{"id":"8aa3187c-8185-47cb-9671-1f54a442ac9b","arxiv_id":"2505.24770","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Under photon loss, the phase-estimation quantum Fisher information of entangled coherent states differs between interferometers with and without a reference beam, and the no-reference configuration matches NOON-state scaling at high photon numbers.","lead":"This paper calculates the ultimate phase-measurement precision of entangled coherent states in a lossy optical interferometer, comparing setups that use or omit a reference beam. It finds that photon loss breaks the usual equivalence between the two setups, and that dropping the reference beam can help in an intermediate photon-number range.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation (7) incorrectly assumes QFI additivity over input NOON components; after loss their supports overlap, so Eq. (8) overestimates the no-reference QFI and the claimed NOON-like advantage reverses.","rationale":"The paper's central no-reference result, Eq. (8), is derived by an invalid additivity step. The phase-averaged ECS in Eq. (6) is a direct sum over NOON components before loss, and Eq. (7) applies QFI additivity as though the lossy components remained orthogonal. They do not: e.g., the lossy n=1 and n=2 NOON states both have support on |1,0⟩ and |0,1⟩. A correct block-diagonal treatment by total photon number after loss yields F_rho^exact = 2N^2 e^{-2(1-η)|α|^2}(η^2|α|^4+η|α|^2). This changes the central comparison with the with-reference result: at large N the reference-free ECS decays as e^{-2N(1-η)}N^2, which is worse, not comparable, to the NOON scaling e^{-N(1-η)}N^2. The claimed disappearance of ECS inferiority relative to NOON in the reference-free case is therefore reversed. The reader's conditional verdict focused on presentation and the phase-averaging model, but the more load-bearing flaw is the internal additivity error. Since Eqs. (7)–(8) and the associated qualitative conclusions are central to the paper, the verdict should move to REJECT.","tokens_in":9421,"tokens_out":33789,"duration_ms":377616,"concrete_test":"Numerically diagonalize the lossy phase-averaged ECS: truncate the Fock basis at n_max = |α|^2 + 10, apply amplitude-damping channels with transmittance η to each mode, then compute the QFI for G = (n1−n2)/2 directly from the spectral decomposition. Compare with Eq. (8). Equivalently, evaluate the block formula F = Σ_K C_K^2 K^2 / A_K. For |α|^2 = 5, η = 0.9, the exact value is about 9.04 whereas Eq. (8) gives about 14.91, a ratio e^{-0.5}; at |α|^2 = 10, η = 0.9, the ratio should be e^{-1}, confirming the missing factor.","verdict_should_be":"REJECT","load_bearing_attack":"The no-reference QFI in Eq. (8) rests on Eq. (7), which sums QFIs of individual lossy NOON components with weights p_n. This additivity is valid only if the phase-averaged ECS remains a direct sum over n after loss. It does not. The loss channel maps different NOON components into overlapping supports within the same total-photon-number sectors. Writing p_n = 2N^2 e^{-|α|^2}|α|^{2n}/n!, and A_K = Σ_{n≥K} p_n B(n,K) with binomial B(n,K), C_K = p_K η^K, the lossy state in each sector K is σ_K = (1/2)[A_K(|K,0⟩⟨K,0|+|0,K⟩⟨0,K|) + C_K(|K,0⟩⟨0,K|+|0,K⟩⟨K,0|)]. The QFI of that block is A_K(C_K/A_K)^2 K^2. Summing over K gives F_rho^exact = 2N^2 e^{-2(1-η)|α|^2}(η^2|α|^4+η|α|^2), not Eq. (8). Eq. (8) is larger by a factor e^{(1-η)|α|^2}. Consequently the corrected reference-free ECS scaling is e^{-2N(1-η)}N^2, which is worse than the NOON scaling e^{-N(1-η)}N^2, so the abstract's claim that the inferiority of ECSs relative to NOON states disappears without a reference beam is not supported.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes phase estimation in a lossy two-mode interferometer using entangled coherent states (ECSs), contrasting the scenario with a common reference beam against the reference-free scenario modeled by phase averaging the probe. The authors claim two main results: (1) the lossless metrological equivalence between the two scenarios for the two-phase-shifting configuration breaks down under photon loss, and (2) the known inferiority of ECSs relative to NOON states in the presence of a reference beam disappears when no reference beam is used. The central technical content is the evaluation of the quantum Fisher information (QFI), with Eq. (8) for the reference-free