Quantized memory of measurement records near the light cone of quantum circuits

Modern quantum computers are beginning to do something that was once impossible: they can measure their own internal parts while a calculation is still running. Instead of waiting until the very end to see the result, these machines peek at the state of their tiny quantum bits, or qubits, in the middle of a process. This stream of data is the raw output of the machine, containing the answers to complex problems like error correction or the preparation of exotic quantum states. However, for the most powerful and chaotic types of circuits, predicting what this stream of data should look like is just as hard as simulating the entire computer itself. Once a device becomes too large for classical supercomputers to model, scientists lose their reference point. They have no way to know if the machine is working correctly or if the strange data it produces is just noise, because there is no "correct" answer to check against.

The researchers in this study tackled a fundamental question: is there any part of this chaotic data stream that remains predictable, even for the largest machines and with any type of quantum gate used? They discovered that the answer is yes, but only if you look at the data from a very specific perspective. By focusing on the measurements taken at the very edge of the "light cone"—the fastest possible path information can travel through the circuit—they found that the data follows a simple, exact rule. This rule holds true regardless of how large the computer is or what specific gates are used to manipulate the qubits. It turns out that the chaotic behavior of the rest of the machine acts like a perfect, forgetful background, leaving the edge measurements to follow a clear, mathematical pattern that can be calculated exactly.

To understand how this works, imagine a long line of people passing a message down the chain. In a chaotic system, the message usually gets scrambled beyond recognition. But if you stand at the very end of the line and listen only to the person who just arrived, you find that the message they bring depends only on the person who stood next to them a moment ago. The rest of the crowd, no matter how large or noisy, does not affect what that specific person says. The researchers showed that in a quantum circuit, the measurements taken along this fastest path behave exactly like a simple two-state system. The outcome of the next measurement depends only on the previous one, forming a predictable chain. This is surprising because the rest of the circuit is doing something incredibly complex, yet the edge of the light cone remains simple and exact.

The team took this idea further by moving their observation point slightly inside the circuit, away from the very edge. They found that as they moved deeper, the memory of the system grew in a very specific, quantized way. For every step they moved inward, the amount of information the system could remember increased by exactly one bit. This means the complexity of the data is not random; it is strictly limited by geometry. If you are one step inside the edge, the system remembers one extra piece of information. If you are two steps inside, it remembers two. This rule applies to almost any type of gate used in the circuit. The researchers proved that for generic gates, the system hits this limit, remembering exactly as much as the geometry allows, and no more.

To prove these ideas were not just theoretical, the scientists tested them on a real quantum processor with 156 qubits, accessed through the cloud. They ran a series of experiments where they prepared a specific starting state and then measured the qubits as they moved through the circuit. First, they confirmed that the measurements taken right at the edge of the light cone followed the simple, predictable pattern they had derived. They then tested how the system reacted to noise. They found that if they let the qubits wait in a "resting" state before they reached the measurement line, the data became corrupted. However, they discovered a simple fix: by randomly flipping the state of the qubits just before they entered the measurement line, they could erase the corruption and restore the exact law, all without needing extra complex gates. This "late twirl" acted as a perfect reset, ensuring the data remained clean.

The researchers also used this setup to check if the machine was performing a special type of gate known as "dual-unitary." These gates are rare and have unique mathematical properties. The theory predicted that if the machine used a dual-unitary gate, the measurements taken one step inside the edge would show absolutely no memory of the past; the data would be completely random. When they ran the test, the data for the dual-unitary gate showed exactly zero memory, confirming the gate was working as intended. In contrast, when they used a standard, chaotic gate, the data showed the predicted memory, proving the system was remembering the past as the theory said it should.

Finally, the team looked for the "memory" of the system deeper inside the circuit. They predicted that for chaotic gates, the data would remember the outcomes of measurements from two or three steps ago, but for dual-unitary gates, it would remember nothing. On the real hardware, they found exactly what they expected. The chaotic gates showed a clear memory of the past, while the dual-unitary gates showed none. They repeated these tests on two other different processors, and the results held up every time. The size of the memory matched their predictions perfectly, provided they accounted for the specific noise of the machine. This work provides a new, exact reference for checking quantum computers. It shows that even in the most chaotic systems, there are pockets of perfect predictability that can be used to verify the machine is working correctly, offering a reliable way to test hardware without needing to simulate the entire impossible task.