{"id":"1dbfae12-0165-4ecc-b96f-1f43aaf32dd9","arxiv_id":"1908.03786","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"An approximate factorization method recovers the support of an unknown acoustic scatterer from phaseless total-field data, using a large-radius asymptotics that makes a phaseless data operator converge to the classical far-field operator.","lead":"A new numerical method reconstructs the shape and location of an unknown object from intensity-only sound wave measurements at a single frequency, using only the magnitude of the total field on a large circle. It is the first factorization-type algorithm for phaseless inverse scattering, offering a fast, non-iterative alternative that does not need to know whether the object is sound-soft, sound-hard, impedance, or penetrable.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The inversion guarantee rests on an unproved spectral-inheritance step: operator-norm convergence of (N~_{R,M})# to a scalar multiple of F# (Remark 4.8, eq.","rationale":"I read the paper in good faith and find the asymptotic core sound: the oscillatory-integral estimates in Section 3 lead to the stated operator-norm relations, and the range identity for F# in Theorem 4.7 is a standard factorization-method application. The numerical experiments provide qualitative support across obstacle types and noise levels. However, the central inversion claim depends on the unproved step that the eigensystem of (N~_{R,M})# inherits the range characterization of F#. This is not an external consensus disagreement; it is an internal gap that the authors themselves disclose in Remark 5.3. The reader identified exactly this assumption, and I agree with the CONDITIONAL verdict. My proposed test would at least show whether the heuristic is numerically reliable in a semi-analytically tractable case; it would not, by itself, supply the missing proof, but it would settle whether the concern manifests in practice. Since the reader's verdict already reflects this conditional status, no verdict change is needed.","tokens_in":25976,"tokens_out":9333,"duration_ms":115144,"concrete_test":"Use a sound-soft disk, where the far-field operator F# and its eigenpairs can be computed semi-analytically via Mie/Hankel series, and compute the corresponding (N~_{R,M})# from synthetic |u| data for the same disk. For fixed k and M, increase R (e.g., 10, 20, 40, 80) and compute, on a fixed sampling grid, both W(z) from F# and W_M^PW(z) from phaseless data; record the contrast between a fixed interior point and a fixed exterior point and the Hausdorff error of the extracted boundary. If the contrast and boundary error do not improve monotonically with R at fixed discretization, the spectral-inheritance heuristic fails in the simplest resolvable case. If they do improve, the gap remains a missing proof rather than a demonstrated failure.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's proved results stop at operator-norm approximation: Theorem 4.2 and Remark 4.8 give ||(N~_R)# - (1/sqrt{8kπR})F#|| ≤ C R^{-α} for α∈(1/2,1), and Theorem 5.2 with Remark 5.3 extends this to the truncated (N~_{R,M})# actually used in computation. The inversion claim requires more: W_M^PW(z), defined from the eigensystem of (N~_{R,M})#, must inherit the range test z∈D iff B*_{1/2}φ_z ∈ R(F#^{1/2}) proved in Theorem 4.7. Operator-norm convergence of compact operators does not, by itself, control the weighted eigenseries sum |(φ,ψ_j)|^2/λ_j that defines the indicator, because the λ_j tend to zero and the ψ_j are stable only when spectral gaps are controlled. Small norm perturbations can mix eigenvectors of nearby eigenvalues, and the tail of the series—precisely the part that distinguishes z∈D from z∉D through divergence of the Picard sum—is not controlled by the O(R^{-α}) bound. The authors state this explicitly in Remark 5.3: 'we are currently not able to give a rigorous theoretical analysis on the property of the indicator function W_M^PW(z)'. Thus the central reconstruction guarantee is a heuristic, supported by qualitative numerical examples, not by the theorem stated. This is the load-bearing gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper considers the inverse acoustic scattering problem at fixed frequency with phaseless total-field data |u(x,d)| on a circle ∂B_R surrounding the unknown scatterer. It introduces the phaseless total-field operator N_R^PW in (3.1) and proves, in Theorem 3.7, that N_R^PW minus (e^{iπ/4} e^{ikR}/√(8kπR)) times the far-field operator F is O(1/R) in the H^{1/2}(S^1)→H^{-1/2}(S^1) operator norm. After multiplying by the Sobolev weight operators B_{1/2}, B*_{1/2}, it obtains the L2→L2 estimate in Theorem 4.2 for the modified