STMicroelectronics is set to drive SiC-device competition

"We feel that the time is right to start investing significantly in SiC," explained Salvo Coffa of STMicroelectronics. Coffa is research director of the SST (soft computing, silicon-optics and post-silicon technologies) group within the company s R&D division.

STMicroelectronics is the world s sixth-biggest semiconductor manufacturer, and in 2003 it had a global market share of 4%. The company, which has its headquarters in Geneva, Switzerland, already has an established portfolio of silicon-based power devices. Over the last few years it has investigated aspects of SiC technology, such as epitaxial growth, metallization, and junction formation. Its reason for doing this is straightforward - SiC can overcome the limitations associated with silicon (see Comparative table).

"What has changed, as was shown in the meeting in Bologna [the 5th European Conference on Silicon Carbide and Related Materials, which took place in September], is that there is an increase in the number of providers of substrates, epitaxial reactors, and so on. This is the only way to start having economies of scale in the manufacturing of these [SiC] devices," said Coffa. He explained that the company has made significant progress in its ability to process SiC, using mostly 3 inch substrates in its existing lines.

STMicroelectronics has yet to produce SiC SBDs commercially, but it has sent samples to customers to be evaluated. Short-term goals include the production of 600 and 1200 V Schottky diodes, and the development of 600 V power MOSFETs.

The temperature dependence of the company s SiC SBDs, which Coffa claims could operate at temperatures of up to 300 °C, is shown in figure 1. These devices will compete directly with Infineon s thinQ! range, which includes 11 different SBDs that are claimed to operate at up to 600 V and at temperatures of -55 to 175 °C.

The temperature stability of SiC SBDs is superior to that of equivalent silicon-based devices, thanks to the wider bandgap. Coffa explained that Schottky devices that can perform at high temperature are important in two applications. First, in devices that operate at high powers and are subject to self-heating; and second in devices that operate in harsh environments, such as the combustion chamber of an automobile engine.

STMicroelectronics has also investigated the dependence of the leakage current from its SiC SBDs on temperature and applied voltage. Coffa explained that it is difficult to give a definitive value for an acceptable leakage current because it depends so much on the intended application. The main problem associated with leakage currents is power dissipation in the "off state", when the diode is under reverse bias. Coffa noted that Schottky diodes have higher leakage currents than bipolar devices, and that individuals would have to decide which device was appropriate.

The dynamic characteristics of SiC SBDs have been compared with silicon diodes, and found to be superior. Silicon bipolar devices, which have similar breakdown voltages, respond more slowly than SiC devices because they require the recombination of minority carriers.

In collaboration with an undisclosed car manufacturer, STMicroelectronics is investigating the use of SiC in pressure sensors in combustion engines. SiC is incorporated into a membrane, and changes in pressure lead to deformation and variations in the resistance of the material. To increase the sensitivity, four piezo resistors are configured in a Wheatstone bridge circuit, forming a package measuring 4 mm x 4 mm x 850 μm. This device, which can withstand high temperatures and pressures, has already been tested in an engine and provided pressure readings with the engine running at speeds of 500, 1050, 1400 and 2000 rpm. Coffa believes that these sensors may be employed in cars in the next 5-10 years, depending on the price. He pointed out that one SiC device could replace several of the sensors that are used in cars today.

Despite these recent advances, Coffa warned against complacency: "Sometimes silicon surprises you. For example, for years people have considered silicon to be unsuitable for photonics, but my group has clearly shown that you can have some applications of silicon in the optical domain. You can have LEDs that are completely made with silicon processes. This doesn t mean that you can cover all the applications of compound semiconductor devices, but in some cases you can use silicon, especially if you want to integrate the optical part with the electronics part."

Another company that manufactures SiC-based power electronics is SiCED. It was established in 2000 as a joint venture between Siemens and Infineon Technologies, and now acts as an R&D center for both of its parent companies. Dietrich Stephani, senior director of the Germany-based company, believes that the two major technological breakthroughs made in the SiC-based electronics industry occurred in 1993, when SBDs fabricated on 6H-SiC substrates showed blocking voltages of 1000 V, and 1 inch 4H substrates became available from Cree. However, these two achievements didn t lead immediately to the commercialization of SiC-based devices, and Stephani believes that only the very optimistic could have foreseen that they would have applications in industry.

