Grown in an MOCVD reactor, nitride films tend to yield fewer defects than when they are formed in MBE chambers. But this gap in material quality can disappear with high-temperature ammonia MBE, which produces epitaxial structures with outstanding electrical characteristics, argues Alexey Alexeev and Stanislav Petrov from SemiTEq.

SemiTEq’s approach to slashing defect densities in nitride films

There is a fundamentaldifference between the growth of most nitride devices and those based on the arsenide and phosphide families. While the latter are grown on native substrates with very low defect densities, nitride devices tend to be formed on the likes of sapphire, silicon and SiC, because GaN substrates are so difficult to make, pricey, limited in availability, and – in relative terms – not of that high a quality.

The biggest downside of growth on mismatched substrates is that it leads to high dislocation densities. If the more widely used deposition technology, MOCVD, is deployed, dislocation density in GaN films is typically in the range 108-109 cm-2. However, if this approach is combined with an epitaxial overgrowth technology, dislocation density can fall towards 107 cm-2. In comparison, due to the lower temperature traditionally associated with MBE, this growth technology leads to a lower surface mobility of adatoms and ultimately an inferior material quality. Dislocation densities for nitride samples grown by MBE are typically 109-1010 cm-2.

Although MBE tends to yield inferior dislocation densities, it does offer several advantages over MOCVD. Some of them occur for all material systems: the opportunity for in-situ monitoring offered by reflection high-energy electron-diffraction, which can be used to identify the transition from three-dimensional to two-dimensional growth and provide surface reconstruction; the growth of sharper hetero-junctions; and a safer, easier route for the research and development of material systems and devices.

In addition, there are benefits associated with MBE growth that are unique to nitrides. In its most common form of MBE, plasma-assisted (PA) MBE, growth is relatively simple and hydrogen is absent from the growing surface. What’s more, deposition at low temperatures can be advantageous – incorporating indium in binary and ternary nitrides is easier at low temperatures, so this technique is good for the growth of InN, InGaN and InAlN with a high indium content. In addition, low temperatures aid effective p-type doping.



MBE System STE3N in SemiTEq’s Application Lab

If epiwafer growers select PAMBE, they are benefitting from the continuous development over the last decade of not only this technology, but also that of the plasma sources. By optimising growth conditions at the initial stage and during the deposition process, it is possible to improve  material quality.

Examples of this approach, which includes using migration-enhanced epitaxy, have paid dividends in several groups, including Jim Speck’s team at the University of California, Santa Barbara, and Valentin Jmerik’s group at Ioffe Physical Technical Institute. However, the growth temperature is still limited with PAMBE.

Consequently, dislocation density is high, due to insufficient surface mobility of adatoms and inferior coalescence of nucleation blocks at the initial growth stage. What’s needed is to increase the growth temperature in order to slash the defect density, while retaining the advantages that MBE has over MOCVD.

At SemiTEq, a brand within the Russian Joint Stock Company Semiconductor Technologies and Equipment, this has been the goal for more than a decade. We are now offering equipment that is capable, in MBE terms, of growth of GaN and AlN at very high temperatures.

This range of tools is selling well into our domestic market, enabling engineers to produce epistructures that combine excellent electronic attributes with very impressive levels of crystal quality.

Working with ammonia

Our tools are not based on the most common form of nitride MBE growth, PAMBE, but are based on an alternative approach known as ammonia MBE. We are not the sole pioneers of this deposition technology, but use far higher temperatures and III-V ratios than is normal with this technique.

The typical approach is to use growth temperatures that are 100-200 °C higher than those associated with PAMBE and III-V ratios a little higher. With these conditions, growth rates are 1 µm/hr or more. In this regime, researchers don’t tend to crank up the III-V ratios, which restricts growth temperatures to levels considerably lower than those found in MOCVD tools, where they are often above 1000 °C.

A major downside of keeping the temperatures below this is the high density of dislocations arising in nitride films – they tend to be more than an order of magnitude higher than those produced in epiwafers grown by MOCVD.

