Microsoft Quantum and Quantinuum report major advancements in quantum error correction, while IBM Research explores new codes using AI.
Recent research published in the journal Nature highlights significant advancements in quantum error correction by Microsoft Quantum and Quantinuum. Their collaboration showcases improvements in logical error rates ranging from 11 to 800 times over existing physical schemes, marking a crucial step in overcoming one of the primary obstacles to scalable quantum computing.
IBM Research has also made strides by utilizing large language models to identify new quantum error correction codes. Their system, based on OpenEvolve, has proposed 465 potential candidates, although further testing is necessary to confirm their practical applications.
Details of the Microsoft and Quantinuum Collaboration
Error correction is vital for the functionality of quantum computers, as current qubits are highly susceptible to noise, which leads to the accumulation of errors. Logical qubits, which are more reliable than their physical counterparts, are essential for maintaining computational integrity over extended periods.
The study titled Improved quantum processor logical error rates via correction and detection describes two optimized designs for Quantinuum's ion processor: a 12-qubit code derived from the Knill scheme and a 16-qubit tesseract color code. The 12-qubit design encodes two logical qubits while the 16-qubit design encodes four.
Notably, when generating a Bell state, the logical error rate improved drastically from approximately 0.8% to 0.001%, representing an 800-fold enhancement. Additionally, repeated error correction achieved a 51-fold reduction compared to the physical baseline.
“Our results show that modern quantum devices are already capable of leveraging fault tolerance and error correction to significantly suppress errors in non-trivial quantum circuits,” the abstract notes.
Microsoft also referenced past achievements with Quantinuum, including over 14,000 experiments without error, the successful demonstration of 12 logical qubits, and a hybrid chemical simulation utilizing AI and high-performance computing.
IBM's Innovative Approach to Error Correction Codes
IBM Research has introduced the OpenEvolve library to explore new quantum error correction codes. This open-source tool employs large language models to enhance software code evolution.
The focus has been on bivariate bicycle codes, which are characterized by low parity-check density and are significant for IBM's fault-tolerant quantum computing roadmap. The parameters for these codes are denoted as [[n,k,d]], with n representing physical qubits, k the logical qubits, and d the code distance, indicating the maximum number of errors tolerated.
From initial evaluations, 465 candidate codes were proposed, including a standout code [[288,50,8]] featuring 50 logical qubits, surpassing the previous record of 16. Other notable candidates include a compact code [[72,4,8]] and codes [[288,16,12]] and [[360,12,≤24]].
IBM anticipates that some of these candidates may rival the [[144,12,12]] gross code under certain noise conditions, although practical application of these codes requires further validation. The source code for this project is publicly available on GitHub, and IBM's plans for a large-scale fault-tolerant quantum computer, IBM Quantum Starling, are set for completion by 2029, aiming for 200 logical qubits and 100 million quantum gates.
Additionally, Quantum X Labs and the Quantum Machines IQCC research platform are planning to test an AI decoder for quantum error correction, further pushing the boundaries of this field.