operators. The paper then invokes the factorization results of [31,32] to characterize D via the range of (tilde F#)^{1/2} (Theorem 4.7) and defines the indicator W_M^PW from the eigensystem of the positive part (tilde N_{R,M})# of the truncated modified phaseless operator. The algorithm is tested numerically for sound-soft, sound-hard, impedance, and penetrable media with up to 20% noise.","tokens_in":26268,"tokens_out":6356,"duration_ms":73186,"significance":"If the inversion claim were fully established, this would be a useful first factorization-type method for phaseless data, adding to the small set of non-iterative phaseless reconstruction algorithms. The rigorous asymptotic analysis of N_R^PW—the oscillatory-integral estimates in Lemmas 3.2–3.4 and the interpolation argument in Theorem 3.7—is a solid contribution, as is the combination of these estimates with the independently established factorization and range-identity results of [31,32] to justify the approximate factorization structure. The numerical study is reasonably broad (five obstacle types, noise levels, varying R). The central limitation is that the step from operator-norm approximation to the eigensystem-based indicator is explicitly unproved (Remark 5.3), so the reconstruction guarantee currently rests on a heuristic; this limits the significance to a conditional result.","major_comments":[{"comment":"The sentence 'we are currently not able to give a rigorous theoretical analysis on the property of the indicator function W_M^PW(z)' identifies the exact load-bearing step of the paper. The exact range characterization (4.11)–(4.12) is proved only for the modified far-field operator tilde F#, while the algorithm replaces tilde F# by (tilde N_{R,M})# without a theorem showing that the weighted eigenseries in (5.8) inherits the divergence property of W(z). Since the abstract and Section 1 claim that the method 'reconstructs' the scatterer, this missing justification is a central gap; the numerical examples illustrate but do not prove the claimed inheritance.","section":"§5, Remark 5.3"},{"comment":"Operator-norm convergence in (4.13) and (5.2) does not control the indicator (5.8). The summands |(B*_{1/2,M}φ_z, ψ_j)|^2/λ_j involve reciprocals of eigenvalues λ_j that tend to 0 for a compact operator, and the eigenvectors of a compact operator are not stable under small norm perturbations unless spectral gaps are controlled. The tail of the series is precisely what distinguishes z∈D from z∉D through divergence of W(z), and the O(R^{-α}) bound gives no estimate for that tail. A rigorous or at least a regularized version of the indicator, for example with an eigenvalue cutoff depending on R, would be needed to make the reconstruction claim load-bearing.","section":"§5, Eq. (5.8) and Remark 5.3"}],"minor_comments":[{"comment":"The proof of Lemma 3.1 is only a reference to [17] and well-posedness; for reproducibility, please spell out why the C^1 bound (3.4) and the decay (3.5) hold uniformly in d for both obstacle and medium scattering.","section":"§3, Lemma 3.1"},{"comment":"The phase factor e^{-i(kR+π/4)} in the definition of tilde N_R^PW is introduced without explanation; a brief remark that it cancels the phase appearing in Theorem 3.7 would improve readability.","section":"§4, Eq. (4.3)"},{"comment":"Step (5) instructs the user to locate sampling points where W takes 'a large value' but gives no threshold or quantitative stopping rule; a reproducible criterion or a statement that the indicator is used only relatively would help.","section":"§5, Algorithm 5.1"},{"comment":"The parametrization table lists centers (c1,c2) symbolically, but the actual center values are not given for Examples 1, 2, 3, and 5; please include them for reproducibility.","section":"§6, Table 6.1"}],"recommendation":"major_revision","confidential_remarks":"The core difficulty is the unproved spectral-inheritance step in Remark 5.3, and the authors are honest about it. I do not see circularity: the factorization of the far-field operator is taken from [31,32], and the phaseless operator is not fitted to the reconstruction target. The priority claim ('first attempt to develop a factorization type method for phaseless data') should be checked carefully against the literature on approximate/asymptotic factorization methods, since the method is closely related in spirit to [3,20,21,58,59]. For a major revision, I would ask either for a stability or regularization analysis of the indicator under the known operator-norm error, or for a clear reframing of the contribution as a heuristic numerical scheme with the rigorous asymptotic expansion as motivation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Send this one. Zhang