Benchmarking Gaussian and non-Gaussian States Demonstrates Performance Costs for Quantum Sampling Platforms

Quantum computation relies on increasingly complex states of light, but maintaining these states presents a significant challenge, often leading researchers to simplify them, potentially sacrificing computational power. Michael Stefszky, Kai-Hong Luo, and Jan-Lucas Eickmann, working with colleagues at various institutions, now demonstrate a new method for directly comparing the performance of quantum samplers using both simplified and more complex light states.

The team developed the Paderborn Quantum Sampler, a unique platform capable of running experiments with a variety of input states within a single setup, allowing for a fair comparison of their computational abilities. This innovative approach, coupled with a robust verification method, confirms that utilising more complex, non-Gaussian light states delivers a clear performance advantage, representing a crucial step towards realising the full potential of quantum computation and establishing a definitive demonstration of quantum advantage.

Photonic Chip Demonstrates Quantum Sampling Potential Scientists have developed a photonic chip capable of performing Gaussian Boson Sampling (GBS), a complex quantum computation aimed at demonstrating quantum advantage, the ability of a quantum computer to solve problems intractable for even the most powerful classical computers.

The team generates complex quantum states using the chip and samples from the resulting output, attempting to create a process difficult for classical computers to simulate. The experiment utilizes Gaussian states of light, relatively easy to create and manipulate using standard optical components. The core of the system is a silicon nitride photonic integrated circuit, a tiny chip containing a complex network of waveguides and beam splitters that acts as a programmable interferometer. The chip features twelve optical modes, representing a significant step towards building larger and more powerful quantum samplers. The experiment employs a pulsed laser to generate photons, using a heralded single-photon source to ensure that only single photons enter the chip. A twin beam source, generating correlated photon pairs, further enhances the experiment’s efficiency. Single-photon detectors, known as superconducting nanowire detectors, measure the photons at the chip’s output, providing crucial data about the quantum state. Careful calibration of these detectors is essential to account for imperfections and ensure accurate results. Precise control over the laser pulses, chip parameters, and detectors is paramount. A sophisticated data acquisition system records the detection events, and coincidence counting identifies simultaneous photon detections, allowing scientists to characterize the quantum state. Minimizing optical loss is a major challenge, addressed through the use of the silicon nitride chip and careful design of the optical path.

The team also verifies that the generated states exhibit genuinely non-classical behavior, confirming that they cannot be explained by classical physics. The researchers utilize the Walrus library to calculate complex mathematical functions related to GBS and simulate the sampling process. Sophisticated data analysis techniques extract information about the quantum state and compare the experimental results with theoretical predictions. This work represents a significant advance in demonstrating quantum advantage using Gaussian Boson Sampling, paving the way for more complex quantum computations.

Integrated Photonic Platform for State Benchmarking The Paderborn Quantum Sampler (PaQS) represents a significant advance in quantum sampling technology, engineered to directly compare the performance of different input states within a single experimental run. Scientists constructed a hybrid platform capable of performing sampling experiments with either Gaussian or non-Gaussian input states in a 12-mode interferometer, enabling side-by-side benchmarking under identical conditions. The core of the system utilizes an integrated photonic platform, housing both the squeezed-light source and a programmable interferometer, to maximize integration and stability. This design allows dynamic switching between squeezed vacuum states and heralded single-photon states through an electro-optic modulator and polarizing beam splitter, facilitating rapid acquisition of data from multiple sampling configurations. To generate the necessary light sources, the team employed a Menlo laser system, consisting of a stabilized frequency comb operating at 1544nm, coupled with a frequency doubling stage to produce 772nm light for pumping the squeezed-light source. Pulses are spectrally shaped to optimize pump efficiency before entering a single-pass, type-II periodically poled potassium titanyl phosphate waveguide, where squeezed light is generated. Subsequent filtering, utilizing silicon windows and 2nm bandpass filters, removes unwanted spectral components, and a compensation crystal corrects for group delay dispersion. The resulting picosecond squeezed-light pulses enable intrinsic photon-number resolution when detected by superconducting nanowire single-photon detectors. The programmable interferometer, also realized on an integrated photonic platform, offers complete control over the implemented unitary transformation with an average insertion loss below 3 dB. Scientists implemented a sophisticated benchmarking framework based on normally ordered moments of the photon-number operator to certify the presence of quantumness in the measured data, a crucial prerequisite for demonstrating quantum computational advantage. This approach directly identifies genuinely non-classical features, contrasting with previous methods that only assessed closeness to theoretical expectations. The system’s modular design emphasizes the integration of these key subsystems, allowing for precise control and efficient data acquisition. Gaussian and Non-Gaussian Boson Sampling Benchmarks Achieved The Paderborn Quantum Sampler (PaQS) represents a significant advance in quantum computation, delivering a hybrid platform capable of performing boson sampling experiments with both Gaussian and non-Gaussian input states within a single experimental run. This unique architecture allows for direct, side-by-side benchmarking of different sampling regimes under identical conditions, a crucial step towards realizing practical quantum advantage. Researchers successfully demonstrated the ability to switch between generating eight Gaussian states, eight non-Gaussian states, or a combination thereof, all within a 12-mode interferometer. Experiments revealed stark differences in the quantum behavior of datasets generated using Gaussian boson sampling and single-photon-state boson sampling. The quantumness of single-photon-state data consistently increased with the mean photon number of the input states, demonstrating a clear enhancement of non-classical behavior. In contrast, Gaussian boson sampling data exhibited strong non-classical signatures at low photon numbers, but this behavior diminished as the brightness increased, highlighting the limitations of Gaussian states for sustaining quantum advantage. This divergence underscores the fundamentally different behaviors arising from distinct input-state resources and emphasizes the need for a conceptual shift in expanding sampling architectures. The PaQS system achieves Klyshko efficiencies exceeding 6. 5% for all signal modes, with an average of 8. 7% ±1. 5%, demonstrating high-performance signal transmission. Researchers implemented intrinsic photon-number resolution using superconducting nanowire single-photon detectors, enabling the detection of up to three photons per channel by analyzing the rise time of the detector signal. The integrated interferometer exhibits an average insertion loss below 3 dB, ensuring minimal signal degradation during the quantum computation. This combination of high efficiency, photon-number resolution, and low loss positions PaQS as a powerful tool for exploring the potential of boson sampling and advancing the field of quantum computation. PaQS Demonstrates Quantum Sampling Advantage This research presents the Paderborn Quantum Sampler (PaQS), a novel platform designed to benchmark boson sampling regimes using both Gaussian and non-Gaussian input states within a single experimental run.

The team successfully demonstrates that utilising non-Gaussian states yields improved performance in sampling experiments, confirming their importance as a quantum resource for achieving computational advantages. PaQS achieves this by employing a hybrid architecture and a semi-device-independent framework, allowing researchers to verify that observed data cannot be replicated by classical models, a crucial step in demonstrating quantum advantage. The platform addresses a key challenge in the field, namely the trade-off between experimental feasibility and the preservation of non-Gaussian characteristics, which are essential for outperforming classical computation. Previous implementations often replaced demanding single-photon sources with more readily available Gaussian states. 👉 More information 🗞 Benchmarking Gaussian and non-Gaussian input states with a hybrid sampling platform 🧠 ArXiv: https://arxiv.org/abs/2512.08433 Tags: