Out of the Lab: Light From Silicon
The driving force for this research is a problem that's expected to crop up in about four or five years' time. As transistors shrink in size, the electrons that power them spend proportionately more time in the wires and connections between components. In the end, speed and processing power hit a brick wall.
Of course, we all know that photons are the information carriers supreme. And just as light carries information down an optical fiber, it can be used to transport signals around electronic circuitry. Instead of having multiple layers of metal to transport electrons from one part of the chip to another, light could travel along optical "highways" to connect critical processing functions. "It's like waving across a crowded room," says Homewood. "You get the information across without having to fight the crowds".
To make this a viable proposition, a light source (laser) and a detector are needed. The only problem is, bulk silicon doesn't emit light, so other materials have to be brought into play -- and that makes the processing far too complicated.
For the past ten years, researchers have tried to coax light out of silicon, with varying degrees of success. "There are lots of things that appear to do the job, but in actual fact they don't, because either they don't operate at room temperature, or they're not compatible with standard ultra-large-scale integration [ULSI] manufacturing in silicon," says Homewood. Surrey's technology meets both requirements, he adds.
The emission wavelength is about 1100 nanometers, which lies outside the low-loss regions of optical fiber. But for the short distances involved in optical interconnects, it doesn't matter.
There's a second, and potentially more explosive agenda. Homewood believes he could tweak the wavelength of the silicon so that it lases in the telecom bands around 1550nm. Ultimately, because silicon is so plentiful and so cheap, this might result in silicon lasers replacing today's transmitters. He doesn't want to say much about this, because it's commercially sensitive. But he adds "There's no reason, if we get a [silicon] laser working, that it shouldn't be better [than today's telecom lasers]."
But the sceptics still need convincing. Breakthroughs in light-emitting silicon have been announced countless times before. To understand why this one is different entails going back to the basics of how light is emitted in a semiconductor.
Semiconductors contain two types of charge carrier: electrons, which are negative, and holes (locations in the crystal that ought to contain electrons, but don't), which are positive. To get light out of a semiconductor, an electron needs to meet a hole. It pops into the hole, restoring electrical neutrality and giving up energy as a photon.
As noted, bulk silicon doesn't emit light. It has an "indirect" band gap. That means that the electron can't drop into the hole without also changing its momentum, and that's not very likely to happen. Instead, electrons wander around the crystal until they find defects, where they can give up energy more easily as heat, not light. Even though there are very few defects in single crystal silicon, the heat-generating process for recombination dominates over the light-generating one.
To get light out of silicon, it's necessary to prevent electrons from wandering 'round the crystal, so that they can't meet up with the defects. Cooling everything down helps because it slows down the electrons. Physically confining the electrons, by manufacturing the silicon as tiny particles or wires (porous silicon) also works, but that's not compatible with semiconductor processing.
Surrey's approach is to implant ions in the silicon. That means shooting ions at the silicon with enough energy so they get embedded below the surface. Each ion creates a "dislocation loop" around it -- rather like a ripple around a pebble thrown in a pond -- which creates local stress in the material, and that's what localizes the electrons.
In the work described in Nature, boron ions were used both for creating dislocation loops and for doping to create the LED. Boron implantation is a standard technique used in the fabrication of silicon integrated circuits. Crucially, though the dislocation loops hamper the long-range movement of electrons, "there's no problem with current flow," according to Homewood. Proof of this lies in the fact that the silicon-based LED has an efficiency that's comparable to existing gallium-arsenide LEDs, he says.
Surrey University filed a patent on this work a few weeks ago. It claims that the principle can be applied to other materials, like germanium or silicon carbide, to coax light out of them.
— Pauline Rigby, senior editor, Light Reading http://www.lightreading.com