In the recent decades, GaN and its alloy compounds have attracted much studies for applications in our daily life due to their superior properties such as chemical inertia, wide direct energy bandgap (3.4 eV), high breakdown used【1】. The common applications of GaN based alloys are optoelectronic devices such as light emitting diodes (LED), laser diodes (LD)【2-3】, and power amplifiers such as high voltage Schottky rectifiers, heterojunction bipolar transistors (HBTs) and high electron mobility transistors (HEMT).
Owing to the excellent material properties of GaN, the applications of AlGaN/GaN high electron mobility transistors (HEMTs) are focused on high frequency, high power and high temperature applications. The improvement in material growth and processing techniques has produced the GaN devices with performances superior to that of GaAs and InP power devices.
However, the transconductance property which is important for low-power consumption driving has not been improved dramatically【4】. Therefore, recessed gate technique which is generally used in GaAs and InP devices
can be employed to develop to improve the performance of the GaN devices.
For the conventional III-V recess etching process, two major methods were used; one is chemical wet etching and another is plasma dry etching.
However, due to high bonding energy and low chemical reactivity of GaN compound semiconductors, wet etching is not easy for GaN as compared to the conventional GaAs semiconductors. Although photoenhanced chemical (PEC) wet etching【5】 has been demonstrated for GaN etching, alkaline solution can etch or dissolve the photo-resist easily. This kind of wet etching is therefore not suitable for the GaN HEMT fabrication, especially for the critical process such as gate recess. For these reasons, the plasma dry etching is considered as a better approach for the etching of GaN materials.
But surface damage caused by plasma dry etching results in the deterioration of the device performances such as lower breakdown voltage and higher leakage current. Therefore, the reduction of plasma damage becomes a main issue for gate recess process. Recently, there are many methods proposed for the reduction of residual plasma damages, such as adding CH4 and O2 into conventional Cl2/Ar plasma【6】, using CF4/O2 to reduce plasma damage【7】, and pre/post annealing process【8-10】. In this study, we focused on SF6/O2 plasma post-treatment after dry etch recess at the recess region to reduce plasma damages.
1.2 Thesis Content
The contents of this thesis include: literature review, fundamentals of electrical characteristics, experiment, results, discussions and conclusions. In Chapter 2, the literature survey on the characteristics of recessed AlGaN/GaN HEMT and fluorine plasma post-treatment are reviewed. In
Chapter 3, the fundamentals of electrical characteristics and XPS analysis are addressed. In Chapter 4, the recess etching process, the samples preparation for material analysis, and the BCl3 recessed AlGaN/GaN HEMT device process flow and the SF6/O2 plasma post-treatment process are described. In Chapter 5, Schottky characteristics after different plasma exposures and the recessed HEMT device performances with SF6/O2 plasma post-treatment are discussed. Finally, the conclusions will be given in Chapter 6.
Property Si GaAs SiC GaN
Energy Gap (eV)
1.11 1.43 3.2 3.4Critical Breakdown Field (MV/cm)
0.6 0.65 3.5 3.5Thermal Conductance (W/cm/K)
1.5 0.5 4.9 1.5Mobility (cm2/V-s)
1300 6000 600 1500Power Density (W/mm)
~0.8 ~1.0 2 to 4 >2Saturation Velocity (cm/s)
1x107 1.3x107 2x107 2.7x107FET Technology
MOS HEMT MESFET HEMTTable 1-1 Comparison between material parameters for Si, GaAs, SiC and GaN at T= 300 K
Chapter 2 Literature Review
2.1 Dry Etching of Ш Nitrides Materials
Fabrication of the wide range of new electronic optical-electronic devices envisioned for the gallium nitride material system will require the development of entirely new processing procedures. The group Ш nitrides are distinguished by their unusual chemical stability, a characteristic that has posed unique challenges for device processing. Due to the extreme resistance of the group Ш nitrides to chemical attack by conventional wet etchants, the etching rate of the dry etching is much higher than the wet etching. Hence, etching nitride films has been carried out almost entirely using dry etching methods. These techniques include reactive ion etching (RIE), magnetron enhanced RIE (MERIE), electron cyclotron resonance (ECR), chemically assisted ion beam etching (CAIBE), and inductively coupled plasma (ICP). Fig 2-1 shows the basic scheme of ICP system.
The basic principle of these methods is similar, i.e. etching occurs through a combination of both chemical and physical means. Reactive neutral atoms of chlorine, iodine, bromine or other elements produce chemical etching by forming volatile products. Physical removal of material by sputtering is produced by impinging ions. The combination of the two mechanisms is able to produce anisotropic etching at practical rates but with varying degrees of damages. A schematic representation of the reaction around the sample during plasma etching is shown in Fig 2-2.