case and Eq. (21) for the reference case, followed by a comparison with NOON states.","tokens_in":9713,"tokens_out":36546,"duration_ms":308981,"significance":"The question addressed is timely and relevant: practical phase-estimation schemes often lack a stable reference beam, and the behavior of non-classical states under loss in that setting is of genuine interest. The paper uses standard QFI methods and correctly identifies that the loss channel is phase-covariant, which is a useful structural observation. However, the main advertised conclusion is overturned by a specific error in the reference-free QFI derivation. The corrected calculation shows that, at large photon numbers, the reference-free ECS performs much worse than NOON states, not comparably to them. Since the second key result in the abstract is false, the paper in its current form cannot be accepted.","major_comments":[{"comment":"The step from Eq. (6) to Eq. (7) is not valid for a lossy interferometer. Equation (6) is the phase-averaged probe before loss; after the loss channel, the supports of different NOON components overlap, so the QFI is not the weighted sum of the single-NOON QFIs n^2 eta^n. Because the loss channel is covariant under total phase shifts, the correct reference-free state is the phase average of the lossy ECS of Eq. (10), which is a direct sum over n of two-level blocks with diagonal weight a_n = N^2 exp(-eta|alpha|^2)(eta|alpha|^2)^n/n! and coherence b_n = a_n exp(-(1-eta)|alpha|^2). The QFI per block is 2 n^2 b_n^2/a_n, yielding F_rho^correct = 2N^2 exp(-2(1-eta)|alpha|^2)(eta^2|alpha|^4 + eta|alpha|^2). This is smaller than Eq. (8) by the factor exp(-(1-eta)|alpha|^2). Consequently, for mean photon number Nbar ~ |alpha|^2, the reference-free ECS QFI scales as exp(-2(1-eta)Nbar) Nbar^2, not as the NOON-like exp(-(1-eta)Nbar) Nbar^2 claimed in the paper. This error is load-bearing because Eq. (8) is the basis for the paper's second main conclusion.","section":"Sec. III A, Eqs. (6)-(8)"},{"comment":"The comparison of the reference-free ECS with NOON states and with the reference-beam ECS is based on the incorrect Eq. (8). With the corrected F_rho, the reference-free ECS at large Nbar decays exponentially faster than NOON states and is also exponentially smaller than the reference-beam ECS, whose QFI approaches the shot-noise floor eta Nbar according to Eq. (21). Therefore the statements that for N > N1 the reference-free ECS 'resembles NOON states' and that 'omitting the reference beam can be advantageous' are not supported. The abstract's claim that the inferiority of ECSs relative to NOON states disappears without a reference beam is reversed by the corrected calculation: the inferiority becomes much more pronounced in the reference-free case.","section":"Sec. III C, Fig. 2 and Eqs. (22)-(24)"}],"minor_comments":[{"comment":"The expansion of the ECS as a superposition of NOON states should use the coefficients c_n, not |c_n|^2; as written, |ECS> = sqrt(2) N sum_n |c_n|^2 |n::0> is dimensionally and algebraically incorrect.","section":"Eq. (5)"},{"comment":"For a two-dimensional density matrix with trace one, the eigenvalues are gamma_+- = (1 +- sqrt(1 - 4 det sigma))/2. The manuscript writes sqrt(1 - det sigma) in Eq. (13) and Eq. (B5), which is inconsistent with the expression for zeta_+- in the main text and with the spectral decomposition.","section":"Eq. (13) and Appendix B"},{"comment":"The full expression for F_sigma is never displayed; only the p -> 0 limit is given in Eq. (21). Please provide the exact closed form or a reproducible derivation in the appendix, so that the limit and the subsequent comparison can be checked.","section":"Sec. III B"},{"comment":"The symbol N is used both for the normalization constant of the ECS and for the mean photon number, leading to confusing expressions such as 'with N = N for ECSs' in the caption of Fig. 2. Please distinguish the two, e.g. with calligraphic N for the normalization and \\bar N for the mean photon number.","section":"Notation throughout"}],"recommendation":"reject","confidential_remarks":"The central advertised result of the paper is refuted by the corrected calculation of the reference-free QFI: the no-reference ECS is exponentially worse than NOON states at large photon numbers, not comparable to them. This is not a local fixable issue; it changes the main message of the paper. A resubmission could be considered if the authors reframe the work as a negative result on reference-free ECS interferometry and carefully redo the comparison, but the present manuscript should not be published as is."