{"id":"a0c093ad-72df-46bc-aa3a-2e9c1337153b","arxiv_id":"2507.17961","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"At the parametric oscillation threshold, an exceptional-point sensor with squeezing reaches a Cramér-Rao bound scaling as θ^{2n}, which is quartic (θ^4) for a second-order exceptional point.","lead":"Quantum sensors that use squeezed light and sensors that exploit exceptional points are usually studied separately. This paper derives a unified framework showing that when both effects are combined at the parametric oscillation threshold, the precision limit for estimating a weak perturbation can scale as the fourth power of the perturbation strength for a second-order exceptional point.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed θ^{2n} scaling is derived in a linearized model that diverges at the PO threshold, so the physical validity of the quartic precision is not established.","rationale":"The reader's weakest assumption already identified the same load-bearing concern: the linearized quantum-noise model is assumed valid at the PO threshold where the photon number diverges and the bandwidth narrows. My analysis confirms that this is the most critical point and that it directly undermines the physical interpretation of the headline claim. The paper itself acknowledges the divergence in Supplemental Sec. IV.C.a, but frames it as a resource-scaling issue (I_θ ∝ N, SQL-like) rather than as a breakdown of the linearization. That distinction matters: even if the in-model derivation of QFI ∼ θ^{−4n} is algebraically correct, the model's own predictions make it self-referential — the large QFI arises from a divergent operating point that cannot be reached in a real parametric oscillator. The reader's CONDITIONAL verdict is appropriate: the mathematical derivation may be internally sound, but the claim as stated in the abstract ('the sensing precision exhibits a unique quartic scaling') requires the nonlinear regularization to be addressed. I agree with the reader's identification and therefore do not change the verdict. A concrete saturation check would settle whether the scaling is robust or merely a linearization artifact.","tokens_in":22711,"tokens_out":5217,"duration_ms":59410,"concrete_test":"Extend the quantum Langevin equation (6) with a minimal gain-saturation term, e.g., replace ϵ_j a_j† with ϵ_j a_j† (1 − |a_j|²/n_sat) or add −(η_j/2)|a_j|² a_j, and compute the steady-state quantum Fisher information for the resulting nonlinear (non-Gaussian) system as θ→0. If δθ_CRB no longer follows θ^4 but flattens or crosses over at a scale θ* set by η_j/n_sat, then the quartic scaling is an artifact of the unregularized linear model. Concretely, use a truncated-Wigner or exact-diagonalization simulation for n_sat = 10², 10³, 10⁴, extract the exponent of δθ_CRB versus θ, and check whether the exponent remains 4 for θ values smaller than the saturation-induced cutoff; saturation of the exponent would falsify the physical claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that at the parametric-oscillation threshold and an exceptional point, the Cramér-Rao bound scales as δθ_CRB ∼ θ^{2n}, with the quartic θ^4 case for n=2. The derivation uses the quantum Langevin equations (Eqs. 6–7) and the Hamiltonian in Eq. (1), which assume an undepleted pump and a constant squeezing amplitude ϵ. However, the same linearized model predicts that the mean intracavity photon number diverges as θ^{−4n} (Supplemental Sec. IV.C.a) and that the bandwidth narrows as θ^2 (Supplemental Eq. S94). These divergences are the very mechanism behind the Green's function scaling G_θ ∼ θ^{−2n} in Eq. (10), which yields QFI ∼ θ^{−4n}. Thus the paper applies the linearized model exactly in the limit θ→0 where its own assumptions fail: gain saturation, pump depletion, and nonlinear terms that regularize the divergence are omitted. Because the claimed precision improvement is an asymptotic statement as θ→0, the result is not a validated prediction for a real parametric-oscillator sensor; it is a property of an unregularized linear system. This is not an internal inconsistency, but it is a correctness risk for the headline claim, which is presented as a physical sensing capability.