Periodic Reporting for period 1 - GTGBS (Gate-teleported Gaussian boson sampling)
Periodo di rendicontazione: 2023-10-01 al 2025-09-30
Photonic quantum technology promises to bring unprecedented quantum advantage—doing a specific task that no classical system will ever be able to do—and revolutionize human’s computational power. In this line of research, Gaussian boson sampling (GBS) has emerged as a leading protocol. However, challenges such as photon loss and setup scalability pose significant obstacles to the study of GBS. These issues have led to the claimed quantum advantage being questioned by recent advances in classical simulations and complicated the task of harnessing photonic quantum technology for practical applications. The ambition of project GTGBS is to achieve quantum advantage in two approaches: via learning a multi-point time-domain displacement process and via implementing ultra-large-scale GBS in a measurement-based manner. Crucially, both approaches leverage continuous-variable (CV) entanglement to create large-scale optical circuits with a constant setup depth and high transmissivity, thereby delivering robust and reliable results. Our findings show that provable quantum advantage with photons is readily achievable with the current technology, and the decent scalability of the photonic system will enable us to pursue even stronger quantum advantage.
We develop optical parametric oscillators that can create CV entangled photonic states with a noise power of only 32% of the vacuum. In the quantum learning approach, we use the entangled state as the probe state and the quantum memory to learn the properties of a multi-point displacement process. Our entanglement-assisted approach uses almost a trillion times less sample to reconstruct the characteristic function of the process compared to a classical, entanglement-free one. Furthermore, we promote the quantum advantage to a provable one—an advantage that does not rely on any assumption (e.g. from the computational complexity theory)—via a hypothesis testing protocol. In the GBS approach, we use time-domain multiplexing to build a 60,000-mode CV cluster state. We cast homodyne measurements on half of the modes; this induces gate operations on the remaining half of the modes via quantum teleportation. We develop a novel FPGA-based feedforward platform and use it to remove the undesired displacements accompanying the gate operations and then count the photon number of the output state. Our novel approach enables a GBS experiment that has an efficiency independent of the number of modes and contains more than 1000 squeezed photons, creating unprecedented quantum computational advantage.
Project GTGBS has created the first photonic quantum learning advantage and the first proved photonic quantum advantage over any task. The method, once adapted to different systems, will dramatically accelerate the learning of physical processes spanning in time, facilitate applications such as quantum sensing, gravitational wave detection and Raman spectroscopy. Project GTGBS is also expected to create the strongest photonic quantum computational advantage so far and revolutionize the way GBS experiments are implemented. In this direction, it is expected that a follow-up project will develop an algorithm to systematically study the CV measurement-based circuit, making the setup to also be suitable for solving practical and exciting problems with significant real-world applications in fields such as machine learning, optimization, and statistical modeling. Ultimately, project GTGBS constitutes a leap in the study of quantum advantage in different forms and contributes to applying the quantum advantages in more practical tasks.