Photonic systems offer advantages for quantum computation and simulation due to their high coherence and robustness. While the Knill-Laflamme-Milburn (KLM) scheme is a milestone in linear optical quantum computation, its non-deterministic nature and resource cost limit practical applications. Boson sampling and its variants, such as Gaussian Boson Sampling (GBS), have demonstrated quantum advantage but lack universal programmability. Current photonic processors face limitations in either programmability (bulk optics) or loss (integrated optics). This research introduces a novel von-Neumann-like architecture that addresses these challenges by employing temporal-mode encoding and a looped structure, enabling multimode-universal programmability with resource efficiency and software scalability. The architecture's effectiveness is showcased through its application in studying the quantum signature of chaos.

The paper reviews the evolution of photonic quantum computing, starting with the KLM scheme and its limitations due to non-deterministic gates. It discusses alternative approaches such as using entanglement and memory for deterministic gate implementation. The advancements in Boson sampling and its variants, which demonstrate quantum advantage, are also highlighted. The trade-off between achieving quantum advantage and maintaining multimode-universal programmability in existing architectures is critically examined, noting the challenges posed by loss in both bulk and integrated optics systems. The limitations of current processors in handling large-scale, universally programmable circuits are emphasized as the motivation for the proposed new architecture.

The researchers designed a von-Neumann-like photonic processor with a looped structure and temporal-mode encoding. This architecture resembles a Harvard architecture, with separate instruction and quantum data memories. The processor unit comprises electro-optic modulators (EOMs): EOM1 and EOM5 for address control and EOM2-4 for creating arbitrary unitary gates. The quantum data memory is a delay line. The system operates by fetching instructions, processing data in the processor unit, and storing intermediate results in the data memory. Temporal-mode encoding enhances phase stability and resource efficiency. To demonstrate the processor's capabilities, they chose the quantum kicked-top model to study quantum chaos. Two quantum programs were developed: one to visualize phase space distributions using the Husimi distribution (13 modes) in regular and chaotic regions; and a second one to quantitatively verify the Fermi golden rule (26 modes) by measuring the average fidelity between perturbed and unperturbed evolutions using a Hadamard test. The Trotter expansion was employed to decompose the evolution operator for implementation.

The researchers successfully implemented a von-Neumann-like photonic processor demonstrating multimode-universal programmability. Using this processor, they investigated quantum signatures of chaos. The first program visually distinguished regular and chaotic regions in the quantum kicked-top model by showing the different dispersions in Husimi distributions. The second program, utilizing 26 temporal modes, quantitatively verified the Fermi golden rule for the first time experimentally, demonstrating a good agreement between experimental results and theoretical predictions considering the limitations from Trotter expansion. The high fidelity of the processor (over 0.992 for a single period) enabled these detailed investigations. The results showed the exponential decay of average fidelity in the chaotic regime, a key characteristic of the Fermi golden rule.

The development and successful implementation of the von-Neumann-like photonic processor represent a significant advancement in photonic quantum computing. The architecture's multimode-universal programmability, coupled with its resource efficiency and phase stability, addresses key limitations of existing approaches. The experimental verification of the Fermi golden rule highlights the processor's potential for studying fundamental physics questions and for benchmarking quantum simulators. The ability to execute different programs with varying resource demands on the same platform showcases its versatility. The results suggest this architecture is a promising candidate for tackling complex real-world problems that are intractable for classical computers.

This paper presents a novel von-Neumann-like photonic processor architecture with enhanced programmability and resource efficiency. Its successful application in studying quantum chaos, including the experimental verification of the Fermi golden rule, showcases its potential for broader applications in quantum computation and simulation. Future work could explore scaling up the processor's capabilities and applying it to more complex quantum algorithms and real-world problems.

The study's findings are based on the quantum kicked-top model. While this model is widely used for studying chaos, its applicability to other systems needs further investigation. The use of the Trotter expansion introduces a systematic error, although this error is accounted for in the analysis. The size of the processor is currently limited, and further scaling up is crucial for tackling larger problems.