Xanadu

Xanadu’s quantum computers are based on photonics and use quantum light sources that emit squeezed-light pulses. Our Borealis photonic quantum computer provides a programmable loop-based interferometer with over 216 squeezed-state qubits, and true photon-number-resolving detectors, allowing for quantum computational advantage. Borealis is the first photonic quantum hardware accessible in the cloud through Amazon Braket.

While still within the noisy intermediate-scale quantum (NISQ) era, the squeezed-state qubit technology used by Xanadu addresses two main challenges surrounding commercial quantum computing:

• Fault-Tolerance: While Borealis is a NISQ device, a number of the technologies developed for and employed within it are critical components for a universal, fault-tolerant photonic quantum computer. In particular, Borealis incorporates time multiplexing, stabilized fiber-optical buffer memory for photonic states, and synthesis of three-dimensional entanglement.
• Scalability: Borealis addresses scalability by leveraging time-domain multiplexing (TDM) for hardware-efficient experimental setups and the feasibility to encode and process information with light.

Our Borealis device demonstrates quantum computational advantage by tackling a challenging mathematical problem: generating random numbers that obey specific patterns expressed through a mathematical function called the Hafnian, via a process known as Gaussian Boson Sampling (GBS). Calculating the Hafnian, even using the most powerful classical computer, would take a prohibitively long time, and so too would generating numbers from Hafnian-linked distributions. Using a photonic quantum computer can speed up execution of this task exponentially.

Borealis is able to sample GBS instances orders of magnitude faster than any known classical algorithm run on the most powerful supercomputer using its hardware-efficient TDM architecture. Specifically, over 216 squeezed-state qubits, or squeezed modes, can be programmed and entangled (up to a maximum of 288). In our peer-reviewed paper, we reported up to 219 photons detected and a mean photon number of 125, estimating that it would take over nine thousand years for the best available algorithms and supercomputers to produce, using exact methods, a single sample. With Borealis on Amazon Braket, you can run circuits like these yourself.

Figure 1 shows a schematic of the Borealis device. Borealis was designed to generate and efficiently process a large number of squeezed-state qubits. The setup consists of three interferometers, referred to as loops, each able to hold up to 1, 6 and 36 modes each at a time. The squeezed states, or pulses are produced at the beginning in an optical parametric oscillator (OPO), and are continuously sent into the loops. Each of these loops incorporates some of the most important building blocks of a photonic quantum computer: a programmable beam-splitter, a programmable phase shifter, and an optical fiber delay line playing the role of a buffer memory for light.

The loops are the key components of the system, allowing each incident pulse to be stored in the delay line until another pulse arrives. When a pulse arrives at a beam-splitter gate, the gate can either switch the pulse into the fiber (store in memory), switch the pulse out of the loop (retrieve from memory), or perform an entangling gate between the pulse being retrieved from the loop and the next incoming pulse. Due to the different round-trip lengths of the loops, entanglement can extend beyond temporally adjacent pulses and create an entangled state with a three-dimensional topology, something that would be difficult to achieve in a purely spatial architecture. This is important for the future where more error correction codes can be potentially utilized.

Finally, at the output of the interferometer, the optical pulses are detected by a photon-number-resolving (PNR) detection module operating at the Borealis clock rate of 6 MHz. This high-speed PNR detection is facilitated by a temporal-to-spatial demultiplexing unit, partially distributing incoming pulse trains to an array of 16 detector channels.

The Borealis hardware is readily accessible via Amazon Braket, and can be programmed directly through Strawberry Fields, an open-source Python library for building, simulating, and executing photonic quantum algorithms pre-installed with Amazon Braket notebooks. Using Strawberry Fields, you can construct time-domain multiplexing programs by specifying the loop-interferometer gates (beam-splitters and phases) and input squeezed states (brightness and number of modes), before submitting it for execution on Borealis.

For more details on how to execute programs on Borealis, be sure to check out our tutorial on operating Borealis, and recreate the results of our quantum computational advantage experiments over the AWS cloud.