UCLA team cuts neutral atom errors up to 25 times with an atom-loss decoder

[ColdStorage77]

Neutral atom quantum computers have a particular weakness: atoms can simply disappear. Quantum Zeitgeist reports on a preprint from Jacob Murri at UCLA and University of California collaborators presenting a decoder that handles this atom loss far better than before

The decoder, called maximally matched mixture decoding, weights possible error scenarios according to the chance of correlated atom loss and then runs standard belief propagation. It was tested on Gross codes, a type of quantum low density parity check code with parameters 144, 12 and 12

Compared with decoders that ignore atom loss, logical error rates fell by up to around 25 times. Against earlier loss aware methods, the improvement was about two times

The key insight is that correlated losses, where two connected qubits fail together, are far more likely than independent failures, so a decoder that expects them performs much better. This is a simulation study tested on one code family, and scaling to larger machines remains an open question

Atom loss is one of the main problems holding back neutral atom platforms, so decoders tailored to it are very valuable. Quantum LDPC codes are attractive because they need fewer physical qubits than surface codes. Combining the two could be an important step for companies like QuEra, Atom Computing and Pasqal

How big a problem is atom loss for neutral atom computers? And will quantum LDPC codes replace surface codes?

Researchers Cut Quantum Errors 25x With New Atom-Loss Decoderquantumzeitgeist.com