by Jörn Epping
In our last article, we already learned about qumodes and their difference to the more well-known qubits. The topic of today’s article is how to actually let the qumodes do what we want them to do.
The best way to do a computation with qumodes is to use so-called integrated photonic circuits. To follow up on the picture of the previous article, just imagine photonic circuits as rails for photons.
The analogy from rails is due to so-called waveguides in which electromagnetic waves such as light and, hence, photons can be guided as trains on rails. The most familiar of waveguide for light is the optical fiber that is today part of the infrastructure for high-speed data transmission. In waveguides, light is trapped in such a way that it cannot move left or right and only follows the direction of the waveguide.
But how does this trick work? It can be understood by an effect called “total internal reflection”. Here, light is reflected at the interface between two materials with different optical properties as shown in the picture below. Here, light can get reflected when it has to go from material where it is slower to a material where it is faster (under a certain angle). Light can go faster and slower? Yes, that is right. We say that materials in which light is slower have a higher refractive index than materials with a lower index. By combining two material with different indices and using the internal reflection effect, we can make waveguides in which the light follows the path given by the material with the higher index. At QuiX, we use silicon nitride to define the path the light travels and surround it by silicon dioxide (glass). So as mentioned, this works because silicon nitride has a higher refractive index than silicon dioxide.
The next thing to ensure is that the photons of the qumodes actually arrive at their final destination by being measured at a detector. Since with qumodes, we are dealing already with the tiniest fraction of light, hence, we cannot allow to lose anything without losing information. To have the best probability of detecting our qumodes and obtaining a successful calculation we have to remove anything that hinders our qumodes and losses have to be reduced at all costs.
What are the main causes of losses in photonic circuits? In our platform where we use glass-like materials (silicon dioxide and silicon nitride), these mainly come from coupling losses, circuit damping and possibly components of the circuits. Coupling losses arise from the effect that the photons take up space that depends on their surroundings. In our circuits, we need them to be small to have a high density of components allowing for more complex photonic processors. Hence, we couple the photons from optical fibers where they are rather big (about 10µm) to our photonic circuits where we need them to be small (about 1µm). But first we must shrink down in size. This is done with what you can imagine to be a funnel for photons to not spill them, co-called spot-size converters.
The main second cause of losses is the damping that occurs while the photons are travelling through the maze of the circuit. Here they scatter due to the roughness of the channels defined in our materials. Again, to stay in the picture of rails you can imagine surface roughness causing friction. Since roughness is only experienced that the interface of two materials, in our circuits silicon nitride and silicon dioxide, we want keep much as possible of the qumode from this interface. This can be achieved in two ways by choosing the cross-section of your photonic channel: you either shrink the mode to reduce the interaction with the interface or you distribute the mode between the two materials by using multiple layers. The latter is done in our case since it also makes the spot-size converters easier since the qumode is already a bit bigger.
The last challenge is the exact steering of the qumodes, so to say where the modes change tracks. To get an accurate calculation with qumodes you need to be able to accurately program the photonic processor and control the probabilities of qumode interference. For these junction points it is important to know that the photons of our qumodes not only consist of one single color but a variety, a bandwidth. This means that on-chip switches must accurately work for that bandwidth of colors. Achieving this accuracy can be really tricky and means that you need to have good control over your fabrication processes.
By controlling the coupling, losses, and switches in your photonic circuits it is possible to build reliable and large-scale photonic processors for qumodes. The larger the processor gets, the more complex calculations with qumodes are possible and the reliability of the photonic processor will get more important.