Objective of the study and organization of this thesis

Chapter 1 Introduction

1.5 Objective of the study and organization of this thesis

In this study, the growth of high quality GaN material by MBE for the HEMT fabrication is investigated. All the materials were grown on the 2-inch c-plane (0001) sapphire substrate which is relatively cheap and available with high quality.

The thesis is divided into six chapters. The first chapter gives the background of the study.

In the second chapter, detail descriptions on the MBE system, experimental procedures and the characterization methods are presented. Chapters 3 to 5 are the experimental results. The dependence of the defect structure of GaN film on the AlN buffer growth parameters was first investigated (Chapter 3). This understanding is important for improving the GaN crystal quality. With this understanding, an effective method by combining the Ga-lean GaN buffer

and migration enhanced epitaxy was invented and the growth of high crystal quality GaN film by MBE using this approach will be described in Chapter 4. In Chapter 5, the roles of different dislocation types on the electrical properties of GaN HEMT grown by MBE are investigated. Chapter 6 concludes the study.

Table 1.1 Advantages of GaN material for electronic applications.

Material property Advantages

Wide bandgap 3.42 eV

 Great endurance for high device operating temperature

 Suitable for high power applications.

 Working under high temperature environment High breakdown field

~5x10⁶V/cm

 Larger power density

High thermal conductivity

~1.3W/㎝ K

 Better heat dissipation, enhanced device performance

 Easier device packaging High saturate electron velocity

~2.7*107 cm/sec

 Suitable for high frequency applications

Table 1.2 Material properties of GaN, AlN, Si, SiC and sapphire.

Material a (Å ) c (Å )

Table 1.3 Comparison of 2DEG mobility and sheet carrier concentration of AlGaN/GaN structure grown by MOCVD and MBE on different substrates. MOCVD-GaN stands for GaN template grown by MOCVD; HVPE-GaN stands for GaN template grown by HVPE;

Dislocation free-GaN stands for very high quality free-standing GaN template; the carrier mobility and concentration are measured at 300, 77, 4.2 or 0.3 K unless specify in the bracket;

x is the Al content in AlGaN layer.

(Sheet carrier concentration (n_s (cm^-2)) Reference

x 300K 77K 4.2K 0.3K

(2.610¹²) 109000 Skierbiszewski el

al., (2005) [34]

N N

Net fixed polar charges at the AlGaN/GaN interface:

Ppz (AlGaN) + Psp (AlGaN) – Psp (GaN)

(AlGaN)

(GaN)

Fig. 1.1 (a) Polarization effect on the AlGaN/GaN structure. Psp and Ppz are spontaneous and piezoelectric polarizations, respectively. (b) Carriers at the 2DEG are induced by the surface traps (charges) and polarization field in the AlGaN/GaN structure.

(a)

(b)

Chapter 2 Plasma-Assisted Molecular Beam Epitaxy System, Experimental Procedure and Characterization Methods

In this chapter, the plasma-assisted molecular beam epitaxy (PA-MBE) system used for GaN materials growth is introduced. The general procedure for the epitaxial growth will also be described. Finally, various measurement methods that used to characterize the material crystal quality and electrical performances of AlGaN/GaN devices are presented.

2.1 Plasma-assisted MBE system

The PA-MBE system was designed by ULVAC Inc. for the epitaxy growth of group III-nitride materials. Both binary and ternary compound materials are grown using the group III sources of gallium (Ga), aluminum (Al) and indium (In) and group the group V source of nitrogen. The following sections describe briefly the important parts of the MBE system.

2.1.1 Chambers and vacuum pumps

The MBE system consisted of two chambers (Fig. 2.1): a sample preparation chamber or loading chamber (LC) and a growth chamber (GC) connected by a gate valve. A dry rotary pump (DRP) and a turbo molecular pump (TMP) are used in the LC. Pressure in the LC has to reach below 5x10^-6 Pa before the sample is transferred into the GC. The GC, which is equipped with a rotary pump (RP), a TMP, a sputter ion pump and a titanium getter pump, has

a base pressure <2x10^-8 Pa. However, during the material growth, only the RP and TMP are used to maintain the growth pressure at the range of 5-8x10^-3 Pa depending on the flow rate of the nitrogen gas.

2.1.2 Sample manipulator and growth sources

Sample holders (molybdenum) are available for 2, 4 and 6 inch substrates. The substrate surface is facing down when it is placed on a sample manipulator which is rotated by an external motor. Substrate rotation up to 10 rpm is applied to improve the deposition uniformity across the wafer. A substrate heater is located just above (behind) the sample holder to control the material growth temperature (up to about 850^oC). Four effusion cells are installed for high purity solid sources of Ga, Al (2 cells) and In. Source fluxes emerging from the effusion cell are controlled by the cell temperature (up to 1250^oC). Nitrogen source in the form of nitrogen radical is generated by two RF plasma generators. The amount of generated nitrogen radical is affected by the nitrogen gas flow rate as well as the RF power. Beside the plasma generator, nitrogen source from ammonia gas is also available by the installation of a hot-wired „cracking cell‟. All the sources are deployed at the bottom part of the chamber and are co-focused onto the substrate holder (Fig. 2.1). A pneumatically controlled shutter is placed in front of each cell to control the sources used in material growth.

