Dopants in GaN

John T. Torvik

1 Introduction

GaN and the related alloy system (Al-In-Ga-N) has tremendous potential for both optoelectronic and electronic devices due to superior materials parameters such as a wide and direct bandgap energy, high breakdown fields, high saturated electron velocity and adequate electron mobility and thermal conductivity. However, when Maruska and Titjen succeeded in growing GaN on sapphire substrates in the late 1960s using chemical vapor deposition [1], it quickly became obvious that doping and defects would play a vital role in the future development of GaN. The early unintentionally doped GaN was invariably n-type, which at the time was believed due to nitrogen vacancies. The high n-type background carrier concentration on the order of 1018 cm—3 proved difficult to minimize and the absence of a shallow acceptor dimmed the prospects of a production-scale GaN-based device effort. Nevertheless, the early work using zinc-compensation led to the first demonstrations of blue, green, yellow and red metal-insulating-n-type GaN light-emitting diodes (LEDs) [2] but, further device development was still stifled by the seemingly insurmountable problem of making conducting p-type GaN. The search for p-type GaN was not successful until Akasaki and Amano demonstrated this feat in 1989 [3]. This remarkable achievement was actually a result of two significant milestones. First, the crystalline quality and the background n-type earner density in unintentionally doped GaN films was significantly reduced by the use of a low temperature AIN buffer layer [4,5]. Second, p-type GaN was demonstrated with Mg-doping followed by an ex situ low energy electron beam irradiation (LEEBI) treatment. Conducting p-type GaN had previously remained elusive despite other Mg-doping efforts because it was found that hydrogen passivates the Mg-acceptors [6], similar to the effect of hydrogen on acceptors in Si [7]. Thus, it was theorized that the LEEBI treatment, which was accidentally discovered while studying the cathodoluminescence of a Mg-doped sample, disassociated the H-Mg complex allowing the Mg to form a quasi-shallow acceptor level. This theory was later confirmed by producing p-GaN by annealing GaN:Mg in a hydrogen-free ambient such as N2 [8]. The process was also reversible rendering GaN:Mg insulating by annealing in a hydrogen-rich environment such as ammonia (NH3). These remarkable discoveries eventually led to the demonstration of a variety of bipolar devices such as blue and green pn junction LEDs and violet laser diodes (for example, see [9]), solar-blind and ultraviolet sensitive p-i-n photodiodes [10–13] and high-power, high-temperature bipolar transistors [14,15].

Even though much progress has been made in doping GaN there still exists significant challenges; especially with p-type doping. The low hole mobility and low achievable free hole concentration result in large sheet resistance preventing the fabrication of reliable Ohmic contacts with low contact resistivities. These material challenges have prevented the use of the AlGaN/GaN system to its full potential in electronic applications such as microwave heterojunction bipolar transistors (HBTs). Furthermore, the immature p-type doping technology has led to degradation (lifetime) problems and required that InGaN laser diodes operate at a higher than expected bias voltage.

The aim of this chapter is to describe the state-of-the-art undoped and doped GaN by comparing select optical and electronic properties. This is not an attempt to exhaustively cover all material properties or varieties, but rather to compare fundamental phenomena, such as luminescence, absorption, and conduction. Furthermore, we explore the electrical and optical characteristics using relatively simple and inexpensive measurement techniques, which are readily available in most semiconductor laboratories.

2 Background

Doping control is a prerequisite for the fabrication of most optical and electrical devices. For example, n- and p-type layers are the basic building blocks for bipolar devices, such as light-emitting diodes, laser diodes, photodiodes, and bipolar transistors, while undoped (or high resistivity) layers are needed for field effect transistors and photodiodes. Ideally, one should be able to grow intrinsic GaN prior to intentionally introducing dopants into GaN for conductivity modulation. However, as previously mentioned, unintentionally doped GaN tends to be n-type, which has been attributed to nitrogen vacancies [1,16] and residual oxygen [17] impurities. Recent electron irradiation experiments suggests that the background electron concentration cannot be due to nitrogen vacancies as this defect forms a level at 65 meV below the conduction band edge [18]. For comparison, the thermal activation energy(s) of electrons from the donor level(s) to the conduction band edge in unintentionally doped GaN is typically less than 37 meV [19]. Another challenge with GaN growth is the lack of a cheap and lattice matched substrate. C-plane (0001) sapphire and a-SiC are the most common substrates with a lattice mismatch of 16 and 3%, respectively [20]. The resulting strain necessitates the use of GaN or AIN nucleation (buffer) layers prior to growth of high quality GaN epitaxial layers with high electron mobility, specular surface morphology and low background electron concentration.

