BARC Codes as Algebraically Restricted Coherent-State Constellations: Geometry, Separation Bounds, and a Worked Single-Mode Benchmark

In the paper's words

Bosonic algebraically-restricted constellation (BARC) codes encode quantum information in finite superpositions of coherent states whose constellation points are solution sets of multivariate complex polynomial systems, with the polynomial constraints chosen to reflect photon-gain and photon-loss error symmetries [1],[2]. This paper develops a self-contained geometric analysis of the single-mode BARC construction. We restate the polynomial-to-symmetry pipeline, derive explicitly the upper bound on minimum pairwise separation for N equally weighted points on a circle of fixed mean photon number, and evaluate it for the hexagonal BARC family: for N = 6 points at mean photon number n̄, the maximal minimum chord is d = √n̄, giving worst-case coherent-state overlap |⟨α|β⟩|² = exp(−n̄/2). [...]

The abstract, verbatim; every line below is quoted from the text.

What it claims

for N points on a circle of radius r, the minimum chord satisfies d_min ≤ 2r sin(π/N).

Numbers it reports

1. We then compute, with full arithmetic, the exact pairwise overlaps and the loss-error Gram matrix entries for a concrete hexagonal instance at n̄ = 8, obtaining $|\langle \alpha _{i}|\alpha _{j}\rangle |^{2} = e^{-4} \approx 0.0183$ and post-loss overlaps $e^{-3.8} \approx 0.0224$ at transmissivity $\eta = 0.95$, and we verify the approximate Knill–Laflamme diagonal condition, finding that the off-diagonal-to-diagonal deviation for the loss error is ~13.5% (ratio $0.1353353, O(10^{-1})$), while the next-nearest and opposite-pair contributions are individually bounded at $O(10^{-2})$ or below.

   from Abstract

2. The QNFO corpus study "Bosonic Codes as the Native Encoding" compares cat, GKP, binomial, and surface codes using a photons-per-logical-qubit metric at logical error rate $p_{L} = 10^{-6}$, finding that bosonic codes require 5–40× fewer photons and roughly 100× fewer modes than surface codes.

   from 2. Background and Related Work

3. Relative to the diagonal value $c = 8$, the off-diagonal-to-diagonal ratio is $1.0826824/8 = 0.1353353$ — i.e., about 13.5%, which is not negligible.

   from 4. Analysis

4. However, the KL-relevant object for the normalized code includes the recovery's ability to project; the standard figure of merit is the post-loss overlap ratio: $|\langle \sqrt{\eta} \alpha _{i}|\sqrt{\eta} \alpha _{j}\rangle |^{2} / |\langle \sqrt{\eta} \alpha _{i}|\sqrt{\eta} \alpha _{i}\rangle |^{2} = e^{-3.8} / 1 = 0.022370 (A6)$ for adjacent pairs at $\eta = 0.95$ — a 2.24% distinguishability ceiling, which is the quantity an optimal recovery must work with.

   from 4. Analysis

5. The off-diagonal KL violation, however, is 13.5% of the diagonal value at n̄ = 8 (ratio 0.1353353), which is large by qubit-code standards; whether approximate KL at this level yields acceptable logical error rates is an empirical question that only the optimal-recovery entanglement fidelity can settle.

   from 6. Discussion

6. The four-point contrast (Result 6) shows that at fixed photon budget, fewer constellation points buy dramatically better separation — 54.6× smaller worst-case overlap squared.

   from 6. Discussion

How it is done

1. We restate the polynomial-to-symmetry pipeline, derive explicitly the upper bound on minimum pairwise separation for N equally weighted points on a circle of fixed mean photon number, and evaluate it for the hexagonal BARC family: for $N = 6$ points at mean photon number n̄, the maximal minimum chord is $d = \surd \bar{n}$, giving worst-case coherent-state overlap $|\langle \alpha |\beta \rangle |^{2} = \exp(-\bar{n}/2)$.

   from Abstract

2. The promise of this approach is architectural rather than instantaneous: polynomial solution sets are a vast, well-understood space of geometric objects, and symmetry is precisely the property that makes KL conditions tractable, since error operators that commute with the symmetry act uniformly on the code space.

   from 1. Introduction

3. It also highlights the dual use of the same overlap quantities we compute in Section 4: small pairwise overlaps aid both error correction and Bell-state distinguishability.

   from 2. Background and Related Work

4. A single bosonic mode with annihilation operator â; coherent states $|\alpha \rangle$ = $e^{-|\alpha |^{2}/2}$ e^(αâ†) $|0\rangle$.

   from 3. Methods

What would falsify it

1. (iii) The projection in Result 7 uses an undetermined proportionality constant; it should not be cited as a fidelity prediction.

   from 6. Discussion

Where the claim stops

1. The off-diagonal KL violation, however, is 13.5% of the diagonal value at n̄ = 8 (ratio 0.1353353), which is large by qubit-code standards; whether approximate KL at this level yields acceptable logical error rates is an empirical question that only the optimal-recovery entanglement fidelity can settle.

   from 6. Discussion

2. (ii) Our $\eta = 0.95$ benchmark is an assumption, not a measurement.

   from 6. Discussion

3. As argued in Section 2, BARC codes arrive as a v1 preprint whose benchmark claims cannot be verified from the abstract alone, in a corpus where withdrawn and removed arXiv records [3],[5],[6],[7],[8],[9] are a standing caution.

   from 6. Discussion

Where to read

1. Abstract253 words · 1 min
2. 1. Introduction322 words · 1 min
3. 2. Background and Related Work771 words · 3 min
4. 3. Methods403 words · 2 min
5. 4. Analysis1,511 words · 7 min
6. 5. Results366 words · 2 min
7. 6. Discussion444 words · 2 min
8. 7. Conclusion137 words · 1 min