E0 · 7-qubit embedding

Clebsch–Gordan register.

2⁶ = 64 is too small. 2⁷ = 128 is the information-theoretic minimum. Three reserved codewords. The kernel sits in two contiguous coupled-basis blocks, so every E47 operation is a sandwich U_CG D₄₇ U_CG†.

qubits

7

min embedding

Hilbert

128

3 invalid

waste

2.3%

vs 75.6% naïve 9q

E47 blocks

2

10–34 · 90–111

Address map

J=0
J=1
J=2
J=3
J=4
J=5
J=6
inv
010–34 E₆90–111 E₃₀125–127
|10E47 · J=2

product |m₁ m₂ m₃⟩ = |2, 0, 2⟩ · base-5 idx = 25·0 + 5·2 + 0 = 10

M₂ / M₅ multiplicity

Inside the kernel, E₄₇ ≅ (ℂ⁵ ⊗ ℂ⁵) ⊕ (ℂ² ⊗ ℂ¹¹). Collective SU(2) noise acts on the representation factors R₂ and R₅. The multiplicity spaces are noiseless for that declared algebra:

  • J = 2 · addresses 10–34

    M₂ ≅ ℂ⁵ · 5-level qudit · log₂ 5 ≈ 2.32 bits

    Casimir C = 6. SU(5) logical gates on the multiplicity.

  • J = 5 · addresses 90–111

    M₅ ≅ ℂ² · binary logical qubit

    Casimir C = 30. Internal gap ΔE = 24 from M₂.

This is not a claim of generic Knill–Laflamme QEC on all 47 dimensions.

Three-stage sandwich

  1. Stage 1

    Uncoupling U_CG† — product basis to coupled CG basis.

  2. Stage 2

    Block-diagonal D₄₇ on 10–34 and 90–111. Reflection uses +1 on E47, −1 on complement.

  3. Stage 3

    Recoupling U_CG back to the physical product basis.

R₄₇ = U_CG D₄₇ U_CG†

Generic 7-qubit unitary synthesis is O(4⁷) ~ 16k CNOTs — a design estimate, not a measured transpile. Structured CG networks are the intended research target. OpenPulse M₂ SU(5) schedule on transmons is listed at 31.68 µs, under 32% of typical T₁.