2.1.3 Analytical instruments and others

The ULVAC MBE system is equipped with a Reflection High Energy Electron Diffraction

(RHEED) for the real-time monitoring of epitaxial growth. High energy electron beam (20 keV), which is directed onto the sample surface at a grazing incidence (~2 degree), allows a surface-sensitive analysis during the growth.

An ionization gauge mounted at a flange on the chamber wall can be brought forward and place in front of the substrate to measure the source fluxes when necessary. In this way, the ionization gauge acts as a beam flux monitor (BFM) to calibrate the molecular beam intensity from time to time. A quadrupole Residual Gas Analyzer (RGA) is also placed at the rear part of the GC to detect the presence of the impurity gases in the chamber.

Both the growth chamber wall and source flanges are surrounded by liquid nitrogen (LN₂) cryopanels. LN₂ is flowed into the cryopanels during the growth process to prevent re-evaporation of impurities from parts other than the hot cells. The crypanels also provide thermal isolation among different cells.

2.2 Growth procedures

In this section, the general procedure for the growth of GaN material on sapphire substrate is described. The typical growth steps includes: sample preparation, nitridation, growth of buffer layer and growth of GaN epi-layer.

2.2.1 Sample preparation

The sapphire substrates used in this study were the commercial epi-ready 2-inch (50.8

mm) substrates with C-place orientation (<0001>). After loading into the growth chamber, the substrate was first annealed at ~820^oC for about an hour to clean the surface. The cleanliness of substrate surface was monitored using RHEED. After the thermal treatment, a clear and sharp RHEED pattern appeared as shown in Fig. 2.2a.

2.2.2 Nitridation

After the thermal annealing, the sapphire surface was treated with nitrogen plasma in the so-called „nitridation‟ process. During the nitridation, the sapphire surface (Al₂O₃) was converted slowly into a thin AlN layer. This layer acted as a starting layer for the epitaxial growth of AlN buffer layer. The effect of nitridation process on the crystal quality of GaN epilayer grown by the same MBE system was discussed in a previous study [35]. Without the nitridation process, the grown GaN material has low crystal quality and rough surface morphology. Furthermore, the nitridation temperature and duration are also important for the GaN epilayer growth. A rougher sapphire surface occurred after nitridation at high temperature (800^oC) and it could generate more screw dislocations in the GaN film. On the other hand, nitridation performed at low temperature (200^oC) would need longer time (90min) to transform the sapphire surface into AlN layer. In this study, substrate temperature of

~600^oC and duration of 60min were chosen for the optimum nitridation result. The change of the surface structure of sapphire upon nitridation could be observed using RHEED. As shown in Fig. 2.2, the RHEED pattern of the original structure of Al2O3 (Fig. 2.2 (a)) changed to that

of AlN structure (Fig. 2.2(c)) after one hour of nitridation.

2.2.3 Growth of buffer layer

Buffer materials generally used for the growth of GaN on sapphire substrate include low temperature GaN (LT-GaN), LT-AlN and high temperature (HT-AlN). In this study, AlN buffer layer was used. AlN not only improve the material quality of GaN but also help to achieve Ga-face GaN film grown by MBE on sapphire [18, 36]. Ga-face GaN provides polarization charges at the AlGaN/GaN interface for high electron mobility transistor (HEMT) fabrication. The effect of AlN growth parameter on the defect structure of GaN will be investigated in the Chapter 3. RHEED pattern on AlN grown on sapphire after the nitridation is shown in Fig. 2.3 (a).

2.2.4 Growth of GaN layer

The crystal quality of GaN film grown by MBE is determined by several factors such as the growth temperature, III/V ratio and the quality of buffer layer. The growth temperatures of GaN using MBE fall in a wide range from 650-800 C [12, 16, 20, 34]. A higher growth temperature is preferable for better crystal quality but will enhance the desorption rate of Ga adatoms from the substrate. As a result, a larger Ga flux is needed to maintain a reasonable growth rate. On the other hand, the Ga/N ratio is used to control both the material quality and surface morphology of GaN. While a rough and low quality material is formed under the nitrogen rich growth condition, a smooth but with Ga metal droplets left on the GaN surface if

too Ga-rich growth condition is used [20, 37]. Therefore, the optimum growth conditions for GaN with good quality and smooth surface (but without the Ga-droplet) can be achieved by the careful control of both the growth temperature and the Ga/N ratio. In this study, the

growth parameters of GaN were initially optimized using a homoepitaxial growth approach.