Today that progress is evident, with SiC-based devices appearing in everyday appliances. For example, a 2000 W transformer containing Infineon s 600 V SiC SBDs is half the size of its conventional 650 W counterpart; and SiC SBDs are also used in Siemens 2.3 kW solar-converter.

Stephani explained that today s commercially available SiC SBDs operate with blocking voltages of 300-1200 V and currents of 2-10 A. He said that "today s winning combination" is power factor correction in switched-mode power supplies, where a 600 V SiC SBD is twinned with a 600 V low-loss silicon MOSFET. Looking ahead, he expects higher-voltage SiC SBDs, with blocking voltages of 1200-1700 V, to penetrate the low-power inverter sector and DC-AC conversion for solar cells.

Stephani believes that over the next few years silicon-based power switches operating in the 1200-1700 V range will face increasing competition from SiC-based alternatives. Agreeing with Coffa, he pointed out that silicon-insulated gate bipolar transistors suffer from dynamic losses at high switching frequencies, and suggested that SiC junction field effect transistors (JFETS) might offer a high-frequency elevated-temperature solution, with an unparalleled on-resistance. SiCED is developing a hybrid switch by combining a vertical SiC JFET with a low-voltage silicon MOSFET, for use in resonant converters and auxiliary power supplies.

Low channel mobility and long-term stability of the gate oxide are cited as the two key issues hampering the development of SiC n-channel inversion MOSFETs. Stephani noted, however, that recent progress using N2O in the thermal oxidation process had enabled the fabrication of 3 kV SiC MOSFETs with a room-temperature on-resistance of 3 Ω, for a 2.1 mm2 chip area. He believes that within the next three years these devices, which have an attractive cost:performance ratio, could impact the market, although their precise application is not yet clear.

Stephani s optimism does not extend to the commercialization of SiC pin diodes, though. Here SiC has to compete with its well established silicon-based counterparts, which have lower turn-on voltages thanks to their lower bandgap. One area where SiC devices can compete is in operations demanding very high blocking voltages, said Stephani. He thinks there is a need for 6.5 kV, 25 A single chips. High-voltage SiC pin diodes require high current-handling capabilities, and consequently large chip areas. Engineers at SiCED have demonstrated a SiC pin with an emitter area of 5.7 mm2 and a blocking voltage of almost 5 kV, but Stephani explained that today the yield of large-area chips is low because of poor-quality SiC material. This causes high defect densities, which lead to an uncompetitive cost:performance ratio.

Improved material quality and a reduced cost are also required before bipolar SiC power devices can impact the market. Stephani does not expect these devices to appear within the next three years, but he is optimistic about their long-term viability: "High dynamic performance combined with high-temperature operation and blocking 10 kV or more is the ultimate challenge for electronic power switches. There is just one candidate today to solve this task - SiC."

Engineers at SiCED are already addressing the cost of SiC-based power devices by investigating the quality of epilayers grown on a multiwafer reactor. Bernd Thomas, also from SiCED, has set targets that are suitable for the production of 300, 600 and 1200 V SBDs: doping and thickness homogeneities better than 10 and 5% respectively; a doping range (the difference between the densities of donor and acceptor atoms, ND-NA) of 4 x 1015-2 x 1016 cm-3; and drift layer thicknesses of 3-18 μm.

In trials, SiCED s engineers have used 4° off-axis (0001) 4H-SiC 3 inch wafers, which are more affordable than their 8° counterparts. They have also employed a multiwafer hot-wall Epigress VP2000HW CVD system, which is capable of a 7 x 2 or a 5 x 3 inch configuration. Wafer-to-wafer uniformity has been evaluated by growing 18 µm-thick layers, with a doping of 4 x 1015 cm-3, suitable for blocking over 2.5 kV.

Intrawafer values of sigma/mean of 1.6 and 3.3% for the thickness and doping were calculated using an exclusion zone of 2 mm, which Thomas says "meet the requirements for production process management".