Loading...
Technical Insight

Magazine Feature
This article was originally featured in the edition:
Volume 29 Issue 3

Taking Ga₂O₃ to the next level

News

Agnitron’s MOCVD reactors are underpinning tremendous advances in the material quality of Ga2O3, as well as leaps in the performance of various power devices made from this ultra-wide bandgap semiconductor.

BY FIKADU ALEMA, AARON FINE AND ANDREI OSINSKY FROM AGNITRON, ARKKA BHATTACHARYYA AND SRIRAM KRISHNAMOORTHY FROM UCSB, AND CAMERON GORSAK AND HARI P. NAIR FROM CORNELL UNIVERSITY

On the back of continuous rapid strides, Ga2O3 devices are poised to shape the future of the power semiconductor market. This ultra-wide bandgap material has all the key characteristics necessary to produce better devices and, at the same time, the potential for cost-competitiveness, thanks to the capability to make high-quality melt-grown bulk substrates.

It is now a decade since device and materials research into Ga2O3 began in earnest. Over that time tremendous progress has been made, in terms of both the quality of the epitaxial material and the performance of Ga2O3 diodes and transistors. Recent reports include β-Ga2O3-based devices with a breakdown voltage over 8 kV and critical breakdown fields beyond 5 MV/cm, a regime that exceeds the theoretical limits of SiC and GaN. Such success establishes β-Ga2O3 as the most promising candidate for next-generation solid-state power-switching applications.

Now is the time to build on all this progress and deliver further improvement to the quality of β-Ga2O3 epitaxial films, as progress on this front will open the door to practical power electronic devices that fully exploit the potential of this oxide. That’s the key message at the heart of this feature, which summarises the latest MOCVD process optimisation that has ensured high-quality materials and ultimately high-performance devices. It is clear that when β-Ga2O3 is grown by MOCVD using optimal process conditions, the resulting material is of unrivalled quality.


Figure 1. 2D thickness map of a Ga2O3 thin film grown on a 50 mm sapphire substrate using the Agnitron Agilis 100 MOCVD reactor. A 2 mm edge exclusion is applied to the map. Film thickness is very uniform, with a non-uniformity of only about 2 percent.

At Agnitron Technology of Chanhassen, MN, we have a well-deserved reputation for being the world’s leading provider of MOCVD reactors for the growth of β-Ga2O3. One of our most well-known tools, the Agilis, is being used in both our laboratory and in various world-class institutions, including: the University of California, Santa Barbara (UCSB); Cornell University; the US Naval Research laboratory (NRL); Iowa State University, the University of Utah, the Ohio State University (OSU), and Bristol University, UK. Researchers from these institutions are greatly satisfied with the top performances of our reactors and continue to produce ground-breaking material and devices results. Here we highlight the results of Ga2O3 grown on our reactors in our laboratory and in the labs of our co-authors at UCSB and Cornell University.

Growing highly uniform films

Thanks to continuous improvements in the supply of large-area Ga2O3 substrates, now available in diameters of up to 100 mm, there’s a compelling need to be able to grow highly uniform thin films, in terms of both thickness and doping uniformity. While our Agilis 100 MOCVD is primarily intended for R&D, using far smaller substrates, it is capable of accommodating wafers up to 50 mm in diameter. Due to this, it is crucial that this reactor can realise uniform films on wafers of such sizes.

Our Agilis 100 employs a unique remote-injection showerhead design. Oxygen enters the reactor from a single central injector with an inert gas push flow, while the gallium metalorganic source, typically tri-ethyl-gallium, is introduced from the showerhead area that surrounds the central oxygen injector, again with an associated push flow. Merits of this design include laminar flow conditions, and prevention of mixing of the oxygen and metalorganic sources until they reach the substrate, thereby mitigating gas phase reactions and an associated reduction in growth rate. Another advantage of separate injection is that it provides an additional knob for fine-tuning the reactor’s gas flow profile.

Following installation of our Agilis 100 MOCVD reactor at Cornell University, we worked together to identify process conditions for increasing the uniformity of Ga2O3 films on 50 mm substrates. We found that through careful tuning of the flows and the reactor pressure, we could produce Ga2O3 films with a thickness non-uniformity of around 2 percent (see Figure 1). To realise the highest uniformity, we combined high total gas flows with low reactor pressures, conditions that are known to mitigate buoyancy associated with flow instabilities [1]. These conditions are conducive to the growth of films with high mobilities and smooth surfaces – measurements have revealed a root-mean-square roughness below 1 nm. The combination of a high carrier mobility and a low surface roughness is essential for the fabrication of high-performance devices over the entire 50 mm substrate.

Another strength of our design of reactor is that thanks to its unique injection system, it is amenable to further modifications. For example, adding a nozzle with a reduced diameter to the central injector increases the velocity of gas flow, and thus ensures a higher growth rate.



Figure 2. Capacitance-voltage measurements determine the doping concentration dependence of the silane molar flow rate, for the growth of β-Ga2O3 films at Agnitron Technology using optimal process conditions (a). 2D atomic force microscopy (AFM) images of 5.1 µm thick layers grown on (010) β-Ga2O3 substrates with miscut angles of 1º (b) and 0º, also known as on-axis (c). The features on the film grown on the on-axis substrate are bigger than those for the film on the substrate with miscut angle of 1º.

Realising low doping
Due to an absence of p-type doping, most Ga2O3 device architectures are unipolar. This limitation means that there is a need to grow thick epitaxial films with a low doping concentration to realise efficient voltage-blocking power devices. If there is a high doping concentration in the Ga2O3 epitaxial layer, which is also known as the drift layer, the depletion region narrows, resulting in a strong electric field at the junction between the metal and Ga2O3. This high electric field exerts a substantial force on electrons, propelling them across the junction. In turn, the electrical current through the device increases, leading to a threat of electrical breakdown. Due to these concerns, it is far better to have a low doping concentration in the drift layer. When that’s in place, the depletion width increases, reducing the electric field at the junction and thus increasing the breakdown voltage. Note that it’s important that the drift layer is thick enough to either match or exceed the depletion width.

We recommend MOCVD for the growth of devices with a thick drift layer, because this technique offers high-speed growth – a rate of around 10 mm/hr is possible – alongside nanometre surface roughness and uniform doping [2]. Last year, in the pages of this publication, we reported that with this growth technology we had demonstrated free-carrier concentrations that range from around 7 × 1015 cm-3 and around 2.5 × 1016 cm-3, for unintentionally doped and lightly silicon-doped β-Ga2O3 films, respectively. Recent process optimisation has improved these figures, with the realisation of even lower doping concentrations that have a tremendous advantage for vertical power devices. Our latest breakthrough came from film growth at a reduced growth pressure of less than 20 Torr, which drives down gas phase reactions and lowers defects and inclusions in the film.

Capacitance-voltage measurements (see Figure 2 (a)) on our epitaxial films unveiled the relationship between their doping concentration and the molar flow rate for silane (SiH4/N2). This study showed that the doping concentration varied linearly with silane molar flow rate, as expected. However, the key breakthrough is that when we dialed the silane molar flow rate down to around 3 × 10-12 mol/min, the doping concentration in the film fell to around 1.5 × 1015 cm-3.