Ars Technica’s Chris Lee has spent much of his adult life playing with lasers, so he’s a big fan of photon-based quantum computing. Even as various forms of physical hardware like superconducting wires and trapped ions progressed, it was possible to find it springing from an optical quantum computer being developed by a Canadian startup called Xanadu. But, in the year since Xanadu first described its hardware, companies using this other technology have continued to make progress by reducing error rates, exploring new technologies, and increasing the number of qubits.
But the advantage of optical quantum computing hasn’t gone away, and now Xanadu is back with a reminder that it hasn’t gone away either. Thanks to some tweaks to the design described a year ago, Xanadu is now able to perform operations with more than 200 qubits at times. And it has been shown that simulating the behavior of just one of these operations on a supercomputer would take 9,000 years, whereas its optical quantum computer can do them in tens of milliseconds.
This is an entirely artificial benchmark: much like Google’s quantum computer, the quantum computer is simply itself while the supercomputer tries to simulate it. The news here is more about the potential for Xanadu’s hardware to evolve.
stay in the light
The benefits of optics-based quantum computing are considerable. Almost all modern communications depend on optical hardware at some point, and improvements in this technology have the chance to be directly applied to quantum computing hardware. Some of the manipulations we might need can be done with hardware miniaturized to the point of being able to burn it onto a silicon chip. And all hardware can be kept at room temperature, avoiding some of the challenges of inputting or outputting signals from equipment near absolute zero.
Xanadu seems confident that these perks are substantial enough that building a business around them makes sense. The hardware Lee described last year relies on a single chip to put photons into a specific quantum state and then force the pairs of photons to interact in such a way as to entangle them. These interactions form the basis of qubit manipulations that can be used to perform calculations. The photons can then be sorted according to their state, with the number of photons in each state providing an answer to the calculation.
Scaling this technology presents challenges. Since photons can only interact in pairs, adding another photon means you need to include enough hardware functionality for its necessary interactions. This means that scaling the processor to a higher qubit count involves scaling all that hardware on the chip. It’s not a problem now, but it could easily be one as things go from hundreds to thousands.
Choose your own adventure
This scaling is likely why Xanadu’s new system, called Borealis, involves a significant architectural overhaul. His previous machine used a bunch of identical photons that all entered the chip in parallel and passed through it simultaneously. In Borealis, photons enter the system sequentially and follow a path that is a bit like a “choose your own adventure” game.
The first piece of hardware hit by the photons is a programmable beam splitter, which can perform two functions. If two photons hit it simultaneously, they can interfere with each other and become entangled. And depending on its state, the beamsplitter can deflect photons out of the main path and into a fiber optic loop. Moving around this loop adds a delay to the photon’s motion, allowing it to exit the fiber at the same time as a new photon arrives at the beamsplitter, allowing it to become entangled with a subsequent photon.
After passing the first beamsplitter, photons travel to a second, with a longer fiber optic loop that introduces a longer delay for any photons sent there. And then on a third with an even longer loop. Optional delays allow photons to become entangled with other photons that only arrive at the material long after them. As Xanadu presents, each of Borealis’ three beamsplitters amounts to adding an extra dimension to the entanglement matrix, going from no entanglement to three dimensions of potential entanglement.
Once through, the photons are sorted according to their properties and sent to a series of detectors. The detectors keep track of how many photons arrive and when, which will provide an answer to any calculations it performs. As configured, it can handle over 200 individual photons in a calculation.