GaAsing Up Cellphones
Gallium arsenide transistors could power tiny, blazingly fast multimedia handsets
A practically unnoticed announcement in January from Freescale Semiconductor Inc., in Austin, Texas, set the small community of researchers who study exotic semiconductor transistors buzzing. A group of Freescale researchers led by Matthias Passlack had fabricated metal oxide semiconductor field-effect transistors (MOSFETs), the types that drive just about every silicon integrated circuit, using gallium arsenide (GaAs) and a novel gate dielectric.
The rest of us should be buzzing, too. If Freescale and other research groups can overcome some significant manu--facturing challenges, this inno---vation could lead to a cellphone-on-a-chip and instant analog-to-digital conversion. It may even enable chip makers to improve processor speed and performance when transistors on silicon chips can be miniaturized no further.
Improbably, this groundbreaking work originated in the semiconductor backwater of East Germany in the 1980s. That's where Passlack [see photo, " "], then at the Dresden University of Technology, began his obsessive pursuit of gallium arsenide digital ICs. After the Berlin Wall fell in 1989 and he was free to emigrate, Passlack eventually joined Bell Laboratories' optoelectronic device research group in Murray Hill, N.J., which was at the time focusing on III-V semiconductors.
Gallium arsenide and other III-V semi-conductors are a better choice of materials than silicon for lots of things, including light-emitting diodes and lasers. These compounds, which combine elements from the third and fifth columns of the periodic table, conduct electrons up to 20 times as fast as silicon does.
Despite its advantages, gallium arsenide has never made it into commercial-grade microprocessors or memory
circuits. That's because until the Free--scale breakthrough, the material couldn't be used to fabricate MOSFETs, which form the basis of the complementary metal oxide semiconductor (CMOS) circuits used in microprocessor, RAM, and microcontroller chips. In a MOSFET, a voltage on the gate sets up an electric field in the channel, allowing current to flow. Prior to Freescale's discovery, no material had been found to provide an effective insulating layer between a gate and the channel through which current flows.
Upon his arrival at Bell Labs in 1993, Passlack and his colleagues decided to build a gallium arsenide MOSFET. "We [agreed] that's the real killer application," Passlack says. "GaAs MOSFETs had been pursued for 30 years, and every-one just miserably failed. And I thought, that's just the right challenge I want to take on."
As Passlack and co-workers discovered, they could evaporate single-crystal gadolinium gallium oxide (Gd 3 Ga 5 O 12 ) to deposit gallium oxide (Ga 2 O 3 ) molecules onto the GaAs surface to act as a dielectric. But stray gadolinium molecules would inevitably sneak in, which caused defects at the GaAs/Ga 2 O 3 interface.
Passlack left Bell Labs for Motorola in 1995. There he found a pure source of Ga 2 O 3 : polycrystalline ingots of gallium oxide. When evaporated onto GaAs substrates, the gallium oxide formed the high-quality interface that he had been searching for. But by itself, Ga 2 O 3 is a poor insulator. To enhance the dielectric's insulating qualities, Passlack turned to gadolinium. He used ingots of Ga 2 O 3 in a tightly controlled process to create a stack of pure Ga 2 O 3 monolayers less than 1 nanometer thick, topped by almost 20 nm of gadolinium gallium oxide deposited from beams of Ga 2 O 3 , Gd, and oxygen.
Passlack's group managed to make transistors (the MOS-heterojunction FET, a kissing cousin to the MOSFET) using Ga 2 O 3 as the gate dielectric in 1997. And his former colleagues at Bell Labs, including Minghwei Hong and J. Raynien Kwo, also produced some experimental GaAs MOSFETs around the same time. But none of these had commercial performance characteristics. Despite those steady if modest successes, around 2001 Motorola put additional research into GaAs MOSFETs on the back burner, only to revive the program in 2004. Still, it remained niche research until Intel Corp. put III-V transistors on its digital CMOS road map one year later. "All of a sudden," Passlack says, "people started listening up."
Inspired by the renewed interest in III-V MOSFETs, in October 2005 Passlack and his colleagues reported in IEEE Electron Device Letters that they had for the first time successfully paired a Ga 2 O 3 /GdGaO dielectric stack with an indium-gallium-arsenide channel layer to create a GaAs MOSFET structure. Soon after, Passlack's team added ohmic contacts and a gate electrode to form a complete transistor that was about 20 times more efficient than any GaAs "enhancement" MOSFET ever made, prompting the January announcement. (In an enhancement device, when the gate voltage is zero, the device is off. Voltage on the gate enhances the channel, turning the device on.)
The Freescale GaAs MOSFET exhibits carrier mobility in the conducting channel that is 30 times higher than that of silicon MOSFETs, even those using a next-generation gate insulator, hafnium dioxide.
Freescale's Ga 2 O 3 /GdGaO dielectric stack has electrical characteristics similar to those of hafnium dioxide, and that got the attention of Jack Lee, a professor in the department of electrical and computer engineering at the University of Texas at Austin. "Right now we're working on hafnium dioxide simply because of our experience," says Lee, who leads a team working on III-V MOSFETs. "But I think the oxide that [Passlack's] working on is something that we definitely want to look at, too."
Lee's not alone. The Semiconductor Research Corp., based in Research Triangle Park, N.C., is launching a new research program aimed in part at finding ways to fabricate III-V MOSFETs using CMOS technology. Then there's Intel. Wilman Tsai, who is involved in III-V research at Intel, would say of Freescale's innovation only that some of the content in this article "is very sensitive to what Intel is doing, so we are not in a position to make public comment."
So far, only discrete devices (just transistors, as opposed to transistors incorporated into circuits) have emerged from Passlack's team. But if Freescale can integrate GaAs MOSFETs into circuits--by all accounts a difficult but not insurmountable task--several intriguing possibilities open up.
"One of the applications could be as a replacement for the silicon LDMOS [laterally diffused MOS] microwave power transistors that are used in power amplifiers for base stations," says Stan Bruederle, research vice president of Gartner Semiconductor Research, San Jose, Calif.
Ultra-low-power handsets that you charge once a month could also be in the offing, according to Karl Johnson, director of Freescale's microwave and mixed signal technology laboratory in Tempe, Ariz. In fact, GaAs MOSFETs could completely revolutionize handset design by integrating power amplifiers, transmit/receive switches, and power controls on one chip.
Furthermore, a cellphone's typically sluggish multimedia experience might be amped up considerably using GaAs integrated circuits. As Johnson explains, if you could directly convert the analog signal as it comes into the antenna at a higher frequency into a digital signal, you would eliminate a lot of the components that sit between the base band and the RF section of the phone.
As analysts and Freescalers alike caution, there's a long way to go before GaAs MOSFETs make it to market. Even so, Passlack has given Freescale a huge head start on something a lot of other chip makers now want to develop themselves.