Europe Improves Reliability Of GaN Microwave Devices For Space Applications
GaN is the most promising semiconductor since silicon. Its attributes enable it to create bigger, brighter TVs, and better performing mobile phones and wireless base stations. And its potential is not just limited to terrestrial applications – in fact, it actually holds particular appeal for the space sector. In this environment, devices made from GaN promise to deliver reliable operation at far higher voltages and temperatures than current equivalents built from silicon and GaAs, and initial results indicate that this wide bandgap variant is also far better at withstanding the high levels of radiation found in space. The attractions of deploying GaN in space have spurred development of microwave devices designed for that environment. Benefits of switching from the incumbents to this wide bandgap semiconductor include a major improvement in radiation hardness, a five-to-ten-fold increase in RF power output and the opportunity to reduce the mass and size of cooling systems. If high reliability GaN microwave devices are produced, they could be widely deployed in space because microwave signals form the backbone of space communications, thanks to their suitability for carrying broadband data that can pass through the Earth’s atmosphere. In addition, the microwave region of the electromagnetic spectrum can also be used to transmit spacecraft telemetry data back to Earth and deliver vision to radar imaging missions, such as Europe’s Envisat. Commercial production of GaN power transistors has been spearheaded by the US firm Cree and the Japanese chipmaker Eudyna (now part of Sumitomo Electric Device Innovations Inc.), who both commercially introduced devices on to the market about five years’ ago. However, evaluating these overseas devices’ performance and scrutinizing their reliability data is not always easy for any organization outside the borders of the manufacturer, due to International Traffic in Arms Regulations (ITAR) controls and restrictive end-user licence agreements on these components. When microwave products from Eudyna and Cree were unveiled, the primary developers of GaN microwave components within Europe were research institutes and universities. Funding for these programmes came through national agencies, with efforts supported by innovative component development activities at the European Space Agency (ESA) and EU defence initiatives, such as the Korrigan project. Although these activities produced a portfolio of devices with very impressive performance figures that were reported in the 2002-2006 timeframe, European commercial suppliers of GaN microwave devices still had to overcome reliability-related issues before establishing themselves on a commercial basis. Reliability requirements If GaN devices are to be deployed in space applications,they must deliver high levels of reliability withoutcompromising performance. Typical satellite operatinglifetimes are eight-to-nine years for Earth Observationmissions and upwards of 18 years for telecommunicationsatellites with little room for failure. To address this, in2008 the ESA launched its GaN REliability And TechnologyTransfer initiative (GREAT2), with the aim of establishinga European supply chain for the manufacture of highreliability, space compatible, GaN-based microwavetransistors and integrated circuits that would be free fromany ITAR or end-user licence restrictions. The first phaseof the project, which focuses on device reliability, has secured €8.6 million of funding and there are plans in place for a follow-on phase that will concentrate on validating and qualifying the manufacturing process for the special requirements of the space environment. Seven academic and industrial partners from four European countries are involved in the programme, and each has a specific responsibility (see Figure 1 for details). Figure 1: The project partners within GREAT2include foundries, research institutions and universities The German firm Tesat-Spacecom (Tesat) is coordinating and managing the project, which is divided into five primary work packages (see figure 2). The aim of the largest of these, WP2000, involves undertaking process development trials to establish a space-compatible foundry process. Performance validation is being undertaken using L-band and X-band technology evaluation structures. Figure 2: Activities within the €8.6 million GREAT2 project have been divided into several work packages This work package is led for ESA by United Monolithic Semiconductors (UMS). It is responsible for establishing the final commercial foundry process. Additional support for this effort comes from: IAF and FBH, who are undertaking specific processing trials under the guidance of UMS to help improve the reliability performance of the foundry process; Tesat, which is responsible for device packaging and reliability assessment in RF packages; and the University of Rome, which is providing support for simulating and understanding the physics of GaN HEMT failure. Two types of foundry process are being produced by UMS in this programme: a 0.5 μm gate length process (GH50) for fabrication of discrete GaN HEMTs for operation up to 6 GHz, and a full MMIC process using 0.25 μm gate length technology (GH25) for operation to about 20 GHz. Two of the other work packages, WP3000 and WP4000, are focusing on optimising device processing and epitaxial growth on silicon and SiC substrates. One of the primary aims of this work is to confirm whether device passivation produces