Processing Techniques for III-V Nitride Semiconductors

In order to utilize the excellent properties of III-V nitride semiconductors and fulfill the devices, various processing techniques have been developed specifically for these materials. In this section, these techniques including epitaxy, ohmic contacts formation, and high-density-plasma etching will be introduced.

2.2.1 Epitaxial Growth

III-nitrides decompose into the group-III element and nitrogen before they start to melt because of the extremely high melting temperature. This would make it difficult to grow crystals from the nitrides in the melt. The growth of GaN crystals from gallium solution requires again, high temperature (1400 – 1500^oC) and elevated nitrogen vapor pressure (10 kbar). The lateral size of bulk single crystals of GaN is therefore limited to up to several millimeters. The difficulty in the growth of bulk substrate material results in epitaxial growth on foreign substrates like sapphire (α-Al2O3) and silicon carbide (6H-SiC).

The nitride semiconductors grown on sapphire or SiC substrates by metal-organic chemical vapor deposition (MOCVD) are commercially available [21]-[23]. The major precursors include tri-methyl or tri-ethyl forms of Ga, Al, and In. Silane (SiH4) and Cp2Mg are used as n-type and p-type dopant sources, respectively. The growth condition is set at the temperature of around 1000^oC and under the pressure of about 100 torr. Owing to the lattice mismatch between sapphire substrates and GaN semiconductors, a thin buffer layer is usually grown directly on the substrate at a low temperature of 500^oC to serve as a template of nucleation [24]. However, the densities

of the threading dislocations in these materials are still in the range of 10⁹ ~ 10¹⁰ cm^-2, which is on the order of million times higher than that of other semiconductors, like Si or GaAs. The lateral epitaxial overgrowth (LEO) can be employed to further reduce the dislocation density by about 5 orders [25]-[27]. In addition, a few of bulk GaN substrates with low defects are successfully produced by metal organic hydrogen chloride vapor phase epitaxy [28]-[31].

The growth of the ternary nitrides is more complex than that of GaN. In particular, the growth of InGaN is complicated by numerous problems. Due to the thermal instability of InN, In incorporation is expected to be elevated by the reduction of the growth temperature, which can be achieved at the expense of a diminished crystalline quality. Further more, the large lattice mismatch between InN and GaN produces considerable internal strain in the InGaN alloy due to a crystalline lattice distortion, which leads to phase separation and immiscibility [32]-[35]. The existence of large compositional fluctuation may be encouraged by the miscibility cap in this system [32].

The as-grown p-type GaN layer has very few carriers because the Mg-dopants are trapped by hydrogen atoms which come from the reactive sources and carrier gases.

To obtain the real p-type GaN layers, these Mg-H bonds must be broken after an activation process, which is performed by a post thermal annealing at 500 ~ 700^oC under a pure nitrogen atmosphere [36]. However, there is only about 1% of the Mg atoms ionized at room temperature owing to the deep acceptor level of around 170 meV above the valence band edge [37]. The typical mobility of holes is as low as 20 cm²/Vs, just allowing the realization of p-n junctions. A record value of 150 cm²/Vs was ever obtained by compensating the scattering atoms [38].

2.2.2 Metal Contacts

Metal-Semiconductor (MS) junctions are of great importance since they are present in every semiconductor device. They can behave either as a Schottky barrier or as an ohmic contact dependent on the characteristics of the interface.

Low-resistance, thermally stable ohmic contacts to GaN are crucial for obtaining good performance of light emitting diodes. This section will primarily focus on the formation of ohmic contact on GaN.

Unlike the cases of Si and GaAs, the Fermi level at the interface between the GaN semiconductor and the metal would be unpinned due to the substantial ionic component of the bonds in GaN [39]. Therefore, the Schottky barrier height (eΦb), which is the difference between the semiconductor band edge and the Fermi level at the junction, can be evaluated as follows.

eΦb = eΦm – eχs, for n-type GaN, (1.3) eΦb = Eg – (eΦm – eχs), for p-type GaN (1.4)

Where eΦm represents the work function of the contact metal, and eχs (= 4.1eV) is the electron affinity of GaN.

A metal-semiconductor junction results in an ohmic contact (i.e. a contact with voltage independent resistance) if the Schottky barrier height, Φb, is zero or negative.

