Electrical Properties of ICP Plasma-Damaged n-GaN

Group-III nitrides are important semiconductors. Success in the etching of Group-III nitrides has been achieved predominantly by dry plasma etching. As GaN is highly inert chemically, a high-density plasma technique such as Inductively Coupled Plasma (ICP) etching is used. ICP offers etch rates beyond 200nm/min and fine surface morphology. Despite these favourable characteristics, there are bound to be near-surface changes, which will affect the electrical and optical properties of the GaN substrate. Plasma damage to the material will hinder the fabrication of GaN based devices. For example the fabrication of a GaN/InGaN/AlGaN laser diode typically involves etching that terminates at an n-GaN layer, on which an ohmic contact is deposited. Knowledge of the electrical properties of plasma damaged n-GaN would be valuable. In this work we examined the changes to the chemical composition of the GaN surface after exposure to ICP plasma and subsequent annealing. Electrical properties, including the formation of ohmic contacts, were evaluated.

Figure 1: AFM image of plasma damaged GaN surface.


Figure 2: XPS scans of samples: (a) unexposed (b) exposed to ICP and (c) exposed to ICP and RTA. 

The physical damage caused by the plasma etching process is depicted in Figure 1, which was obtained by atomic force microscopy (AFM). On the other hand, the measured resistance changes to plasma-damaged semiconductor samples provides an indication of the extent of ion-induced damage. 

Our measurements showed that the resistance of GaN planar resistors was reduced by approximately 30% after exposure to the ICP, contrary to findings for InGaP and SiC. A drop in the contact resistance was also recorded. Rapid Thermal Annealing (RTA) treatment was used to restore the resistance to the original levels. Since the deposited contacts are non-ohmic, it is difficult to draw any inference from the variations in the contact resistance. Instead the current/voltage characteristics were investigated to gain further insight into the nature of the contacts. The un-annealed contacts were non-ohmic regardless of whether they were deposited on a sample exposed or unexposed to ICP. This is partly due to the low level of doping in the epitaxial layer. Annealing of the contacts resulted in the unexposed sample approaching ohmic behaviour, while the ICP-exposed sample exhibited a perfect linear I-V relationship. Hence, the exposure of GaN surfaces to ICP resulted in a reduction in resistance and it also facilitated the formation of ohmic contacts. 

In order to explain the observations, the near surface stoichiometry changes due to plasma exposure was studied by XPS (X-ray Photoelectron Spectroscopy). A wide XPS scan (depicting the Ga3s, N1s and O1s peaks) is shown in Figure 2, for samples that were unexposed to ICP, exposed to ICP, and exposed to ICP and RTA. Immediately apparent is the enhanced oxygen content and suppressed nitrogen content on the surface of the etched sample. Annealing reverses this phenomenon. Narrow XPS scans were performed for evaluation of Ga:N ratios. Comparing the Ga3d:N1s ratio between the unexposed sample (Ga/N ratio = 2.65:1) and the sample exposed to ICP (Ga/N ratio = 5.76:1), we can ascertain that there is a loss of nitrogen atoms during the ICP process. 

These phenomena can be understood by considering the mechanism of ICP etching, which involves physical sputtering, followed by chemical reactions that take place on the sputtered surface. Being a high-density plasma technique, the effect of the latter is more dominant. During this stage GaN dissociates to react with Cl2 and BCl3 to form chlorides, bromides and nitrogen. Since nitrogen is the most volatile it is preferentially removed, leaving a N2 depleted surface. Such N2 loss occurs in the very near-surface region.

Nitrogen loss and the formation of N vacancies is the key mechanism that gives rise to changes in resistance and ohmic contact formation. Experimental and theoretical evidence suggests that N vacancies are responsible for excess electrons in GaN, hence resulting in a n+ region at the surface. The excess electrons increase conductivity, and hence the resistance drops. Contact formation on an n+ surface also improves its ohmic behaviour, since the tunnelling current through the metal-semiconductor interface is dependent on the doping level of the semiconductor region, and a higher current results in an improved ohmic contact. Based on the N vacancies theory, the return of the resistance to the initial level upon annealing could be explained by the restoration of a stoichiometric surface (Ga/N = 1.45:1). 

We believe that a second mechanism operates concurrently to produce the high conductivity surface region. The XPS data in Figure 2 also reveals an increase of oxygen content to the GaN surface after exposure to the ICP. We attributed this to the exposure of reactive Ga dangling bonds, whereby Ga-O bonds could be formed easily upon exposure to the ambient. It is well known that the incorporation of oxygen in GaN introduces a shallow donor level, which readily ionizes to increase the conductivity of the material. Following the same argument the increased resistance was brought about by the reduction in oxygen content after RTA treatment.

This work was performed by Professor S J Chua and graduate student HW Choi.