Silicon waveguides move us closer to faster computers that use light

Do you want your gadgets to be even faster?

Most of us do – but with current technology, they aren’t likely to get much faster than they are today. For the past decade and a half, the clock rate of single processor cores has stalled at a few Gigahertz. We can no longer push the boundaries of Moore’s law, which has predicted we’d be cramming ever more transistors on a chip — at least not without consuming too much power.

Enter optical circuits.

With results published in Light Science & Applications,¹ our Zurich-based team of researchers has just managed to efficiently guide visible light through a silicon wire — an important milestone towards faster, more efficient integrated circuits. Our low-loss silicon waveguide could enable new photonic chip designs for applications that rely on visible light, and could lead to more efficient lasers and modulators used in telecoms.

Making a fence to confine light

In optical circuits, information is encoded in light rather than electronics. In 2019, together with partners from Skolkovo Institute of Science and Technology and University of Southampton, we built the world’s first ultrafast all-optical transistor that can operate at room temperature. Our latest work is a build-up on that: a silicon waveguide that can be used to connect such transistors, carrying light between them with minimal losses.

Wiring up transistors of an optical circuit with silicon waveguides is crucial for compact, highly integrated chips. And it’s easier to integrate other components such as electrodes if the waveguide is made of silicon — a cheap and abundant material that happens to be an excellent semiconductor.

One challenge, though, has been the ability of silicon to absorb visible light — for which it’s called a ‘dark’ material. While important for capturing sunlight in solar panels, this is not great for a waveguide where light absorption means signal loss.

To deal with the absorption issue, we’ve opted for nanostructures called high contrast gratings. Such a grating consists of nanometer-sized ‘posts’ lined up to form a ‘fence’ that prevents light from escaping. The posts are 150 nanometers in diameter and are spaced so that light passing through them interferes destructively with light passing between them. Destructive interference is a phenomenon where waves – including electromagnetic waves such as visible light – that oscillate out of sync cancel each other out. This way, no light can “leak” through the grating and most of it gets reflected back inside the waveguide.

Our team has also shown that absorption of light inside the posts is minimal. Together, these two features lead to losses of only 13 percent along the light’s travel path of one millimeter inside the waveguide. For comparison, in a pure silicon waveguide without the grating, losses would amount to 99.7 percent in just 10 micrometers.

Simulations for precise grating design

The idea behind the high contrast gratings may look simple. Still, we were surprised to find that gratings could keep light from being absorbed by a ‘dark’ material like silicon. We first observed the grating effect in 2010 in a laser microcavity – which helped because the light amplification by the laser compensated for the losses. Also, then the light hit the gratings at almost 90 degrees – a sweet spot for the grating effect to kick in.

But keeping the losses low in a waveguide without the benefit of the laser gain, and at an almost-grazing angle of the incoming light, is much more challenging.

To ensure the grating design would work, we ran simulations to see how light propagation inside the waveguide would change with different grating dimensions. We found that the grating would provide efficient guiding over a broad band of wavelengths.
We just needed to determine the right spacing between the grating posts and make the posts the right thickness, within a precision margin of 15 nanometers. We did that — and our tests confirmed the simulations when we obtained low loss for visible light in the range between 550 and 650 nanometers.

In simulations, we also showed that the design could be used to guide the light around corners. In the future, we aim to confirm this experimentally. If it works, we’ll need to optimize the technology further to keep the additional losses low.

The next step is to engineer the efficient coupling of the light out of the waveguides into other components. That’s a crucial step in our research, with the ultimate goal of integrating the all-optical transistors into integrated circuits that would be able to perform simple logic operations.