Katia Hakem of Quandela and Télécom Paris, along with colleagues, have demonstrated that the (2,2) Shor code does not always improve performance in photonic quantum computing when accounting for imperfect photon distinguishability. The work reveals a fundamental limitation of evaluating logical encodings with idealized resource states, showing that how initial quantum states are created can critically determine fault tolerance. Researchers modeled noise in a 6-ring fusion network, comparing all-photonic and emitter-assisted repeat-until-success schemes for resource state preparation. For all-photonic construction, encoding reduced the distinguishability threshold from 0.136% to 0.080%.
Photonic FBQC & Dual-Rail Encoding Fundamentals
The all-photonic construction of a quantum circuit reduced the required photon distinguishability threshold to 0.080%. This refinement demonstrates that logical encoding, specifically the (2,2) Shor code, does not universally enhance performance in the face of hardware imperfections, a finding that challenges common assumptions about error correction strategies. The research, focused on photonic qubits, reveals that the benefits of encoding are contingent on the method used to generate the initial quantum states.
Researchers modeled a 6-ring fusion network to explore the interplay between encoding and resource-state generation, focusing on the impact of partial photon distinguishability during both state creation and measurement. Dual-rail encoding, where a qubit’s information is distributed across two optical modes, was central to the investigation, defining |0⟩ and |1⟩ states based on photon occupancy in those modes.
A valid dual-rail state contains exactly one photon across the two modes, while photon loss maps the qubit to a vacuum state outside the logical subspace. This approach allows for the detection of errors through the identification of missing photons, but its effectiveness is diminished by imperfections in photon indistinguishability. In contrast to the all-photonic approach, a construction using quantum emitters, specifically, a repeat-until-success (RUS) scheme, showed comparable unencoded and encoded thresholds of 0.256% and 0.271% respectively.
This result highlights a fundamental limitation in evaluating logical encodings using idealized resource-state assumptions, as the method of resource state generation critically determines the overall fault tolerance of photonic architectures. The team’s analysis demonstrates that simply adding error correction layers does not guarantee improved performance if the underlying physical states are not sufficiently well-defined.
The study’s methodology involved comparing the performance of unencoded and Shor (2,2)-encoded resource states prepared using both all-photonic and emitter-assisted RUS schemes. Graph states, serving as the resource states for FBQC, were represented as networks where vertices denote photonic qubits and edges represent controlled-Z (CZ) operations. The work acknowledges that photon loss remains a primary hardware limitation in photonic quantum computing, influencing the detected photon number and potentially leading to errors.
However, the findings extend beyond photon loss, demonstrating that partial photon distinguishability, the inability to perfectly differentiate between photons, can significantly impact the effectiveness of error correction. This is particularly true when employing all-photonic methods for resource state preparation, where even small levels of distinguishability can degrade performance.
The implications of this research extend to the design of future photonic quantum computers, suggesting that a comprehensive approach is needed, considering both error correction codes and the physical implementation of resource states. The team’s results emphasize that optimizing resource-state generation is as important as implementing sophisticated error correction schemes.
Distinguishability Modeling Impacts Fusion Measurements
Researchers developed hardware-motivated noise models to capture the effects of partial photon distinguishability during both resource state generation and the essential fusion measurements that underpin fusion-based quantum computation (FBQC). These models were then integrated into fault-tolerance simulations of a six-ring fusion network, a common architecture in FBQC, and its encoded counterpart using the (2,2) Shor code. The simulations compared resource states prepared using two distinct methods: a fully photonic approach and an emitter-assisted repeat-until-success (RUS) scheme, revealing differing tolerances to distinguishability.
This indicates that encoding can indeed improve performance in this scenario by lowering the required quality of photons. However, the RUS scheme exhibited a different behavior, maintaining comparable distinguishability thresholds for both unencoded (0.256%) and encoded (0.271%) resource states. The differing results stem from the distinct preparation circuits inherent in each method, leading to different effective error channels.
The simulations considered the impact of partial photon distinguishability on Type-I and Type-II fusion operations, key components of FBQC. A Type-II fusion, a destructive joint measurement, is reduced to separate measurements in the Z-basis if it fails, with the XX outcome erased but the ZZ parity potentially recoverable.