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a unified quantum-noise framework for bosonic sensors that combine single-mode squeezing with non-Hermitian exceptional-point (EP) physics. The central claim is that when a sensor is operated simultaneously at the parametric-oscillation (PO) threshold and at an nth-order exceptional point, the Cramér-Rao bound for estimating a weak detuning perturbation θ scales as δθ_CRB ∼ θ^{2n}, with the second-order case giving a quartic scaling θ^4. The authors derive this scaling from a linearized quantum Langevin treatment, evaluate quantum and classical Fisher information for single-mode and coupled-mode configurations, discuss imperfect-threshold effects, and outline experimental implementations in photonic and circuit-QED platforms.","tokens_in":22885,"tokens_out":15339,"duration_ms":158303,"significance":"If the central result holds, it would identify a qualitatively new precision scaling in quantum metrology, going beyond both the linear scaling of passive sensors and the θ^n scaling previously reported for EP sensors at the lasing threshold. The paper is also valuable for proposing a concrete mechanism—the joint, non-additive interplay between squeezing-induced amplification and EP response—and for providing explicit parameter conditions and numerical QFI/CFI curves. The authors are honest about several limitations, including the divergence of the intracavity photon number and the narrowing bandwidth near the threshold. However, the analytic derivation contains a potentially load-bearing gap in the Green's function expansion, and the linearization validity in the θ→0 limit is not established; these issues must be resolved before the central scaling claim can be accepted.","major_comments":[{"comment":"Equation (S85) correctly states that (aθ^2 I_2 + M_EP)^{-1} = a^{-1}θ^{-2} I_2 - a^{-2}θ^{-4} M_EP, i.e. the leading term is θ^{-2}, not θ^{-4}. Substituting this into Eq. (S84) yields diagonal θ^{-2} contributions from the f1 blocks. Equation (S86) and the corresponding Eq. (10), however, retain only the -a^{-2}θ^{-4} M_EP terms and omit the θ^{-2} identity terms. Since P is stated to have constant nonzero leading-order elements, the θ^{-2} terms cannot be assumed to vanish without a proof. The claimed leading behavior G_θ ∼ θ^{-2n} and the resulting QFI ∼ θ^{-4n}, δθ ∼ θ^{2n} therefore do not follow from the expansion as written. The authors need to either correct the expansion or demonstrate explicitly that the θ^{-2} terms cancel in the matrix products defining dμ/dθ and V_θ.","section":"Main text, Eq. (10); Supplemental Sec. IV.A, Eqs. (S84)-(S86)"},{"comment":"The same linearized Green's function that produces the claimed precision scaling also predicts that the mean intracavity photon number diverges as θ^{-4n} and that the response bandwidth vanishes as θ^2. This means the undepleted-pump approximation, which treats ϵ_j as a fixed parameter in Eq. (1) and Eqs. (6)-(7), breaks down precisely in the asymptotic limit θ→0 where the scaling is claimed. Gain saturation, pump depletion, and other nonlinearities will regularize the divergence. The paper should state a quantitative validity condition (e.g., θ^2 must remain large compared with the saturation-induced linewidth) or extend the model to include the leading nonlinearity; without this, the result is a property of an unregularized linear system rather than a validated prediction for a physical parametric-oscillator sensor.","section":"Supplemental Sec. IV.C.a and IV.C.d (also Eqs. S89, S94)"},{"comment":"The Cramér-Rao bound in Eq. (3) counts N_m measurement rounds but does not include the duration of a single round. Because the bandwidth narrows as θ^2 (Eq. S94), the time required to reach steady state and perform a measurement grows as θ^{-2} per round. When this is accounted for, the precision per unit time scales less favorably than the per-round bound; at minimum, the authors should specify whether the quoted scaling is for a fixed integration time per round and justify why the diverging measurement time is not counted as a resource. This is especially important because the paper emphasizes that the scheme relies on linear steady-state response without state-preparation time.","section":"Main text, Eq. (3); Supplemental Eq. (S94)"}],"minor_comments":[{"comment":"In Eq. (S54), the coupling to the intrinsic loss channel of mode 1 is written as √γ_02 b_in1, but it should presumably be √γ_01 b_in1 to match the notation used elsewhere; please check.","section":"Supplemental Eq. (S54)"},{"comment":"The Green's function G_θ[ω] in Eq. (2) uses the Hamiltonian M, but M is not defined before this equation; defining M explicitly in the main text (rather than only in the Supplemental Material) would improve readability.","section":"Main text, Eq. (2)"},{"comment":"The caption lists the dashed, dotted, and dash-dotted lines as linear, quadratic, and quartic scalings, but the colors and line styles in the figure are not described in the body text; a brief identification of each curve in the caption would help the reader.","section":"Main text, Fig. 2 caption"},{"comment":"The statement that the θ^{2n} scaling has 'no analog in existing CQS protocols' is