better, i.e. more reliable, devices when this process is carried out in-situ, rather than exsitu. The in-situ nitride devices are being fabricated at imec, and the ex-situ nitride devices fabricated at Fraunhofer IAF. In the WP3000 programme, accurate measurements of channel temperature, which are needed to determine the mean-time-to-failure in accelerated lifetime tests, are being obtained through Raman micrography measurements at the University of Bristol. The GaN microwave devices made during the project are being assessed under space environmental operating conditions by Tesat as part of the work package WP5000. By the end of the first phase of the project, which is scheduled for completion in November 2012, more than 170 wafers will have been fabricated and assessed for reliability and suitability in a space operating environment. The successes of these efforts will be judged against performance-related milestones, which feature incremental increases in performance and reliability over time, and were drawn up by the ESA. Meeting the interim milestones M3 and M5 requires device operation in excess of 1000 hours and 10,000 hours, respectively, and fulfilling the final goal, M7, requires demonstration of device operation of at least 20 years at a junction temperature higher than 230 °C. In addition, the device must be resilient to space environmental effects, including radiation, hydrogen poisoning and electro-static discharge. This level of performance must be realised for both an L-band discrete power transistor and an X-band MMIC for the GH50 and GH25 foundry processes respectively. What’s more, specific performance and manufacturing yield targets must be met at each milestone. When the project kicked-off in 2008, engineers at UMS put together a technology development plan that defined the methodology for technology transfer within the consortium. This included a patent survey and the signature of an intellectual property agreement to allow exploitation of key results. At the outset, a great deal of effort was devoted to establishing a common methodology for routine device assessment, reliability testing and the analysis and display of measurement data. To allow a fair and sensible crosscomparison of data, all mask sets featured common test structures, such as 1 mm gate-width RF transistors, process control monitoring structures and dedicated radiation test cells. All the partners within the consortium evaluate these structures consistently – this was confirmed to be the case with a ‘round-robin’ test campaign on an evaluation wafer. The multiple batches of epitaxial wafers that have been used for experimental processing trials within the consortium were grown by IAF and imec (see figure 3 for an example of the growth tools used by IAF). SiC and silicon substrates provided a foundation for these epiwafers, which featured intentional variations in AlGaN composition and GaN buffer structure and were fabricated with both in-situ and ex-situ nitride passivation. UMS, a firm with considerable manufacturing experience, provides feedback concerning the optimum choice of epitaxial structures and processing recipes to be considered. Figure 3. Engineers at Fraunhofer IAF in Germany are fabricating GaN-on-SiC epiwafers with an MOCVD multi-wafer reactor By the end of 2010, two wafer-batch-processing trials had been successfully completed at FBH and imec, plus three wafer batch trials by Fraunhofer IAF. In addition, UMS had finished processing 30 wafers focusing on ‘pre-process freeze’ engineering variations of its GH50_10 production process and development trials on the GH25_10 production process dedicated to GREAT2. In total, over 74 wafers had been fabricated and assessed, providing enough information to understand the impact of the epitaxial design on the reliability of the structures. Initial processing batches were screened for DC, pulsed DC and RF performance, plus on-wafer reliability. Wafers delivering the most promising performance were diced and used to form packaged test samples that were evaluated by Tesat. Both the L-band HEMT and the X-band MMIC passed M3, the first critical milestone (see Figures 4 and 5 for details). Figure 4: Devices made for the M3 milestone of the GREAT2 project include an L-band discrete GaN HEMT. Operating at 50 V and a baseplate temperature of 50°C, this transistor delivers approximately 10W of output power for >15 dB associated gain and a power added efficiency (PAE) >40 percent at a frequency of 1.7 GHz. The PAE performance was not optimised for the reliability test campaign, but with appropriate matching and a deep class AB quiescent bias point the typical PAE value easily exceeds 60 percent with the UMS GH50 process Figure 5: The X-band evaluation vehicle built for the M3 milestone is a two-stage MMIC. Operating at 30 V and a baseplate temperature of 50°C, this device delivers more than 6W of RF output power over the 8 to 8.5 GHz frequency band, with around 18 dB associated gain and a PAE of typically 30-40 percent depending upon the quiescent bias condition To assess the RF reliability criteria within the M3 milestone, the consortium assembled 68 L-Band and 67 X-band parts in hermetic packages and a further 64 test cells that were housed in DIL24 ceramic packages, specifically for performing radiation and hydrogen poisoning tests. UMS GH50 transistors and IAF GaN25 MMIC devices passed accelerated RF life tests carried out by Tesat in December 2010. Confirmation of reliability to 1000 hours of operation came from independent, internal testing at ESA. Figure 6: As