REVIEW
Bosonic Loss Codes with Bounded Occupation
Quantum Physics · Submitted 25 Oct 2026 · 56 references
Pith's verdict: Solid, honest rate paper for constant-occupation Fock loss codes: modest headline gains, real machinery, and certified numerics that deserve a serious referee.
Why Pith says this
Highlight any text to comment on it.
Referee report
The paper
The paper claims
For every fixed occupation cap, bosonic codes can correct a fixed fraction of photon losses at explicit positive rates; the cap-two balanced case reaches about 0.281 logical qubits per excitation.
The reading
What carries the argument
The engine is a two-step reduction: choose a classical set of occupation vectors inside the capped simplex $\Sigma_{q,N}^{(B)} = \{n \in \mathbb{Z}_{\ge 0}^q : \sum_i n_i = N,\ n_i \le B\}$ with pairwise deletion distance greater than $\lfloor \delta N \rfloor$, then lift it to a quantum code. The lift, a cap-aware specialization of the lifting theorem the paper inherits from the cited literature, maps each occupation vector $h$ to a moment point whose coordinates are normalized diagonal Knill–Laflamme loss moments; pairwise separation removes every off-diagonal error term, and a Tverberg partition of these moment points — a point common to the convex hulls of $K$ parts — determines the logical states. The cap enters through the count $m_B(q,t) = [z^t](1 + z + \cdots + z^B)^q$ of loss patterns that remove at most $B$ photons per mode, which replaces the unrestricted pattern count and is never larger; projecting the moment points onto these coordinates preserves their affine dimension because the omitted coordinates vanish identically on the capped simplex. The quantum dimension exponent is then the classical constant-composition Gilbert–Varshamov exponent minus the moment exponent $\Lambda_B$, and the cap-two closed form $k_2(\delta)$ evaluates this difference explicitly.
Extended reading notes
Core claim
The paper's central claim is that a hard cap on per-mode occupation does not prevent Fock codes from correcting a linear number of photon losses at positive rate. For every fixed cap $B \ge 1$ and every ratio $\alpha = q/N > 1/B$, there exist exact $q = \lfloor \alpha N \rfloor$-mode Fock codes of total excitation $N$, per-mode occupation at most $B$, and loss distance $\lfloor \delta N \rfloor + 1$ whose dimension exponent is at least $k_B(\alpha,\delta) = R_{cc}^{B}(\alpha,\delta) - \Lambda_B(\alpha,\delta)$, the difference between a constant-composition Gilbert–Varshamov exponent for classical capped multiset codes and the exponential size of the cap-respecting loss patterns. At $(B,\alpha) = (2,1)$ the exponent has the closed form $k_2(\delta) = \log_2\!\bigl(\tfrac{3-u}{1+u}\bigr) + 2\delta \log_2 u - \Lambda_2(\delta)$, is positive for $0 < \delta < \delta_2^* \approx 0.1355$, and equals about $0.281$ at $\delta = 0.1$, exceeding the previously established qutrit exact-code rate of about $0.275$. As a converse applying even to degenerate codes, the paper proves $K h_B(N,r) \le [z^{N-r}] F_B(z)^q$ for every $0 \le r \le t$. For the pure-loss channel truncated to at most $\lfloor \delta N \rfloor$ losses and renormalized, uniformly random partitions of separated supports yield approximate codes with exponentially small worst-case error, and for loss probability $0 < \gamma \le 1/(B+1)$ their rates strictly improve the exact-code rates, staying positive at $(B,\alpha,\gamma) = (2,1,0.1)$ up to $\delta \approx 0.158$. The paper also establishes that deletion-ball volume in the capped simplex is Schur-concave in its center, smallest at the saturated point and largest at the balanced point.
Load-bearing premise
The load-bearing premise is the affine-hull form of the lifting theorem the paper inherits from the cited literature — that the dimension of an exact code scales with the affine dimension of the projected moment points rather than with the full pattern count — which the paper takes as implicit in the proof of a cited theorem rather than proving from scratch, together with the companion claim that projecting onto cap-restricted loss patterns preserves the affine dimension needed for Tverberg balancing.
Editorial extensions
If this is right
- Positive-rate, linear-distance exact loss correction holds under every fixed occupation cap with $\alpha B > 1$, so a bounded photon budget per mode is not a barrier to asymptotic loss protection (Theorem 13).
- At $B=2$, $q=N$, the optimized histogram layer gives about $0.281$ logical qubits per excitation at $\delta = 0.1$, above the qutrit rate of about $0.275$, and exceeds the tabulated dual-rail and concatenated inner-code constructions for every $\delta \le 0.130$.
- Every exact code in the capped sector, degenerate or not, obeys the explicit bound $K h_B(N,r) \le [z^{N-r}]F_B(z)^q$, which at $(2,1,0.1)$ yields an upper exponent of about $1.188$ against the achieved $0.281$.
- For the renormalized channel truncated to $\lfloor \delta N \rfloor$ losses, random equipartitions produce approximate codes with exponentially small worst-case error; for $\gamma \le 1/(B+1)$ their rates exceed the exact-code region and at $(2,1,0.1)$ stay positive up to $\delta \approx 0.158$.
