Microsoft And Quantinuum Report on Major Gains in Quantum Error Correction

Insider BriefQuantum computers have long promised to solve problems beyond the reach of conventional machines. Yet the field has faced a persistent obstacle. The quantum bits, or qubits, that store and process information are extremely fragile. Even minor disturbances from their environment can introduce errors that accumulate and overwhelm a calculation before it finishes.A study published in Nature reports a significant step toward addressing that challenge. Researchers from Microsoft Quantum and Quantinuum demonstrated that carefully designed error-correction techniques can reduce the rate of computational errors well below those observed in the underlying hardware itself. This study builds on a series of Microsoft and Quantinuum demonstrations first reported in 2024, including the creation of record-setting logical qubits, but extends those earlier results by showing that quantum error correction can repeatedly suppress errors during computation itself.In a Microsoft blog post on the work, the team writes: “Previously, Microsoft and Quantinuum reported the most reliable logical qubits on record, involving more than 14,000 individual experiments without a single observed error and active syndrome extraction without destroying the logical qubits. We then extended that progress by creating 12 highly reliable logical qubits and demonstrating a hybrid, end-to-end chemistry simulation that combined logical qubits, AI, and HPC to estimate the ground-state energy of an important catalytic intermediate within chemical accuracy. Together, those milestones showed not only that reliable logical qubits are possible, but that they can be used in workflows that begin to solve real scientific problems.” According to the study, the team achieved improvements ranging from 11 times to 800 times compared with equivalent calculations performed directly on physical qubits. The experiments were conducted using Quantinuum’s trapped-ion quantum processors.The findings address whether today’s quantum hardware benefit from quantum error correction strongly enough to justify the substantial overhead required to implement it.“Performing quantum algorithms for critical problems in physics and chemistry requires substantially lower error rates than the physical error rates of present quantum computers,” the researchers wrote in the study, adding, “Our results show that state-of-the-art quantum devices are already able to make use of fault tolerance and error correction to strongly suppress errors in non-trivial quantum circuit computations.”Quantum error correction is often described as the bridge between today’s experimental systems and future practical quantum computers. Unlike classical computers, which can simply copy information to protect it, quantum mechanics prohibits direct duplication of an unknown quantum state. Instead, information must be distributed across many physical qubits in a way that allows errors to be detected and corrected without destroying the underlying data.That approach comes at a cost because protecting a single logical qubit typically requires many physical qubits, along with additional operations to monitor for mistakes. The challenge has been demonstrating that the benefits outweigh the added complexity.Previous experiments had shown promising signs with some groups demonstrating that larger error-correcting codes could reduce the rate at which errors accumulated. Others performed operations on encoded qubits. But uncertainty remained over whether error correction could consistently outperform unencoded calculations across increasingly complex computations.The researchers developed two error-correction schemes tailored to Quantinuum’s trapped-ion architecture. The first, known as the carbon code, used 12 physical qubits to protect two logical qubits. The second, called the tesseract subsystem code, used 16 physical qubits to protect four logical qubits while enabling more efficient correction procedures.The experiments took advantage of several characteristics of trapped-ion systems. In Quantinuum’s processors, ions can be physically moved within the device, allowing interactions between distant qubits rather than restricting operations to nearest neighbors.The team first tested their approach using Bell states, a fundamental form of quantum entanglement. In the unprotected version of the experiment, Bell-state preparation produced an error rate of approximately 0.8%. When the carbon code was applied without aggressive filtering techniques, the error rate fell to roughly 0.17%.The most dramatic result came when the researchers used a strategy combining correction with selective rejection of runs that showed ambiguous error signatures. Under those conditions, they observed no errors in more than 15,000 accepted trials. Statistically, that corresponded to an estimated error rate of 0.001%, representing an approximately 800-fold improvement over the physical baseline.The researchers also examined repeated error correction during an ongoing computation.This capability is essential for practical quantum computers. Correcting errors only at the beginning or end of an algorithm offers limited protection. Large-scale applications will require errors to be detected and corrected repeatedly as calculations proceed.Using the carbon code, the researchers performed as many as 10 rounds of error correction. They reported an error rate per correction cycle of 0.006%, compared with physical baselines ranging from roughly 0.37% to 0.59%, depending on the comparison method. That represented an improvement of more than 50 times.The study described this as the first demonstration — at least to the teams’ knowledge — of beneficial mid-computation error correction reducing logical error rates during repeated computational operations.The team then turned to more complex tasks using the tesseract code. They prepared several forms of graph states, highly entangled arrangements of qubits that serve as important benchmarks for logical operations. The experiments included structures involving four, eight and 12 logical qubits.In each case, the protected calculations outperformed their unprotected counterparts.The Path-4 experiment achieved a 15-fold reduction in errors. The Cube-8 experiment produced an 11-fold improvement despite involving deeper circuits and multiple rounds of correction. The Cat-12 experiment showed a 22-fold improvement.The results suggest that error correction can provide meaningful advantages even before quantum computers reach the scale required for commercially useful applications.While the work makes significant progress, the researchers suggest it does not eliminate the challenges that remain.The demonstrations relied partly on post-selection, a technique in which runs showing certain error patterns are discarded during analysis. While the researchers report that scalable alternatives exist, including passing information about detected errors to higher layers of encoding, future fault-tolerant systems will need to operate continuously without depending heavily on rejecting results after the fact.The experiments also did not incorporate full real-time decoding and feed-forward correction during computation. Instead, many decisions about how to interpret syndromes — the patterns used to identify errors — were handled in post-processing.Scaling the approach further will require increasingly sophisticated control systems capable of making those decisions quickly while computations remain in progress.The study also points out the importance of hardware improvements. The dominant source of noise in the Quantinuum devices arose from dephasing associated with transporting ions within the processor. The researchers reported that compiler optimizations and the use of dynamical decoupling techniques helped reduce those effects.Additional refinements in hardware performance could amplify the gains demonstrated in the current work.Despite the remaining limitations, independent observers are likely to view the study as another indication that quantum computing is transitioning from proof-of-concept experiments toward more robust architectures capable of supporting practical algorithms.The field increasingly appears to be entering an era defined not merely by adding more qubits, but by using them more effectively.Fault tolerance has long been regarded as the threshold that separates experimental quantum devices from machines capable of solving real-world problems in chemistry, materials science and optimization.The new results do not mean that large-scale fault-tolerant quantum computers have arrived. The experiments involved relatively small logical systems, and enormous engineering hurdles remain before millions of physical qubits can be coordinated reliably.As the researchers write, “Our results show that quantum error correction with present quantum processors already reduces error rates in small circuits.”Future studies will focus on expanding these demonstrations to larger systems, incorporating real-time decoding and developing broader sets of fault-tolerant operations.Share this article:Keep track of everything going on in the Quantum Technology Market.In one place.