Last Updated on Sunday, 08 April 2012 07:51 Published on Thursday, 04 February 2010 20:08
Researchers are running into the physical limits of speed and scaling in silicon transistor technology, forcing them to look elsewhere for next-generation devices.
The leading candidate to replace silicon being pursued by, well, pretty much everyone, is graphene. Graphene, single sheets of graphitic carbon, is exciting because it is a single atom thick and has remarkably high electron mobilities (100 times greater than silicon), making it ideally suited to atomic-scale, high-speed operation.
Also, graphene's electrical properties can be controlled, switching it among conducting, semiconducting and electrically insulating forms. That means graphene-only (or, more likely, graphene-mostly) devices are, in principle, possible.
In this week's Science, researchers from IBM demonstrate graphene-based field effect transistors (FETs) that may operate at much higher speeds (100GHz) than Si FETs. Graphene layers were thermally grown on two-inch SiC wafers and the FETs were formed using standard Si fabrication techniques with HfO2 as the gate oxide. That's a rather significant point—the researchers actually created an entire wafer of these devices.
The smallest gate length demontrated in the paper was 240nm, quite large compared to current generation Si (32nm), but the graphene was one or two layers (meaning one or two atoms) thick in all the tested devices—a considerable improvement over Si.
High frequency operation, colloquially referred to as the speed of the transistors, was the key property examined in the paper. As operating frequency increases, electrons have less time to respond to the electrical fields that drive transistors, which will eventually cause the transistor to fail because the electrons simply can't conduct across the material fast enough.
The graphene FETs in this work were tested up to 30GHz and, extrapolating those results, the authors showed that the FETs would operate, albeit poorly, up to 100GHz. Similarly sized Si devices are limited to 30GHz operation. Assuming these devices can be scaled, they will undoubtedly present a dramatic speed increase over current generation Si.
Because the graphene used in this study was conductive (i.e. no band gap), the demonstrated voltage-current characteristics were strange compared to Si. Specifically, current continued to increase linearly with drain voltage up to device breakdown. Si-based transistors typically have a point, called the threshold, at which a current cannot increase despite increasing drain voltage.
This study is a mixed bag of promise and hype. The 100GHz speed touted in the article's title is an extrapolation—no such properties were actually measured. Also, the electron mobilities, the key property for high frequency operation, that the authors measured in the fabricated devices were pedestrian compared to graphene's potential, probably due to the thermal process used to synthesized the graphene layer. Future devices could dramatically outperform these FETs if wafer-scale fabrication can replicate some of the better electron mobility measurements of graphene.
Graphene devices have grown by leaps and bounds over the past few years, and they are probably the best bet to eventually replace silicon. Demonstrations like this are important because they show that wafer-scale production is possible, and the properties, while not ideal, are truly impressive, in that they're already beginning to push the limits of Si technology. [arstechnica]