Europe Turns To AlInN To Push The Limits Of Transistor And Sensor Performance
The AlGaN/GaN heterostructure has its pros and cons. Its strengths include the natural formation of a two-dimensional electron gas (2DEG) at its interface thanks to piezoelectric effects, which greatly simplifies HEMT fabrication. But this pairing also creates high epilayer strain that might impact reliability. Turning to AlInN can eliminate this weakness. Switching to this ternary not only does away with stress, because it allows lattice-matching to GaN – it also leads to a superior 2DEG density. Thanks to far stronger spontaneous polarization, the AlInN/GaN heterojunction produces twice the charge density of its Al0.25Ga0.75N/GaN cousin.
One team that is looking to exploit the strengths of AlInN is a European consortium called MORGaN – Materials for Robust Gallium Nitride. This three-year effort that kickedoff in November 2008 is backed by €9.2 million of funding from the European Commission, and involves 24 partners from 11 nations. One of its goals is the development of material for AlInN/GaN HEMTs that can deliver an output of 1 kW at 2 GHz, and another of its aims is the fabrication of chemical and pressure sensors that are based on the same material and can operate in very harsh environments.
If the program is successful, it could aid a very wide variety of applications. It could help the construction of high-power amplifiers based on AlInN HEMTs, which could be employed in radar systems and base-stations for mobile communications, or in power electronics for consumer applications. MORGaN could also lead to the development of pressure sensors capable of operating at temperatures of up to 700oC that could aid oil exploration, and help to determine the environment in automobile and jet engines. And robust chemical sensors could be fabricated in the project, which increase the pH detection limit for a compound semiconductor device from 12 or 13 to values of 15 or more.
Producing devices capable of operating in these extreme conditions requires innovation on many fronts, and this is reflected in the wide range of activities in the MORGaN project. Efforts are being directed at the growth of nitrides on a variety of platforms; optimization of the growth of AlInN/GaN heterostructures; new sensor architectures; and radical packaging technologies that help devices to operate in these harsh environments.
One of the striking aspects of this effort is its incredibly broad agenda. That’s because MORGaN is actually the amalgamation of two projects proposed to the European Commission. One was focused on the development of AlInN-based pressure and chemical sensors; and the other, a successor to a project called UltraGaN, aimed to demonstrate AlInN HEMTs’ appropriateness with microwave operations. All members of original UltraGaN team are participating in the MORGaN project together with additional partners from sensor fields and advanced material manufacturing.
Building on UltraGaN
“We got excellent results in UltraGaN. It was a major success," says Sylvain Delage, a researcher from the Alcatel-Thales III-V Laboratory in France who is now leading the MORGaN project. “In America they are now launching many projects using AlInN, based on all these results we obtained in UltraGaN." He claims that one of the highlights of this project was the demonstration of an AlInN/GaN HEMT working at 1000oC, which is 200oC higher than an AlGaN-based equivalent. And another success was the realization of a continuous-wave output of more than 10 W/mm at 10 GHz, alongside a poweradded efficiency of 56 percent.
These results were produced with a very simple, normallyon transistor architecture that is now being employed in the MORGaN project. By lattice-matching the AlInN/GaN heterostructure, it is possible to form a 2DEG with properties that are independent of the thickness of the ternary layer. “That means that we can make a HEMT with about 3 nm of AlIN," claims Delage. This allows a metal gate to change the surface potential of the material, and opens the door to the fabrication of a normally-off transistor, because the Schottky contact removes the 2DEG below it.
Producing AlInN/GaN heterostructures is challenging, because the ternary tends to phase separate and it is difficult to form a high-quality interface between the two materials. So to overcome these barriers to high-quality AlInN growth, four strong epitaxial partners were set loose on these issues: the MOCVD toolmaker Aixtron, Germany; EPFL (Ecole Polytechnique Fédérale de Lausanne), Switzerland; the Alcatel-Thales III-V lab, France; and FORTH (Foundation for Research and Technology – Hellas), Greece. Between them, they have one MBE reactor, which has provided excellent results for AlN growth to demonstrate an AlN/GaN HEMT, and three MOCVD tools, which have produced the better InAlN/GaN device results. The UltraGaN project used sapphire substrate for heterostructure optimization, while best power demonstrations were obtained on SiC substrates thanks to their superior high thermal conductivity.
These teams - plus the University of Bath, England - are now focusing on the development of growth technologies for realizing highquality InAlN/GaN heterostructures on a variety of platforms. These include composites that feature diamond, an incredibly efficient heat spreader. Some progress has already been made, including the deposition of relatively smooth epitaxial layers directly on all three orientations of the material.
Sensing the environment
Other partners in MORGaN are developing sensors that employ an AlInN HEMT to measure values of pH or pressure. HEMTs can reveal levels of acidity, because liquids impinging on the surface alter the chemical potential within the device, leading to a variation in current flow through the transistor. And measurements of pressure are possible, because the piezoelectric nature of the material means that strain impacts the concentration of the 2DEG. Variations can be determined with great precision by connecting this HEMT to three others in a Wheatstone bridge configuration.
One company that’s leading the development of both types of sensors is a start-up called MicroGaN, which spun out of the University of Ulm in 2002. The company has already developed GaN-on-silicon cantilever structures that are 200-400 μm wide, and 1 mm in length, and GaN actors, where cantilevers can be moved by electrical stimulation.
