Advanced Processing of Group III-Nitrides

S.J. Pearton* ^a, F. Ren ^b, B.P. Gila ^a and C.R. Abernathy ^a

^aDepartment of Materials Science and Engineering, University of Florida, ^bDepartment of Chemical Engineering, University of Florida

ABSTRACT

Recent advances in developing process modules for GaN photonic and power electronic devices are reviewed. These processes include damage removal in dry etched n- and p-GaN, implant doping and isolation, novel gate dielectrics, improved Schottky and ohmic contacts.

Keywords: GaN, processing, etching, contacts, doping, ion implantation, dielectrics

1. INTRODUCTION

GaN device research is now shifting from photonics to electronics with the commercialization of laser diodes. GaN electronic devices are attractive for high voltage, high temperature applications, including microwave power sources, power switches and communication systems.⁽¹⁾ Very impressive progress has been achieved in AlGaN/GaN heterostructure field effect transistors (HFETs) in a relatively short time.^(2-6) Over the past year or so, there have also been demonstrations of AlGaN/GaN heterojunction bipolar transistors (HBTs)⁽⁷⁾, GaN Metal-Oxide Semiconductor Field Effect Transistors (MOSFETs)⁽⁸⁾ and AlGaN power rectifiers (both Schottky and p-i-n versions.⁽⁹⁾ With these new devices come demands for edge and surface termination techniques, gate dielectrics and more exacting requirements for minimizing disruption to the surface via thermal or energetic ion-driven processes.

An example of the difference in sensitivity to dry etching of photonic and electronic devices is shown in Figure 1. In a typical laser or light-emitting diode, a mesa is formed by dry etching in order to contact the n-side of the junction. In this case, residual etch damage (which is n-type in character) is actually beneficial in improving the ohmic contact resistance by increasing the n-type doping level. By sharp contrast, in an HBT structure, the base mesa terminates on a p-GaN layer. In this case, residual etch damage can severely degrade the p-ohmic contact resistance. In addition, the base layer is relatively thin (1000-2000Å) and therefore a very controlled etch rate must be employed, whereas in photonic devices the n-layer is many microns thick.

In this paper we will review recent progress in developing advanced process modules for GaN devices.

Figure 1. Schematics of prototypical photonic (left) and electronic (right) devices.

2. RESULTS AND DISCUSSION

2.1 Implantation doping

Table I summarizes the maximum achievable doping levels, ionization level and diffusivity at high temperatures for implanted donor and acceptor impurities in GaN. In general, lower doping levels are achieved in Al_xGa_1-xN, with the fall-off dependent on the Al composition. In GaN, the best choice for n-type doping by implantation is Si, due to its low ionization level, high activation efficiency (over 90% in most cases) and low diffusivity. For the case of p-type doping, the highest hole concentrations are achieved using either Mg⁺ or Ca⁺ implantation. Both of these acceptors have high ionization energies, leading to low ionization efficiencies at room temperature (a few percent of the acceptor concentration). At elevated device operating temperatures, this ionization efficiency increases to ~60% at 300^oC.

Table I. Characteristics of different implanted dopants in GaN.

Donors	Max achievable doping level (cm^-3)	Diffusivity (cm^2×s^-1)	Ionization Level (meV)
Si	5x10^20	<2x10^-13 (1500^oC)	28
S	5x10^18	<2x10^-13 (1400^oC)	48
Se	2x10^18	<2x10^-13 (1450^oC)	---
Te	1x10^18	<2x10^-13 (1450^oC)	50
O	3x10^18	<2x10^-13 (1200^oC)	30
Acceptors
Mg	~5x10^18*	<2x10^-13 (1450^oC)	170
Ca	~5x10^18*	<2x10^-13 (1450^oC)	165
Be	<5x10^17	Defect-Assisted	---
C	n-type	<2x10^-13 (1400^oC)	---

*Acceptor Concentration

For high dose implants (³5x10¹⁴ cm^-2), it is generally necessary to anneal at temperatures ³1400^oC to remove all of the lattice damage. Under these conditions, the GaN surface must be encapsulated with a stable dielectric – such as AlN, or annealed under a high N₂ overpressure.⁽¹⁾

2.2 Implant Isolation

Figure 2 (top) shows a schematic of a structure for measuring the sheet resistance of implant isolated GaN, while the lower part of the figure shows simulated atomic profiles for Ti⁺ implanted into a 0.3mm thick GaN epitaxial layer. We typically obtain sheet resistances up to ~10¹² W/square with virtually any implanted species, with this maximum value being achieved after post implant annealing at 450-650^oC.⁽¹⁰⁾ Beyond these temperatures, the sheet resistance decreases to its initial (unimplanted) value. This behavior is characteristic of damage-induced isolation – we have not yet found an implanted species that creates thermally stable isolation in GaN. The species investigated have been Ti, V, Cr, Fe and O in both n- and p-type material.

Figure 3 shows the energy level positions of electron and hole traps induced in GaN by implant damage, as determined by the temperature dependence of the sheet resistivity of the implanted isolated material. Even though the levels are not at midgap, they are sufficiently deep to create high-resistivity material.

Most isolation studies in GaN reported in the literature have been done using keV ion implantation. However, in the case of keV ion implantation, the implanted species stop inside the conductive GaN layer, and the damage profile is highly non-uniform throughout the GaN film. This makes the separation of the effects of defect and chemical isolation (and, therefore, the interpretation of data) difficult. Another problem of electrical isolation by keV ion bombardment is that many GaN-based devices are fabricated from quite thick (³1.5 mm) GaN epilayers. In this case, the achievement of adequate electrical isolation using conventional keV ion bombardment requires a large number of overlapping implants, which is undesirable.

To overcome the problems of electrical isolation of GaN with keV ions, MeV light ion irradiation has been applied. In this case, projected ion ranges are greater than the thickness of typical GaN epilayers, and the profiles of generated atomic displacements are essentially uniform throughout the conductive GaN film. A single MeV implant, therefore, is sufficient to isolate a relatively thick GaN film. Moreover, irradiation with MeV ions allows effects of defect isolation to be separated from chemical isolation and to avoid the formation of a layer with substantial defect-induced (hopping) conduction, which inevitably forms at the ion end-of-range region in the case of keV implants and usually complicates the interpretation of data.

Figure 4 shows the evolution of sheet resistance (R_s) of resistors with different original free electron concentrations (n) exposed to irradiation with 6.6 MeV C ions at RT. This figure reveals that each of the curves has three distinct dose regions. The first region comprises the lowest doses, where R_s increases only slightly with increasing ion dose. The second region is characterized by a very fast increase in the value of R_s (by 9-10 orders of magnitude) in a relatively narrow dose interval. Such and increase in R_s is caused by the trapping of carriers at defects created by ion irradiation and damage-induced degradation of carrier mobility. Figure 4 (a) also shows that, with further increasing ion dose, R_s values reach their highest levels after a certain dose has been accumulated (D_th: the so-called threshold dose). The third dose region is for ion doses larger than D_th. With increasing ion dose above D_th and up to the maximum doses used, the measured values of R_s remain approximately constant, forming a plateau. The levels of R_s at the plateaus shown in Figure 4 are of the order of 1-2x10¹¹ Wsq^-1. However, the real maximum values of R_s of GaN layers are even larger since the R_s values measured have a contribution from the parasitic resistances of the experimental set-up, which are of
the same order of magnitude.