{"id":"e2dafa1a-c87f-4aeb-b142-f2b5bf42bc7a","arxiv_id":"2505.03290","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A 50.6%-efficient photonic quantum switch yields geometric-phase precision below the global Heisenberg limit without postselection, for n=29,30 displacement pairs.","lead":"A high-efficiency photonic quantum switch measures a geometric phase with precision claimed to beat the Heisenberg limit, even when all lost photons are counted. If the resource accounting is accepted, it is the first loss-inclusive demonstration of an indefinite-causal-order metrology advantage.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Resource-counting ambiguity is decisive: if each displacement operation is a process (as Ref. [45] describes '2N processes'), the loss-inclusive HL criterion at n=30 gives ~0.94 < 1, so the demonstrated violation disappears.","rationale":"Good-faith reading: the experiment is carefully executed, with quantified efficiency, visibility, multi-pair probability, and bootstrap error bars, and the measured RMSE appears to follow the predicted 1/(nu n^2 sqrt(m)) curve. The decisive issue is not whether the data fit the ICO curve but whether the benchmark line is the correct global HL. The reader's weakest assumption is exactly the resource-counting convention, and I agree. Because the factor of two enters the comparison at the largest achieved n, where the theoretical margin is about 3.7 under pair-counting versus about 0.94 under displacement-counting, the headline 'unconditional' claim is contingent on an undefended convention. This is addressable by a clear statement of the process unit and, if pair-counting is defended, a citation or derivation of the fixed-order bound for that unit; if not, larger n or stronger efficiency would be needed. The Fig. 3 fitting-function inconsistency and the unpropagated nuisance parameters are secondary but reinforce that the RMSE analysis is not yet fully closed. These are conditional corrections, not evidence of misconduct; hence the reader's CONDITIONAL verdict is appropriate and my stress-test does not move it.","tokens_in":9435,"tokens_out":24353,"duration_ms":260286,"concrete_test":"Recompute the violation criterion using N_proc = 2 * m_consumed * n with m_consumed = m(1+xi)/eta: evaluate R = eta^2 * nu^2 * n^2 / (4 * m * (1+xi)^2) at n=29 and n=30 with the published eta=0.506, nu=0.989, xi=0.0004, m=60. If R < 1 at both n, then the ideal precision curve itself does not cross the global HL under displacement-counting, so the claimed violation is unsupported; if R > 1, the counting objection is settled.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim compares measured RMSE with the global HL 1/(mn), treating one position-momentum displacement pair as a single process. But each photon physically traverses 2n independent displacement elements, and the quantum switch superposes the order of the aggregate position and aggregate momentum operations; the unknowns x_j and p_k are encoded one per element. The paper's own account of Ref. [45] says 'coherent control over 2N processes', i.e. the established ICO resource count treats each displacement as a process. Under that count, total processes per detected photon are 2n, and with the paper's loss/multi-pair accounting the HL bound becomes eta/[2mn(1+xi)]. The theoretical RMSE 1/(nu n^2 sqrt(m)) then beats HL only for n > 2 sqrt(m)(1+xi)/(eta nu), about 31. The experiment reaches n=30, so the ideal curve lies above the HL at every demonstrated point; the criterion of Eq. (6) changes from about 3.76 to about 0.94. The paper gives no argument that a pair, rather than its two constituent displacements, is the indivisible resource unit, and the achieved n is within one step of the threshold, so the claimed 3-sigma violation is not robust under this counting.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports an experimental implementation of a high-efficiency quantum switch for estimating the geometric phase A = x̄ p̄ associated with n position and n momentum displacements superposed in causal order. Using m = 60 detected photons per trial and up to n = 30 displacement pairs, the authors measure the control-qubit fringe P_-(n), estimate A by maximum likelihood, and compare the bootstrapped RMSE against a fixed-order Heisenberg limit that accounts for detection efficiency η = 0.506, visibility ν = 0.989, and multi-pair probability ξ = 0.04%. They report that for n = 29, 30 the measured RMSE lies below the loss-inclusive global Heisenberg limit by at least 3 standard deviations, claiming the first unconditional (postselection-free) violation of the Heisenberg limit using indefinite causal order.","tokens_in":9690,"tokens_out":13601,"duration_ms":134958,"significance":"If the claim holds, this would be a milestone: an experiment that includes all losses in the resource count and still exceeds the Heisenberg limit, with a clean noise model (Eqs. (4)-(6)) and bootstrapped statistics. The experimental effort is substantial: high-efficiency heralded source, optimized collection, and careful calibration of the displacement elements. The derivation of Eq. (6) from the noisy fringe model is straightforward, and the quoted parameters are plausible. However, the