The models account for how partial distinguishability introduces errors in the reported parity outcomes, influencing the overall fidelity of the computation. The analysis builds upon existing work that considers photon loss and fusion failure, but goes further by directly linking photon distinguishability to specific error probabilities within the fusion network. Previous studies modeled these errors as abstract parameters, but this work derives circuit-dependent error models, providing a more realistic representation of the physical imperfections present in real-world hardware.
The team’s models detail how partial photon distinguishability introduces a phase-flip error on surviving qubits during Type-I fusions, with a probability proportional to the degree of distinguishability. For Type-II fusions, the reported outcome of the XX parity measurement is flipped with a similar probability, while the ZZ outcome remains unaffected.
These subtle differences in error channels contribute to the observed variations in fault tolerance between the all-photonic and RUS schemes. The choice of preparation scheme, therefore, can critically determine the fault tolerance of photonic architectures, and a seemingly beneficial encoding strategy may not always deliver the expected improvements if the initial resource states are not sufficiently high quality.
All-Photonic Resource State Preparation
Simulations of six-ring fusion networks reveal that the benefits of error correction are not universal, particularly when accounting for imperfections in photon distinguishability during resource state creation and measurement. The work details how the effectiveness of the Shor (2,2) code, a common technique for bolstering resilience against photon loss, is contingent upon the method used to prepare the initial quantum states. Researchers compared all-photonic and emitter-assisted repeat-until-success (RUS) schemes for generating these states, finding divergent outcomes in fault tolerance thresholds.
The study modeled optical circuit-dependent error channels for both fusion measurements and resource-state preparation, incorporating these effects into simulations of unencoded and encoded six-ring networks. This approach moves beyond treating errors as abstract parameters, instead deriving circuit-specific error models to more accurately reflect hardware limitations.
The team’s analysis focused on a six-ring graph state, a resource state defined by six photonic qubits initialized in the | + ⟩ state and connected by CZ operations. Each vertex in the graph is associated with a qubit, and edges represent these controlled-Z gates, with stabilizer generators defining the entangled state.
This indicates that the RUS scheme is less sensitive to distinguishability errors, and that encoding does not provide a comparable advantage in this scenario. All-photonic approaches rely on successive interference and postselection stages, where preparation errors can propagate and correlate across multiple fusion outcomes.
In contrast, the RUS scheme uses sequentially controlled quantum emitters, offering a deterministic path to generating entangled states but demanding sustained coherence throughout the emission sequence. The choice between these schemes, therefore, is not simply a matter of achieving the same ideal encoded resource state. Instead, the specific preparation circuit introduces unique error channels that can either amplify or mitigate the impact of photon distinguishability. These channels were then incorporated into fault-tolerance simulations, providing a more subtle understanding of how imperfections affect the overall system performance.
RUS Scheme & Shor (2,2) Encoding Comparison
These models move beyond abstract error parameters to consider the optical circuit-dependent nature of distinguishability errors, detailing how they manifest during both fusion measurements and the preparation of resource states. The team’s approach considered that preparation errors can propagate through multiple fusion outcomes, creating correlations that complicate error correction. The distinction between the two schemes arises from the specific preparation circuits and the resulting error channels they introduce.
The simulations revealed that the interplay between error channels and preparation methods can significantly alter the effectiveness of error correction. The study’s focus on the six-ring fusion network provides a concrete example of how these effects manifest in a specific quantum architecture.
Simply achieving an ideal encoded resource state is not enough; the method used to create that state must also be carefully considered. The findings emphasise the importance of developing hardware-motivated noise models that accurately capture the imperfections present in real-world quantum devices. This requires a detailed understanding of the optical circuits used to prepare and manipulate photons, as well as the characteristics of the photon sources themselves. The team’s work provides a framework for evaluating the performance of different resource state preparation schemes and identifying those that are most robust to imperfections.
👉 More information
🗞 Performance of Fusion-Based Quantum Computing with Imperfect Photon Indistinguishability
✍️ Katia Hakem, Grégoire de Gliniasty and Stephen C. Wein (Quandela, Massy, France); Nathan Shettell (MajuLab, CNRS-UCA-SU-NUS-NTU International Joint Research Laboratory and Centre for Quantum Technologies, National University of Singapore, Singapore, Singapore); Paul Hilaire (LTCI, Télécom Paris, Institut Polytechnique de Paris, Palaiseau, France)
🧠 ArXiv: https://arxiv.org/abs/2611.23083




See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