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper has a real question and one correct half, but the headline no-reference result is wrong, and the error is load-bearing. I re-derived the no-reference QFI after loss and the stress-test note is correct. Eq. (7) assumes QFI additivity across the input NOON components, but the loss channel maps different n components into overlapping total-photon-number sectors. For each sector K the block has diagonal population A_K = Σ_{n≥K} p_n C(n,K)η^K(1-η)^{n-K} and off-diagonal C_K = p_K η^K, so the QFI is Σ_K K^2 C_K^2/A_K, which evaluates to 2N^2 e^{-2(1-η)|α|^2}(η^2|α|^4 + η|α|^2), not Eq. (8). The missing factor e^{-(1-η)|α|^2} is exponentially large, so the paper's high-N 'NOON-like' scaling is wrong: the corrected reference-free QFI decays as e^{-2N(1-η)} N^2, while NOON is e^{-N(1-η)} N^2. Hence the abstract's claim that ECS inferiority disappears without a reference is reversed. The with-reference calculation (Eq. 21) looks right, and the lossless-equivalence discussion is a fair summary of prior work, so there are salvageable parts if the authors recompute the no-reference case and change the conclusions. There are also smaller presentation problems: the full F_sigma expression is not displayed, Eq. (13) has typos, and the new results are not cleanly separated from Ref. [16]. I would not take the current version as a reliable guide to reference-free ECS metrology; if it crossed our desk I would send it back for major revision rather than desk-reject, because the error is subtle and the corrected calculation is worth having. The likely audience is quantum metrology researchers working on reference-free limits; for them the corrected negative result is still useful, but only after a rewrite.","headline":"The no-reference QFI in Eq. (8) is missing a factor e^{-(1-η)|α|^2}; the corrected scaling kills the paper's main claim.","tokens_in":10293,"tokens_out":22611,"would_cite":false,"duration_ms":225848,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Photon loss breaks the metrological equivalence between interferometers with and without a reference beam, and the paper derives the quantum Fisher information formulas showing how the two settings diverge.","keywords":["quantum Fisher information","phase estimation","entangled coherent states","photon loss","optical interferometry","reference beam","NOON states","quantum metrology"],"falsifier":"Send the entangled coherent state $|\\alpha,0\\rangle+|0,\\alpha\\rangle$ through a Mach-Zehnder with per-path transmittance $\\eta\\approx0.9$, and measure the optimal phase variance both with and without a reference beam at mean photon numbers around $\\bar N=20,50,100$. The paper predicts that at $\\bar N=100$ the no-reference QFI is about $e^{-10}\\cdot10^4\\approx0.45$, giving $\\delta\\phi\\approx1.5$, while the reference-assisted QFI stays near $\\eta\\bar N=90$, giving $\\delta\\phi\\approx0.1$; observing a reference-free sensitivity near the shot-noise floor at such photon numbers, or an exponential decay in the reference-assisted case, would contradict the claim.","tokens_in":9176,"feed_emoji":"⚛️","tokens_out":19642,"duration_ms":170757,"temperature":0.7,"pith_summary":"This paper asks whether an external reference beam matters for the ultimate phase sensitivity of a lossy optical interferometer when the probe is an entangled coherent state (ECS). In the ideal lossless case the two settings are metrologically equivalent for the two-phase-shifting configuration; the paper derives the quantum Fisher information with photon loss and shows this equivalence fails. Without a reference beam, the probe must be phase-averaged into a mixture of NOON components, giving a Fisher information that contains a Heisenberg-scaling term decaying as $e^{-|\\alpha|^2(1-\\eta)}$ and, at high photon number, tracking the exponential loss degradation of NOON states. With a reference beam, the coherent superposition survives partial loss and the Fisher information retains a shot-noise term, so the two scenarios diverge as loss grows. The claimed practical consequence is that the known disadvantage of ECSs relative to NOON states in reference-assisted measurements disappears