Starting next year, computers will be available with three-dimensional transistors - these will incorporate vertical components, unlike the flat chips that we're used to seeing. This structure will allow them to have shorter gates, which are the components that allow the transistors to switch the electrical current on and off, and to direct its flow. The shorter the gate, the faster the computer can operate. While the new 3D transistors will have a gate length of 22 nanometers, as opposed to the present length of about 45, the use of silicon as a construction material limits how much shorter they could ultimately get. That's why scientists from Purdue and Harvard universities have created prototype 3D transistors made out of indium-gallium-arsenide - the same compound recently used in a record-breaking solar cell.

Computers implementing 3D silicon transistors will not only be able to run faster, but should also weigh less, and generate less heat than their present-day flat-transistor-using counterparts. Their new-and-improved shorter gates are made from dielectric-coated silicon nanowires, and it is estimated that such gates could be further shortened to about 14 nanometers within a few years. In order to go any shorter, however, a material is needed that can move electrons faster than silicon is able to.

Studies of the indium-gallium-arsenide gates suggest that they should be able to move electrons five times faster than silicon gates, allowing for gate lengths in the neighborhood of just 10 nanometers.

At any length below 14 nanometers, the silicon dioxide insulating layer currently used on transistor gates no longer works properly, allowing the electrical charge to leak out. To that end, the Purdue/Harvard transistors instead utilize a thinner layer of aluminum oxide. It appears to serve as a better insulator at such a small scale, which in turn should allow the transistors to run faster, using less power - they are still being tested.

The production process for the 3D indium-gallium-arsenide transistors could be easily implemented into existing manufacturing processes, the scientists report, so adoption of the technology on a wide scale ought to be feasible.