Programmable photonic chip lights up quantum computing
Computers are made of chips, and in the future, some of those chips might use light as their main ingredient. Scientists from the Ontario, Canada-based quantum computing firm Xanadu and the US National Institute of Standards and Technology have taken a big step towards that future by building a light-based chip that can be programmed through cloud access.
While conventional computers use electricity to create the ones and zeros that are their lifeblood, quantum computing experts have multiple options when developing their qubits. Some rely on superconductors, some start with extremely cold atoms, and some, like the researchers at Xanadu, use light.
But not just any light. The light that travels through the thumbnail-sized Xanadu chip, or circuit, has been “squeezed” – that is, its quantum uncertainty has been minimized. Squeezing light is possible because of the Heisenberg uncertainty relation that says that trying to make any microscopic object very narrow is like squashing a piece of clay: the narrower it gets in one direction, the more it bulges in another. Squeezing light produces precisely shaped photonic states that can be used for very accurate measurements in optical physics. Xanadu researchers, however, had other ideas: they used these squeezed states as qubits.
Xanadu’s chip , which is described in Nature (591 54), works in three stages. First, laser light is fed into four microring resonators – tiny circular tracks in which light loops around and changes shape as it, in effect, catches its own tail. These resonators act as “squeezers” that smush many photons into a single squeezed state.
Next, a network of optical elements manipulates the photons’ properties in a way that is analogous to changing their direction by bouncing them off a mirror or changing their colour by passing them through a filter. Sequences of these light manipulations are the equivalent of computer code. Whenever the network bounces or rotates light, it executes operations similar to adding ones and zeroes in a classical computer.
In the final stage, the light enters a detector that counts how many photons are within each squeezed state. The result of the computer’s calculation lies in these photon numbers. “Some particular integer pattern of photon counts for a particular circuit that you dialled in will tell you something about the problem that you encoded in the device,” says Zachary Vernon, a physicist at Xanadu and a co-author on the study.
Vernon explains that this approach makes it possible to perform some computations that are new even to other quantum computers. “It lets you access a space of problems which are different than the ones that are accessible by matter-based qubit devices,” he says. In one particularly novel calculation, squeezed states encoded the shape of two graphs. The photon numbers detected at the end of the computation reflected how much structure those graphs had in common. This graph similarity analysis would not be easy to implement on any other quantum computer, Vernon says.
The small size of the Xanadu chip is another key advantage. According to Shuntaro Takeda, a physicist at the University of Tokyo, Japan, who was not involved with the study, previous squeezed-light experiments required large tables full of bulky optical elements like mirrors and lenses. In Takeda’s view, on-chip integration technology like Xanadu’s will be indispensable for building large-scale, general-purpose optical quantum computers in the future.
Being able to perform more than one calculation is already a leap forward for light-based quantum computing, says Zheshen Zhang, a quantum information researcher at the University of Arizona in the US who was also not part of the study. He notes that similar devices could, in the past, execute only one type of code, and could not be programmed to perform different tasks for different users. The Xanadu chip’s accessibility through a cloud service is a further benefit, he says.
To make their devices useful for a broad base of future quantum programmers, Xanadu’s scientists still need to overcome some scientific and engineering challenges. In the current set up, for example, many photons are lost as they travel through the chip due to small flaws in the chip’s structure. Engineering more perfect chips and developing codes that take photon loss into account could be important for future generations of these devices, Zhang says. Future chips will also have to handle more information – and thus more light – before they can outperform classical computers.
One example of a problem where a classical and an optical quantum computer could go head to head would involve simulating the behaviour of many molecules. “Can you show that the classical algorithm of simulating such a problem becomes intractable whereas the quantum algorithm would still allow you to actually get the answer?” Zhang asks.
The Xanadu team say that addressing this question is the next item on their agenda. They have, however, already measured the quantum-ness of their device by demonstrating that approximating its mechanisms by some classical model would be extremely difficult. “If everything else stays the same, and you scale the [chip] system up, it will still be very quantum,” Vernon says. “Of course, a lot of things have to come together to make that work.”