when no reference beam is used.","feed_headline":"Photon loss breaks reference-beam equivalence in interferometry","feed_subtitle":"Without a reference beam, entangled coherent states decay like NOON states under photon loss.","key_machinery":"The central object is the phase-averaged ECS, $\\varrho_{\\mathrm{ECS}}=2\\mathcal{N}^2\\bigoplus_n |c_n|^2 |n::0\\rangle\\langle n::0|$, an incoherent direct sum of weighted NOON states; because the QFI is additive on this block-diagonal mixture, the no-reference calculation reduces to summing the per-NOON lossy Fisher information $n^2\\eta^n$. For the reference-assisted case the machinery is the spectral decomposition of the lossy two-component coherent state: after loss the two components $|\\alpha\\sqrt{\\eta},0\\rangle$ and $|0,\\alpha\\sqrt{\\eta}\\rangle$ are non-orthogonal with overlap $p=e^{-\\eta|\\alpha|^2}$, and Gram-Schmidt orthogonalization plus the formula $F=4(\\lambda_+\\Delta^2G_++\\lambda_-\\Delta^2G_--4\\lambda_+\\lambda_-|G_{+-}|^2)$ for the generator $G=(a_1^\\dagger a_1-a_2^\\dagger a_2)/2$ yields the compact asymptotic expression Eq. (21). The NOON-state comparison uses $F_{\\mathrm{NOON}}=\\eta^{\\bar N}\\bar N^2$ at equal mean photon number.","core_discovery":"For the two-phase-shifting operation $U^T_\\phi=\\exp[-i\\phi(a_1^\\dagger a_1-a_2^\\dagger a_2)/2]$, the paper computes the quantum Fisher information (QFI) of an ECS probe $|\\mathrm{ECS}\\rangle=\\mathcal{N}(|\\alpha\\rangle|0\\rangle+|0\\rangle|\\alpha\\rangle)$ in a lossy interferometer with equal per-path transmittance $\\eta$. With no reference beam, phase-averaging makes the probe the mixed state $\\varrho_{\\mathrm{ECS}}=2\\mathcal{N}^2\\bigoplus_n |c_n|^2 |n::0\\rangle\\langle n::0|$, and additivity of the QFI gives $F_\\varrho=2\\mathcal{N}^2 e^{-|\\alpha|^2(1-\\eta)}(|\\alpha|^4\\eta^2+|\\alpha|^2\\eta)$. With a reference beam, diagonalizing the two-component lossy state and applying the mixed-state QFI formula gives, in the limit $\\eta|\\alpha|^2\\gg1$, $F_\\sigma=2\\mathcal{N}^2(e^{-2|\\alpha|^2(1-\\eta)}|\\alpha|^4\\eta^2+|\\alpha|^2\\eta)$. At $\\eta=1$ both reduce to the same lossless expression, reproducing the earlier equivalence; for $\\eta<1$ the no-reference Heisenberg term is damped by one exponential factor and the reference-assisted term by two, and at large mean photon number the no-reference QFI scales like $\\exp[-\\bar N(1-\\eta)]\\bar N^2$, the NOON-state scaling, while the reference-assisted QFI approaches the shot-noise form $\\eta\\bar N$. The paper therefore concludes that photon loss breaks the equivalence and reverses the ECS-versus-NOON performance comparison depending on the reference.","pith_inferences":["Applied to other coherent-superposition probes, the phase-averaging argument suggests that loss will erase coherence between photon-number sectors and push their QFI toward the corresponding number-state mixture scaling, a consequence the paper does not develop.","Equation (21) is derived in the large-$\\eta|\\alpha|^2$ limit; the paper's exact two-state diagonalization could be pushed to finite sizes, which would sharpen the crossing points $N_1$ and $N_2$ in the comparison with NOON states.","Since Eq. (3) is a uniform prior over the global phase, a reference-free experiment with a peaked prior or adaptive phase tracking would be expected to interpolate between the no-reference and reference-assisted results.","A direct photon-counting or parity measurement of the output would realize the reference-free scenario by construction, making the predicted NOON-like decay of Eq. (8) experimentally testable without needing an external phase reference."],"forward_implications":["At any nonzero photon loss, the two-phase-shifting interferometer has different ultimate sensitivity with and without a reference beam; the lossless equivalence is restored only at $\\eta=1$.","At high mean photon number, reference-free ECS probes follow the NOON-state exponential loss scaling $\\exp[-\\bar N(1-\\eta)]\\bar N^2$, so the Heisenberg-like scaling term is effectively killed by loss.","Reference-assisted ECS probes keep a shot-noise-limited Fisher information $\\eta\\bar N$ at high photon number, so the presence of a reference beam becomes decisive for lossy metrology.","Because reference-free ECSs track NOON-state scaling, the intermediate-photon-number advantage of NOON states over reference-assisted ECSs