The power of pseudomorphic nitrides
Pseudomorphic AlGaN/AlN promises to enhance the performance of UVC lasers, far-UVC emitters and high-power solid-state devices.
BY LEO SCHOWALTER FROM NAGOYA UNIVERSITY AND LIT THINKING AND MAKI KUSHIMOTO, ZIYI ZHANG, AKIRA YOSHIKAWA AND HIROSHI AMANO FROM NAGOYA UNIVERSITY
Recently, and for good reason, our community has been showing much interest in ultrawide bandgap semiconductors. Part of their appeal is their potential to create optoelectronic devices emitting in the UVC, a spectral domain associated with wavelengths below 280 nm (see Figure 1). But in addition, this class of semiconductors is a strong candidate for creating higher performance, high-power RF devices, and for increasing the power density and performance of power devices and power switches.
Figure 1. Nagoya University, working in partnership with Asahi Kasei/Crystal IS, has demonstrated the first UVC laser diodes
Unfortunately, ultrawide bandgap semiconductors are challenging for the very reasons that make them attractive. It’s the norm that very large bandgaps go hand-in-hand with very strong chemical bonds, leading to very high temperatures for crystal growth. Likewise, the large bandgaps hamper controlled doping of p-type and n-type carriers.
This is so severe that in the not so distant past, what we now call ultrawide bandgap semiconductors were listed as insulators! And even now, aside from a few special cases, it is incredibly challenging to realise reliable hole and electron conductivity in these materials.
One example of an ultrawide bandgap semiconductor is high-aluminium-content Al1-xGaxN. When the aluminium content in this alloy exceeds about 50 percent – that is, when the value of x is less than 0.5 – its bandgap monotonically increases with aluminium concentration from about 4.4 eV to 6.1 eV, when pure AlN is reached.
Figure 2. Nagoya University, working in partnership with Asahi Kasei/Crystal IS, has demonstrated the first UVC laser diodes.
Related to high-aluminium-content Al1-xGaxN is GaN, as well as the alloys of Al1-xGaxN and AlInGaN with bandgaps less than 3.5 eV. These materials have enjoyed much success, so it’s natural to try and builid on what has been accomplished by progressing the capabilities of ultrawide bandgap alloys of Al1-xGaxN with values for x that are less than 0.5.
While this may sound simple, the reality is far from that, with two major obstacles blocking the path to success. One is that the epitaxial growth of high-quality, high-aluminium-content AlGaN is relatively difficult on foreign substrates. It is much easier to undertake epitaxial growth on native bulk AlN, particularly at high aluminium concentrations, but progress has been held back by the lack of high-quality substrates.
The other major issue is the difficulty in doping high-aluminium-content AlGaN. Until recently, the most common approach to doping has been the addition of silicon, to produce n-type material. This dopant works quite well at aluminium concentrations below 85 percent and, very recently, researchers at Georgia Tech have even shown successful n-type doping of AlN. However, producing p-type doping is much more problematic.
Breaking new ground by overcoming both these issues is our team at Nagoya University, which has been working in partnership with Asahi Kasei/Crystal IS. Our recent demonstration of UVC laser diodes highlights that we have established a clear path past these issues that can unlock the door to allowing high-aluminium-content AlGaN to play a much bigger role in ultrawide bandgap device applications (see Figure 1).
Our work is an important milestone in the development of the UVC laser diode, a device that has been intensely pursued for more than 20 years. The commercialisation of a low-cost laser diode in the UVC will revolutionise portable, cost-effective chemical and particle detectors, medical instrumentation, and surface monitoring and machining. All these applications draw on short-wavelength high-energy photons and their very short absorption length in most materials.
However, despite the great demand for such a source that inspired numerous development efforts, the shortest wavelength laser diode developed prior to our successful demonstration of a 271 nm laser diode in October 2019 only reached down to a wavelength of 315 nm (see Figure 2).
There are two keys to our success. One is the availability of pseudomorphic Al1-xGaxN on high-quality AlN substrates, and the other is the use of distributed polarisation doping. With this form of doping it’s possible to realise low-resistivity hole injection without the use of impurity doping in the p-type laser cladding layer.
Figure 3. In high-aluminium-content AlGaN, misfit dislocations may form along the interface.
What is pseudomorphic AlGaN/AlN?
Due to progress in bulk crystal growth of AlN, several companies, including Crystal IS, are now supplying substrates. This foundation for epi-growth comes from slicing material from AlN crystals with very low defect densities.
During the growth of nitride semiconductor heterostructures, one defect that arises in the epilayers is the so-called threading dislocation. High densities of this type of dislocation tend to arise during growth on foreign substrates, such as sapphire, SiC or silicon, due to lattice mismatch. As these threading dislocations degrade device performance, the nitride community has devoted much effort to reducing their density, as well as mitigating the impact that they have on devices. But despite all this work, it still takes a great deal of care to bring the threading dislocations in high-aluminium-content AlGaN below 108 cm-2, leading to a high cost.
The good news is that 2-inch AlN substrates with threading dislocation densities of less than 103 cm-2 are now available from a couple of suppliers, and there are more producers on the horizon. When the surfaces of these substrates are properly prepared (most typically, the Al-polar c-face), AlN homoepitaxy with threading dislocation densities below 104 cm-2 are routinely produced.
Crucially, AlN substrates can also provide the foundation for AlGaN alloys with very low threading dislocation densities. Since the crystal structure of GaN, AlN and related alloys is identical, one would hope that a well-prepared surface of native AlN would act as a good template for the growth of all AlGaN alloys. Unfortunately there is a significant lattice mismatch between bulk GaN and AlN – it is as high as 2.4 percent. Thus, if GaN is grown on AlN using the c-face crystal surface as a template, the epitaxial GaN layer is compressively strained in the c-plane by 2.4 percent. This results in an enormous strain energy, increasing linearly with the thickness of the GaN layer. Eventually, this strain has to be relieved in some way.
As epitaxial growers are well aware, there are a number of mechanisms for strain relief, including forming islands and roughening the surface. However, the mechanism that is most significant for high-aluminium-content AlGaN is the formation of a misfit dislocation along the interface (see Figure 3). As shown in that figure, an edge misfit dislocation forms when a plane of atoms is skipped along the interface, leading to a reduction in the compression of the GaN epitaxial layer that fits onto the underlying AlN lattice.