The body of plasma is neutral, containing equal numbers of electrons and
ions (~5x1011 cm-3), but is at a positive potential with respect to the the surface. Volatile etch products formed by adsorption and reaction of these species with the sample materials are removed by ion assisted processes at more rapid rate than would occur in the absence of ion bombardment. Sputter assisted removal of these etch products exposes a fresh surface for the following chemical process to occur, and in this fashion there is a synergy between chemical and physical etch mechanisms.
2.2 Recessed AlGaN/GaN HEMT
Following the advancement in the nano-technology, reducing gate length is one of the most effective ways to increase the HEMT speed.
However, the main problem of scaling down the gate length is short channel effect. In HEMT devices, if the barrier layer is thick or the gate length is extremely short, the aspect ratio between gate length and barrier thickness becomes critically important. In general, the aspect ratio becomes a problem when it is less than 5. And when this happens, it will shift the threshold voltage to more negative part which means harder to turn off the channel. This is so-called ―short channel effect.‖ To scale down the gate length and avoid short channel effect, gate recess technique
is applied to solve the problem. The so-called ―gate-recess‖ technique is to reduce the thickness of the barrier layer under the gate metal.
For conventional III-V compound semiconductor, such as GaAs- based HEMT, there are many chemical wet-etching recipes that can be applied to recess etching. The major advantage of wet etching is low damages. However, it is difficult to find a compatible wet etching method for AlGaN/GaN HEMT. As an alternative approach, a chloride-based inductively coupled plasma reactive ion etching (ICP) has been employed to fulfill such tasks by several groups. 【 6-8 】 This approach can effectively modify the threshold voltage of AlGaN/GaN HEMT to positive direction. However, the induced damages and the associated defects lead to an increase in Schottky reverse leakage current. Larger Schottky reverse leakage current would lower the breakdown voltage of the HEMT devices.
2.3 Improvement on Plasma Damage of Recess Etching
Because there are limited wet etching possibilities on GaN, a significant effort has been devoted to various dry etching techniques. As reported, plasma-induced damage has degraded the device performances such as more leakage current. Because of the methods have been suggested to reduce plasma-induced damage such as pre/post Rapid Thermal Annealing (RTA) and post-plasma treatment【7-10】. However, annealing at high temperature during the RTA is not compatible with the gate metal (Ni/Au, for example) and has to be carried out prior to the gate deposition or the high temperature would cause gate metal to react with
Schottky layer. As a result, photoresist has to be removed after the recess etching and followed by a second photolithography step for the gate electrode. Thus, the gate electrode and the recess etching are not self-aligned. To avoid a large access resistance that could be caused by the ungated recess region, the gate electrode is required to be larger than the recess window. For this reason, plasma post-treatment seems to be a better method to recover plasma-induced damage. It has been reported that CF4/O2 plasma treatment between the processes of the gate recess and gate metallization can reduce leakage current induced by plasma damage.【7】 But the reason why CF4 plasma can recover plasma damage is not yet clarified.
2.4 Trap Effect and Pulse I-V Measurement
Since the demonstration of the first GaN based transistor, rapid progress has been made in the development of GaN-based HEMT devices.
Output powers of over 100 W have already been reported up to 6 GHz
【 11 】 . However, the existence of dispersion effects observed in GaN-based devices has limited the initial expectations. The presence of trapping centers in GaN based transistor which is related to surface, material, or interface states, has been considered as the main cause of these effects【12-13】. Some of the observed effects are current collapse, transconductance frequency dispersion, gate-lag, drain-lag transients, and limited microwave power output. Up to now it has not been possible to obtain concise conclusions of the inner mechanisms explaining the
behavior of the devices due to the presence of surface states.
Pulse measurements have already been applied to electronic device characterizations such as pulsed I-V and pulsed power measurement. The main reasons for using the pulse characterizations can be attributed to the need to investigate the trapping carriers and device heating effects. In the consideration of trapping carrier, the response time of these trapping carriers is the main reason to affect the performance of pulse measurement. Generally speaking, the response time of these trapping carriers are in μsec ~ msec level, which is obviously slower than the nsec level of electron carriers transportation especially for the high speed devices. Therefore, it is needed to check the behaviors of trapping carriers to prevent the slow response time of the trapping carriers that affecting the RF performance. Because high mobility electron transistor (HEMT) is a kind of horizontal devices and its channel is also very close to the surface of the device. It always shows more serious surface trapping effect than BJT devices, and pulse measurements are important characterization methods to detect the trapping effect in HEMT 【14-16】.
Additionally, the heating effect in high speed electronic is also an measurement especially for the microwave power device, where there is always a serious heating effect on the device under test (DUT).