headline result is contingent on a resource-counting convention that is not defended against the standard convention of the underlying theory (Ref. [45]), and the demonstrated margin is not robust under the alternative count.","major_comments":[{"comment":"The central violation claim counts a position-momentum pair as one independent process, so the total resource is mn. The implementation, however, applies n position displacements D_xj and n momentum displacements D_pk to each photon, i.e., 2n independent displacement elements per photon. Under the alternative count of 2n processes per photon, which matches the paper's own description of Ref. [45] as 'coherent control over 2N processes', the loss-inclusive Heisenberg bound becomes η/[2mn(1+ξ)] and the criterion in Eq. (6) is replaced by η²ν²n²/[4m(1+ξ)²] > 1. Substituting the quoted η=0.506, ν=0.989, m=60, ξ=0.04% gives ≈0.94 at n=30, so the demonstrated violation disappears. The manuscript provides no argument for why the pair, rather than its two constituent displacements, is the indivisible resource unit; this is decisive because the achieved n=30 is just one step below the alternative threshold n≈31.","section":"Eq. (6) and Fig. 4"}],"minor_comments":[{"comment":"The word 'effciency' in the title is a typo and should be 'efficiency'.","section":"Title"},{"comment":"The resource count mn is introduced via the Mach-Zehnder phase-shift example, but in the quantum switch each photon traverses n position and n momentum displacement elements; the analogy should be justified or the count redefined explicitly.","section":"Introduction and Fig. 1"},{"comment":"The fitting function in Fig. 3 includes terms c n and φ0 inside the cosine, while Eq. (2) has only cos(n²A); the origin of these nuisance terms and their irrelevance to the scaling should be stated in the main text.","section":"Eq. (2) versus Fig. 3"},{"comment":"The theoretical curve labeled '1/(√ν N²)' does not match the stated RMSE formula δA = 1/(√(mν) n²) unless N and the y-axis normalization are defined; please clarify the definition of N and the plotted quantity.","section":"Fig. 4"},{"comment":"The main text refers to an 'orange dashed line' in Fig. 4, while the caption mentions a 'yellow dashed line'; the color labels should be made consistent.","section":"Fig. 4 colors"},{"comment":"Equation (6) is derived at the optimal operating point (cos n²A = 0), but the data points for n=25-30 are not at that point; since the headline demonstration relies on the direct RMSE comparison in Fig. 4, clarify the supporting (not decisive) role of Eq. (6).","section":"Eq. (6)"},{"comment":"The nuisance parameters c and φ0 are estimated from the same dataset used to estimate A; a brief description of how this is handled in the bootstrap would strengthen the statistical analysis.","section":"Statistical analysis"}],"recommendation":"major_revision","confidential_remarks":"The experimental work is careful and the data are likely reliable, but the headline claim is not robust under the resource-counting convention of Ref. [45]. I recommend that the editors require the authors either to provide a convincing argument for counting a displacement pair as one process, or to reframe the claim accordingly. If the latter, the paper would no longer support the title's assertion."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe experiment is genuinely well engineered, and the loss-inclusive analysis is a real step beyond the postselected version from Yin et al. But the central claim—unconditional violation of the global Heisenberg limit—does not hold up under the resource count that the theory itself uses. The paper quotes Zhao et al. as achieving super-Heisenberg scaling via coherent control over 2N processes. Here each photon traverses n position and n momentum displacements, so the total number of independent process applications is 2n. If you count that way, the Heisenberg bound with loss and multi-pair effects is η/[2mn(1+ξ)], and the measured RMSE beats it only for n > 2√m(1+ξ)/(η ν) ≈ 31. The experiment stops at n=30. The violation margin disappears. The paper instead counts n displacement pairs as the resource, giving mn, which makes the threshold ~15.5 and the demonstrated violation comfortable. No argument is given for why the pair, rather than the individual displacement, is the correct unit. This is a factor-of-two issue that is decisive exactly where the data are.\n\nWhat the paper does well: the source efficiency (71.5% coincidence, 50.6% end-to-end) is a substantial improvement, the visibility and multi-pair suppression are documented, and the measured RMSE data do follow the predicted 1/(ν n²√m) curve. If the resource-counting question were resolved in their favor, this would be a strong result.\n\nOther soft spots are minor by comparison: the Fig. 3 fitting function includes extra terms (c n and φ0) that are not derived from Eq. (2), and the uncertainty in these nuisance parameters is not propagated into the reported RMSE. Both are addressable.\n\nWho is this for: people working in indefinite causal order metrology and experimental quantum metrology. The engineering and the loss-inclusive framework are worth a careful look. But as a claim to have beaten the global Heisenberg limit, the paper needs to defend the resource accounting or amend the claim. A serious referee should see it—the question is important and the experiment is competently done—but acceptance should hinge on the counting issue.