November 28, 2025 by Ingrid Fadelli, Phys.org
Collected at: https://phys.org/news/2025-11-google-quantum-ai-dynamic-surface.html
Quantum computers are computing systems that process information leveraging quantum mechanical effects. These computers rely on qubits (i.e., the quantum equivalent of bits), which can store information in a mixture of states, as opposed to binary states (0 or 1).
While quantum computers could tackle some computational and optimization problems faster and more effectively than classical computers, they are also inherently more prone to errors. This is because qubits can be easily disturbed by disturbances from their surrounding environment, also referred to as noise.
Over the past decades, quantum engineers and physicists have been trying to develop approaches to correct noise-related errors, also known as quantum error correction (QEC) techniques. While some of these codes achieved promising results in small-scale tests, reliably implementing them on real circuits is often challenging.
Researchers at Google Quantum AI recently demonstrated the implementation of the surface code, a well-known QEC technique, using three distinct dynamic circuits. Their paper, published in Nature Physics, opens new possibilities for the real-world application of the surface code and could contribute to the development of more dependable quantum computers.
“For quantum computers to work reliably, they need a way to fix errors that inevitably occur,” Matt McEwen, co-first author of the paper, told Phys.org. “This is called QEC, and it’s essential for building fault-tolerant quantum computers. However, implementing QEC is a significant challenge because the error-detecting and correcting circuits are complex and demand extremely precise operations.”
The practical implementation of the team’s theoretical proposal
The surface code, the code that McEwen and his colleagues implemented, has been widely studied and tested in the past. This code works by organizing qubits on a 2D grid and repeatedly checking for faults, performing tests that can detect error without disturbing the quantum data stored in a system.
“I’d previously worked on a theory proposal showing that there are multiple ways to implement the surface code, which we called dynamic circuits,” said McEwen. “In particular, that work proposed three variations of the surface code‘s implementation that had different strengths and weaknesses.”
This recent study thus builds on McEwen’s earlier work, which theoretically demonstrated the feasibility of three distinct dynamic surface code implementations. His analyses suggested that these implementations would better adapt to the complex error mechanisms observed in experimental settings, yet they did not conclusively prove that they would work in real-world conditions.
“At the same time, there is a big experimental effort at Google on RCS and OTOC, using iSWAP gates,” said Alexis Morvan, senior author of the paper. “One of the new constructions enables doing QEC with the iSWAP gate. So, we thought that was a great opportunity to combine our knowledge from the NISQ experiments and QEC experiments into a single demonstration.”
When they started running their experiments, the researchers realized that they should be able to demonstrate all three of the dynamic surface code implementations initially described by McEwen. Their hope was to prove that theoretically demonstrated surface code implementations typically also work in experiments under real-world conditions.
Three successful circuit implementations
The three circuit implementations demonstrated by this team at Google Quantum AI are called hex, iSWAP and walking circuits.
“The hex circuit recompiles the standard surface code circuit to work on a hexagonal grid rather than a square grid,” explained Alec Eickbusch, co-first author of the paper.
“In a hexagonal grid of qubits, each qubit only needs to have 3 neighbors, which reduces complexity of fabrication and operation, allowing us to squeeze better performance out of the fewer couplers we are using.”
The hex circuit’s underlying could potentially enable the development of quantum devices with qubits arranged in a hexagonal grid, as opposed to a square grid. This could in turn simplify their fabrication and the control of qubits in the devices.
The team’s second circuit implementation was dubbed the iSWAP circuit. This is an adaptation of standard circuits that employs a different entangling gate, called iSWAP, instead of the standard CZ gate.
“iSWAP gates are simpler operations to perform, and don’t directly produce leakage errors, like CZs do,” said Eickbusch.
“However, they do suffer from a new kind of error (CPHASE error), so it is not obvious whether they could produce better performance than the CZ based circuit. We demonstrated that the iSWAP works well even on a device that was designed and optimized for CZ gates, so we hope that future devices can focus on designing for iSWAPs.”
The third and final implementation outlined in the researchers’ paper is called the walking circuit. This design essentially makes it possible for the physical qubits making up the code to switch roles, ‘walking’ protected quantum information across a quantum device.
“This walking leaves leakage errors behind, where they can be cleaned up before we walk the code back again, reducing the impact of leakage error events,” said Eickbusch. “It also enables a new way of moving logical qubits around the device, opening up new opportunities for logical algorithm compiling. “
Inspiring more dynamic surface code implementations
When they tested the three dynamic surface codes, the team found that they achieved remarkable results. The hex circuit enhanced the suppression of errors by 2.15 times, the walking circuit by 1.69 times, and the iSWAP by 1.56 times.
“The biggest takeaway from our work is confirming that these dynamical circuit implementations work in reality,” said McEwen. “We hope that this gives other groups confidence to try implementing codes using circuits that are adapted to perform best on their hardware and encourages people to come up with even more interesting circuit implementations now that they know they’re on solid footing. “
Overall, the results of this study highlight the potential of dynamic QEC circuits for realizing fault-tolerant quantum computing. A further novelty of the team’s work is the new detector budgeting technique they employed in their experiments.
“This is a novel and lightweight way of computing the impact of various component error rates on the overall logical performance and helps indicate which errors are important to focus on improving,” explained Morvan.
“Previously, error budgeting a QEC experiment has been a fairly complex affair, so providing this simple analytical way of arriving at an error budget should put this kind of information within reach of a lot more experimental QEC work.”
McEwen, Morvan, Eickbusch and their colleagues are now planning further experiments aimed at demonstrating even more complex circuits. One of these is a new dynamic circuit rooted in the so-called LUCI framework, a recently presented scheme for building fault-tolerant circuits.
“These are new circuits designed to adapt the surface code around dead components, like uncooperative qubits or couplers on the device,” added Morvan. “These circuits promise much better performance on devices with imperfect yield than previous state of the art, but are even more time-dynamic and spatially complex than the circuit we’ve demonstrated so far.”
More information: Alec Eickbusch et al, Demonstration of dynamic surface codes, Nature Physics (2025). DOI: 10.1038/s41567-025-03070-w.
Journal information: Nature Physics
Leave a Reply