and Zhang construct a phaseless total-field operator from intensity measurements on a large circle and prove that, after a fixed phase rotation and Sobolev scaling, it converges in operator norm to 1/sqrt(8kπR) times a modified far-field operator, with rate controlled. That asymptotic relation (Theorems 3.7 and 4.2) is the real new mathematics, and it is credible: the estimates follow from standard oscillatory-integral bounds and an interpolation argument. The paper is also honest. Instead of pretending to prove a full factorization range identity for phaseless data, it introduces a modified far-field operator F# whose range test comes from the known Kirsch-Liu theory, and then proposes to use the nearby operator (N~_R)# as a stand-in. This is the first factorization-type non-iterative algorithm for phaseless total-field data, and the numerical examples for sound-soft, sound-hard, impedance, and penetrable obstacles support the method visually.\n\nThe weak point is exactly where the stress-test note points. The reconstruction guarantee requires that the eigensystem of (N~_R,M)# inherit the range characterization of F#. The paper only proves norm convergence, and norm convergence of compact operators does not control the weighted Picard series that defines the indicator. The authors know this; Remark 5.3 says explicitly they cannot provide a rigorous analysis of W_M^PW. So the central inversion claim is heuristic. It is a reasonable heuristic, and the numerics are suggestive, but no amount of operator-norm asymptotics closes that gap. To tighten the paper, I would want either a spectral-inheritance theorem under modest assumptions, or at minimum a convergence study: fixing the obstacle, increasing R and M, reporting quantitative reconstruction errors, and comparing against the direct sampling methods in [26,28,66]. The absence of code and quantitative metrics is minor but contributes to the conditional feel.\n\nThe citation pattern is clean: the factorization of F is taken from Kirsch/Grinberg and Kirsch/Liu, not from the authors' prior work, and the self-citations are to their own direct-sampling phaseless papers, which are the relevant baselines.\n\nNet: I would send this to peer review. It is a genuine first step with a proven asymptotic core and an openly admitted gap. A good referee can push the authors to either close the gap or market the method as an empirical heuristic with a convergence study. Both are salvageable. Not a desk reject.","headline":"A credible first factorization-type method for phaseless acoustic data, with a proven operator asymptotics and an honestly admitted spectral-inheritance gap that keeps the full inversion guarantee heuristic.","tokens_in":26784,"tokens_out":3304,"would_cite":true,"duration_ms":39055,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R30","35Q60","65R20","65N21","78A46"],"pacs":[],"model":"deepseek-v4-flash","headline":"The location and shape of an acoustic scatterer can be recovered from intensity-only measurements through an approximate factorization method.","keywords":["inverse acoustic scattering","phaseless total-field data","factorization method","approximate factorization","far-field operator","oscillatory integrals","non-iterative reconstruction","Helmholtz equation"],"falsifier":"For a sound-soft disk with known radius, compute the discretized matrix $(\\widetilde N_{L,M})_\\#$ from forward data and compare its first several eigenvectors with those of $(1/\\sqrt{8k\\pi R})\\widetilde F_\\#$ as $R$ increases. If the angle between the dominant eigenspaces does not tend to zero, or if the indicator $W_{L,M}^{PW}(z)$ stops being large inside and small outside the disk, the spectral-inheritance step on which the method depends is false.","tokens_in":2147,"feed_emoji":"📡","tokens_out":3464,"duration_ms":109096,"temperature":0.7,"pith_summary":"This paper tries to establish that the location and shape of an unknown acoustic scatterer can be recovered from phaseless (intensity-only) total-field data at a fixed frequency, measured on a circle enclosing the scatterer, without knowing whether the scatterer is sound-soft, sound-hard, impedance-type, or a penetrable medium. The route is an approximate factorization method: an operator is built from the measured intensities $|u(R\\hat x,d)|^2-1$, and the paper proves that, as the measurement radius $R$ grows, this operator converges in operator norm to a constant multiple of a modified far-field operator. Because the modified far-field operator admits an exact factorization whose spectral system characterizes the scatterer, the same spectral indicator applied to the measured-data operator is expected to recover the scatterer. The practical value is