Google announced that its Quantum AI team has successfully overcome a key bottleneck in quantum computing: the feasibility of the "magic state preparation" scheme has been demonstrated on the Willow processor, resulting in an error reduction of approximately 40 times compared to not using this technology. This groundbreaking achievement was published today in the journal Science and represents a crucial step toward the scalable practical application of quantum computing.
Google's Quantum AI Team Successfully Demonstrates the Feasibility of Magic State Preparation Technology
As early as 2004, researchers predicted a challenging bottleneck. To achieve universal quantum computing and run programs of practical value, it is essential to perform a specific set of quantum operations. The method involves preparing what is known as a "magic state," a term derived from the ability of these quantum states to generate quantum computations that surpass classical computing. However, there is a challenge: preparing high-quality, error-corrected magic states is extremely resource-intensive. Some estimates suggest that quantum computers may need to utilize up to 95% of their total computational power to generate magic states, leaving very little resource for actual computation.
Moreover, once a magic state is used, it cannot be reused, making it a scarce resource. Consequently, researchers have been continuously working to reduce the resource expenditure for magic state preparation.
In 2024, Google Quantum AI researchers Craig Gidney, Noah Shor, and Cody Jones drew inspiration from existing literature and optimized it to develop what is now referred to as magic state preparation technology. This scheme is expected to significantly reduce resource consumption, operating like a filter. A noisy quantum state is injected into a small area of the quantum chip, followed by a series of highly coordinated measurements to identify errors and eliminate subpar samples. Craig's team demonstrated that this method can "cultivate" quantum states of extremely high purity and low error, which can then be simultaneously converted into larger, practical coding forms.
If successfully implemented on quantum hardware, this could drastically reduce the scale of quantum computers required for practical tasks.
During the initial testing of the cultivation protocol, Google Quantum AI researcher Emma Rosenfeld observed data that appeared to contradict physical laws. After months of troubleshooting, simulations, peer discussions, and consistency checks, she discovered that the issue stemmed from erroneous assumptions in the team's commonly used measurement methods. Specifically, noise from laboratory equipment drift and temperature fluctuations interfered with the validity of the benchmark results. The team needed to synchronize adjustments between the theoretical framework and engineering solutions, adding extra control measures to mitigate the effects of this noise, and innovated a benchmarking method that uses the cultivation technology itself as a measurement tool to address the problem at its root.
Building on this achievement and further algorithmic innovations and engineering solutions, Google has elevated quantum metrology technology to a new pinnacle. On Google Quantum AI's Willow processor, this protocol achieves up to 40 times error compression compared to unprocessed quantum states; the resulting T-state logical fidelity reaches a record 99.99%. This achievement has now been published in Science, providing the first proven solution to one of the most severe resource bottlenecks in quantum computing, demonstrating that Google can efficiently prepare this critical magic state on real hardware.
Google has also preliminarily verified that the theories proposed by Craig and others can be adapted to its quantum computing architecture. The team achieved a switch between "color codes" (a new type of quantum error correction code particularly suitable for magic state preparation) and "surface codes" (a more mature error correction scheme), integrating a distance-3 color code quantum state into a distance-5 surface code. This integration protocol first appeared in a theoretical paper in 2024, and this experiment confirms that purified quantum states can stably exist and be smoothly transmitted within the hardware.