Advances in computing have revolutionised many aspects of our lives. However, increasingly new models of computation are required, as shrinking feature sizes in processors approach the limit where quantum effects become relevant, and some important problems remain intractable to solve. In this context, quantum computing has far-reaching implications – it holds the promise of exponentially greater computing power for particular tasks, such as optimization of a quantity depending on many variables or simulation of molecules to enable the design of new drugs and materials. Optical quantum computing with single photons is a good candidate for demonstrating a near-term computational speed-up. Photons are well isolated from thermal noise and maintain their quantum coherence even at room temperature, avoiding the cost and complexity of cryogenic or vacuum conditions. The propagation of single photons in a linear optical circuit is inefficient to simulate using classical methods (so-called Boson Sampling), so by extension this experimentally accessible architecture can provide a quantum advantage for specific problems.
Gaussian boson sampling is a variant on Boson Sampling which replaces the single photon inputs with 'squeezed' light - highly non-classical states which are readily generated using nonlinear optics. Gaussian boson sampling exhibits a quantum advantage, and has a variety of useful applications - these include calculating the vibronic spectra of molecules, and identifying densely connected sub-graphs in networks. Hence Gaussian boson sampling is a practical and useful model for near-term quantum computation. However, current implementations based on free-space optics are bulky, labour-intensive to align, and lack programmability. Approaches based on integrated photonics offer compactness and the potential for programmability, but they are currently limited by lossy components.
This project aims to develop compact and programmable approaches which can quickly and practically be scaled to a useful size - with a focus on frequency-encoding of information onto photons, to reduce the number of separate components required. By leveraging the frequency entanglement generated by a single squeezed-light source powered by a short pump laser, complex multi-mode states can be generated without the need for a large optical circuit. Further, applying pulse-shaping techniques to the pump pulse adds control over the generated state without the need to manipulate the actual quantum light - and hence avoids introducing loss and other imperfections. An experimental demonstration verified the quantum interference between 8 photons distributed across 16 frequency channels - the largest such experiment to be performed in frequency-encoding. The use of commercially available components in a compact setup will make this scheme convenient to scale to still larger experiments.
Another focus has been understanding the classical complexity of Gaussian boson sampling - when assessing the advantage achieved by quantum devices, they are often compared to a simulation on classical hardware, so it is important to have an accurate idea of the classical runtime. New classical algorithms were developed for the simulation of Gaussian boson sampling outperforming previous efforts by nine orders of magnitude, which were benchmarked on high-performance computing hardware up to a scale of 100 modes - comparable to the largest experiments. This will be valuable in assessing and verifying future experiments.