\n\nMy advice: send it to review, but expect the authors to either justify why a pair is a single process or soften the claim to a demonstration of super-Heisenberg scaling under a specific resource convention.","headline":"Strong experiment, but the resource-counting choice decides the result, and the natural count erases the violation.","tokens_in":10275,"tokens_out":6156,"would_cite":false,"duration_ms":60975,"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":"A quantum switch experiment beats the Heisenberg limit with photon losses counted.","keywords":["indefinite causal order","quantum switch","Heisenberg limit","quantum metrology","geometric phase","postselection-free","photon loss","super-Heisenberg scaling"],"falsifier":"Re-run the estimation with n=31 and m=60 under the same loss accounting and check whether the RMSE stays below the fixed-order bound 1/(mn); if it rises above the bound, the claimed scaling is not real. Alternatively, recompute the precision bound with each of the 2n displacement operations counted as an independent resource and test whether any n <= 30 still violates it.","tokens_in":9182,"feed_emoji":"⚛️","tokens_out":4788,"duration_ms":38444,"temperature":0.7,"pith_summary":"Estimating a geometric phase produced by conjugate position and momentum displacements, this experiment uses a photonic quantum switch that superposes the order of n displacement pairs with m=60 detected photons per trial. The paper argues that once all experimental inefficiencies are included—source coincidence efficiency near 71.5%, interferometric visibility above 0.989, and multi-pair emission—the measured root-mean-square error still falls below the fixed-order Heisenberg bound 1/(mn) for n=29 and n=30, by at least three standard deviations. If true, this is the first postselection-free, loss-inclusive violation of the Heisenberg limit using indefinite causal order. The established precision scaling, close to 1/($n^{2}$ $\\sqrt$(m)), exceeds what any definite causal order strategy could achieve with the same resources.","feed_headline":"Quantum switch beats the Heisenberg limit with losses counted","feed_subtitle":"A 60-photon geometric-phase measurement with indefinite causal order lands below the global Heisenberg bound at n=29,30.","key_machinery":"The load-bearing object is a polarization-encoded quantum switch whose control qubit |+> superimposes two orders of operation, D_{n x-bar} D_{n p-bar} and D_{n p-bar} D_{n x-bar}, acting on a transverse spatial mode. The conjugate displacements are the generators of position and momentum translations, D_x = $e^{{-i x P}}$ and D_p = $e^{{-i p X}}$; their commutator produces a geometric phase $n^{2}$ A in the interference signal. The argument is carried by the Fisher-information identity F_A = $n^{4}$ and by the loss-accounting criterion of Eq. (6), which converts raw photon counts into a threshold involving eta, nu, and xi. A high-efficiency heralded single-photon source and antireflection-coated switch keep these overhead factors small enough that the criterion is exceeded.","core_discovery":"The central claim is that indefinite causal order changes the resource scaling of phase estimation. With n pairs of displacements (position D_x and momentum D_p) placed in a quantum switch, the control-qubit interference probability is P_± = 1/2(1 +/- cos($n^{2}$ A)), giving Fisher information $n^{4}$ per photon, so the ideal precision is 1/($\\sqrt$(m) $n^{2}$). The paper shows that after including detection efficiency eta, visibility nu, and multi-pair emission probability xi, beating the Heisenberg limit 1/(mn) only requires $eta^{2}$ $nu^{2}$ $n^{2}$ / [m(1+xi)^2] > 1. With m=60, n=30, nu > 0.989, and a worst-case detection efficiency of 0.506, this quantity is about 3.76, and the observed RMSE lies below the Heisenberg bound by at least three standard deviations. The paper therefore claims an unconditional, postselection-free violation of the global Heisenberg limit.","pith_inferences":["The decisive resource-accounting convention is that a position-momentum displacement pair counts as one independent process. If instead each displacement is counted separately, the total resource doubles to 2mn and the paper's criterion requires n > 2 sqrt(m)(1+xi)/(eta nu) approximately 31, so the demonstrated violation at n=30 disappears; standardizing this convention would settle whether the ad","A direct extension would test the predicted 1/(n^2 sqrt(m)) scaling at larger n or different m, and check whether the 3-sigma violation persists when n exceeds the stricter threshold of 31 under the doubled resource count.","The same switch geometry could estimate products or sums of other conjugate displacement pairs, such as rotations about orthogonal axes, transferring the super-Heisenberg scaling to a wider class of parameters.","The loss-accounting method used here, including the xi correction for multi-pair emission, could be applied retroactively to earlier quantum-switch experiments to compare them on equal