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

Method

Research

From the mathematical method to the photonic implementation: how AGE-TGI-2S3/IDS + QEC enables fault-tolerant quantum computing

Method

AGE-TGI-2S3/IDS: symmetric composition of operators (S₂) and Yoshida composition (S₄)

The AGE-TGI-2S3/IDS formula, in notation S₂(h/3)³, is the symmetric Strang formula S₂ applied three times with step h/3 (three partitions of the interval h). This construction approximates the target unitary operator

U_target = eh(A+B)

through the sequence of 7 physical segments alternating operators A and B. The name "2S3" reflects that it is a symmetric order-2 formula (local Trotter error O(h³)) with 3 subdivisions of the interval h, resulting in the concatenation of 7 exponentials:

S₂(h/3)³ = A(h/6) · B(h/3) · A(h/3) · B(h/3) · A(h/3) · B(h/3) · A(h/6)

Genuine order 4 with the same 7-segment pattern (Yoshida composition, one negative weight):

S₄(h) = S₂(γ₁h) · S₂(γ₀h) · S₂(γ₁h)    γ₁ ≈ 1.3512, γ₀ ≈ −1.7024

The symmetry of each S₂ guarantees cancellation of odd-order error terms; repeating S₂ three times does not raise the order, but reduces the local error exactly 9-fold (= 3²) versus a single S₂(h) — numerically verified (slope 3.00). Genuine order 4 is obtained with the Yoshida composition S₄ = S₂(γ₁h)·S₂(γ₀h)·S₂(γ₁h), with γ₁ ≈ 1.3512 and γ₀ ≈ −1.7024 (one negative weight: a reverse rotation in SU(2)); after merging adjacent A exponentials it keeps the same 7 segments and achieves local error O(h⁵) (slope 4.99). Both checks are reproduced by the published script.

Key Properties

Order:
2 (S₂(h/3)³, local error O(h³)) · genuine 4 with Yoshida S₄ (O(h⁵))
Segments:
7 physical
Sequence:
A-B-A-B-A-B-A
Weights:
S₂(h/3)³: alternating h/6 and h/3 · S₄ (Yoshida): γ₁ = 1.3512, γ₀ = −1.7024
Symmetry:
Temporal (palindromic)
Architecture

The 7 Physical Segments

Each segment corresponds to an elementary unitary operator implemented as a photonic component

Seg 1
A(h/6)

Directional coupler with partial coupling.

Seg 2
B(h/3)

Thermal/electro-optic phase shifter.

Seg 3
A(h/3)

Directional coupler with full coupling.

Seg 4
B(h/3)

Central phase shifter at the symmetric midpoint.

Seg 5
A(h/3)

Coupler symmetric to segment 3 (temporal reflection).

Seg 6
B(h/3)

Phase shifter symmetric to segment 2.

Seg 7
A(h/6)

Closing coupler, symmetric to segment 1, completing the palindromic sequence.

Model

SU(2) Model

In the SU(2) group, the elementary operators are constructed from Pauli‑like matrices. We define:

A = ½σx    B = ½σz

The target operator is U_target = e^{h(A+B)}, where h is the time step (Trotter parameter). In SU(2), the exponential of any linear combination of Pauli matrices has a closed analytical form, facilitating exact numerical verification of the Symmetric composition of operators approximation.

The choice of SU(2) as a test model is strategic: any single-qubit quantum gate can be decomposed into SU(2) rotations, and universality is achieved by adding entanglement. Therefore, optimizing single-qubit gate fidelity in SU(2) is the critical first step toward universal quantum computing.

Gate Fidelity

Gate fidelity is defined as F = |Tr(U_target† · U_approx)|² / d² where d = 2 for SU(2). In the absence of optical losses, AGE-TGI-2S3/IDS achieves near‑unity fidelity for typical h, demonstrating the power of the high‑order formula.

Note: In practice, fidelity is limited by optical loss in components (couplers and phase shifters), not by the Trotter error. This is the fundamental motivation for combining AGE-TGI-2S3/IDS with QEC codes.
Theory

Symmetric composition of operators

The higher-order Symmetric composition of operators formula allows approximating the exponential of a sum of non-commuting operators as a product of individual exponentials. For two non-commuting operators A and B ([A,B] ≠ 0), the first-order formula (Lie-Trotter) is simply e^{h(A+B)} ≈ e^{hA} · e^{hB}, with error O(h²).

eh(A+B) ≈ ehA · ehB   (error O(h²))

