To Distill, or Not to Distill

 

Networking quantum processors via photon-mediated entanglement is a promising approach to large-scale quantum computing. The remote entanglement is shared as Bell pairs, which are noisy, and it has long been assumed that they must be distilled to produce higher-fidelity pairs before being used in quantum error correction.

 

Recent results show that lattice-surgery operations at surface code boundaries can tolerate substantially higher error rates than those in the interior. Building on this, we derive an explicit condition for choosing between distillation-free and distillation-assisted strategies. Below a crossover fidelity, distillation is resource-optimal. Above it, the distillation-free approach wins. We further show that this conclusion remains robust under finite entanglement generation rates and memory decoherence, which together define three distinct operating regimes. We discuss implications for both trapped-ion and neutral-atom platforms, noting that state-of-the-art trapped-ion links already fall within the distillation-free regime.