Back to age-quantum.org AGE-TGI-2S3 + QEC · Photonic quantum research · ES/EN

Results

Results

Error rates, gate fidelity, and quantum scalability on integrated photonic platforms

Segments
7
1 - Feff
4.76×10-2
εlogical (d=5)
7.45×10-3
εlogical (d=13, η=0.95)
7.24×10-5
Physical Errors

1 - Feff: Physical Error per Segment

The physical error per segment, quantified as 1 - Feff, is the probability that a photonic qubit suffers an error (photon loss or coupling error) when traversing a single segment of the AGE-TGI-2S3/IDS decomposition. This parameter is calculated from the component's optical loss and coupling/phasing fidelity.

For a typical platform with competitive loss per segment, the resulting physical error is at a moderate level. The QEC code reduces it exponentially: with d=13 and high ηdet, the logical error drops by several orders of magnitude.

It is important to note that the physical error has two components: passive optical loss (attenuation in waveguides and junctions) and coupling errors (deviations in the coupling ratio from the nominal value). Both are modeled as erasures in the QEC framework, allowing efficient correction.

Component Typical loss Contribution
Waveguide (Si)~0.5 dB/cmPropagation
Directional coupler~0.05 dBOperator A
Phase shifter~0.02 dBOperator B
Bending loss~0.01 dBRouting
Total/segment0.103 dBReference
Logical Errors

εlogical: Logical Error of the QEC Code

Code Distance (d) Failure condition εlogical (moderate η) εlogical (high η) Reduction vs. no QEC
No QEC 1 Any error 4.76×10-2 4.76×10-2 —
Mini-QEC (Repetition Code) 5 ≥ 3 / 5 (majority vote) 7.45×10-3 1.71×10-3 ~6×
Strong QEC (Repetition Code) 13 ≥ 7 / 13 (majority vote) 7.24×10-5 2.51×10-6 ~660×
Key observation: with the published model, d=13 and ηdet = 0.95 give εlogical ≈ 7.24×10⁻⁵ (≈660× below the physical error); at ηdet = 0.99, ≈ 2.51×10⁻⁶. The previous values (1.4×10⁻⁶ and ~10⁻⁹) mixed different erasure probabilities across rows and were withdrawn on 2026-11-16 (see the errata log).
Repetition Code

Error Evolution with Distance

The repetition code for erasures shows an exponential reduction of the logical error with distance d. For erasures, the code failure probability is the probability that more than d/2 physical qubits are lost, which decreases exponentially with d as long as the physical error probability is below the code threshold.

Our simulation data confirm this exponential trend. The decay rate depends directly on ηdet: with high η, the slope is much steeper than with moderate η, explaining the large improvement observed in εlogical.

d εlógico (η=0.95) εlógico (η=0.99)
32.55×10-29.42×10-3
57.45×10-31.71×10-3
72.27×10-33.24×10-4
97.10×10-46.31×10-5
137.24×10-52.51×10-6
Evidence

Evidence status

Every published figure is generated by an accessible script. This table summarises what is numerically verified (VER) and what is still pending modelling (PENDING).

Claim Status Source
S₂(h/3)³ is order 2 and reduces the local error ×9 versus Strang VER assets/data/verify_integrator_order.py
The Yoshida composition S₄ reaches genuine order 4 with the same 7 segments (slope 4.99) VER assets/data/verify_integrator_order.py
Repetition-code εlogical table (d=3–13) VER assets/data/repetition_code_model.py
Temporal-mode selectivity Σ 39→646 (×16.5) VER data/04_age_correction.json · TM banks (age-quantum.org)
Two-qubit gates, GHZ states and logical-qubit counts PENDING bank QC-02 (under construction)
Scalability

Scalability and Perspectives

The scalability of AGE-TGI-2S3/IDS + QEC depends on three critical factors: optical loss per segment (determines physical error), detection efficiency ηdet (determines QEC code effectiveness), and integration density of the chip (determines how many physical qubits fit in a given area).

Logical error and integration density define the scaling path. The reference configuration combines AGE-TGI-2S3/IDS (7 segments) with the d=13 repetition layer and high-efficiency detectors (η > 0.99). Per-chip logical-qubit counts will be published once the model includes two-qubit gates (PENDING, bank QC-02).

Next steps include: (1) 3D FDTD simulations with MEEP to individually optimize optical components, (2) experimental verification of predictions on fabricated chips, and (3) extension to topological codes (surface codes) adapted to the erasure structure of photonics.

0.103 dB
Loss/segment (reference)
16.5
εlogical improvement (model)
Open Data

Simulation Data

Raw data available for independent replication

Repetition Code

Logical error rates for d=3–13 and ηdet = 0.90/0.95/0.99, generated by the published model script. CSV format.

Download CSV

Model script (Python)

Single-source repetition-code model: regenerates the published εlogical table and CSV. Standard library only.

repetition_code_model.py