Quantum Random Number Generators (QRNGs) play an essential role in modern quantum technologies, being at the backbone of Quantum Key Distribution systems [1]. Besides cryptography, QRNGs find broad applicability in other fields like Monte Carlo simulations, industrial testing, massive data processing, financial industries. A state-of-the-art hardware approach to generate QRNGs is based on measuring laser phase noise, which is a source of quantum randomness from spontaneous emission [2,3]. In these generators, laser diodes are gain-switched with a bias current below threshold in order to obtain a series of pulses with random phases. After passing through an unbalanced Mach-Zehnder interferometer, the pulse-encoded phase noise is converted to random fluctuations of intensity, which can be measured and digitized. This scheme is simple, fast (Gbps) and it has recently been demonstrated in a photonic integrated platform [4].  

However, current QRNGs based on gain-switched lasers are scalar (one laser per channel) and its parallelization keeping the full bandwidth is only possible by equally scaling the number of photonic elements. This PhD studentship will develop a novel approach to QRNG based on the coupling of gain-switched multimode lasers to photonic integrated dynamic billiards. Particularly, the chaotic nature of the wave dynamics in photonic billiards adds an additional physical layer of randomness to the phase encoded in each of the laser’s modes. The combination of photonic billiards with the multimode gain-switched lasers will multiplex the number of extracted random bits per laser pulse. The key advantages of this scheme are its low fabrication cost, compactness, simplicity (coherent detection is not required) and scalability.  

Recently, photonic integrated billiards have been proposed as a hardware platform for machine learning, showing high performance in classification tasks based on the projection of the input information to a high-dimensional phase space enabled by the chaotic light dynamics [5]. Here, a similar machine learning approach will be used to analyse and optimize the randomness of the numbers generated in each output channel. In addition, exploration of different billiard shapes and excitation protocols will be supported by the expertise of the supervisory team on wave chaos theory, which is analogue to quantum chaos [6].  This will push the boundary of knowledge in these systems by analysing optimal regimes of operations at the edge of full chaos, and potentially leading to quantum-inspired machine learning approaches in this platform.