Moore’s Law lives: UCLA researchers solve power dissipation problem in chip design

Published 30 June 2006

Moore’s Law states that complexity of integrated circuits, with respect to minimum component cost, doubles every 24 months; the very advances depicted by the law, however, threatened to invalidate it at some point (the point is here, in fact), owing to the power dissipation in traditional silicon semiconductors; an innovative team of UCLA researchers found a way around the problem, and in the process also brought closer the day of convergence of photonics and electronics

Gordon Moore, one of Intel’s founders, predicted in 1965 that innovative research would allow for a doubling of the number of transistors in a given space every year. In 1975 he adjusted this prediction to a doubling every two years. This became known as Moore’s Law.

There was a built-in tension in Moore Law, however: The very advances in silicon design which Moore’s Law describes also created the circumstances which make it more and more difficult to see the law’s applicability continue in perpetuity. One key challenge to Moore’s Law was the problem of power dissipation in traditional silicon semiconductors. This power dissipation problem is already regarded as so severe that it threatens to slow down, if not halt altogether, the continued advance of the technology described by Moore’s law.

This is where researchers at the UCLA Henry Samueli School of Engineering and Applied Science come in. Building on a series of recent breakthroughs in silicon photonics, these researchers have developed an innovative approach to silicon devices which combines light amplification with a photovoltaic — or solar panel — effect. UCLA researchers report that optical amplification in silicon can be achieved with zero power consumption, but also that power can now be generated in the process. More specifically, the team’s research shows that silicon Raman amplifiers possess nonlinear photovoltaic properties, a phenomenon related to power generation in solar cells. In 2004 the same UCLA group demonstrated the first silicon laser, a device that took advantage of Raman amplification. Professor Bahram Jalali, the UCLA engineering professor who led researcher Sasan Fathpour and graduate student Kevin Tsia in making the recent discovery, says: “After dominating the electronics industry for decades, silicon is now on the verge of becoming the material of choice for the photonics industry, the traditional stronghold of today’s semiconductors.”

Background: The amount of information which can be sent through an optical wire is directly related to the intensity of the light. Thus, to perform the key functions in optical networking — amplification, wavelength conversion, optical switching — silicon must be illuminated with high intensity light to take advantage of its nonlinear properties (one example is the Raman effect). The challenge in silicon photonics is that material stops being transparent at high optical intensities, making light unable to pass through. The reason: “As light intensifies in silicon, it generates electrons through a process called two-photon-absorption. Excess electrons absorb the light and turn it into heat. Not only is the light and the data-carrying capacity lost, the phenomenon exacerbates one of the main obstacles in the semiconductor industry, which is excessive heating of chips. The optical loss also makes it all but impossible to create optical amplifiers and lasers that operate continuously,” Jalali explained.

Earlier efforts to address two-photon-absorption used a diode attached to the chip to “vacuum” out the electrons which block light. The vacuum itself, however, adds an additional watt of heat onto the chip, which is almost a million times the power that a single transistor consumes in a digital circuit.

The UCLA researchers’ new development will allow the recycling of power which would otherwise be lost. Dr. Robert R. Rice, senior scientist at Northrop Grumman Space Technology’s Laser and Sensor Product Center, says: “In space and military laser systems, the impact of device efficiency on electrical power and thermal management is a prime consideration.”

The fundamental significance of the Jalali team discovery is that it brings closer the day of convergence of photonics and electronics. If this happens, many applications that silicon photonics has promised will materialize.

Which brings us back to Moore’s Law: The ability of the UCLA team to resolve the power dissipation problem will allow engineers to continue to increase the power of microchips. Silicon photonics technology has the potential to use the power of optical networking inside computers and to create new generation of miniaturized and low-cost photonic components, among other applications.

Jalali’s research at UCLA has been funded by DARPA, and was also cosponsored by the Northrop Grumman Corporation.