that phase information is often hard to measure accurately, and the algorithm is non-iterative and needs no prior knowledge of the boundary condition.","feed_headline":"Intensity-only acoustic data recover scatterer shape","feed_subtitle":"A spectral trick turns intensity-only circle measurements into obstacle shape, no boundary condition needed.","key_machinery":"The load-bearing object is the modified phaseless total-field operator $\\widetilde N_R^{PW} = e^{-i(kR+\\pi/4)} B_{1/2}^* N_R^{PW} B_{1/2}$, where $N_R^{PW}$ is the integral operator with kernel $|u(R\\hat x,d)|^2 - 1$ on the measurement circle, and $B_{1/2}$, $B_{1/2}^*$ are the Fourier-multiplier operators $(1+m^2)^{-1/4}$ that renormalize the Sobolev scales so that everything acts on $L^2(\\mathbb S^1)$. The identity carrying the argument is the operator-norm proximity of the symmetrized absolute value $(\\widetilde N_R^{PW})_\\# = |\\operatorname{Re}(\\widetilde N_R^{PW})| + |\\operatorname{Im}(\\widetilde N_R^{PW})|$ to $(1/\\sqrt{8k\\pi R})\\widetilde F_\\#$, obtained by estimating the residual scattered-field terms with oscillatory-integral bounds. The paper then feeds the factorization of $\\widetilde F = B_{1/2}^* F B_{1/2}$ through a range identity, yielding the exact criterion $z\\in D \\iff B_{1/2}^*\\varphi_z \\in R(\\widetilde F_\\#^{1/2})$, and uses that criterion to define the indicator function from the eigensystem of $(\\widetilde N_{L,M})_\\#$.","core_discovery":"The central claim is that the symmetrized phaseless total-field operator $(\\widetilde N_R^{PW})_\\# := |\\operatorname{Re}(\\widetilde N_R^{PW})| + |\\operatorname{Im}(\\widetilde N_R^{PW})|$ has the asymptotic expansion $(\\widetilde N_R^{PW})_\\# = (1/\\sqrt{8k\\pi R})\\widetilde F_\\# + O(R^{-\\alpha})$ in $L(L^2(\\mathbb S^1))$ for every $\\alpha\\in(1/2,1)$, where $\\widetilde F_\\#$ is the symmetrized modified far-field operator. For $\\widetilde F_\\#$ the paper proves an exact range characterization: $z\\in D$ if and only if $B_{1/2}^*\\varphi_z$ lies in the range of $\\widetilde F_\\#^{1/2}$, where $\\varphi_z(\\hat x)=e^{-ik\\hat x\\cdot z}$ and $B_{1/2}^*$ is the adjoint of the Sobolev renormalization operator $B_{1/2}$. Consequently, the paper asserts that the eigensystem of $(\\widetilde N_R^{PW})_\\#$ approximately reconstructs both the location and the shape of $D$ when $R$ is sufficiently large, and it presents this as the first factorization-type inversion method for phaseless scattering data. Numerical experiments with peanut, kite, rounded square, rounded triangle, and two-component obstacles are used to support the claim.","pith_inferences":["The paper leaves open whether operator-norm proximity transfers to the spectral systems; if eigenvectors are unstable, the indicator could fail for scatterers with clustered eigenvalues, so a numerical eigenvalue-stability study would be a direct test of the method's core heuristic.","Plotting $W_{L,M}^{PW}(z) - W(z)$ at fixed points inside and outside a known scatterer as $R$ grows would quantify the actual convergence rate and expose whether the $O(R^{-\\alpha})$ bound is sharp in practice.","The symmetrization $|\\operatorname{Re}(\\cdot)| + |\\operatorname{Im}(\\cdot)|$ is a transferable device: it could be applied to other phaseless inverse problems, such as electromagnetic scattering or near-field measurements with point sources, whenever the underlying far-field operator admits a range-identity factorization.","The paper gives no $k$-dependent constants in the error bound (Remark 3.8), so the practical resolution limits at high frequency remain unknown; deriving such constants would require a wavenumber-explicit analysis of the far-field pattern."],"forward_implications":["The same algorithm runs unchanged for sound-soft, sound-hard, impedance, and penetrable scatterers, because the boundary condition never enters the construction of the indicator.","Once the measurement radius $R$ and truncation $M$ are large enough, the indicator $W_{L,M}^{PW}(z)$ should be large inside the obstacle and small outside it, matching the exact far-field characterization.","Larger measurement radii improve the reconstruction in the paper's numerical study of an impedance obstacle with $R=4,8,12$, consistent with the $O(R^{-\\alpha})$ approximation error.","The method tolerates additive noise: the paper's examples with 10% and 20% noise still locate the scatterers.","The reconstruction is non-iterative and requires only one eigensystem computation of the discretized symmetrized matrix $(\\widetilde N_{L,M})_\\#$."],"supporting_citations":[{"why":"Supplies the oscillatory-integral lemma used to control the residual terms in the asymptotic expansion of the phaseless