footing."],"forward_implications":["If the claim holds, fixed-order interferometric strategies using the same total resources mn cannot reach the precision reported here, so indefinite causal order provides a genuine metrological advantage under realistic loss.","The postselection loophole that affected earlier quantum-switch metrology demonstrations is closed: losses and detector inefficiencies are included in the resource accounting.","The experiment implies that the Heisenberg limit 1/(mn) is not the ultimate bound when causal order is indefinite, and that the relevant scaling becomes 1/(n^2 sqrt(m)) for this geometric phase.","The criterion eta^2 nu^2 n^2 / [m(1+xi)^2] > 1 gives a practical recipe for other laboratories to test ICO-enhanced metrology without requiring entangled multiphoton states.","For n > 25 the measured precision crosses below the Heisenberg bound, indicating the effect appears across a range of n, not as a single marginal point."],"supporting_citations":[{"why":"Supplies the theoretical result that coherent control over 2N processes in infinite dimensions gives N^-2 scaling and that fixed-order strategies are limited to Heisenberg scaling.","marker":"[45]"},{"why":"Previous photonic quantum-switch demonstration of super-Heisenberg geometric phase estimation under idealized, postselected conditions; the present experiment aims to remove that postselection.","marker":"[49]"},{"why":"Demonstration of unconditional, loss-included violation of the shot-noise limit, providing the template for full resource accounting used here.","marker":"[18]"},{"why":"Gives the SQL-violation criterion eta nu^2 n > 1 that the paper contrasts with its own HL-violation criterion.","marker":"[57]"},{"why":"Defines quantum Fisher information, the precision measure used to derive the n^4 scaling and the loss-corrected bound.","marker":"[17]"},{"why":"Bootstrap methods used to assign standard deviations to the measured RMSE values.","marker":"[61]"},{"why":"Analysis of heralding efficiency and correlated-mode coupling in SPDC sources, informing the efficiency optimization.","marker":"[58]"}],"fun_headline_variants":["Quantum switch beats Heisenberg limit, losses counted","Indefinite causal order breaks Heisenberg limit, losses included","Loss-inclusive quantum switch surpasses global Heisenberg bound","Beating Heisenberg limit with a 60-photon quantum switch","Quantum switch: Heisenberg limit violated with all losses accounted"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The violation rests on counting one position-momentum displacement pair as a single independent process; if each displacement counts as its own process, the same data no longer clear the Heisenberg-beating threshold at n=30.","fun_headline_variants_meta":{"raw":{"variants":["Quantum switch beats Heisenberg limit, losses counted","Indefinite causal order breaks Heisenberg limit, losses included","Loss-inclusive quantum switch surpasses global Heisenberg bound","Beating Heisenberg limit with a 60-photon quantum switch","Quantum switch: Heisenberg limit violated with all losses accounted"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000742,"raw_usage":{"total_tokens":3293,"prompt_tokens":913,"completion_tokens":2380,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":529,"completion_tokens_details":{"reasoning_tokens":2298}},"tokens_in":529,"tokens_out":2380,"duration_ms":17015,"temperature":1.0,"reasoning_tokens":2298,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:58:13.703073+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the estimation with n=31 and m=60 under the same loss accounting and check whether the RMSE stays below the fixed-order bound 1/(mn); if it rises above the bound, the claimed scaling is not real. Alternatively, recompute the precision bound with each of the 2n displacement operations counted as an independent resource and test whether any n <= 30 still violates it.","supporting_citations":[{"cited_title":"Demonstration of superior communication through thermodynamically free channels in an optical quantum switch","cited_arxiv_id":"2406.02236","evidence_quote":"Supplies the theoretical result that coherent control over 2N processes in infinite dimensions gives N^-2 scaling and that fixed-order strategies are limited to Heisenberg scaling."},{"cited_title":"Noisy quantum metrology with the assistance of indefinite causal order","cited_arxiv_id":"2104.06284","evidence_quote":"Previous photonic quantum-switch demonstration of super-Heisenberg geometric phase estimation under idealized, postselected conditions; the present experiment aims to remove that postselection."},{"cited_title":"Yang, Memory effects in quantum metrology, Physical review letters 123, 110501 (2019)","cited_arxiv_id":null,"evidence_quote":"Gives the SQL-violation criterion eta nu^2 n > 1 that the paper contrasts with its own HL-violation criterion."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines quantum Fisher information, the precision measure used to derive the n^4 scaling and the loss-corrected bound."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Analysis of heralding efficiency and correlated-mode coupling in SPDC sources, informing the efficiency optimization."}],"review_version":1}