In such case, the carriers are free to flow in or out of the semiconductor so that there is a minimal resistance across the contact. For n-type GaN, this means that the work function of the metal must be close to or smaller than the electron affinity of GaN (~

4.1 eV). For p-type GaN, it requires that the work function of the metal must be close to or larger than the sum of the electron affinity and the bandgap energy. However, the work function of most metals is less than 6 eV, and the sum of the electron affinity and the bandgap energy (~ 3.4 eV) is about 7.5 eV. It can be problematic to find a

metal that provides a good ohmic contact to p-type GaN.

For n-type GaN, choosing the metals of low work functions and increasing the doping concentration of n-type GaN can provide a good ohmic contact. At first, Al and Au ohmic contacts to GaN were used. These contacts yielded specific contact resistances of 10^-4 and 10^-3 Ωcm², respectively. [40] The use of Ti in ohmic contacts to GaN resulted in much smaller contact resistance. Lin et al. [41] described an Al/Ti ohmic contact to n-GaN with a specific contact resistance of 8x10^-6 Ωcm². Later, Fan et al. [42] reported on the Al/Ni/Al/Ti contact to n-GaN and obtained the specific ohmic resistance as low as 9x10^-8 Ωcm² after alloy at a proper temperature. The mechanism of obtaining such a low contact resistance was shown to be the formation of TiN, which leads to a large concentration of N vacancies (that behave as donors in GaN) near the surface [43]. The dependence of the specific contact resistance on doping was studied by Khan et al. [44]. Wolter et al. [45] studied ZrN/Zr ohmic contact to GaN that showed a promising thermal stability with a reasonable specific contact resistance of 2x10^-5 Ωcm² for n-GaN with the electron concentration of 7x10¹⁷ cm^-3. These contacts exhibited excellent thermal stability in evacuated quartz tubes at 600^oC for 1000 hours. Holloway et al. [46] reviewed the results obtained for ohmic contacts to GaN. A low contact metallization for ohmic contacts was reported in [47].

As for p-type GaN, it is a great challenge to form low-resistivity ohmic contacts due to the following reasons.

(a) According to the equation of (1.4), the work function of the metal should be close to 7.5 eV. However, most metal exhibits a work function being lower than 6 eV.

(b) The carrier concentration in p-type GaN is low due to the deep ionization level of the Mg acceptor. There are only 1% of dopants ionized in p-GaN and this low hole concentration can not lead to a tunneling junction at the metal/p-GaN

interface.

(c) There is a tendency for the preferential loss of nitrogen from the GaN surface during processing, which may produce surface conversion to n-type conductivity.

A bilayer metal film of Ni/Au on p-type GaN is most common structure adopted for GaN-based optoelectronic devices. Ni/Au layer is deposited on p-GaN by an e-beam evaporator and subsequently annealed at 500-700°C. The typical contact resistance is around 10^-3-10^-2 Ωcm² [48], [49]. Although the contact resistivity is not low enough, it is allowable for application in LEDs. It was also found that the specific contact resistance of Ni/Au could be reduced to 4x10^-6Ωcm² after annealing in an oxygen ambience [50]-[52], but both the reliability and the reproducibility were disputable [53], [54]. Furthermore, Al0.15Ga0.85N/GaN strained superlattices were fabricated to enhance the surface carrier concentration, so the contact resistance was reduced effectively with the traditional Ni/Au-contact method [55], [56]

2.2.3 High-density-plasma Dry Etch

Due to the inert properties of III-Nitrides, it is difficult to etch GaN with wet chemicals. It was only found that GaN could be etched at practical rate with molten salts such as KOH or NaOH at temperatures above 250°C. Although the technique of photochemical etching has been developed for nitrides [57], the difficult procedure of adding electrodes to spread current and selectivity over polarity of GaN restricts its application. It is much easier to etch N-polarity than Ga-polarity GaN owing to the much lower photo threshold energy of N-polarity GaN as mentioned in section 1.1.4.

However, the epitaxial quality of N-polarity material is always poorer than

Ga-polarity one, and it is necessary to utilize Ga-polarity materials to fabricate devices with high performance.

Dry etching is the most practical and feasible method. Especially, the high-density plasma (HDP) etchers, which use inductively coupled plasma (ICP) or electron cyclotron resonance (ECR) techniques to generate the plasma sources with 10¹¹~10¹²cm^-3 densities, can provide higher etching rates than the typical reactive ion etcher (RIE) without serious damages on the GaN surface [58], [59]. A variety of reactive gases have been investigated for GaN etching. Some special recipes can improve the etching selectivity among different epi-layers [60]. The chlorine-based gas mixtures are usually adopted owing to fair volatilities of gallium chlorides.

Methane (CH4) is also added in the mixture in order to etch the epi-layer containing indium content.