stronger than the supporting discussion, which compares only I ∝ N versus I ∝ N^2; a more careful comparison with the measurement-time-normalized precision of critical quantum sensing would be appropriate.","section":"Main text, Discussion"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses an important problem and contains promising numerical evidence, but the analytic derivation of the central scaling contains a concrete gap: Eq. (S86) drops the θ^{-2} leading term that follows from Eq. (S85). If the θ^{-2} terms do cancel in the QFI computation, the authors should show that explicitly; if they do not, the quartic claim is not supported. The linearization/saturation concern raised by the stress-test is real and should be addressed by a validity condition or a nonlinear treatment, though I would not by itself reject the paper if the scaling is restricted to the ideal linear model. The numerical curves in Fig. 3 may be correct, but the manuscript needs a corrected derivation before it is suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this paper for the θ^{2n} scaling: it is a genuine, clean result that answers a pointed open question, and the derivation is in-model coherent. The paper shows that at the parametric oscillation threshold and an nth-order EP, the Cramér-Rao bound scales as θ^{2n}, which is not in the cited literature—ref. [47] gives θ^n for EP sensors at the lasing threshold without squeezing, and critical quantum sensing refs give θ^2. The mechanism is also well explained: squeezing mixes the ±ω quadratures, so the EP and squeezing effects are non-separable, giving θ^{2n} rather than θ^{n+2}. I did not find a mathematical error in the Jordan-block expansion or the determinant conditions (S81)-(S82). The numerical CFI/QFI curves support the θ^4 case.\n\nThe soft spot is real and load-bearing for the physical claim. The linearized quantum Langevin model is used exactly at the PO threshold, where it predicts a diverging intracavity photon number (θ^{-4n}) and a narrowing bandwidth (θ^2). That means the asymptotic θ→0 result is a property of an unregularized linear system; in a real parametric oscillator, gain saturation, pump depletion, and nonlinearities will regularize the divergence. The paper is honest about the photon number scaling and mentions critical slowing down in the supplemental, but the abstract and main-text claim are stated without these caveats. A related issue: the CRB is per measurement round and does not include the growing measurement time from critical slowing down, so the practical precision advantage is weaker than the asymptotic scaling suggests. The imperfect-threshold analysis is good, however, and shows the scaling degrades gracefully.\n\nA minor point: the QFI covariance formula for noisy Gaussian states is quoted from the literature without derivation, and I have not verified every step of the 8x8 Hamiltonian construction. That is standard, but the referee should check it.\n\nWho is this for: theorists working on EP-enhanced sensing, critical quantum sensing, and quantum metrology in open systems. It deserves a serious referee. The scaling result is worth publishing even if the practical claims need tempering. Send it to peer review; the referee should push for a clear statement of the linearization validity at threshold and a proper accounting of measurement time. With those additions, the paper would be solid.","headline":"A genuinely new θ^{2n} precision scaling for squeezing-enhanced EP sensors, derived cleanly but with a real caveat about the linearized model at threshold.","tokens_in":23546,"tokens_out":2594,"would_cite":true,"duration_ms":26047,"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 sensor operated at both an exceptional point and the parametric oscillation threshold achieves a precision limit that scales as the fourth power of the perturbation strength.","keywords":["exceptional point","parametric oscillation threshold","squeezed light","quantum Fisher information","Cramér–Rao bound","non-Hermitian Hamiltonian","quantum sensing","precision scaling"],"falsifier":"Measure, for a two-mode sensor satisfying the PO-threshold/EP conditions (Supplemental Eqs. S81–S82), the output quadrature variances and the estimation error as a function of θ over at least two decades below the loss imbalance; the central claim is falsified if the empirical log-log slope of δθ vs θ is shallower than 4 (or than 2n for an nth-order EP), or if the anti-squeezed variance fails to diverge as $θ^{{-4n}}$, while the system remains in the linear regime.","tokens_in":22417,"feed_emoji":"🔬","tokens_out":8609,"duration_ms":84860,"temperature":0.7,"pith_summary":"Quantum sensing normally gains from two separate resources: squeezed light, which lowers noise in one quadrature, and exceptional points, non-Hermitian degeneracies where a small perturbation produces a large spectral response. This paper claims that combining both resources in one open bosonic system yields a precision that scales as the fourth power of the perturbation strength when the sensor is tuned to the parametric oscillation (PO) threshold. In concrete terms, for an nth-order exceptional point at the PO threshold, the Cramér–Rao bound behaves as δθ_CRB ∼ $θ^{{2n}}$, so a second-order EP gives $θ^{4}$. That is sharper than the θ^n scaling of EP sensors at the lasing threshold and than the $θ^{2}$ scaling of a single squeezed mode at the PO threshold, which suggests a new operating point for ultraweak-signal metrology.","feed_headline":"Squeezing plus an exceptional point gives quartic sensing precision","feed_subtitle":"A unified quantum-noise framework shows the Cramér–Rao bound improves as the fourth power of the perturbation.","key_machinery":"The central object is the non-Hermitian Hamiltonian matrix M in the quadrature basis, whose eigenvectors and eigenvalues encode the gain/loss balance and the squeezing. The argument hinges on two structural facts. First, at the parametric oscillation threshold the eigenvalue of the amplified mode is purely real and its response to a frequency perturbation θ is quadratic, λ ≈ -iγ/2 ± i|ϵ|(1 - $θ^{2}$/2|ϵ|^2), so the Green's function G_θ = -(ωI - M)^{-1}(I⊗Ω) picks up terms scaling as $θ^{{-2}}$. Second, when the modes are coupled to form an nth-order EP, the Hamiltonian is similar (via a P matrix) to a Jordan normal form with nilpotent blocks M_EP,n satisfying M_EP,n^n = 0. Expanding ($aθ^{2}$ I_n + M_EP,n)^{-1} using the Neumann series terminates at the (n-1)th term because of the nilpotency, leaving a leading contribution proportional to $θ^{{-2n}}$ M_EP,$n^{{n-1}}$. This $θ^{{-2n}}$ term enters the derivative dμ_θ/dθ and, through the quantum Fisher information, produces the δθ_CRB ∼ $θ^{{2n}}$ precision bound. The specific conditions for reaching the EP and PO threshold simultaneously, such as Eqs. (S81)–(S82) for the two-mode case, are what make the degenerate eigenvalues scale as $θ^{2}$ while the eigenstates merge.","core_discovery":"The central discovery is that an exceptional point (EP) and single-mode squeezing act jointly, not additively, to determine the ultimate sensing precision. The paper analyzes a generic chain of N coupled bosonic modes, each with degenerate parametric amplification, dissipation, and phase-insensitive gain, described by a non-Hermitian Hamiltonian in the quadrature basis. When the system is tuned so that it sits at an nth-order EP and simultaneously at the parametric oscillation threshold—where net loss balances the squeezing-induced amplification—the Green's function, and hence the response to a perturbation θ, acquires a leading contribution scaling as $θ^{{-2n}}$. Substituting this into the quantum Fisher information for Gaussian states gives I(θ) ∼ $θ^{{-4n}}$ and therefore a Cramér–Rao lower bound δθ_CRB ∼ $θ^{{2n}}$, i.e., quartic scaling for n=2. The mechanism is that squeezing creates a large coherent displacement along the anti-squeezed quadrature, the perturbation rotates the squeezing eigenbasis, and the Jordan-block structure of the EP (with M_EP^n = 0) causes the series for G_θ to terminate, leaving the enhanced $θ^{{-2n}}$ term. The same scaling is reached by homodyne and heterodyne detection, is invariant to external attenuation, and holds for higher-order EPs with appropriately engineered coupling.","pith_inferences":["A resource-corrected comparison that includes measurement time could convert the θ^{2n} precision into a constant advantage rather than a divergent one, because the bandwidth narrows as θ^2 near the threshold.","In any real device, gain saturation and pump depletion will set a floor on the achievable perturbation, so the θ^{2n} scaling is best tested at an intermediate perturbation range where the linearized model still holds but the photon number has not yet diverged unstably.","The mechanism suggests that the sensing advantage is tied to the rotation of the squeezing eigenbasis by the perturbation, pointing to a general design principle where the parameter of interest is mapped onto a rotation of the noise ellipse rather than onto its amplitude."],"forward_implications":["A sensor at a second-order EP and the PO threshold can in principle estimate a weak perturbation θ with an absolute uncertainty proportional to θ^4, improving without bound as θ → 0 within the linearized model.","Higher-order EPs extend the scaling to θ^{2n}, so an EP3 sensor at the PO threshold is predicted to reach δθ_CRB ∼ θ^6.","The