Conductivity modulation can be achieved in GaN by doping with donors such as Si, O and Ge or with acceptors such as Mg, Zn and Be. Si and Mg are primarily the n-type and p-type dopants of choice, respectively. Si-doping has been successfully used to produce films with free electron concentrations from the low 1017 to the mid 1019 cm—3 with almost complete room temperature donor activation. Furthermore, Si-doped GaN can routinely be grown with a bulk electron mobility above 300 cm2/V-s at modest doping densities. However, cracking of thicker films (>1 μm) has been observed in heavily Si-doped films when the doping density exceeds 1019 cm—3. The cracking has been attributed to the smaller ionic radius of +0.41A∘ compared to +0.62A∘ [19]. Unfortunately, the story is less encouraging for p-type GaN. The maximum reproducible hole concentration achieved in p-GaN with conventional doping techniques barely exceeds 1018 cm—3 without compromising the surface morphology. Furthermore, the deep nature of the Mg acceptor (Ev > 170 meV) leads to a poor room temperature hole activation of several percent. This leads to the use of excessively high Mg-concentrations above the mid 1019 cm—3 for heavy p-doping. The high Mg concentration degrades the hole mobility; often to below 10 cm2/V-s. This results in quite resistive (typically >2 Ω-cm) p-GaN epilayers and devices exhibiting large series resistances and poor contacts.

3 Experiment

It is important to understand both the electrical and optical properties of GaN, due to the tremendous potential for GaN in both the electronic and optoelectronic arena. The optical and electronic properties of undoped, n-type and p-type GaN are therefore discussed in detail in this chapter. The discussion relies heavily on Hall-effect measurements and photoluminescence and photoconductivity spectroscopy, which are relatively simple and powerful measurements techniques and as shown in this chapter can yield a wealth of information regarding GaN.

3.1 Characterization

Photoluminescence (PL) spectroscopy has been the workhorse of the optical characterization techniques due to its non-destructive nature and ability to yield valuable information about both intrinsic and extrinsic transitions. The latter is important since both defect-related and near bandgap transitions are frequently observed in GaN. The photoluminescence measurements presented in this chapter are performed using single-pass 0.5 m prism monochromator or a 0.32 m grating monochromator. The detectors used were a photomultiplier tube for the visible and UV, while a thermoelectrically cooled InGaAs detector was used for the IR part of the spectrum. The temperature-dependent measurements were performed using a closed-cycle He-cooled cryostat equipped with quartz windows. The UV excitation sources used include a HeCd laser operating at 325 nm, a UV line from an Ar-ion laser at 351.1 nm, and a pulsed (sub ns) N2 laser operating at 337 nm. The IR excitation source (for the Er-doped section) was an InGaAs laser diode operating at 983 nm or a tunable Ti:sapphire laser. The laser spot size diameters were < 1 mm. The PL signals were detected using the lock-in technique and recorded using a computer.

Photoconductivity (PC) spectroscopy is another sensitive and non-destructive optical characterization tool. In PC spectroscopy, one measures the change in conductivity between two Ohmic contacts in response to optical illumination. PC measurements can therefore be sensitive to defect-related absorption allowing the investigation of the defect distribution in the ‘forbidden’ energy gap as well as the absorption near and above the bandgap energy [21]. The PC measurement was performed using co-planar indium contacts spaced about 1 mm apart or interdigitated finger contacts (Ni/Au for p-GaN and Ti/Al for n-GaN) with finger spacing of 3 μm. Up to 50 V was applied across the samples and the photocurrent was measured across a variable load using a lock-in amplifier and...