Insider Brief
- A Chinese research team demonstrated that Zuchongzhi 3.0, a 105-qubit superconducting quantum processor, completed a computational task in seconds that would take the world’s most powerful supercomputer an estimated 6.4 billion years to replicate.
- The experiment used random circuit sampling, a benchmark designed to favor quantum processors, and demonstrated a computational gap six orders of magnitude greater than Google’s 67-qubit Sycamore experiment.
- While this result reinforces quantum advantage claims, the benchmark does not directly translate to practical applications, and improvements in classical algorithms may challenge its long-term significance.
A Chinese research team has officially published findings that their quantum processor, Zuchongzhi 3.0, outperforms Google’s latest quantum computing efforts by a factor of one million. The study, released in Physical Review Letters and detailed previously on arXiv, reinforces China’s growing influence in the race for quantum computational advantage, a milestone where quantum computers outperform classical machines in specific tasks.
Initial details on the experiment had been released prior in the pre-print server for early review.
Findings and Implications
The researchers demonstrated that Zuchongzhi 3.0, a 105-qubit superconducting quantum processor, completed a complex computational task in mere seconds, according to the study. By comparison, simulating the same task on the world’s most powerful classical supercomputer, Frontier, would take an estimated 6.4 billion years. This result marks the most significant leap in quantum computational advantage to date, surpassing Google’s 2019 claim of quantum supremacy with Sycamore, which completed a similar task in 200 seconds—something Google estimated would take the best classical supercomputer 10,000 years.
The results establish a new benchmark in quantum computational advantage, the study suggests emphasizing that the work could open avenues for investigating how increases in qubit count and circuit complexity can enhance efficiency in solving real-world problems.
Quantum computational advantage remains an evolving benchmark, often contested as classical computing techniques improve. Yet, Zuchongzhi 3.0’s scale — running a 32-cycle random circuit sampling experiment, a common benchmarking test for quantum advantage, with 83 qubits — widens the computational gap between quantum and classical machines. However, and speaking of the evolving benchmark, critics would charge that the random circuit sampling approach is specifically designed for quantum processors, making it an apples-oranges benchmark comparison to classical machines working on practical problems.
The study suggests, though, that such improvements are bringing quantum computing closer to practical applications in optimization, artificial intelligence and materials science.
Methods and Experimental Design
The team at the University of Science and Technology of China designed Zuchongzhi 3.0 with 105 transmon qubits, significantly increasing from its predecessor, Zuchongzhi 2.0. The processor’s layout features a 15-by-7 qubit rectangular lattice, integrating 182 couplers to enhance connectivity. The researchers selected 83 qubits for their experiment, optimizing them for reduced error rates and higher stability.
To evaluate performance, the team conducted a large-scale random circuit sampling experiment. This process involves executing a sequence of randomly chosen quantum operations, then measuring the system’s output. Classical supercomputers struggle to replicate this process due to the exponential complexity of quantum states.
Google’s previous record in random circuit sampling involved a 67-qubit system running at 32 cycles. The Chinese team increased this complexity, running their experiment at a depth of 32 cycles with 83 qubits. The classical computational cost to simulate this system was estimated at six orders of magnitude higher than Google’s 67-qubit experiment.
Technical Advancements
Zuchongzhi 3.0’s performance stems from multiple engineering improvements. The team optimized qubit design by refining capacitance and Josephson junction parameters to reduce charge noise. Enhancements in fabrication methods included lithographically defining qubit components using tantalum and aluminum bonded through an indium bump flip-chip process, reducing contamination and improving coherence times.
The improvements led to notable gains in qubit stability. The processor’s relaxation time (T1) reached 72 microseconds, and its dephasing time (T2) improved to 58 microseconds. The single-qubit gate fidelity was measured at 99.90%, while the two-qubit gate fidelity reached 99.62%, surpassing the capabilities of earlier superconducting quantum processors.
Limitations and Challenges
Despite its performance gains, the study acknowledges ongoing challenges in scaling quantum computing. Errors in multi-qubit operations remain a hurdle, particularly as circuits increase in complexity. As mentioned above, the researchers also note that while random circuit sampling serves as a benchmark for computational advantage, it does not directly translate into solving real-world problems.
Classical supercomputing algorithms for simulating quantum circuits continue to improve, potentially narrowing the advantage gap. Advances in tensor network algorithms and other classical techniques may challenge the longevity of any declared quantum advantage.
Increasing Qubit Counts, Improving Fidelity
The study suggests that increasing qubit counts and improving circuit fidelity will be critical for advancing practical applications of quantum computing. The researchers highlight optimization problems, machine learning, and drug discovery as potential near-term beneficiaries of these developments.
The rapid progression of quantum hardware suggests that the next phase will focus on error correction and fault tolerance, key requirements for large-scale, practical quantum computing. Companies and institutions globally, including Google, IBM, and various Chinese research groups, are accelerating efforts in these areas.
Research Institutions
The research was conducted with contributions from multiple institutions, including the Hefei National Research Center for Physical Sciences at the Microscale and the School of Physical Sciences at the University of Science and Technology of China, the Shanghai Research Center for Quantum Science, the CAS Center for Excellence in Quantum Information and Quantum Physics, the Hefei National Laboratory, QuantumCTek Co., Ltd., the Henan Key Laboratory of Quantum Information and Cryptography, the National Institute of Metrology in Beijing, the Jinan Institute of Quantum Technology and Hefei National Laboratory Jinan Branch, the School of Microelectronics at Xidian University, and the CAS Key Laboratory for Theoretical Physics at the Institute of Theoretical Physics, Chinese Academy of Sciences.
Because the Physical Review Letters study was behind a paywall, the arXiv study was used for technical information. Cross-checking with media reports that accessed the Letters version suggest close similarities with the arXiv study published in December.