θ^{2n} precision is attainable with standard homodyne or heterodyne detection and is robust to transmission-line loss and detector inefficiency, since the classical Fisher information matches the quantum Fisher information.","The sensor consumes resources only at the standard quantum limit (QFI ∝ N, sensitivity ∝ √N), requiring neither squeezed probe states nor dynamic state-preparation protocols, only steady-state linear response.","Slightly imperfect tuning (operating below the PO threshold) preserves the advantage as long as the loss imbalance is much smaller than the perturbation θ."],"supporting_citations":[{"why":"Establishes the Cramér–Rao scaling and Jordan-block description for EP sensors at the lasing threshold, the baseline this paper extends with squeezing.","marker":"[47]"},{"why":"Contains the full derivations of the Green's function expansion, the EP+PO threshold conditions, and the homodyne/heterodyne Fisher information that the main-text scalings rely on.","marker":"[50]"},{"why":"Supplies the general formula for the quantum Fisher information of multimode Gaussian states used to compute I(θ).","marker":"[51]"},{"why":"Provides the θ^n eigenspectral-splitting result for an nth-order EP that motivates and supports the projected sensitivity enhancement.","marker":"[30]"},{"why":"Introduces critical quantum sensing protocols near the parametric oscillation threshold, the operating point and comparison benchmark for the new θ^{2n} scaling.","marker":"[52]"},{"why":"Identifies the Petermann-factor noise penalty of gain-assisted EP sensors, the noise constraint the framework must account for when assessing precision.","marker":"[46]"}],"fun_headline_variants":["Quartic sensing boost from squeezing at an exceptional point","Squeezing meets exceptional point: quartic precision gain","Exceptional point plus squeezing: fourth-power sensing","Quartic scaling at exceptional point from squeezing","Squeezing and exceptional point give quartic sensitivity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The scaling assumes the linearized quantum-noise model stays valid exactly at the parametric oscillation threshold, where the mean photon number diverges as $θ^{{-4n}}$; real gain saturation and pump depletion will eventually break this assumption.","fun_headline_variants_meta":{"raw":{"variants":["Quartic sensing boost from squeezing at an exceptional point","Squeezing meets exceptional point: quartic precision gain","Exceptional point plus squeezing: fourth-power sensing","Quartic scaling at exceptional point from squeezing","Squeezing and exceptional point give quartic sensitivity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000413,"raw_usage":{"total_tokens":2112,"prompt_tokens":901,"completion_tokens":1211,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":517,"completion_tokens_details":{"reasoning_tokens":1137}},"tokens_in":517,"tokens_out":1211,"duration_ms":8532,"temperature":1.0,"reasoning_tokens":1137,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T14:40:36.301393+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure, for a two-mode sensor satisfying the PO-threshold/EP conditions (Supplemental Eqs. S81–S82), the output quadrature variances and the estimation error as a function of θ over at least two decades below the loss imbalance; the central claim is falsified if the empirical log-log slope of δθ vs θ is shallower than 4 (or than 2n for an nth-order EP), or if the anti-squeezed variance fails to diverge as $θ^{{-4n}}$, while the system remains in the linear regime.","supporting_citations":[{"cited_title":"Zhang, W","cited_arxiv_id":null,"evidence_quote":"Establishes the Cramér–Rao scaling and Jordan-block description for EP sensors at the lasing threshold, the baseline this paper extends with squeezing."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Contains the full derivations of the Green's function expansion, the EP+PO threshold conditions, and the homodyne/heterodyne Fisher information that the main-text scalings rely on."},{"cited_title":"Wiersig, Physical review letters112, 203901 (2014)","cited_arxiv_id":null,"evidence_quote":"Provides the θ^n eigenspectral-splitting result for an nth-order EP that motivates and supports the projected sensitivity enhancement."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces critical quantum sensing protocols near the parametric oscillation threshold, the operating point and comparison benchmark for the new θ^{2n} scaling."},{"cited_title":"Wang, Y.-H","cited_arxiv_id":null,"evidence_quote":"Identifies the Petermann-factor noise penalty of gain-assisted EP sensors, the noise constraint the framework must account for when assessing precision."}],"review_version":1}