The homoepitaxial growth was carried out by depositing GaN film (~1m) on a high-quality GaN film (~2m) grown on sapphire by metal-organic chemical vapor deposition (MOCVD)

(referred here as the GaN-template). This approach proved useful for achieving high-quality GaN film by MBE [33, 34]. Under the optimized growth condition, the crystal quality and surface morphology of GaN after regrowth should be identical with or better than that of the

GaN-template. For the ULVAC MBE system, the growth parameters used are listed in Table 2.1. The growth rate of GaN is about 0.4 m/hr under this growth condition. After that, the

GaN material is grown directly on the sapphire by using an AlN buffer layer. In situ RHEED analysis during the GaN growth is shown in Fig. 2.3(b). The streaky RHEED pattern indicates that the GaN surface has a smooth surface morphology. After the growth was completed and the substrate was cooled down to below 200^oC, the RHEED pattern shows a 2x2 surface reconstruction (Fig. 2.3(c)) which indicates that the GaN has Ga-face polarity [38]. The effect of the growth conditions of AlN buffer layer on GaN quality will be discussed in Chapter 3.

2.3 Characterization measurements

In this section, the measurement methods used to characterize the material quality and electrical properties of GaN grown by ULVAC MBE system are described.

2.3.1 Field emission scanning electron microscope (FE-SEM)

The FE-SEM consists of a field emission electron source rather than a thermionic emission source used in a Thermal Tungsten wire SEM. A FE-SEM is therefore has a cold cathode. This SEM provides higher resolution imaging with higher beam density (brightness), and longer tip life. The primary electrons enter a specimen surface with energy of 0.5 to 30 keV to generate many low energy secondary electrons. The second electrons emission depends largely on the accelerating voltage of the primary electrons and also the probing incident angle on the specimen surface. In general, a large quantity of secondary electrons is generated from the protrusions and the circumferences of objects on the specimen surface, causing them to appear brighter then smooth portions. Under a constant accelerating voltage, an image of sample surface can thus be constructed by measuring the secondary electron intensity as a function of the position of scanning primary electron beam. In this study, Hitachi S-4700 FE-SEM is used routinely to check the GaN surface morphology as well as the film thickness.

2.3.2 Atomic force microscope (AFM)

AFM is used to check the surface topography of a material by measuring the interaction force between the AFM‟s probe and surface structure. The AFM probe consists of a sensitive

cantilever and a sharp tip (tip radius <10 nm) capable of obtaining the detail surface morphology of a material surface on the atomic scale if the scanning condition is carefully controlled. In general, the AFM is reliable to achieve lateral and vertical resolutions down to about 2 nm and 0.05 nm, respectively, in the ambient air environment. In this study, AFM Dimension 3100 system from Digital Instrument was employed to examine the surface information, such as morphology, roughness and density of pits, from the GaN material grown by MBE. This information provides the indications of the material quality and is a useful means to tune the growth condition.

2.3.3 High resolution X-ray diffraction (HRXRD)

HRXRD is the most powerful and commonly used characterization method to check the crystal quality of epitaxial grown materials. Besides being a non-destructive method, it also requires no sample preparation prior to the analysis. In this study, the Bede D1 HRXRD system (Fig. 2.4) equipped with a high power (2kW) X-ray source of Cu K_ line was used to determine the crystal quality of GaN grown by MBE. In additional to the crystal quality, other information of the GaN, such as lattice parameters, Al composition in the ternary compound of AlGaN and the AlGaN thickness were also determined by HRXRD. The following sections describe the frequently used scanning modes of the HRXRD in this study and its applications for epi-layer analysis.

2.3.3.1 Rocking curves (-scan)

Rocking curves were measured by scanning on the sample angle (or Omega) only. Such a scan is suitable for investigating the epi-layer crystal quality. In this study, it is assumed that the broadening of the rocking curve of a certain diffraction plane is mainly caused by the presence of threading dislocation (TD) in the GaN material. Three types of TDs are normally occurred in GaN: screw, edge and mixed TDs. These TDs can be distinguished by having different defect structures or Burgers vector, b. The b for screw, edge and mixed TDs are

<0001>, 1/3 <11-20>), and 1/3 <11-23>, respectively. Due to this specific defect structure, the screw dislocation can distort all the (h k i l) planes with l non-zero, while the edge TD can distort only the (h k i l) plane with either h or k non-zero. Therefore, the GaN (0002) plane rocking curve is broadened by screw- and mixed-type TDs, while the GaN (10-12) plane rocking curve is broadened by all TDs [7, 39].

In order to estimate the TD density in the GaN grown by MBE, the rocking curves for both the symmetric plane (0002) and asymmetric plane (10-12) were scanned for each sample.

