Qubits prime? Amazon uses quantum cat states with error correction The company shows off a mix of error-resistant hardware and error correction. John Timmer Feb 26, 2025 2:15 pm | 6 Credit: ALFRED PASIEKA/SCIENCE PHOTO LIBRARY Credit: ALFRED PASIEKA/SCIENCE PHOTO LIBRARY Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreFollowing up on Microsoft's announcement of a qubit based on completely new physics, Amazon is publishing a paper describing a very different take on quantum computing hardware. The system mixes two different types of qubit hardware to improve the stability of the quantum information they hold. The idea is that one type of qubit is resistant to errors, while the second can be used for implementing an error-correction code that catches the problems that do happen.While there have been more effective demonstrations of error correction in the past, a number of companies are betting that Amazon's general approach is the best route to getting logical qubits that are capable of complex algorithms. So, in that sense, it's an important proof of principle.Herding catsThe basic idea behind Amazon's approach is to use one type of qubit to hold data and a second to enable error correction. The data qubit is extremely resistant to one type of error, but prone to a second. Those errors are where the second type of qubit comes in; it's used to run an error-correction code that's effective at picking up the problems the data qubits are prone to. Combined, the two are hoped to allow error correction to be handled by far fewer hardware qubits.In a standard computer, there's really only one type of error to worry about: a bit that no longer holds the value it was set to. This is called a bit flip, since the value goes from either zero to one, or one to zero. As with most things quantum computing, things are considerably more complicated with qubits. Since they don't hold binary values, but rather probabilities, you can't just flip the value of the qubit. Instead, bit flips in quantum land involve inverting the probabilitiesgoing from 60:40 to 40:60 or similar.But bit flips aren't the only problems that can occur. Qubits can also suffer from what are called phase flip errors. These have no equivalent in classical computers, but they can also keep quantum computers from operating as expected.In the past, Amazon demonstrated qubits that made it trivially easy to detect when a bit flip error occurred. For the new work, they moved on to something different: a qubit that greatly reduces the probability of bit flip errors.They do this by using what are called "cat qubits," after the famed Schrdinger's cat, which existed in two states at once. While most qubits are based on a single quantum object being placed in this sort of superposition of states, a cat qubit has a collection of objects in a single superposition. (Put differently, the superposition state is distributed across the collection of objects.) In the case of the cat qubits demonstrated so far by companies like Alice and Bob, the objects are photons, which are all held in a single resonator, and Amazon is using similar tech.Cat qubits have a distinctive feature compared to other options: bit flips are improbable, and get even less probable as you pump more photons into the resonator. But this has a drawback: more photons mean that phase flips become more probable.Flipping catsThose phase flips are why a second set of qubits, called transmons were brought in. (Transmons are a commonly used type of qubit based on a loop of superconducting wire linked to a microwave resonator and used by companies like IBM and Google.) These were used to create a chain of qubits, alternating between cat and transmon. This allowed the team to create a logical, error-corrected qubit using a simple error-correction code called a repetition code. The layout of Amazon's hardware. Data-holding cat qubits (blue) alternate with transmons (orange), which can be measured to detect errors. Credit: Putterman et. al. Here, each of the cat qubits starts off in the same state and is entangled with its neighboring transmons. This allowed the transmons to track what was going on in the cat qubits by performing what are called weak measurements. These don't destroy the quantum state like a full measurement would but can allow the detection of changes in the neighboring cat qubits and provide the information needed to fix any errors.So, the combination of the two means that almost all the errors that occur are phase flips, and the phase flips are detected and fixed.In more typical error-correction schemes, you need enough qubits around to do measurements to identify both the location of an error and the nature of the error (phase or bit flip). Here, Amazon is assuming all errors are phase flips, and its team can identify the location of the flip based on which of the transmons detects an error, as shown by the red flags in the diagram above. It allows for a logical qubit that uses far fewer hardware qubits and measurements to get a given level of error correction.The challenge of any error-correction setup is that each hardware qubit involved is error-prone. Adding too many into the error-correction system will mean that multiple errors are likely to occur simultaneously in a way that causes error correction to become impossible. Once the error rate of the hardware qubits gets low enough, however, adding additional qubits will bring the error rate down.So, the key measurement done here is comparing a chain that has three cat qubits and two transmons to one that has five cat qubits and four transmons. These measurements showed that the five qubit chain had a lower error rate than the smaller one. This shows that the hardware is now at a state where error correction provides a benefit.The characterization of the system indicated a couple of major limits, though. Cat qubits make bit flips extremely unlikely, but not impossible. By focusing error correction only on phase flips, any bit flips that do occur inescapably trigger the failure of the entire logical, error-corrected qubit. "Achieving long logical bit-flip times is challenging because any single cat qubit bit flip event in any part of the repetition code directly causes a logical bit flip error," the authors note. The other issue is that the transmons used for error correction still suffer from both bit and phase flips, which can also mess up the entire error-corrected qubit.Where does this leave us?There are a number of companies like Amazon that are betting that using a somehow less error-prone hardware qubit will allow them to get effective error correction using fewer total hardware qubits. If they're correct, they'll be able to build error-corrected quantum computers using far fewer qubits, and so potentially perform useful computation sooner. For them, this paper is an important validation of the idea. You can do a sort of mixed-mode error correction, with a robust hardware qubit paired with a compact error-correction code.But beyond that, the messages are pretty mixed. The hardware still had to rely on less robust hardware qubits (the transmons) to do error correction, and the very low error rate was still not low enough to avoid having occasional bit flips. And, ultimately, the error rate improvements gained by increasing the size of the logical qubit aren't on a trajectory that would get you a useful level of error correction without needing an unrealistically large number of hardware qubits.In short, the underlying hardware isn't currently good enough to enable any sort of complex calculation, and it would need radical improvements before it can be. And there's not an obvious alternate route to effective error correction. The potential of this approach is still there, but it's not obvious how we're going to build hardware that lives up to that potential.As for Amazon, the picture is even less clear, given that this is the second qubit technology that it has talked about publicly. It's unclear whether the company is going to go all-in on this approach, or is still looking for a technology that it's willing to commit to.Nature, 2025. DOI: 10.1038/s41586-025-08642-7 (About DOIs).John TimmerSenior Science EditorJohn TimmerSenior Science Editor John is Ars Technica's science editor. He has a Bachelor of Arts in Biochemistry from Columbia University, and a Ph.D. in Molecular and Cell Biology from the University of California, Berkeley. When physically separated from his keyboard, he tends to seek out a bicycle, or a scenic location for communing with his hiking boots. 6 Comments