Silicon photonics

More than just cool-sounding, silicon photonics is the future of computing. Ashton Mills sheds some light on the subject.

Optical communications is nothing new. Much of our communications infrastructure relies on it. Optical fibre is fast, high volume and can handle long distances.

But it’s also bulky as all buggery. Not that we need industrial-strength optics when it comes to designing the next-generation of PCs, but one fact remains true: We are pushing the limits of traditional metal conductors, and the high-volume bandwidth the future demands will require processors and pathways between components to handle far greater volumes of data than they do today.

The concept of an optical computer, one whose inner workings are powered by light and not electricity, with data contained in photons and not electrons, is ultimately where we are heading. However, before this can even reach fruition we first need to work out how to use light at the silicon level. Which we’ll do – just as soon as the boffins figure out how to make infinitesimally small lasers.

Infinitesimally?
As it happens, this is exactly what they’ve done. The biggest hurdle to date with sticking optics into integrated circuits has been getting a light source small enough to sit on a silicon substrate. Even the smallest of lasers with current technology would need to be attached and powered externally – but silicon photonics gets around this by embedding the laser directly into the silicon itself, and does it cheaply too, ultimately impacting the feasibility of mass production.

Intel has already achieved this feat, using a series of Star Trekian-sounding technologies including the ‘Raman Effect’ and ‘Two Photon Absorption’. Actually, the latter is an effect observed in quantum physics, and a problem that had to be overcome – but more on this later.

In silicon photonics, indium phoside – one of the few materials that emits light when voltage is applied – is bonded directly to the silicon through heating the materials in a plasma and ‘gluing’ them together. This ‘glass glue’ as it’s called is a tiny 25 atoms thick.

The material and electrodes are bonded to a channel, called a waveguide, in the silicon – when power is applied, light is generated in the waveguide.

At infrared wavelengths silicon is as transparent as glass and can be used to guide the path of light. Source: Intel

Waveguides are literally carved into the silicon and are used to channel, focus or split light. Note that waveguides aren’t a ‘pit’ in the silicon – they are the silicon. To the human eye silicon is opaque, but to laser light in the infrared spectrum silicon is as transparent as glass. Waveguides are literally thin channels of silicon, much like optical fibre, to direct the resultant light from the laser.

But the output of the laser isn’t particularly strong, so amplification is needed. In conventional optics, lasers are amplified by introducing a ‘pump beam’ to the data stream travelling over the glass fibre. Thanks to a process called the ‘Raman Effect’, as the photons from the pump beam collide with atoms in the carrier material (here, glass fibre) energy is released to photons of longer wavelengths. If the data beam is of the appropriate wavelength, it will pick up these photons and be strengthened over distance.

For traditional optics, this effect occurs over kilometres of fibre. In silicon the effect is pronounced – in fact, it’s about 10,000 times stronger. This means that a signal can be amplified over incredibly small distances through a silicon carrier. Which just happens to be exactly what the doctor ordered when building infinitesimally small lasers.

But there’s always a challenge. Because the carrier material also intercepts or interferes with photons due to its physical structure, the goal is to have the amplification of a beam to be greater than the loss incurred by travelling through the silicon waveguide.

When current is applied diodes on either side of the waveguide draw electrons to one side to reduce the impact of the two-photon absorption.