications. A power device has to withstand a larger voltage and current amplitude as compared to a small-signal device, to provide high output power [19]. Small-signal and power HEMTs play an important role in the wireless communication hardware. It consists in infrastructures (base stations) and the user parts (handsets) having receiver and transmitter sections. In the handset, supply voltage as low as 3 V is used to adapt to battery cell voltages. Small-signal low noise transistors are required to amplify incoming signals in receiver front ends, and power transistors with low on-resistance and high on-current are required in the transmitter section. Small-signal devices typically have a gate length below 0.15 µm and breakdown voltage of 5 V, since they have to be optimized mainly with respect to their RF noise performance. On the other hand, power HEMTs typically have larger gate lengths (0.4-1µm) and breakdown voltages higher than 10 V. Gate lengths higher than 1µm may be used for several hundred volt breakdown voltage required in power switching applications. As the gate definition represents a critical technology step and accounts for a significant fraction of the total fabrication cost, the use of uncritical gate-lengths in power HEMTs facilitates their cheap large volume production, so that they compete successfully with various other technologies in power applications, such as Si/Ge-HBTs, III-V HBTs, and GaAs-MESFETs. The following extracted equations will include short- and long-channel case.
1.3.1 Basic equations The HEMT is a three terminal device and it can be characterized by the gate length Lg , the gate width Wg and the Source-Gate (LSG ) and Source-Drain distances (LSD ). The electron transport in the 2DEG takes place between the Drain and Source ohmic contacts. The current flow is controlled and modulated by the bias applied between the gate Schottky contact (fig. 1.7). The application of a negative bias to the gate and the source electrode reduces the positive charge density in the metal close to the metal-semiconductor interface which empties the 2DEG. The total pinch-off of the channel can be obtained by increasing
1.3. Basic principles of HEMTs
11
the negative voltage VG to VGS = Vth (threshold voltage)(fig. 1.8 ).
Figure 1.7: Exemple of HEMT structure
The dependance of the 2DEG sheet carrier concentration ns on the applied gate-source VGS voltage at small drain-source biases can be expressed by [20]: ns =
ε(VGS −Vth ) q(di + ∆d)
(1.2)
with Vth : threshold voltage di , ε : thickness and dielectric permittivity of AlGaN ∆d : effective thickness of the 2DEG In equation 1.2 the threshold voltage is defined as the gate voltage when the conductance of the channel drop to the zero: Vth = Φb −Vp − where
Φb Vp ∆Ec
∆Ec q
(1.3)
: Schottky barrier height : pinch-off voltage : discontinuity of the AlGaN/GaN heterojunction
. Assuming c0 = ε/(di + ∆d) the capacitance per unit of area between the gate and the 2DEG channel and letting x to change along the channel, eq. 1.2 can be rewritten as: qns (x) = c0 [(VGS −Vth −V (x)]
(1.4)
12
1. GaN-based High Electron Mobility Transistors
(a)
(b)
Figure 1.8: Band diagram of HEMT in thermodynamical equilibrium (a) and after nega-
tive biasing of the gate (b) The drift current at any point along the channel is given by I(x) = Wg µ0 q ns E(x)
(1.5)
with µ0 the low-field mobility and E(x) the electric field along the channel. Using the eq. 1.4
dV (x) (1.6) dx Since the current is constant along the channel, integrating this equation from source to drain (0 < x < LDS ) gives 2 VDS Wg IDS = µ0 c0 (VGS −Vth )VDS − (1.7) LDS 2 I(x) = Wg µ0 c0 [(VGS −Vth −V (x)]
The output characteristics of the GaN HEMT are shown in fig. 1.9. As usually for FET transistors, the output characteristic can be divided in an ohmic (linear) region and in a saturation region. • ohmic(linear) region The ohmic region is defined pour VDS