Full-width at half-maximum (FWHM) of the rocking curves was determined by non-linear least-square fitting to a pseudo-Voight function. After that, an approach suggested by Gay et al. [40] as shown in equation (2.1) was used to estimate the TD density in GaN film:

Ddis ~ ²/ 9b² (2.1)

where Ddis (unit = cm^-2) is the TD density in the material,  (in radian) is the FWHM of a given

XRD peak, and b (Å ) is the length of Burgers vector of their corresponding dislocation. For

GaN grown at the <0001> direction, the b of the screw TD is the GaN a-axis lattice constant, i.e. 3.189 Å while the b for the edge TD is the GaN c-axis lattice constant, while is equal to 5.186 Å . To simplify the calculation of dislocation density, the mixed TD was divided into screw- and edge-type TDs [7]. So the “screw TD‟ used in this study was the sum of pure screw-type and the screw component of mixed-type TDs, while the „edge type‟ was the sum of pure edge-type and edge component of the mixed-type TDs. In this way, the rocking curve width caused by the screw and edge TD in the GaN films can be estimated from the (10-12)

rocking curve scan using the following equation:

(10-12)2=s2+e2

(2.2)

where (10-12) is the FWHM of (10-12) plane, and s and e are the contributions of screw and

edge TDs to the FWHM of (10-12) plane, respectively. Using equations (2.1) and (2.2), the screw and edge TD densities of GaN can be calculated. For example, if the GaN film grown by MBE showing the (0002) and (10-12) rocking curves of 100 and 1000 arcsec, respectively,

then the estimated screw and edge TDs are 9.71x10⁶ and 2.54x10⁹ cm^-2, respectively.

2.3.3.2 Omega-2theta (-2) scan

-2 scan was performed by scanning the sample and detector angles together in a 1:2

ratio. Such a scan is usually used to check the sample which contains thin and lattice mismatched layers. Diffraction signals from layers with different lattice constants along the diffraction plane will appear at different scanning angles. -2 scan is therefore useful to

compare the relative change in lattice constant due to residual stress or composition variation.

In this study, -2 scan was normally used to determine the Al composition in the AlGaN

layer grown on GaN layer. Thickness of AlGaN can also be calculated from the simulation and data fitting from the -2 scan.

2.3.3.3 X-ray reflectivity (XRR) scan

XRR is performed at a grazing incidence beam is suitable for studying surface and layers near the sample surface. By fitting the scanned patterns, layer information such as film thickness, surface and interface roughness can be obtained. This is especially useful for the top layers with thickness less than a few hundreds of nm and roughness less than 35 Å (root-mean-square). Even though the thin layer can either be single crystal, polycrystalline or amorphous material, XRR is always used in this study to determine the thickness of the thin AlGaN or AlN/AlGaN layers (<30 nm) which grown on top of the GaN film.

2.3.3.4 Reciprocal space mapping (RSM)

RSM is used to distinguish the strain and tilt in a sample. A series of scans of -2 at different  offset angles is performed and then converted into a map of diffraction intensity in

the reciprocal space. RSM can be performed on both symmetric and asymmetric planes.

When a RSM is performed on the symmetric planes of the substrate and an epilayer, the tilt angle of the epilayer with respect to the substrate can be observed. Besides, the relative out-of-plane lattice parameter of these two materials can also be calculated. Meanwhile, when

a RSM of the asymmetric planes is obtained, both the relative in-plane and out-of-plane lattice parameter of the substrate and epilayer can be calculated. In this study, the RSM of asymmetric plane (such as (10-15) plane) was scanned occasionally to determine the strain and the Al composition of the thin AlGaN layer grown on the GaN film.

2.3.3.5 Triple axis diffraction

The few HRXRD modes described above are performed under the double axis configuration, i.e. without an analyzer crystal being installed at the detector stage. On the other hand, if an analyzer crystal is used as in the triple axis configuration, the acceptance

angle of the detector was reduced to a few arcsec, allowing a finer definition of the scattering angle, 2. With such a high accuracy, the deflection angle from a certain crystal plane can be

used to calculate the lattice constant of the scanned material. In this study, the triple axis configuration was used to determine the c-axis lattice constant of the GaN by obtaining the 2

value of the (0002) plane.

2.3.4 Transmission electron microscope (TEM)

TEM is the powerful tool to characterize the epitaxial layer microstructure. Operating on the basic principles of the optical microscope, the TEM takes the advantage by using electron beam source which has much shorter wavelength as compared to the optical source. Therefore, the resolution of TEM can reach to about 0.2 nm, in the ideal imaging condition, in contrast to

在文檔中以電漿輔助式分子束磊晶成長應用於高電子遷移率電晶體之氮化鋁鎵/氮化鎵異質結構 (頁 20-0)