Researchers affiliated with the AWS Center for Quantum Computing at Caltech have made significant progress in understanding how to minimize errors in quantum computers, a troublesome issue that remains the largest obstacle to developing the technology of the future.
Quantum computers, founded on the seemingly enchanting characteristics of the quantum domain, promise applications in various sectors, such as healthcare, materials engineering, cryptography, and fundamental physics. However, while current quantum computers can be beneficial for probing specific domains of physics, a universal quantum computer that can address more complex problems is not feasible yet, primarily due to its fundamental sensitivity to interference. Vibrations, heat, electromagnetic disturbances from mobile devices and Wi-Fi connections, or even cosmic radiation from space can disrupt qubits—quantum bits—from their quantum state. Consequently, quantum computers are prone to far more errors than their traditional computer equivalents.
In the February 26 issue of the journal Nature, a group of scientists from AWS and Caltech unveil a novel quantum chip architecture aimed at minimizing errors by utilizing a type of qubit referred to as a cat qubit. Proposed initially in 2001, cat qubits have since been developed and enhanced through research. Now, the AWS team has assembled the first scalable cat qubit chip capable of effectively mitigating quantum errors. Dubbed Ocelot, this new quantum computing chip derives its name from the spotted wild feline, while also referencing the internal “oscillator” technology that forms the basis of cat qubits.
“To ensure quantum computers succeed, we require error rates to improve by about a billionfold compared to today’s levels,” explains Oskar Painter (PhD ’01), John G Braun Professor of Applied Physics and Physics at Caltech and lead of quantum hardware at AWS. “Error rates have halved approximately every two years. At that pace, it would take us 70 years to reach our goal. Instead, we are constructing a new chip architecture that might expedite our progress. That being said, this is merely an initial building block. We still have significant work ahead.”
Qubits are founded on 1s and 0s, akin to those in classical computers, yet these 1s and 0s exist in a state of superposition. This means they can simultaneously embody any combination of 1 and 0. However, this trait also makes them delicate, prone to easily falling out of superposition. “The very aspects that grant qubits their strength also render them vulnerable to quantum errors,” states Painter.
Traditional digital computer systems have a straightforward mechanism for managing errors. Fundamentally, the creators of these systems implement extra redundant bits to safeguard data against errors. For instance, a single piece of information is duplicated across three bits, meaning that any individual bit has two backup counterparts. If one of these bits encounters an error (switching from 1 to 0 or vice versa), while the other two remain unchanged, a simple coding system—in this case, a three-bit repetition code—can identify the error and rectify the singular erroneous bit.
Due to the intricate nature of superposition present in qubits, they experience two categories of errors: bit flips, similar to those in classical digital systems, and phase flips, where the qubit states of 1 and 0 become misaligned (or desynchronized) with one another. Researchers have created numerous strategies to address both types of errors in quantum systems, yet these techniques necessitate qubits to have a considerable number of backup partners. Indeed, current qubit technologies may require thousands of additional qubits to achieve the requisite level of error protection. This would be akin to a news organization employing a vast team of fact-checkers to verify the correctness of its publications rather than relying on a small group. The overhead for quantum computers is both excessive and cumbersome.
“We are engaged in a long-term endeavor to develop a functional quantum computer that can perform tasks beyond the capabilities of the finest supercomputers, yet scaling them presents a significant challenge,” states study co-author Fernando Brandão, Bren Professor of Theoretical Physics at Caltech and director of applied science at AWS. “Thus, we are exploring novel methods for error correction that will alleviate the overhead.”
The group’s innovative approach depends on qubits formed from superconducting circuits comprising microwave oscillators, wherein the 1 and 0 states that signify the qubit are defined by two distinct large-scale amplitudes of oscillation. This design confers stability to the qubit states, making them resistant to bit-flip errors. “You can envision the two oscillating states as akin to a child on a swing, swinging at substantial amplitudes, but moving either to the left or the right. A gust of wind may disturb the swing, yet the amplitude is so significant that it cannot rapidly shift from one direction to the other,” Painter illustrates.
Indeed, the term “cat” qubits refers to the characteristic of these qubits to simultaneously occupy two very large, or macroscopic, states—much like the renowned cat in Erwin Schrödinger’s thought experiment, which can exist as both dead and alive at once.
With cat qubits significantly diminishing bit-flip errors, the remaining errors needing correction are phase flip errors. Furthermore, addressing a single type of error allows researchers to employ a repetition code similar to those used to rectify bit-flip errors in traditional systems.
“A classical code like the repetition code in Ocelot indicates that the new chips require fewer qubits for error correction,” states Brandão. “We have demonstrated a more scalable architecture that can decrease the number of additional qubits needed for error correction by up to 90 percent.”
The Ocelot chip achieves this by integrating five cat qubits, along with special buffer circuits to stabilize their oscillations, and four ancillary qubits to identify phase errors. The findings outlined in the Nature article illustrate that the team’s straightforward repetition code is effective in detecting phase flip errors and enhances as the code expands from three cat qubits to five cat qubits. Moreover, the phase error detection mechanism was implemented such that it maintained a high degree of bit-flip error suppression in the cat qubits.
This proof-of-concept demonstration still requires further development, but Painter expresses enthusiasm regarding the performance that Ocelot has exhibited so rapidly, and the team is actively conducting more research to scale the technology. “It’s a challenging problem to address, and we must continue to support fundamental research while remaining connected to, and learning from, the substantial work occurring in academia,” he remarks.
The Nature study titled “Hardware-efficient quantum error correction via concatenated bosonic qubits,” was financed by AWS. Alongside numerous AWS researchers, additional authors from Caltech include John Preskill, Richard P. Feynman Professor of Theoretical Physics and the Allen V. C. Davis and Lenabelle Davis Leadership Chair of the Institute for Quantum Information and Matter, or IQIM, and Gil Refael, the Taylor W. Lawrence Professor of Theoretical Physics.