- Exact samplers for the capped simplex and its histogram layers deliver bounded-occupation classical codeword distributions with positive rate and linear distance after expurgation, answering the sampling question raised in the cited literature; the polynomial-time guarantee is per sample rather than for listing the codebook.
Reading between the lines
- The gap between the achieved $0.281$ and the converse ceiling near $1.188$ at $(B,\alpha,\delta) = (2,1,0.1)$ suggests the cap-two sector can hold substantially more logical qubits than the current construction class reaches; the open problem the paper identifies, jointly optimizing the support-specific affine-hull denominator $r_t(C)+1$ in place of $m_B(q,t)$, is the most concrete route toward cl
- The Schur-concavity of capped deletion balls — smallest at the saturated center, largest at the balanced center — is a purely classical fact about bounded-composition multisets, and it could plausibly transfer to neighbouring constrained-coding settings such as run-length-limited or photon-parity-restricted codes.
- The existential Tverberg step is the main practical bottleneck: any constructive balancing procedure would turn these rates into implementable encoders and recovery maps, and the random-equipartition approximate codes are a partial step because their recovery channel is classical on the logical basis.
- The approximate-code gains appear exactly where the no-loss pattern is the most probable one at every support point; extending the error-intensity evaluation beyond $1/(B+1)$, which the paper leaves open, would likely push the improved region to larger loss probabilities.
Original abstract
We study Fock-state codes on $q$ bosonic modes, with total excitation $N$ and at most $B$ excitations per mode, that exactly correct every pattern of at most $\delta N$ photon losses, for $q$ proportional to $N$. For every fixed cap $B$ and every ratio $q/N>1/B$, we prove explicit achievable rates, positive for all sufficiently small $\delta>0$, in logical qubits per excitation. At $B=2$ and $q=N$, the rate is positive for every $\delta<0.1355$ and is about $0.281$ at $\delta=0.1$, compared with the previously established exact-code qutrit rate of about $0.275$. We also give an explicit upper bound on the dimension of every such exact code, including degenerate codes. For the pure-loss channel truncated to at most $\delta N$ losses and renormalized, we obtain approximate codes with exponentially small worst-case error. When the loss probability is positive and at most $1/(B+1)$, they improve our positive exact-code rate bounds; at $B=2$, $q=N$, and loss probability $0.1$, they remain positive up to $\delta\approx0.158$.
The compiled record
The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
The abstract, sentence by sentence
We study Fock-state codes on $q$ bosonic modes, with total excitation $N$ and at most $B$ excitations per mode, that exactly correct every pattern of at most $\delta N$ photon losses, for $q$ proportional to $N$. For every fixed cap $B$ and every ratio $q/N>1/B$, we prove explicit achievable rates, positive for all sufficiently small $\delta>0$, in logical qubits per excitation. At $B=2$ and $q=N$, the rate is positive for every $\delta<0.1355$ and is about $0.281$ at $\delta=0.1$, compared with the previously established exact-code qutrit rate of about $0.275$. We also give an explicit upper bound on the dimension of every such exact code, including degenerate codes. For the pure-loss channel truncated to at most $\delta N$ losses and renormalized, we obtain approximate codes with exponentially small worst-case error. When the loss probability is positive and at most $1/(B+1)$, they improve our positive exact-code rate bounds; at $B=2$, $q=N$, and loss probability $0.1$, they remain positive up to $\delta\approx0.158$.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | We study Fock-state codes on $q$ bosonic modes, with total excitation $N$ and at most $B$ excitations per mode, that exactly correct every pattern of at most $\delta N$ photon losses, for $q$ proportional to $N$. | contribution 0.98 | 0.91 | 0.35 | asserted 0.99 |
| s1 | For every fixed cap $B$ and every ratio $q/N>1/B$, we prove explicit achievable rates, positive for all sufficiently small $\delta>0$, in logical qubits per excitation. | result 0.93 | 0.94 | 0.47 | asserted 1.00 |
| s2 | At $B=2$ and $q=N$, the rate is positive for every $\delta<0.1355$ and is about $0.281$ at $\delta=0.1$, compared with the previously established exact-code qutrit rate of about $0.275$. | result 1.00 | 0.89 | 0.20 | asserted 1.00 |
| s3 | We also give an explicit upper bound on the dimension of every such exact code, including degenerate codes. | result 0.63 | 0.95 | 0.08 | asserted 1.00 |
| s4 | For the pure-loss channel truncated to at most $\delta N$ losses and renormalized, we obtain approximate codes with exponentially small worst-case error. | result 0.68 | 0.94 | 0.09 | asserted 1.00 |
| s5 | When the loss probability is positive and at most $1/(B+1)$, they improve our positive exact-code rate bounds; at $B=2$, $q=N$, and loss probability $0.1$, they remain positive up to $\delta\approx0.158$. | result 0.95 | 0.87 | 0.13 | asserted 0.98 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
References
Discussion
Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.