The symmetric second-order formula (Strang) reduces the error to O(h³): e^{h(A+B)} ≈ e^{hA/2} · e^{hB} · e^{hA/2}. Higher-order formulas are constructed recursively: the order-2p formula is obtained by composing the order-p formula with a modified step.

eh(A+B) ≈ ehA/2 · ehB · ehA/2   (Strang, O(h³))

AGE-TGI-2S3/IDS = S₂(h/3)³ uses the symmetric S₂ formula with step h/3 repeated 3 times (hence the ³ superscript) and generates the 7 physical segments. It keeps order 2 and reduces the local error exactly ×9 versus S₂(h). Raising the order to 4 requires the Yoshida composition S₄ = S₂(γ₁h)·S₂(γ₀h)·S₂(γ₁h): the same 7 segments with one negative weight, local error O(h⁵), and an additional ×155 error reduction at h=0.05 versus S₂(h/3)³. Verified with the published script.

S₂(h/3)³ → Orden 2   |   7 segmentos   |   Error local O(h³), ×9 vs Strang
S₄ Yoshida → Orden 4   |   7 segmentos   |   Error local O(h⁵), pendiente 4.99
Implementation

Dual-Rail Photonic Qubit

In dual-rail encoding, a qubit is represented by the presence of a single photon in one of two optical modes (waveguides). The state |0⟩ corresponds to the photon in the upper guide and |1⟩ to the photon in the lower guide.

Operator A is implemented as a directional coupler that transfers amplitude between the two guides. Operator B is implemented as a phase shifter that introduces a relative phase difference between the guides.

The fundamental advantage of dual-rail encoding is that the main source of error is photon loss, which manifests as an erasure of the qubit. Unlike Pauli errors (bit or phase flips), erasures are detectable: we know which qubit was lost. This additional information allows much more efficient QEC codes, such as the repetition code for erasures.

Photonic Components

Directional Coupler
Operator A: eσx·θ
Phase Shifter
Operator B: eσz·φ
Optical attenuation
→ Erasure (detectable)
QEC

Quantum Error Correction

How erasure and repetition codes protect photonic qubits

Erasure Code

In photonics, the dominant error is photon loss. When a photon is lost, the qubit disappears but we know it has disappeared (unlike Pauli errors, where the error location is unknown). This property is called an erasure and allows significantly more efficient QEC codes.

The protection layer used in AGE-TGI-2S3/IDS + QEC is a classical repetition code with majority decision, encoding one qubit into d physical copies over the announced-erasure channel: if the majority of copies survive, the qubit is recovered. It tolerates ⌊(d−1)/2⌋ erasures and fails when the majority is lost. It is not a full quantum code: extension to true codes requires two-qubit gates (PENDING).

Erasure ≠ Pauli error: In erasure, the error location is known. This allows more efficient QEC codes.

Mini-QEC (d=5)

d=5 encodes 1 qubit into 5 physical copies with majority decision: it tolerates 2 erasures and fails if 3 or more of 5 are lost. At ηdet = 0.95, the published model gives εlogical ≈ 7.45×10⁻³.

d = 5
εlógico ≈ 7.45×10-3 (η=0.95, modelo publicado)

Strong QEC (d=13)

d=13 tolerates 6 erasures and fails if 7 or more of 13 copies are lost. With the published model: εlogical ≈ 7.24×10⁻⁵ at ηdet = 0.95 and ≈ 2.51×10⁻⁶ at ηdet = 0.99.

d = 13
εlógico ≈ 7.24×10-5 (η=0.95)
εlógico ≈ 2.51×10-6 (η=0.99)
Critical parameter

ηdet: Detection Efficiency

The photon detection efficiency ηdet determines how many erasures enter the code. A value ηdet = 0.95 does not mean 5 % of erasures go unnoticed: in dual-rail, an unregistered photon produces a missing click, i.e. an announced erasure. What low ηdet does is increase the number of erasures: q = 1 − ηdet·(1 − εphys), the quantity entering the code.

Our analysis reveals that ηdet is the main bottleneck for fault-tolerant photonic quantum computing. The relationship is exponential: improving ηdet from typical values to near‑unity reduces εlogical by several orders of magnitude, regardless of the code distance.

This has direct implications for detector design: investing in high-efficiency detection technologies (η > 0.99) has a dramatically greater impact than increasing the code distance or reducing marginal optical losses. In practice, ηdet = 0.99 is the most important technological target for the viability of at-scale photonic quantum computing.

ηdet εlog (d=5) εlog (d=13) Improvement factor
0.902.33×10-29.39×10-41×
0.957.45×10-37.24×10-5~13×
0.991.71×10-32.51×10-6~375×