DARPA sets tough goals for the wide-bandgap community
Military needs are continually evolving and, where electronics is concerned, one of the prime needs is for compact devices that are capable of handling high power densities. Wide-bandgap semiconductors including SiC and the III-nitrides are material systems with the potential to meet this need. Although III-nitride optoelectronic devices and some high-power SiC devices are commercially available, there are pres-sing materials issues that need to be over-come in order to allow SiC and III-nitride electronic devices to perform at the levels required by military users.
The WBSTI is born
Over the years, DARPA has funded a number of programs to develop wide-bandgap semiconductors for military needs. Past programs have focused on areas such as GaN LED technology, solar-blind detectors, AlGaN-based focal-plane arrays, SiC technology for power control and distribution, and also megawatt-class power control devices. In addition, there has been much activity at the Office of Naval Research (ONR) and other areas of the DoD.
"For the last couple of years we have been looking at how to consolidate all of these activities and establish a new initiative that will really push forward this technology into the DoD," said Edgar Martinez, a program manager at DARPA s Microsystems Technology Office (MTO). "As a result of discussions and planning within the DoD, the Wide Bandgap Semiconductor Technology Initiative (WBSTI) was created."
The birth of the WBSTI did not go unnoticed by the compound semiconductor community. On September 5, 2001, DARPA held an industry briefing at which Martinez and MTO colleague and fellow WBSTI manager John Zolper spelt out the aims and objectives of the WBSTI (figure 1). "We received proposals in the fall, then did a first selection and initiated the contract process over the wintertime," explained Martinez.
Work got underway in earnest earlier this year, with a number of companies familiar to those tracking the compound semiconductor industry being awarded contracts. In July Cree announced that it had been awarded $14.4 million to develop high-quality 4 inch SiC substrates, epitaxial processes for SiC MESFETs and GaN HEMTs, and also to develop SiC high-power switching devices. Emcore and Crystal IS were other successful bidders to the WBSTI, receiving contracts for the development of GaN epitaxial processes and large-area AlN substrates, respectively (Compound Semiconductor July 2002 p11). Zolper and Martinez have also been busy getting their message across at industry events, such as this year s GaAs Mantech and the International Microwave Symposium.
Tough targets for researchers
The WBSTI is divided into two areas called Thrust I and Thrust II. Thrust I concentrates on developments that will enable new analog RF, microwave and millimeter-wave applications. Thrust II will endeavor to make advances in solid-state devices for high-power conversion and switching applications, and as such is mainly concerned with SiC.
The WBSTI is being run in phases with the first phase, which began this year, intended to provide a foundation for Thrust I and II by achieving advances in bulk and epitaxial material quality, and by tying these to device performance. Activities in phase one include developing crystalline growth techniques leading to large-area epi-ready semi-insulating substrates, including SiC as well as alternatives such as GaN and AlN. Also included are studies of epilayer growth processes.
Phase one is expected to last two years and has a budget of nearly $40 million. At present about 75% of the funding goes to industry, with the remainder going to universities and federal laboratories. Subsequent phases will concentrate on device and circuit technology and subsystem engineering. These later phases will begin once phase one is completed, and will depend heavily on the advances made in the next two years.
DARPA s wish list for SiC specifies substrates greater than 100 mm in diameter with high resistivity (>107 (Ωcm), a thermal conductivity of at least 4 W/cmK, and less than 10 micropipes/cm2. For epitaxial material, parties funded through the WBSTI will need to aim for homoepitaxy and heteroepitaxy of SiC and/or III-nitrides with a less than ±1% variation in thickness, composition and doping over the entire area of a 100 mm or larger substrate.
Advances in substrate and epiwafer quality on their own are of limited use if they can not be correlated with improvements in device performance. Such correlation helps to ensure that the materials are meeting the requirements of the device types that the military user needs, and also helps to establish the degree to which material improvements are aiding the development of high-performance, manufacturable devices.
The WBSTI includes tight controls at every step, from bulk crystal growth to device processing. Those submitting work proposals are encouraged to work with others in the material-device food chain. For example, organizations working on epitaxial material need to be able to perform device correlation experiments, if not themselves then through partnerships with others. In turn, those developing substrate technologies are expected to work with organizations performing epitaxial growth. Such close co-operation serves as a feedback loop to quickly and efficiently establish whether the desired progress is being made, or if efforts need to be redirected.
High-power electronics technology
The military applications envisaged for the high-power devices that are the subject of Thrust II include motor controllers and power distribution in electric vehicles, and high-frequency power controllers and converters for ship propulsion. The WBSTI defines high-power devices as those that can handle megawatt power levels. This translates to blocking voltages in excess of 10,000 V, conduction currents greater than 100 A and a switch rate of more than 150 KHz.
The material improvements required for the Thrust I RF, microwave and millimeter-wave devices are also needed for high-power devices. However, high-power devices have their own set of material-related challenges. The types of applications described above will require devices with areas in excess of 1 cm2 (figure 2). Reducing defects such as micropipes in SiC substrates is therefore crucial in achieving good yields and reliability for large-area devices. High-power devices can also require thick epilayers (approximately 150 µm). Maintaining crystal quality, good surface morphology and doping control through a thick layer across a large-area substrate, while still growing at a reasonable rate, is challenging but essential to the reliability and manufacturing yield of high-power devices.
Looking ahead to the later phases, process technology is another area that will be investigated in Thrust II. A better understanding of how factors such as surface preparation and treatment, dopant activation, and different insulators impact device performance at high powers is needed. The hybrid and monolithic integration of high-power devices with different component types also needs to be addressed. Such considerations are particularly important to the intelligent operation of power devices. This requires the adjustment of power levels to a system s needs using low-voltage controller ICs. Co-packaging such disparate device types is particularly challenging, and the WBSTI is to address device integration strategies and the management of the high thermal loads generated by high-power electronics.
The next success story?
The wide-bandgap device area is one of the good prospects for high growth in the compound semiconductor industry. The market for III-nitride optoelectronics continues to experience excellent rates of growth compared with the industry as a whole, thanks to high-volume applications such as LED displays and the emerging markets for solid-state lighting and lasers for the forthcoming generations of DVD players.
At present the commercial market pull is limited for high-power GaN- and SiC-based devices. The technology is either too immature to fabricate devices with the required performance and reliability, or is too expensive. Cheap technologies such as silicon LDMOS are firmly entrenched in the market for high-power electronics, despite the inefficiency and the large area of such devices when compared with SiC and GaN. If history repeats itself, all this could change in the coming years. SiC and III-nitride power devices will become much cheaper and reproducibly manufacturable, as did GaAs, thanks to a big push from the military.
