Source-linked AI summary
Scalable boson sampling with a single-photon device
Yu He, Zu-En Su, He-Liang Huang, Xing Ding, Jian Qin, Can Wang, S. Unsleber, Chao Chen, Hui Wang, Yu-Ming He, Xi-Lin Wang, Christian Schneider, Martin Kamp, Sven Höfling, Chao-Yang Lu, Jian-Wei Pan
TL;DR
Boson sampling remains experimentally challenging because reliable, efficient single-photon generation is lacking. This work demonstrates a scalable architecture using solid-state single photons and validates three- and four-boson sampling, including improved four-photon capability.
Problem
Reliable multi-photon boson sampling remains experimentally challenging because existing pseudo-single photons are inefficient and probabilistically generated.
Method
The architecture uses solid-state single photons sent through a passive m-mode network, with equivalent circuits programmed in an electrically controlled loop-based system.
Results
0.991(2) for the 3-boson sampling and 0.953(4) for the 4-boson sampling were achieved, while 3-boson sampling was completed ~200 times faster and 4-boson sampling was demonstrated.
Takeaways & Limitations
The demonstrated architecture supports scalable multi-photon boson sampling with a solid-state single-photon source.
Takeaways & Limitations
The pseudo-single-photon approach described as the prior challenge is inefficient and probabilistically generated, with photons inevitably admixed with additional components.
Abstract
from arXiv · showhide
Boson sampling is a problem intractable for classical computers, but can be naturally solved on a specialized photonic quantum simulator which requires less resources than building a universal quantum computer. The biggest challenge to implement boson sampling with a large number of photons has been the lack of reliable single-photon sources. Here we demonstrate a scalable architecture of boson sampling using a solid-state single-photon source with simultaneously high efficiency, purity, and indistinguishability. The single photons are time-bin encoded and interfered in an electrically programmable loop-based network. We implement and validate boson sampling with input three and four single photons, and track the dynamical multi-photon evolution inside the circuit. With further refinement of the system efficiency, our approach may be feasible to be scaled up to >20-boson sampling to outperform classical computers, and thus provide experimental evidence against the Extended Church-Turing Thesis.
Abstract
The paper addresses the difficulty of scaling boson sampling by combining a high-performance solid-state single-photon source with a programmable loop-based interferometer. It demonstrates validated three- and four-photon sampling and projects scaling toward more than 20 photons.
- Motivation: Boson sampling offers a specialized photonic route to problems believed hard for classical computers, with 20 indistinguishable photons potentially challenging classical computation.This would provide evidence relevant to the Extended Church-Turing Thesis.
- Motivation: Scaling boson sampling is experimentally difficult because prior implementations used inefficient SPDC photons with multi-photon contamination, yielding about two four-photon detections per day.The multi-photon contribution scales approximately as p^2 when the down-conversion probability is p.
- Architecture: The experiment uses a deterministic solid-state source and an electrically programmable time-bin interferometer built from a loop-based architecture.The loop is equivalent to an M-mode beam-splitter network of depth N and is described as stable, rapidly programmable, and resource efficient.
- Experimental results: The experiment records 2015 three-photon and 1219 four-photon events, completing three-photon sampling about 200 times faster than prior SPDC implementations.It also reports the first four-photon sampling with single-photon Fock states in this context.
- Experimental results: The measured sampling fidelities are 0.991(2) for three photons and 0.953(4) for four photons.The output data are further reported to reach 99.8% confidence of being from a genuine boson sampler and to distinguish indistinguishable from distinguishable bosons.
- Scaling outlook: Current system efficiency is limited by the source, interferometric network, and detectors, while approximately 60% source efficiency is projected to enable 20-boson sampling at about 100 coincidences per hour.The reported present efficiencies are approximately 13.9% for the source, 83.4% for one loop, and 33% for detection.