Freescale Puts MOSFETs On The Map
Freescale Semiconductor s MOSFET development team appears to have made a genuine breakthrough in developing a device that could one day be scaled in much the same way as silicon CMOS technology.
If it does prove viable, the development could pave the way for a GaAs equivalent of Moore s law, by which the speed and size of transistors can be advanced rapidly through lithography and other standard processes.
Finding a compatible gate oxide material has turned out to be the key stumbling block in the past. And while Freescale s is not the first GaAs MOSFET to be developed (examples of such devices go as far back as 1965), the company does see it as the first such design to be commercially viable.
Matthias Passlack is a key member of Freescale s MOSFET project group based in Tempe, AZ. "All previously published enhancement-mode GaAs MOSFET data showed performance of less than 1%, with transconductance typically in the microSiemens per millimeter range," he explained. "There has been no progress over the last 30 years, and 1% devices are of no interest for commercial application. Our [80 mS/mm] technology provides a performance that shows promise for applications for the first time."
Passlack highlighted the high defectivity seen at the oxide/GaAs interface - something that causes Fermi-level pinning - as being the main problem with previous MOSFETs. Solving this difficulty was the first of two key breakthroughs that Passlack says were crucial to the development.
To unpin the Fermi level, the Freescale team deposited an amorphous Ga2O3 template by MBE. They then used an amorphous dielectric layer of GdGaO, which provides the required band offsets while neither disrupting the template nor creating a secondary interface. The second crucial part was a new enhancement-mode device concept, which provides the typical terminal behavior of inversion mode devices, but does not use surface inversion. "This allows one to take advantage of the intrinsically high mobility of III-V semiconductors in MOSFET channel layers," explained Passlack.
Freescale s senior executives are also raving over the development. "This remarkable achievement overturns industry assumptions," added Freescale s acting CTO Sumit Sadana. "[It] has the potential to fundamentally change the way high-performance semiconductors are designed, manufactured and deployed."
The biggest silicon house of them all - Intel - has already acknowledged the potential of III-Vs in its own technology roadmap, seeing GaAs as an option beyond the 22 nm mode required around 2012. If III-V materials are ever going to cross over into the digital mainstream, then a high-k dielectric MOSFET is essential, as these are the only devices that are scalable to such small dimensions. "As far as I can tell," Passlack told Compound Semiconductor, "there is a consensus in the industry that III-V MOSFETs must be investigated for future digital applications. Which channel material and device concept [will be used] is still hotly debated, although there is also a consensus that this must happen on a silicon platform."
That would present another difficulty - how to grow lattice-mismatched materials with low defectivity.
Another company that has experience in making GaAs MOSFET material and devices, and that worked with Freescale in the early stages of device development, is the Rochester-based MBE specialist Osemi. Recognizing the importance of Freescale s breakthrough, Dave Braddock from Osemi says that the convergence of compounds and silicon as outlined in Intel s roadmap will fundamentally change the overall semiconductor industry, and that III-V MOSFETs in both GaAs and GaN flavors will open new doors in digital and mixed-circuit electronics.
Freescale says it now has plans to work in collaboration with other companies to develop products that make use of the high-speed computing performance enabled by GaAs MOSFETs. Karl Johnson from the company believes that digital conversion of analog or radio frequency signals would be almost instantaneous with a fully-functional GaAs MOSFET, offering to greatly simplify wireless architectures and improve talk and standby times in handsets.
While wireless power amplifiers are dominated by the GaAs HBT currently, these devices are not suited to front-end switching and power control. A GaAs MOSFET could combine the small-die advantages of the HBT and the efficiency of a PHEMT structure with its own intrinsic integration capability.
Back in 2001, Freescale s parent company courted similar fanfare after announcing that it had developed GaAs-on-silicon wafers, which also promised to revolutionize microelectronics. Motorola even formed a separate spin-off company to exploit that technology, but the initial optimism proved groundless as problems with defects saw the hybrid approach shelved.
But the convergence of what have traditionally been seen as the very separate worlds of silicon and III-V semiconductors has remained a compelling topic, and Freescale s latest work will be widely applauded. As always, however, the proof of the pudding will be in the eating. Let s hope that the GaAs MOSFET proves to be a more satisfying meal than Motorola s attempt at 12 inch GaAs-on-silicon wafers, which left a rather disappointing taste in the mouth.