Boost your quantum smarts and expand understanding of elements that impact the quantum risk timeline with a quick bite on "CAT qubits"

Practical problems and cryptographic attacks on quantum computers make logical qubits essential. Unlike their physical counterparts, which quickly become unreliable, logical qubits offer greater stability. However, creating a logical qubit from multiple physical qubits is a significant challenge. Among the various possible physical implementations of a qubit, "CAT qubits" are a standout option.

Hardware designers strive to develop systems with a built-in physical mechanism that preserves stability, as these are generally more effective than artificial solutions. "CAT qubits" harness this idea. Their stability arises from the physical properties inherent in their design. Other implementations lack this advantage. While the technology itself is not new, recent publications (most recent example - link) and claims by Alice and Bob have brought it into the spotlight. The advancement of this technology is a significant trend to watch in the quantum industry, as it has the potential to impact the quantum risk timeline substantially. CAT qubits offer higher stability than other physical qubits, reducing the overhead for error correction. Current methods for building CAT qubits utilize silicon nanoelectronic devices, which allows forleveraging existing silicon chip production infrastructure.

Building an intuition around complex topics like these is invaluable. Theau Peronnin, the CEO of Alice and Bob, provided an excellent explanation of this subject. For more details about "CAT qubits", below is the enhanced transcript of his speech from a Q2B conference in Paris (link).

P.S. If you've read this far, you can be proud that you won't confuse "CAT qubits" with actual cats, but the name originates from the so-called "Schrodinger cat states".

Building a quantum computer presents significant challenges. While the promise of quantum computers is substantial due to their potential for super-polynomial speed-ups, they still fall short compared to classical machines by several orders of magnitude. Quantum computers typically have at most 1,000 qubits, whereas classical machines handle petabytes. Their gates operate in the nanosecond range compared to the picoseconds of classical machines. Most importantly, quantum computers are highly prone to errors, with an error occurring approximately every 1,000 gates—22 orders of magnitude more than classical machines.

Addressing this high error rate is the core challenge as the industry transitions from noisy machines to fault-tolerant quantum computers.

We need gates that perform five orders of magnitude better than current machines to achieve significant applications like quantum advantage, proven by algorithms in fields such as condensed matter simulation. The same goes for more challenging use cases like finance or breaking RSA encryption, which require even greater performance improvements.

Despite these challenges, we don't need to match classical computers' performance exactly. To get started, we only need five or six orders of magnitude improvements.

Building a quantum computer is paradoxical. On one hand, the computer must behave quantum mechanically and be perfectly isolated, like Schrödinger's cat in a box. However, for the computer to function and be programmable, it needs tobe controlled, which requires opening information channels and inevitably leads to decoherence. This is why quantum error correction is essential.

Classical error correction is straightforward. If Alice wants to send a message to Bob with a 10% chance of corruption, she can use redundancy to encode the message. Instead of sending '0', she sends '000', and instead of '1', she sends '111'. If the received message is '001', it's likely that a single error occurred. This redundancy reduces the error rate. With the redundancy increases, the error rate drops exponentially.

Quantum error correction is more complex. Directly measuring quantum bits would create decoherence. Therefore, we use ancillary qubits to check for agreement without directly measuring the primary qubits. This process requires twice as many qubits since two types of errors need correction: bit flips and phase flips.

In quantum error correction, there's a threshold of redundancy. If adding quantum bits increases more noise than redundancy, then error correction fails. However, if the qubits are good enough, increasing their number will enhance redundancy, reducing logical errors. Currently, Google's paper demonstrates a system just at this threshold, which is crucial for transitioning from noisy intermediate-scale quantum (NISQ) to fault-tolerant quantum computing (FTQC).

CAT Qubits could be good enough to bridge that gap. CAT Qubit is a bipartite system: a pendulum at the bottom and an oscillator with a damping mechanism at the top. Driven at twice the pendulum's frequency, it stabilizes into two states of the same amplitude but opposite phases. This stability is key as it allows the system to remain powered and entropy to be managed effectively.

On a superconducting chips, this concept translates to an LCA oscillator exchanging photons by pairs. This suppresses bit flip errors exponentially, extending the error-free time from microseconds to tens of seconds. However, this also increases phase shift errors linearly, creating a manageable trade-off. CAT qubits reduce bit flips significantly while keeping manageable phase flips.

To correct the phase shift errors, a simple repetition code could be used. Since there is only one type of error to correct, we don't need the two-dimensional redundancy of a surface code. This makes error correction cheaper in terms ofoverhead, control electronics, and wiring, making the final solution more feasible.

CAT qubits are similar to transmon qubits used by companies like Google and IBM but are much more efficient. Transmon qubits require combining many two-level systems to create a large Hilbert space, leading to more errors as redundancy increases. In contrast, Bosonic qubits, including CAT qubits, use a physical system with many levels, like a harmonic oscillator to achieve the same outcome, with only two types of errors — photon loss or frequency shifts.

In terms of performance, a single CAT Qubit performed just as well as the entire Sycamore chip from Google for correcting bit flips.

CAT qubits are part of a family of bosonic codes, including JKP codes, binomial codes, and dual-rail codes. CAT Qubits are the most extreme in this ensemble, handling bit flips at the physical level by suppressing them exponentially and leaving phase flips to redundancy.