disappears when the reference beam is omitted.","Photon-number-resolving measurements, which need no shared reference, should be benchmarked with Eq. (8) rather than with QFIs computed under the assumption of a reference beam."],"supporting_citations":[{"why":"Establish the lossless reference/reference-free equivalence for the two-phase-shifting configuration that the paper re-examines under loss.","marker":"[15]"},{"why":"Give the phase-averaging description of a reference-free probe used in Eq. (3).","marker":"[32]"},{"why":"Supply the superselection and phase-reference rationale for why reference-free states lose inter-mode coherence.","marker":"[37]"},{"why":"Introduce the virtual-beam-splitter loss model with transmittance used throughout.","marker":"[12]"},{"why":"Provide the small-NOON lossy QFI and single-arm-shift ECS baselines.","marker":"[13]"},{"why":"Provide the lossy NOON-state QFI used for the performance comparison.","marker":"[31]"},{"why":"Give the two-component mixed-state diagonalization and QFI formula used to derive Eq. (21).","marker":"[14]"},{"why":"Justify commuting photon loss with phase shifting, simplifying the reference-assisted QFI calculation.","marker":"[38]"},{"why":"Support the unitary-invariance and additivity properties of the QFI used for the phase-averaged mixture.","marker":"[36]"}],"fun_headline_variants":["No reference beam: ECS matches NOON loss scaling","Loss breaks reference equivalence, flips state ranking","Without reference, ECS and NOON scale alike under loss","Reference-less interferometry removes ECS disadvantage"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the absence of a reference beam is equivalent to uniformly averaging the probe over all global phases, as in Eq. (3); if a real reference-free experiment implements the missing phase reference differently, the computed Fisher information and the comparison to the reference-assisted case could change.","fun_headline_variants_meta":{"raw":{"variants":["No reference beam: ECS matches NOON loss scaling","Loss breaks reference equivalence, flips state ranking","Without reference, ECS and NOON scale alike under loss","Reference-less interferometry removes ECS disadvantage"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000379,"raw_usage":{"total_tokens":2046,"prompt_tokens":1008,"completion_tokens":1038,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":624,"completion_tokens_details":{"reasoning_tokens":975}},"tokens_in":624,"tokens_out":1038,"duration_ms":11930,"temperature":1.0,"reasoning_tokens":975,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:28:20.093914+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Send the entangled coherent state $|\\alpha,0\\rangle+|0,\\alpha\\rangle$ through a Mach-Zehnder with per-path transmittance $\\eta\\approx0.9$, and measure the optimal phase variance both with and without a reference beam at mean photon numbers around $\\bar N=20,50,100$. The paper predicts that at $\\bar N=100$ the no-reference QFI is about $e^{-10}\\cdot10^4\\approx0.45$, giving $\\delta\\phi\\approx1.5$, while the reference-assisted QFI stays near $\\eta\\bar N=90$, giving $\\delta\\phi\\approx0.1$; observing a reference-free sensitivity near the shot-noise floor at such photon numbers, or an exponential decay in the reference-assisted case, would contradict the claim.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Give the phase-averaging description of a reference-free probe used in Eq. (3)."},{"cited_title":"Zhong, L","cited_arxiv_id":null,"evidence_quote":"Supply the superselection and phase-reference rationale for why reference-free states lose inter-mode coherence."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduce the virtual-beam-splitter loss model with transmittance used throughout."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provide the small-NOON lossy QFI and single-arm-shift ECS baselines."},{"cited_title":"Chiruvelli and H","cited_arxiv_id":null,"evidence_quote":"Provide the lossy NOON-state QFI used for the performance comparison."},{"cited_title":"Jarzyna and R","cited_arxiv_id":null,"evidence_quote":"Justify commuting photon loss with phase shifting, simplifying the reference-assisted QFI calculation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Support the unitary-invariance and additivity properties of the QFI used for the phase-averaged mixture."}],"review_version":1}