The Signal
Preprint finds photon mismatch can lower photonic correction thresholds
A preprint’s simulations find that Shor encoding lowers the photon-distinguishability threshold in an all-photonic fusion network, while an emitter-assisted scheme has similar thresholds with and without encoding. The results depend on how resource states are prepared and do not include several noise effects the paper identifies as future work.

Encoding results depend on preparation
A preprint posted on November 17 reports simulations in which Shor encoding did not consistently improve tolerance to photon distinguishability in a photonic error-correction network. In the six-ring fusion network studied, the estimated threshold fell with encoding for resource states prepared all-photonically, but stayed similar for an emitter-assisted method, according to the paper by Katia Hakem and colleagues.
Source arxiv.org
The study examines fusion-based quantum computation, where small entangled resource states are connected through linear-optical fusion measurements. The products of those measurements provide parity checks used in error correction. The researchers compared an unencoded network with one using Shor (2,2) encoding, and modeled two ways to prepare the resource states: an all-photonic construction and an emitter-assisted repeat-until-success (RUS) scheme.
Different thresholds in simulation
For the all-photonic construction, the estimated distinguishability threshold fell from 0.136% without encoding to 0.080% with encoding. In the RUS construction, the estimated thresholds were 0.256% for the unencoded network and 0.271% for the encoded network.
The threshold estimates were based on simulations comparing logical error rates at code distances d=9 and d=11, with the threshold located where the difference between those rates changed sign. The authors also modeled the two preparation methods because, despite producing the same ideal encoded resource state, their circuits create different effective error channels.
How distinguishability enters
Partial distinguishability occurs when photons do not perfectly overlap in spectral, temporal or spatial modes. That can reduce two-photon interference and leave information that reveals which photon took a path. The resulting detection may be accepted even when it corresponds to an incorrect fusion parity, creating an error that is not flagged as a failure.
In the paper’s model, partial distinguishability during successful Type-I fusion produces a phase-flip error on the surviving qubit with probability D/2. For Type-II fusion, it can flip the reported XX parity outcome with probability D/2, while the ZZ outcome is unaffected in the unboosted operation modeled. The study also derives circuit-dependent error channels for resource-state preparation, where errors can affect multiple later fusion outcomes.
Limits of the comparison
The authors explain that encoding supplies redundancy for recovering parity information erased by photon loss or fusion failure, but preparing a larger encoded state takes additional operations that can introduce distinguishability errors. In the all-photonic case they studied, that preparation noise outweighed the benefits of encoding; the RUS construction produced a different balance.
The comparison is limited to the six-ring resource state with Shor (2,2) encoding and uses first-order noise channels represented as effective Pauli errors. The paper identifies leakage, multiple distinguishable photons, correlations across successive interference stages and combined photon-loss and distinguishability effects as topics for further study. It is a preprint, not a report of a hardware demonstration.