The new cantilever design will provide superior performance at high temperatures, according to MicroGaN’s Ulrich Heinle. MicroGaN and its partners will need to develop new processing steps for these elevated temperatures. “Traditional ohmic contacts are based on titanium and aluminum, and these are not practical for temperatures above 500oC," says Heinle, who explains that a partner is working on copper-based contacts. Another issue is that the current produced by this type of sensor is not stable at 600oC, and researchers are working to try to understand the degradation mechanisms and devise ways to address them.
The GaN-on-silicon cantilevers that MicroGaN has made tend to bend slightly under no external pressure, due to the internal stresses in the GaN film. “Due to the stress, the cantilever will bend downwards, and this can be 50 μm to 20 μm, depending on the structure," says Heinle. One way to address this issue is to introduce doubleclamped structures that are better at coping with any initial stresses in the material.
The University of Bath is developing a process for producing nitride films on silicon substrates with incredibly low levels of stress. Their approach - pioneered by the academic Wang Nang Wang - involves a nanoscale lateral overgrowth technique that begins with the formation of an array of nano-columns on a GaN template. Adjusting the growth conditions can then lead to the coalescence of the GaN nano-columns into a continuous planar film of a low-stress, high-quality semiconductor.
The Bath team is looking to extend this concept to the growth of thin membranes of GaN clamped either at all edges or along just one edge. In the latter configuration the structure will resemble a microscopic cantilever grown out from the semiconductor surface. “When released these cantilevers will sense applied forces by the piezeoelectric effect as they bend," explains Duncan Allsopp, a researcher at Bath University. “We have already demonstrated the growth of short, stubby cantilevers and we are now extending our research to clamped membrane structures to act as pressure sensors." The Bath team has designed a photolithographic mask and is now optimizing the geometry of these epitaxially grown cantilever and membrane sensors. “Once the geometry is optimized, our collaborators in the MORGaN program will be able to design the most appropriate transducer. We are getting close to that stage now."
A team of suppliers of high-performance, diamond-base substrates is underpinning much of the effort in the MORGaN project. Diamond promises to be a great foundation for these devices, thanks to its incredibly high thermal conductivity of typically 2000 Wm-1K-1. This enables it to act as a very efficient heat spreader that rapidly sucks the heat away from the localized hot spots within devices.
Providing funding for this aspect of the program might raise a few eyebrows, because the epitaxial teams could simply import the diamond composite wafers that are under development by firms such as sp3 Diamond Technologies and Group 4 Labs. However, the Element Six R&D Operations Manager Geoff Scarsbrook, who is heading-up this part of the MORGaN project, says that funding of diamond-based substrates is justifiable, because the technology is far from mature: “Whilst the benefits of incorporating a diamond layer are well understood, the optimum technology for integrating the diamond into the high-power GaN device technology is still under development". He believes that the approaches of others have strengths and weaknesses, and he is confident that Element Six can make a significant contribution, because it is at the leading edge of all aspects of diamond technology. In addition, Scarsbrook suggests that there is a secondary reason for the funding of substrate development - it could lead to an internal supplier of these products for Europe.
Element Six provides partners in the MORGaN project with a variety of diamond-based substrates. These players have the option of growing directly onto single crystalline or poly-crystalline substrates, or depositing their epilayers on a diamond-on-silicon composite. A third option is also available – post-growth bonding of the epilayers to diamond.
Scarsbrook says that all three approaches face the same challenge - managing the strain that is generated at the interface. This arises from the significant differences in thermal expansion coefficients of the materials involved. Element Six produces single crystal diamond by a hightemperature, high-pressure process (HPHT) that involves temperatures of 1400-1600oC and pressures 5-6 GPa, and also by CVD, which offers greater control of purity. For MORGaN the company is developing and supplying both single crystal and polycrystalline diamond substrates produced by the CVD route, as well as some HPHT single crystal. One of the key questions that MORGaN will address is whether it is possible to deposit device quality single crystal GaN on polycrystalline diamond. CVD is also used to make diamond-silicon composites. The diamond layer can have as many as three roles to play in this type of substrate, according to Scarsbrook: it has to act as a heat spreader; it may be needed to provide some mechanical strength, either for processing steps, or in the final device; and it has to provide a backplane to the device. “This backplane is positioned on the opposite side of a dielectric, and the appropriate distance – which determines the thickness of the diamond – is around 100 μm."
This substrate is produced by CVD deposition of diamond on silicon. Element Six can produce 4-inch wafers by this route, but it is only shipping a 2-inch version to its partners in the project at present, because this size is compatible with the majority of processing tools used in the MORGaN effort. Work on the substrate is currently focused on thinning of the silicon layer, in order to stop the silicon from acting as a thermal barrier, and to provide a good interface for nitride deposition.
Putting together a good package
If the sensors and transistors that are developed in the MORGaN project are to reach their true potential, then they need to be housed in well-designed packages delivering excellent thermal management capabilities. Several partners are working towards that particular goal, including fcubic from Sweden. This start-up uses a combination of high-precision ink-jet printing and elevated temperatures to produce three-dimensional packages from a diverse range of materials, including stainless steel, titanium, and silver. Delage says that this approach can make relatively large packages, and it is compatible with medium-sized markets requiring up to 10,000 pieces, making it suitable for microwave product manufacture.
Solders used in these packages must also withstand high temperatures. The Swedish firm Impact Coatings is active in this area, and it has developed a new refractory metal with a very good sheet resistance that can operate at 1000oC.
So it seems that MORGaN has an answer to every potential problem associated with developing AlInN-based devices for extreme conditions. This should serve them well in their quest to raise the performance bar of nitride sensors and transistors during the remaining two years of this project.