part of the assessment for the M3 milestone, researchers studied the change in RF output power for 11 packaged L-Band test vehicles with a peak channel temperature of more than 230°C. Some of these devices were operating at 4 dB of gain compression (Tesat), and others at 6 dB of RF gain compression (ESA). Both devices were also assessed at two different baseplate temperatures – 125°C and 150°C. No burn-in or screening, other than electrical performance and assembly integrity test, has been performed before life testing In fact, tests on 11 packaged L-Band test vehicles show that, in general, the RF output power changes very little over 1400 hours of operation at a peak channel temperature in excess of 230°C. Ten devices had a maximum drift in RF output power of less than 0.5dB – just one showed rapid degradation at the start of the test. However, this particular device had been subjected to failure analysis to identify the physical changes occurring within the device and to better understand any underlying failure mechanisms. No burn-in or screening other than electrical performance and assembly integrity test has been performed before life testing. The radiation tests that have been performed have scrutinized the impact on transistor performance following exposure from gamma radiation, total ionising dose, proton radiation (displacement damage) and single event burnout effects under heavy ions. In assessments made against the M3 milestone, which were carried out using the GREAT2 radiation test cell (see figure 7), no devices from the batch underwent any appreciable DC parameter drift for a total ionising dose of 1Mrad and for a proton fluence of 1.7x1012 p/cm2 (35 MeV proton energy). However, under heavy ion excitation (Xenon: LETGaN=52.93 MeV/mg/cm2) the open channel current and threshold voltage for UMS GH50 devices drifted by about 10 percent. This variation is still comfortably below the target drift specifications of less than 15 percent. Initial tests on devices fabricated with the UMS GH50 process revealed that DC static burnout occurred at 200 V, while under heavy ion excitation single event burnout typically occurred at 150 V. To verify these findings, the consortium will conduct further tests on multiple wafer batches. Further radiation test campaigns, planned for M5 and M7 milestones, will also scrutinize any changes to larger devices under both DC and RF operating conditions. Figure 7: Packaged radiation test cell (left) and the heavy ion test set-up (right) at Lovain La Neuve, Belgium The first wafer batches have also been subjected to 24- hour hydrogen poisoning tests, which involve subjecting unbiased devices at 250°C to a mixture of 5 percent hydrogen gas and 95 percent nitrogen gas. These initial measurements indicate that performance is not particularly sensitive to this effect and will be further validated by incresing the hydrogen test exposure time to several hundred hours. By December 2010, evaluation devices had passed all the tests related to the M3 milestone, an encouraging achievement that suggests there should be no major showstoppers for using GaN technology in space. That is not to say that lessons have not been learnt from this exercise. This programme of work has shown that controlling the strain in the epitaxial layers, appropriate tailoring of the electric field distribution close to the gate, appropriate choice of surface pre-treatments and choosing appropriate metallisation schemes to ensure gate thermal stability are all essential requirements to achieve high reliability, and modifications have been made accordingly to new process batches. The primary goal of these adjustments is further improvement in operating lifetime, while at the same time satisfying performance and manufacturing yield targets. If these efforts are successful, they will play a big part in helping Europe’s space industry to maintain its competitive edge. One of the big hopes behind the GREAT2 initiative is that it will foster an independent GaN supply chain for space, while also being applicable to dual-use markets. Success on both these fronts promises to create sufficient volume demand to sustain European capability over the long term. However, this is not the only benefit of the GREAT2 project. During the first year of this programme there has been an opportunity to demonstrate the use of the X-band test evaluation prototypes, fabricated by IAF, in an experimental X-band telemetry transmitter due to be launched in 2012 on PROBA V, ESA’s small satellite for global vegetation mapping. Syrlinks is carrying out the transmitter design and fabrication, with spacecraft integration and test the responsibility of QinetiQ Space nv (see figure 8). This demonstration offers a unique opportunity to combine extensive test data that has been gathered on the ground with performance and test data achieved in a real-life, in-orbit operating environment. If successful this will be the first launch of European-sourced GaN technology, which will provide a fantastic feather in the cap of the GREAT2 project. Figure 8: MMICs produced by Fraunhofer IAF are assembled into hermetic packages by Tesat, and finally incorporated into an X-band telemetry transistor that will be launched in 2012 on PROBA V, an ESA satellite for global vegetation mapping Resources An overview of GREAT2 project is given at http://www.esa.int/SPECIALS/Technology/ SEMJ1X9BWUF_0.html More information on the PROBA V can be found at http://www.esa.int/esaMI/Proba/SEM7QUYOBFG_0.html © 2011 Angel Business Communications. Permission required.