total-field operator.","marker":"[14]"},{"why":"Provides the well-posedness, far-field asymptotics, and analyticity facts about the scattered field used throughout the estimates.","marker":"[17]"},{"why":"Extends the factorization method to near-field measurements, motivating the modified operator construction used here.","marker":"[22]"},{"why":"Introduces the classical factorization method whose spectral characterization of the scatterer is the template for the algorithm.","marker":"[29]"},{"why":"Supplies the far-field operator factorizations for sound-soft, impedance, and penetrable scatterers and the range-identity framework.","marker":"[31]"},{"why":"Provides the modified range identity that yields the exact characterization $z\\in D$ via the range of $\\widetilde F_\\#^{1/2}$.","marker":"[32]"},{"why":"Supplies the Sobolev interpolation theorem used to pass from $H^1\\to L^2$ and $L^2\\to H^{-1}$ estimates to the $H^{1/2}\\to H^{-1/2}$ bound.","marker":"[39]"},{"why":"Supplies the operator inequality used in Remark 4.8 to convert the convergence of $\\widetilde N_R^{PW}$ into the $O(R^{-\\alpha})$ bound for the symmetrized operators.","marker":"[41]"}],"fun_headline_variants":["Phaseless acoustic data reconstruct scatterer shape","Intensity-only data reveal obstacle shape without boundary info","Factorization trick maps phaseless data to scatterer geometry","First factorization method for phaseless inverse scattering","Shape from phaseless echoes: no boundary condition needed"],"cache_read_input_tokens":28800,"weakest_assumption_plain":"The load-bearing premise is that two operators close in operator norm have spectral systems close enough for the finite-dimensional indicator built from the measured-data operator to inherit the exact range characterization proved for the far-field operator; the paper states in Remark 5.3 that it cannot yet prove this inheritance.","fun_headline_variants_meta":{"raw":{"variants":["Phaseless acoustic data reconstruct scatterer shape","Intensity-only data reveal obstacle shape without boundary info","Factorization trick maps phaseless data to scatterer geometry","First factorization method for phaseless inverse scattering","Shape from phaseless echoes: no boundary condition needed"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000653,"raw_usage":{"total_tokens":3085,"prompt_tokens":1128,"completion_tokens":1957,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":744,"completion_tokens_details":{"reasoning_tokens":1881}},"tokens_in":744,"tokens_out":1957,"duration_ms":14113,"temperature":1.0,"reasoning_tokens":1881,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:03:42.904037+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a sound-soft disk with known radius, compute the discretized matrix $(\\widetilde N_{L,M})_\\#$ from forward data and compare its first several eigenvectors with those of $(1/\\sqrt{8k\\pi R})\\widetilde F_\\#$ as $R$ increases. If the angle between the dominant eigenspaces does not tend to zero, or if the indicator $W_{L,M}^{PW}(z)$ stops being large inside and small outside the disk, the spectral-inheritance step on which the method depends is false.","supporting_citations":[{"cited_title":"Chen and G","cited_arxiv_id":null,"evidence_quote":"Supplies the oscillatory-integral lemma used to control the residual terms in the asymptotic expansion of the phaseless total-field operator."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Extends the factorization method to near-field measurements, motivating the modified operator construction used here."},{"cited_title":"Kirsch, Characterization of the shape of a scatterin g obstacle using the spectral data of the far field operator, Inverse Problems 14 (1998), 1489–1512","cited_arxiv_id":null,"evidence_quote":"Introduces the classical factorization method whose spectral characterization of the scatterer is the template for the algorithm."},{"cited_title":"Kirsch and X","cited_arxiv_id":null,"evidence_quote":"Provides the modified range identity that yields the exact characterization $z\\in D$ via the range of $\\widetilde F_\\#^{1/2}$."},{"cited_title":"Kress, Linear Integral Equations (3rd Ed.), Springer, New York, 2014","cited_arxiv_id":null,"evidence_quote":"Supplies the Sobolev interpolation theorem used to pass from $H^1\\to L^2$ and $L^2\\to H^{-1}$ estimates to the $H^{1/2}\\to H^{-1/2}$ bound."},{"cited_title":"Lechleiter, Factorization Methods for Photonics and Rough Surfaces , PhD thesis, Univ","cited_arxiv_id":null,"evidence_quote":"Supplies the operator inequality used in Remark 4.8 to convert the convergence of $\\widetilde N_R^{PW}$ into the $O(R^{-\\alpha})$ bound for the symmetrized operators."}],"review_version":1}