Exotic Transistors With Silicon Credentials

8 December 2010—Slashing transistor sizes drove the manufacture of ever-faster, lower-cost silicon chips during the last few decades of the 20th century. But more recently, miniaturization has gone hand in hand with incorporating a host of increasingly exotic materials.

The biggest change might be the addition of III-Vs, compounds that combine elements from group III of the periodic table, such as indium and gallium, with those from group V, such as phosphorus and arsenic. Transistor developers have been interested in these materials since the 1960s, because they move electrons around far faster than silicon can. Over the past decade, researchers have made steps toward using silicon and III-Vs to make hybrid circuits. However, nobody has ever made III-V transistors on large wafers that silicon foundries could process into minuscule devices with their state-of-the-art tools—until now.

Sematech, a nonprofit consortium of major semiconductor and chip-manufacturing equipment makers that performs basic research on chipmaking, has put this to rights by producing indium gallium arsenide transistors on 200-millimeter silicon wafers. Sematech engineer Richard Hill detailed the accomplishment this week at the IEEE International Electron Devices Meeting, in San Francisco.

These transistors display very encouraging characteristics, say engineers familiar with the work. Sematech’s use of real foundry tools to produce the transistors improved device uniformity, cutting the variation in the voltage needed to turn on III-V transistors to a level similar to that of silicon devices. What’s more, the electrons in Sematech’s transistors travel four times as fast as those in silicon devices of a similar size.

Purdue University’s Peide Ye, an expert on gallium arsenide transistors, is delighted to see that the platform for fabricating these transistors has advanced from expensive III-V substrates to larger, cheaper silicon ones. "This work gives us more hope that III-Vs in mainstream silicon can happen," he says.

Winning approval to run trials of III-V transistor production on cutting-edge silicon processing tools is tough. There is always the risk of contaminating equipment with microscopic amounts of III-V materials, which linger there and then find their way into silicon devices, where they can compromise reliability and alter electrical characteristics. Cleaning up such contamination could mean millions of dollars in lost production.

To reduce the chances of contamination, Hill and his colleagues modified processing procedures to reduce the risks and used techniques such as total reflection X-ray fluorescence to scrutinize the cleanliness of the tools.

These trials show that silicon foundries can produce III-V transistors without contaminating their tools. "But there needs to be a much larger data set to say this with 100 percent confidence," says Hill.

While this news is encouraging, there is much more work to do before commercial III-V-on-silicon combinations can be launched on the market. Today such transistors are riddled with defects, a problem that stems from a significant difference in the spacing of atoms in the indium gallium arsenide layer and in the underlying silicon substrate. If this crystal compatibility problem were left completely unchecked, the indium gallium arsenide would be so defect ridden that the devices built on it would not work at all. As a solution, Sematech engineers insert a buffer material between these layers. By varying the material’s composition, they can tune the buffer’s atomic spacing to bridge the gap between silicon and indium gallium arsenide.

Although this approach worked, the 2-micrometer-thick buffer must be thinned to no more than a quarter of that value in a commercial device. And there are still too many defects created. These must be reduced from billions per square centimeter to hundreds of thousands to keep transistors from leaking current when they are supposed to be turned off.

To build a successor to silicon CMOS, engineers must also slash the transistor’s gate length—the distance from source to drain—from 0.5 µm to around 10 nanometers. They’ll also have to come up with a III-V transistor that uses holes—positively charged electron deficiencies—instead of electrons as the charge transporter. Germanium is the leading contender for making this type of device, but antimonides offer an "all III-V solution," according to Hill.

All this effort will be in vain unless it is accompanied by the development of manufacturing tools and processes. "If you work back from the projected dates, there’s actually quite a tight schedule to introducing III-Vs," says Hill. "It will take a couple of years to order tools, install them, and ramp up manufacturing, so infrastructure development must start now to get on the correct time scales."