Carbon could enable fastest chips

PORTLAND, Ore. — Carbon transistors could outperform the fastest chip materials, including indium antimonide, according to researchers at the University of Maryland.

The College Park team recently characterized graphene monolayers, sheets of pure carbon just one atom thick. They discovered that graphene appears to be unfazed by temperature, unlike most semiconductors. Usually, speed--called electron mobility--is proportional to temperature (the colder the better since fewer lattice vibrations, called phonons, can scatter flowing electrons).

Instead, pure graphene transistors appear able to achieve their maximum possible speed at room temperature, according to the researchers, if chip makers choose the right substrate.

"We measured the electron mobility of graphene monolayers between 50 Kelvin (minus 370 degrees F) and 500 Kelvin (450 degrees F) and found [electron mobility] to be about 15,000 cm²/Vs regardless of temperature, which is unusual," said team leader Michael Fuhrer of the university's Center for Nanophysics and Advanced Materials, and the Maryland NanoCenter.

In silicon, electron mobility is about 1,400 cm²/Vs, and the highest known electron mobility in any material is about 77,000 cm²/Vs in indium antimonide. By contrast, the lattice vibrations in graphene were measured by the Maryland researchers to be so weak that secondary effects like impurities and substrate choice had a bigger impact than phonons.

"What we now think is that phonon scattering of electrons in graphene is very weak, and that leads us to believe that we are being limited by impurities. If we can remove those impurities, we think we can achieve electron mobilities of 200,000 cm²/Vs at room tempterature--which is more than 100 times better than silicon," said Fuhrer.

By comparison, electron mobility in carbon nanotubes has been measured at about 100,000 cm²/Vs, or half that of graphene monolayers. But to achieve the highest mobility possible for pure carbon transistors fabricated from graphene monolayers, the researchers said they will need a substrate other than ordinary silicon dioxide, which was used in their current tests. Candidates include silicon carbide and diamond.

Also being considered is eliminating the substrate altogether and using air gaps beneath graphene transistor channels.

Since "the lattice vibrations were so weak in the graphene itself, a secondary, little-studied effect seemed to dominate," said Fuhrer. "The phonons in the substrate, which for us was silicon dioxide, appeared to scatter electrons in the graphene, which we believe will limit its electron mobilities to about 40,000 cm²/Vs."

Next, the researchers will try widely available silicon carbide, which is prefabricated into wafers. The team also plans to test graphene deposited atop diamond substrates. They will also try using air gaps, which Fuhrer said will be much more difficult to fabricate in commercial devices.