Draft · v0.2 · Unreviewed
← Index/Project · Nadarasa ReductionDraft · v0.2 · Selene Emulator

Nadarasa · Proofs · Selene/Guppy

Rewriter rules, proved on shots

Each of the five bastard-rewriter rules (F, I, HH, CC, B) is an asserted unitary identity LHS ≡ RHS. We compile both sides into Guppy kernels, run the 6×3 tomography grid (six input states × three measurement bases) on the Selene emulator, and pass the rule iff every cell's |Pₐ(1) − P_B(1)| stays inside the 4σ shot-noise band. The B rule is a classicalpost-measurement identity (Z-basis statistics ignore preceding Z-phases), so it runs on Z basis only — anything else would diverge by design.

rules

5/5

shots/cell

512

wall-time

71s

(F) Spider fusion

LHS RZ(0.3π) · RZ(0.45π) RHS RZ(0.75π)

PASS · 18/18

Thm 8.34 / Cor 8.35 · Z(α)–Z(β) on a wire fuses to Z(α+β).

prep \ basisZXY
|0
0.00/0.00
0.50/0.47
0.50/0.46
|1
1.00/1.00
0.51/0.49
0.49/0.47
|+
0.50/0.48
0.85/0.84
0.19/0.13
|-
0.52/0.50
0.15/0.14
0.87/0.83
|+i
0.53/0.54
0.87/0.88
0.86/0.85
|-i
0.47/0.50
0.15/0.18
0.13/0.13

worst tv = 0.0566 · threshold (4σ) = 0.125 · 15.8s

(I) Identity removal

LHS RZ(0π) RHS ·

PASS · 18/18

Cor 8.35 (special case) · Z(0) on a wire is the wire itself.

prep \ basisZXY
|0
0.00/0.00
0.52/0.48
0.49/0.48
|1
1.00/1.00
0.48/0.49
0.52/0.54
|+
0.48/0.55
0.00/0.00
0.51/0.48
|-
0.50/0.51
1.00/1.00
0.46/0.47
|+i
0.51/0.47
0.51/0.50
0.00/0.00
|-i
0.49/0.52
0.47/0.50
1.00/1.00

worst tv = 0.0645 · threshold (4σ) = 0.125 · 16.2s

(HH) Hadamard cancellation

LHS H · H RHS ·

PASS · 18/18

Cor 9.21 · H·H = identity.

prep \ basisZXY
|0
0.00/0.00
0.52/0.52
0.49/0.49
|1
1.00/1.00
0.46/0.50
0.49/0.51
|+
0.50/0.53
0.00/0.00
0.54/0.50
|-
0.50/0.50
1.00/1.00
0.48/0.49
|+i
0.52/0.51
0.48/0.50
0.00/0.00
|-i
0.48/0.46
0.50/0.51
1.00/1.00

worst tv = 0.043 · threshold (4σ) = 0.125 · 16.1s

(CC) Colour change

LHS H · RZ(0.37π) · H RHS RX(0.37π)

PASS · 18/18

Eq. 9.20 · H·Z(α)·H = X(α).

prep \ basisZXY
|0
0.30/0.28
0.52/0.49
0.97/0.96
|1
0.67/0.70
0.51/0.44
0.05/0.04
|+
0.52/0.50
0.00/0.00
0.47/0.47
|-
0.48/0.52
1.00/1.00
0.48/0.52
|+i
0.03/0.04
0.54/0.52
0.29/0.33
|-i
0.97/0.96
0.51/0.50
0.71/0.68

worst tv = 0.0664 · threshold (4σ) = 0.125 · 16.7s

(B) Bastard absorption (Z-basis)

LHS RZ(0.7π) RHS ·

PASS · 6/6

Thm 8.72 · A classical (post-measurement) Z-spider absorbs adjacent Z-phases: Z-basis readouts are invariant under a preceding Rz(α). Verified by Z-basis-only tomography; X/Y data would diverge and that's the point — bastard is a CLASSICAL rule.

prep \ basisZ
|0
0.00/0.00
|1
1.00/1.00
|+
0.45/0.51
|-
0.53/0.54
|+i
0.50/0.52
|-i
0.52/0.48

worst tv = 0.0605 · threshold (4σ) = 0.125 · 5.7s

Shot-based equivalence tests of the 5 bastard-rewriter rules (F, I, HH, CC, B). Each unitary rule runs the 18-cell tomography grid (6 prep states × 3 measurement bases); the B rule is a classical post-measurement identity so it runs Z-basis only. PASS means worst total-variation distance stays below the 3σ shot-noise band.