Halving the cost of surface codes by running error correction at the logical level

Mike Newman (Google)
simons.berkeley.edu/talks/mik...
Near-Term Quantum Computers: Fault Tolerance + Benchmarking + Quantum Advantage + Quantum Algorithms
We all love the surface code: high thresholds, efficient decoding, universal fault-tolerant logic, all while running on a 2D square grid. But it's also very expensive, requiring around a thousand physical qubits per logical qubit to realize large-scale algorithms. What if, instead, we make the surface codes 'just big enough' to run more simple and efficient error correction at the logical level? We call these 'yoked surface codes', and they can halve (and with added complexity, even third) the cost of a 2D quantum memory. Based on arXiv:2312.04522.