Gaussian Boson Sampling with Pseudo-Photon-Number Resolving Detectors and Quantum Computational Advantage
The paper reports that the Jiuzhang 3.0 quantum computer achieved a significant quantum computational advantage by performing Gaussian boson sampling with up to 255 photon-click events in 1.27 microseconds, a task estimated to take the Frontier supercomputer billions of years using exact classical algorithms.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the quest to build machines that can solve problems beyond the reach of today's most powerful supercomputers, scientists are exploring a specific type of calculation known as quantum sampling. Imagine trying to predict the outcome of a complex game where the rules are governed by the strange laws of quantum physics, the science that describes how light and matter behave at the smallest scales. In this realm, particles like photons can exist in multiple states at once and interfere with one another in ways that create patterns too intricate for classical computers to track efficiently. The goal of quantum sampling is to generate these complex patterns of light and verify that they match the predictions of quantum theory. If a machine can do this faster than any classical computer could possibly calculate, it demonstrates a "quantum computational advantage," a milestone proving that quantum devices have truly entered a new era of capability. This is not just about speed; it is about verifying that the universe behaves in ways that our current classical understanding cannot easily simulate, a concept that has been theorized for decades but only recently tested in the laboratory.
A team of researchers has now pushed this frontier further with a new experiment involving a machine they call Jiuzhang 3.0. This device is designed to perform a task called Gaussian boson sampling, which involves sending streams of light particles through a complex network of mirrors and beam splitters and then counting how many particles arrive at the exit. The researchers upgraded their setup by using a special detection method that can tell not just if a light particle arrived, but how many arrived in a single burst. This allowed them to register up to 255 distinct light-particle events in a single run, a significant increase from previous attempts. By capturing this much more detailed information, the team created a sampling task that is vastly more difficult for classical computers to mimic. They found that their machine could produce a single sample of this complex light pattern in just 1.27 microseconds, a fraction of a second. In stark contrast, they estimated that even the world's most powerful supercomputer, named Frontier, would require approximately 600 years to generate a single ideal sample of the same distribution using the most advanced exact methods available.
The significance of this work extends beyond raw speed; it lies in how rigorously the team proved that their results were genuinely quantum and not just a clever trick by a classical computer. Over the years, skeptics have proposed various "mockups," or classical simulations, that attempt to reproduce the statistical patterns of these experiments without actually using quantum mechanics. Some of these proposed tricks involve assuming the light particles are slightly distinguishable from one another or that they behave like a specific type of noisy thermal light. The researchers in this study systematically tested their data against all the most competitive classical mockups currently known, including a recently proposed "squashed state" model and others designed to exploit limitations in circuit connections. Using statistical tests that compare the likelihood of the data coming from a true quantum source versus a classical imitation, they found that their experimental results consistently aligned with the quantum theory and strongly deviated from every classical alternative. As the size of the data sets grew, the confidence that the results were truly quantum increased, effectively ruling out the possibility that a classical computer could be faking the outcome.
To ensure their findings were robust, the team also developed a more complete model to account for the imperfections inherent in any real-world experiment, such as the fact that not all light particles are perfectly identical. This new model, which included the effects of partial distinguishability between particles, matched their experimental data better than any previous modeling method. They further validated their results by analyzing how the light particles correlated with one another across the system. They compared these correlations against various classical algorithms that try to approximate the experiment by looking at only small parts of the whole picture. The analysis showed that while these classical approximations could mimic simple, low-level patterns, they failed completely when faced with the complex, high-order relationships present in the full quantum system. The experimental data clustered tightly around the predictions of quantum theory, while the classical mockups wandered far off course.
The researchers calculated the computational cost required to simulate their hardest experimental samples on a supercomputer. For the most difficult sample they generated, they estimated that the Frontier supercomputer would require a time far longer than the age of the universe to compute the result using exact methods. This timescale is far longer than the age of the universe, highlighting the immense gap between what their quantum machine achieved in microseconds and what is theoretically possible for classical machines. By overcoming the limitations of previous detection methods and validating their results against every known classical counter-argument, the team has established a new benchmark for quantum computational advantage. Their work demonstrates that as quantum systems grow larger and more complex, they continue to outperform classical simulations by an overwhelming margin, providing strong evidence that the quantum speed-up is a real and measurable phenomenon.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.