In comparison, our tools can get very close to MOCVD growth temperatures. This was a primary goal of ours when we started developing an MBE reactor for nitride growth in the late 1990s. Back then, we already had a great deal of experience in the growth of epiwafers for high power InAlGaAs laser diodes, and we were well aware that higher growth temperatures lead to better devices.

We were also aware of a ‘gap in the market’ when we commenced our development of a GaN tool. At that stage, there were no commercially available MBE systems that were capable of the growth conditions that we believe are best for nitride growth. That led us to think that there were sizeable rewards available for trying to realise an MBE system that could operate reliably in such extreme process conditions.

Initially, we focused on increasing the growth temperature of GaN that is, of course, limited by thermal decomposition on the wafer’s surface. Since nitrogen is more volatile than gallium, thermal decomposition of GaN is governed by the ammonia flux that flows onto the surface of the growing film.

To realise extremely high ammonia flows, we developed a special MBE system that features an increased area of liquid-nitrogen-cooled cryopanels. The pumping system based on these cryopanels, which is used in conjunction with a high-speed turbo molecular pump, can produce a high vacuum in the growth chamber even with very high ammonia flow rates.

Thanks to these features, our prototype system enabled the GaN growth temperature without thermal decomposition to increase to 970 °C, while using an ammonia flow of 400 sccm. This temperature is even significantly higher than the highest values reported for ammonia MBE – about 900 °C – while the vacuum in the growth chamber is typical for ammonia MBE, staying below 5 x 10-3 Pa. Armed with these new growth conditions, engineers can improve the structural quality of GaN. However, the dislocation density is still higher than that for MOCVD-grown layers.

Slashing dislocation densities

To significantly slash the dislocation density with ammonia MBE, we performed a series of experiments that revealed that the key is to grow an AlN buffer layer at an extremely high temperature – more than 1000-1100 °C. This type of buffer improves material quality by increasing coalescence, so nitride film deposition quickly moves to the two-dimensional growth mode.

Our tools show that with ammonia MBE, rather than plasma-assisted MBE, it is possible to grow AlN and high-aluminium-content AlGaN with a high substrate temperature and V/III ratio in excess of unity (N-rich mode).  Alternately, growth at high temperatures in PAMBE is very tricky, because the aluminium-rich mode is essential for two-dimensional growth of the AlN buffer, while aluminium desorption is significant at substrate temperatures of 900 °C or more.

To realise the extremely high substrate temperatures for AlN buffer layer growth we had to develop a specialized MBE system. In this system, ammonia flows are lower for the growth of an optimal AlN layer than for GaN growth, while large area cryopanels enable typical vacuum levels for MBE, even at extremely high substrate temperatures.

This vacuum level, combined with an aluminium effusion cell, rules out the possibility of unwanted, difficult-to-address parasitic reactions that typically occur in the MOCVD reactor between tri-methyl-aluminium and ammonia.

Another of our tool’s features is that it opens the door to the growth of high quality structures for microelectronics or optoelectronics that feature an active region grown by either ammonia or PAMBE. Such structures, which can be grown in a single run, can be used to make ultraviolet emitters and detectors and microwave transistors. These devices demand very high quality layers of AlN and high-aluminum-content AlGaN.

Reactor portfolio

Our STE3N* series of MBE reactors that we released in 2003 combine an extremely high temperature on the substrate – it can be up to 1200°C with a high N/III ratio and an ammonia flow rate of up to 1000 sccm. The heating stage is highly reliable during a long growth run, aiding deposition of structures featuring thick AlN/AlGaN buffer layers and active layers with extremely low dislocation densities.

The STE3N* is available in either a two or three chamber format (the STE3N2 and STE3N3, respectively), with the latter equipped with a buffer preparation chamber.

The STE3N* systems can be configured to combine the ammonia injector with a nitrogen plasma source, which can be used in combination with ammonia MBE for growing active layers of InGaN, InAlN and magnesium-doped AlGaN. Both systems have specially designed indium, gallium and aluminium effusion cells for providing long-term life and growth rates of up to 2 µm/hr, when operating as ammonia MBE tools.

Other features of our STE3N* growth tools are the patented design of the substrate holder and the heating stage of the growth manipulator. This enables high heating uniformity, regardless of substrate material, and wafer diameters of up to 100 mm, allowing this tool to fulfil the requirements of engineers in R&D labs and those responsible for pilot production.

Our latest addition to the nitride MBE portfolio, the STE75, is also capable of the same growth conditions for high-temperature deposition of AlN buffer layers, but features more compact cryopanels for a limited number of samples per growth series.



SemiTEq’s has recently released its advanced compact MBE System, the STE75

Superior samples

We have used our ammonia MBE tools to develop a complex, very special AlN/AlGaN buffer layer that holds the key to the growth of extremely low dislocation density films on several types of mismatched substrates. Using c-plane sapphire, for example, we can form a buffer comprising 200-400 nm-thick AlN grown at a substrate temperature of 1100-1150 °C, followed by an AlGaN/AlN superlattice and AlxGa1-xN (x=0.1-0.3) transition layers deposited at 900-920 °C.

According to scanning transmission electron microscope images, the dislocation density in these samples is

2-4 x 1010 cm-2in the AlN buffer, falling to 4-6 x 109 cm-2 in the AlGaN layer grown after the superlattice (see Figure 1). Meanwhile, dislocation density in the top GaN active layer is just 9-10 x 108 cm-2. The latter value is comparable to GaN grown on sapphire by MOCVD, and far lower than that associated with conventional MBE.



Figure 1. An scanning tunnelling electron microscopy image of an AlN/SLS/AlGaN/GaN heterostructure reveals that the dislocation density in nitride films grown by MBE can rival those found in MOCVD-grown samples

The improvements to material quality have produced a substantial increase in electron mobility, which reaches a maximum value of 600-650 cm2 V-1s-1 in a 1.5-µm-thick, lightly doped GaN top layer (silicon doping of 3-5 x 1016 cm-3). This value is comparable to that of good quality, MOCVD-grown GaN, and in good agreement with calculations determining the relationship between mobility and dislocation density.

The GaN layer can be capped with an AlxGa1-xN barrier layer with variable composition. By changing the value of x from 0.25 to 0.4, electron sheet density can be varied from 1.0-1.8 x 1013 cm-2, while mobility is adjusted from 1300-1700 cm2 V-1s-1. This enables the channel sheet resistance for the two-dimensional electron gas to be tuned between 230-400 Ohm/sq. Higher electron sheet densities are possible by replacing the AlGaN barrier with lattice-matched InAlN – this yields an electron sheet density of 2.3-2.5 x 1013 cm-2 and a mobility of 1200-1300 cm2 V-1s-1.

Historically, only PAMBE could be used to grow the high-quality, lattice-matched InAlN layer, which requires an indium content of 18 percent. The lower temperature limit for ammonia MBE is determined by ammonia cracking efficiency, and this is negligible at temperatures lower than 500 °C.

Increase the temperature above this value and typical ammonia flow leads to very low indium content – it is insufficient for growing InAlN lattice-matched to GaN.

However, Speck’s group have shown recently that if the ammonia flow is increased to 1000 sccm, growth of InAlN that is lattice-matched to GaN is possible. Motivated by this result, we have tried to grow such structures by PAMBE and ammonia MBE, realising good results in both cases. It is worth noting that thanks to the special cryopanels and pumping system design, the vacuum levels are almost unchanged when ammonia flow hits 1000 sccm! This attribute of our tool will make it very attractive for producing many device structures.

Several Russian research centres and companies have employed our MBE tools to grow DHFET epiwafers, which have been used to make prototypes of advanced high power microwave  transistors. Testing of these devices reveals a high quality active region and absence of the current collapse effect, thanks to the adoption of a double heterostructure design that is significantly different from that of a conventional GaN/AlGaN FET.

What is also very encouraging for us is that the high power transistors produced from wafers grown in our STE3N* MBE system are very robust, according to long-term ageing tests.

Our efforts have meant that within a relatively short space of time, we have come a long way from a small R&D company to being the sole Russian MBE manufacturer, competing successfully with world-leading companies. One of the biggest factors behind this success is our continuous work with partners and customers. We offer not only technical training, but enhanced process training, including development and implementation of client-oriented processes. Creating a strong relationship with our customers and the constant monitoring of their work allows us to continuously improve our equipment.