Error-Corrected Memory and Logic on a Heavy-Hex Superconducting-Qubit Processor

The quest to build a useful quantum computer is often described as a race against noise. In the quantum world, information is held in fragile states that can be easily scrambled by the slightest disturbance from the environment. To protect this information, scientists use a technique called error correction, which spreads a single piece of data across many physical components. If one component fails, the others can reveal the mistake and fix it without destroying the data. This approach is essential for scaling up quantum machines, but it requires a massive number of components working in perfect harmony. The challenge is that real-world machines are never perfect; the tiny superconducting circuits used to store quantum information often have slight variations in their performance, and some components are simply broken or much noisier than others. If a computer cannot tolerate these flaws, the manufacturing process would be so wasteful that building a large machine would be impossible.

A team of researchers has demonstrated a way to make a large quantum processor work effectively even when it contains defective parts. Using a 156-qubit superconducting chip known as IBM Boston, they ran a complex error-correction experiment across the entire device. Instead of trying to force the broken parts to work, they developed a strategy to simply ignore them. By mapping out which components were performing poorly and redesigning the measurement schedule to route around them, the team was able to maintain high performance. They found that by excluding the worst-performing parts and using a smart computer program to interpret the results, they could significantly reduce the rate of errors. In tests designed to store information, this approach improved the reliability of the memory by up to 35 percent. In tests designed to perform logical operations, the improvement was even more dramatic, reaching 56 percent. The researchers also showed that by filtering out experiments where the data was likely corrupted by a specific type of failure called leakage, they could achieve extremely high fidelity in creating entangled states, with success rates exceeding 97 percent in post-selected runs.

The experiment was conducted on a chip with a unique layout called a heavy-hexagonal lattice, which connects the qubits in a specific pattern that is easier to manufacture but harder to use for error correction than a simple grid. The researchers used a dynamic code, a method that changes the way it checks for errors over time, allowing it to adapt to the chip's irregularities. They identified two specific areas on the chip where the components were consistently underperforming, with error rates much higher than the rest of the device. Rather than discarding the entire chip, they redesigned the error-correction protocol to avoid these areas entirely. This involved creating new, slightly larger checks that skipped over the bad components, effectively weaving the error correction around the holes in the fabric of the machine.

To make sense of the data coming from the chip, the team employed a sophisticated decoding strategy. They used a method called ACES to measure the specific noise patterns of the circuit on the day of the experiment, creating a detailed map of where errors were likely to occur. This map was fed into a decoder, a classical computer algorithm that acts like a translator, interpreting the raw signals from the quantum chip to determine if an error had happened and how to fix it. The researchers found that using this detailed, real-time noise information was far more effective than relying on generic, average error rates. However, they also discovered that even with the best noise map, the presence of defective components set a hard limit on performance. The decoder could not fully compensate for the bad parts; the only way to overcome them was to physically exclude them from the calculation.

The team tested their approach in three different ways. First, they performed memory experiments, where they stored a quantum state and checked how well it survived over time. They found that by excluding the defective components and using their noise-informed decoder, the logical error rate dropped significantly. In the best cases, the error probability per round of measurement was reduced to between 2.56 and 4.11 percent. This was a substantial improvement over using the defective components or using less accurate noise models. Second, they tested the stability of the system by running the error correction for many rounds to see if the system could suppress errors over time. Without excluding defects, the error rate would stop improving after about 12 rounds, hitting a plateau. With the defective components excluded, the error rate continued to drop all the way to 30 rounds, showing that the system could truly suppress errors when the bad parts were removed.

Finally, the researchers tested the ability of the machine to perform logical operations, specifically a process called lattice surgery, which allows two separate logical qubits to merge and interact. This is a critical step for performing complex calculations. They found that by filtering out runs where the initial state was likely to have leaked out of the safe computational space, they could achieve very high-quality results. When they combined this pre-selection with a post-selection technique that kept only the most confident outcomes, they were able to generate entangled states with a fidelity above 97 percent. This means that the resulting quantum states were extremely close to the ideal theoretical states, a level of quality that is necessary for future large-scale quantum computers.

The study highlights a crucial lesson for the future of quantum computing: perfection in manufacturing is not strictly required to achieve high performance. While the components on a chip will always vary, and some will inevitably be broken, the system can be designed to work around these flaws. The researchers showed that by combining active defect exclusion with smart decoding and careful filtering, a quantum processor can maintain high fidelity even when working with imperfect hardware. This approach suggests that as quantum chips grow larger and more complex, the ability to tolerate and work around defects will be just as important as the raw quality of the individual components. The work provides a practical path forward for building useful quantum computers on current-generation hardware, proving that high performance is possible even when the machine is not perfect.