Fig. 2-1 Schematic diagram of ICP etching system
Fig. 2-2 Schematic of configuration of dry etching of a semi-conducting sample
SHEATH
△V~100V λ D~FEW mm SUBSTRATE
MASK
VOLATILE PRODUCT Plasma
N ＋
+ + +
ELECTRODE
Chapter 3
Fundamentals of Electrical
Characteristics and XPS Analysis
3.1 Schottky Characteristics of Schottky Diodes 3.1.1 I-V Characteristics
Schottky contact characteristics of Schottky diodes are directly related to the gate performances of HEMT devices. One can take account of some Schottky parameters including Schottky barrier height ΦB and the ideality factor n to simulate the gate performance. Calculations of the Schottky barrier height ΦB and the ideality factor n can be achieved using the Richardson Equation below and fitted by the forward I-V characteristic：
Is: saturation current, n: ideal factor, A: diode area, A*: effective Richardson constant, k: Boltzmann’s constant, and T: absolute temperature.
3.1.2 C-V Characteristics
The purpose for C-V measurement is to find out the shift of threshold voltage and the location of 2DEG. The relation between carrier
concentration and capacitance is shown below:
dC
V: Voltage applied on Schottky gate metal, C: the measured capacitance,
ε s: material dielectric constant, ε o: 8.5x10-14 C/V-cm
And the relation between depletion depth and capacitance is shown below
x C
s
oWe could use the above two formulas to estimate the etching depth of BCl3 recess etching.
3.2 Pulse Measurement
This method is described in Fig.3-1. Pulse I-V measurement is acquired with a short pulse (100 ns pulse width) and a pulse separation (1 ms separation). The pulsed IV characteristics are suggested to detect the device performance under the isothermal conditions of device operation【17】.Therefore pulse measurement is always used for observing
the surface state.
A high-density distribution of surface states which act as electron traps located in the access regions between the metal contacts could cause the gate-lag effect. The trapped electrons depleted the 2-DEG in the access regions of the device, thereby limiting the current. The effect of surface trap was characterized using pulse IVs as shown in Fig. 3-2 at two extrinsic quiescent biases equivalent to：
1. Bias Point (gate-off)： VGS0 < VPinch , VDS0 = 0 V 2. Bias Point (gate-on)： VGS0 = 0 V, VDS0 = 0 V
Using these two quiescent bias conditions, the drain current variation could be assumed to be related to the surface trap, since this effect is mainly stimulated by the gate voltage【18】.
Following this, there is a formula brought up to represent the degree of surface trap：
From the phenomenon of surface trap, if there is less surface trap, there is better current recovery. If not, there is worse current recovery.
And the definition of the current recovery is shown in Fig 3-3 and the
3.3 X-ray Photoelectron Spectroscopy Analysis (XPS)
The X-ray Photoelectron Spectroscopy also known as XPS or ESCA (Electron Spectroscopy for Chemical Analysis) has been developed since the fifties by Professor K. Siegbahn. The Physical Nobel Prize was awarded to his work in 1981.
The most interesting thing with this technique is its ability to measure binding energy variations resulting from their chemical environment.
That is, information on the chemical nature and state of the detected elements at the sample surface. The theory of determination of chemical bonding of XPS is based on the equation below:
i
: the electron orbit which emits the photoelectron,) (i
BE
: total binding energy)
(i
l
: binding energy at neutral condition (element)q
: chargeV
: electron potential energy from other atoms within the molecule) (i
Er
: relaxation energyAs the atoms are bonded, valence electrons form the bonds. Then the energy states of the atoms will change to make the atoms deviate from the neutral. The difference in electron affinities will make some atoms more positively charged and the others more negatively charged than neutral.
The positive charged Q is added to the total binding energy and peak
shifts to higher binding energy, and vice versa. The higher oxidation state energy of the photon and released a photonelectron to regain its original stable energy state. The released electron retains all the energy from the striking photon. It can then escape from the atom, and even further from the matter and the kinetic energy keeps it moving which can be described by the following equations:
Z
ph
hv E
E
Eph: The kinetic energy of photo electron;
hv
: Incident X-ray photo energy; EZ: The binding energy of specimen atom.As for solid state material, work functionΦ is added to represent the energy of an electron escape from the surface. Hence the equation is
The incident photon energy of X-ray sources usually used in XPS is Al Kα 1486.6 eV and Mg Kα 1253.6 eV. In our experiment, we use the Mg Kα X-ray source. Upon the released of X-ray induced photo electrons the atoms will be lacking of electrons in the internal shells. To recover from this ionized state the atoms can emit another photon
(fluorescence) or undergo an Auger transition and result in Auger electron signals. We must distinguish the XPS and Auger signals in the spectrum.
