Characterization of electrically active defects at III-N/dielectric interfaces

 
« PreviousUpNext »Contents
Previous: 1.3 Goal of this thesis    Top: Home    Next: 2.2 Measurement setup

2 Experimental aspects

Experimental instrumentation and samples used in this work are the focus of the present chapter. We describe how the test structures under investigation in our studies are prepared. We highlight the significant variations among them and explain the concept behind our choices. Furthermore, we present the experimental equipment that is used for all measurements reported in this thesis. Special attention is given to the new features that have been added in the course of this work. First, a new software for communication with the instruments has been coded in LabView. Secondly, a lock–in amplifier has been installed for impedance measurements, with the special intention of measuring fast transients. Finally, an optical setup has been integrated to the existing electrical characterization system in order to benefit from the advantages of monochromatic light excitation on our test structures.

2.1 Sample preparation

The test structures in use in this thesis are GaN–based devices grown on silicon. The growth process consists in the deposition of an AlN nucleation layer on the Si(111) surface, followed by several AlGaN layers with different aluminum content. This serves to tune the lattice constant until the residual strain is minimized. In this way a low defect density, unintentionally doped (UID) GaN layer can be realized by metal organic vapor deposition (MOCVD), with a state–of–the–art procedure [55]. The AlGaN barrier is also grown by MOCVD by adding the desired aluminum content, which is 20% or 25% in the test structures used in this work, with a thickness between 20 and 30 nm. The dielectric layer is then deposited on top of the barrier. Different materials can be used, for example silica (SiO\( _2 \)), silicon nitride (Si\( _3 \)N\( _4 \), abbreviated as SiN), hafnium oxide (HfO\( _2 \)) or aluminum oxide (Al\( _2 \)O\( _3 \)). In our experiments we use devices with silicon nitride insulation only. Finally, the aluminum gate contact is deposited onto the dielectric, and for MIS–HEMTs the source and drain contacts are made by etching the dielectric and AlGaN with a plasma process. The shape of the gate contact is most of the times circular, or a comb–like structure for having different perimeter–to–area ratios. Fig. 2.1a and b show a schematic cross–section of a GaN/AlGaN MIS–HEMT, and a top view picture taken with an optical microscope.

(image)

Figure 2.1: (a) Typical cross–section of a GaN/AlGaN MIS–HEMT, (b) the top view with an optical microscope (gate in white, source and drain are the circular area and the pad above that of the gate) and (c) the GaN–only MIS structure with a silicon doped layer buried in the buffer. In this illustration the layers are not to scale and the nucleation layer on top of the silicon substrate is not shown.

Most of the devices used for the characterization of interface defects have a uniform UID GaN buffer. However, in order to investigate the role of the barrier, some wafers without the AlGaN layer have been fabricated. In this way the 2DEG cannot be formed, and the carriers are supplied from a 1018/cm3 silicon doped GaN layer buried into the buffer. This particular sample is schematically illustrated in Fig. 2.1c. Other variations in the buffer include doping with carbon (C:GaN) or magnesium (Mg:GaN). These devices have been used for DLOS experiment in the course of the initialization of the optical setup.

Other structures have been developed specifically for opto–electrical characterization. A limiting characteristic of regular GaN–based devices fabricated at Infineon Technologies in Villach is the composition of the gate contact, which consists of about 500 nm of aluminum, which reflects most of the incident light. In fact, this material is used to realize the gratings of monochromators, where the aim is to maximize the reflectance. In order to allow enough light to penetrate through the metal contact and efficiently illuminate the active gate area, gold contacts as thin as 5 nm have been made. To guarantee a better adhesion on the dielectric material, a 3 nm layer of titanium can be inserted below the gold. Both metals can be evaporated on the wafers and structured with a lithography process. We choose these materials because their absorption coefficient is one order of magnitude smaller than that of aluminum [56], as we will discuss in detail in Section 6.3.2. In fact, thin Ti and Au layers are the best choice for this kind of application and are used in many other studies [40, 43, 45, 49].

« PreviousUpNext »Contents
Previous: 1.3 Goal of this thesis    Top: Home    Next: 2.2 Measurement setup