One way to distinguish these two kinds of peaks is to change the X-ray source since the kinetic energy of the Auger electron is independent of the source but kinetic energy of photoelectrons is depending on the source.
There is a severe problem called ―charging‖ that occur among insulator and wide bandgap semiconductors. While the photoelectrons escape from specimen, if there aren’t enough electrons flow into specimen to maintain electrical neutrality, the specimen tends to be positively charged and resulted in a peak shift toward higher binding energy. We can use a flood gun to inject low energy electron to the surface to neutralize the charges. But most of the times the problem can be very complicated. To modify the flood gun current may be not very easy. The worst situation happens when Ar ion sputtering is used for depth profiling of insulators. Ar ions can charge and may also react with the surface, which make it difficult to identify. Coating a thin conducting film on insulator specimen and ground it could be an effective way to avoid charging. We can also take Au 4d3/2= 353 eV as a reference while scanning.
Fig 3-1 The method of pulsed IV measurement.
Fig 3-2 The method of detecting surface trap.
Fig 3-3 The scheme of current recovery calculation
Chapter 4 Experiment
4.1 Schottky Diodes
The studies of Schottky diodes include BCl3 recess etching, SF6/O2
plasma post-treatment and the combination of both of them.
The fabrication of Schottky diodes begins with a standard photolithography which was used to pattern the substrates for ohmic metal. Ti (20 nm)/ Al (120 nm)/ Ni (25 nm)/Au (100 nm) ohmic metal was then deposited on the substrates using an electron-beam evaporator at a pressure of ~1x10-6 Torr. The bulk of the resist and metal were then removed by a wet solvent lift-off process, following by a high pressure DI water rinse to remove the residues. After the lift-off, the samples were annealed using RTA at 800℃ for 60sec. Then, a second photolithography was used to pattern the Schottky gate metal.
4.1.1 BCl
3Recess Etching
For highly sensitive region under the gate, a smooth surface with low damage should be achieved after recess etching. To find the recess etching condition, we fabricated Schottky diodes to observe the Schottky reverse leakage current. Before the Schottky gate metal deposition, we used BCl3 plasma to etch the Schottky gate region. Then the Ni (20 nm)/
Au (50 nm) gate metal was deposited by electron-beam evaporator.
4.1.2 SF
6/O
2Plasma Post-Treatment
SF6/O2 plasma post treatment was applied after the BCl3 recess etching but before the gate metal deposition. There were many different ratios of O2 to SF6 used to observe the change of the Schottky characteristics of O2/SF6 ratio. The other conditions for the treatment are：the applied power was 50 W；the pressure was 10 mT；the flow of SF6 was kept constant at 20 sccm.
4.1.3 I-V, C-V Measurement
I-V and C-V measurements were used to check the Schottky characteristic after the BCl3 recess etching and the SF6/O2 plasma post-treatment. Here, we used HP4156b and HP4280 to measure the I-V and C-V characteristics of Schottky diodes, respectively.
4.2 Device Fabrication
The epitaxial layers of the AlGaN/GaN high electron mobility transistors (HEMT) were grown by metal-organic chemical vapor deposition (MOCVD) on sapphire substrate along the (0001) axis. The schematic cross-sectional view of our AlGaN/GaN HEMT structure is shown in Fig. 4-1. The MOCVD grown AlGaN/GaN HEMT structure consists of a 3μ m-thick undoped GaN buffer layer and a 30 nm undoped AlGaN Schotty layer with Al composition of 30％. Room temperature sheet charge density and Hall mobility were 1×1013 cm-2 and 1300
cm2/V-s, respectively.
The detailed fabrication process on the recessed AlGaN/GaN HEMT device with SF6/O2 plasma post-treatment is described at the following sections. Besides, the conventional HEMTs and recessed HEMTs without post-treatment were also fabricated for comparison. The flow chart of the process flow for device fabrication is shown in Fig. 4-2.
4.2.1 Wafer cleaning
The wafers were immersed in Acetone (ACE) and isopropyl alcohol (IPA), each for five minutes to remove contamination from the wafer surface, and then the wafers were blown dry by nitrogen gas.
4.2.2 Mesa isolation
Here, we used AZ5214E photoresist to define the active region. The mesa isolation was carried out by using Cl2/Ar plasma dry etching in the STS Inductive Coupled Plasma (ICP) system. The ICP dry etching condition was ：Coil power 450 W, platen power 150 W, and Cl2/Ar=20/2 sccm at 10 mTorr. The etching depth was checked by the
α -step measurement.
4.2.3 Ohmic contact formation
After the mesa etching, we also used the AZ5214E photoresist to define the ohmic contact region. Before transferring into the
After the mesa etching, we also used the AZ5214E photoresist to define the ohmic contact region. Before transferring into the