{"id":"0ab6021c-94ef-4762-8a2a-dae6515e1c4c","arxiv_id":"2509.09506","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A localized perturbation makes the differential scattering matrix rank one, so the change in the output wavefront is always the same shape regardless of the input wavefront.","lead":"Changing one small, tunable element inside a complex scattering medium always changes the outgoing wave in a single fixed 'frozen' pattern, no matter how the incoming wave is shaped. The result explains and demonstrates a universal rule for differential wave measurements, with potential uses in 6G wireless, imaging, and thermal-noise control.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central rank-one/freezing result holds only for a true single scalar load; finite-size or multi-mode localized perturbations yield multi-rank ΔS, so 'any complex medium' overstates the theorem.","rationale":"The reader's weakest_assumption identifies the same load-bearing concern: the rank-one/freezing result depends on the perturbed element being a lumped, scalar-load port. The algebraic derivation in Sec. 2.2.2 is clean, and the experimental evidence in Sec. 3 is strong for the specific PIN-diode case, so the concern does not invalidate the paper. But the abstract and Sec. 2.2.2 state the unconditional version, while the SI's MPLC extension replaces sub-wavelength with sub-resolution without an explicit bound, so the universality claim is broader than the derivation. The thermal-noise independence assumption in Sec. 4.2 is a separate limitation but secondary: even if it failed, the rank-one/frozen result would stand. Therefore the conditional verdict is appropriate and no adjustment is needed.","tokens_in":16591,"tokens_out":8173,"duration_ms":94970,"concrete_test":"Use the paper's own MPLC model and perturb a 2×2 block of phase pixels (or two adjacent pixels) instead of a single pixel. From ΔU = Σ_j (e^{iφ_j'}−e^{iφ_j}) f_j g_j^T, compute the SVD of ΔU for random masks and propagators. If the second singular value is generically comparable to the first and the direction of Δy for random inputs varies, then freezing requires the perturbation to occupy a single resolution cell. Equivalently, repeat the Sec. 3 experiment with a meta-atom whose tunable element has two independently biased diodes; if ΔS acquires a second significant singular value, the scalar-load idealization is the load-bearing condition.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Sec. 2.2.2, the rank-one result is derived for a single scalar load ρ in the multiport model: ΔS = p(ρ̂/(1−ρ̂g) − ρ/(1−ρg))q, an exact rank-one update. This is a theorem about a model in which the perturbed object is a point-like, single-port load with no internal degrees of freedom. The paper's universal phrasing ('for a localized perturbation, the differential scattering matrix of any complex medium has rank one') is stronger than what is proven. If a physical perturbation is localized in space but has finite electrical size or multiple independent internal modes, it must be represented by more than one equivalent port (or more than one phase pixel in the MPLC picture), and ΔS becomes a sum of rank-one updates — generically rank > 1. Then Δy = Σ_k β_k(x) s_k and the differential output wavefront is input-dependent. The experiment in Sec. 3 works because the PIN diode is very small, and the paper acknowledges small non-leading singular values. But the SI's sub-resolution (rather than sub-wavelength) condition for MPLCs is itself a nontrivial assumption: a pixel larger than one resolution cell should be subdivided, and each sub-pixel contributes a rank-one term. Thus the central claim is conditionally true for genuinely scalar, point-like perturbations, not unconditionally true for any localized perturbation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces and experimentally demonstrates 'frozen differential scattering': in a linear, passive, multiply-scattering medium, changing a single localized (scalar, point-like) parameter produces a rank-one change ΔS of the full scattering matrix, so the differential output wavefront Δy is collinear with a fixed vector s1 independent of the input x. The authors derive this in a multiport-network model for a single tunable load (Sec. 2.2.2), support it with a coupled-dipole derivation and an MPLC update in the SI, and validate it in a programmable-metasurface-stirred reverberation chamber with 8 antennas, 225 meta-atoms, and 250 random configurations. They further analyze coherence purification and thermal-noise emission, and demonstrate 'customized freezing' by optimizing 224 background meta-atom states to shape the frozen differential output mode or to maximize signal-to-thermal-noise ratio.","tokens_in":1318,"tokens_out":1343,"duration_ms":145088,"significance":"The result is conceptually clean and potentially broadly useful. It turns a known algebraic fact (rank-one update of the scattering matrix) into a physically meaningful, testable wavefront phenomenon, extending the 'frozen wavefront' concept from special static media to generic reconfigurable media. The experimental validation uses four metrics (R, P, C, E) and direct wavefront comparisons; the optimized-frozen-mode demonstration (Sec. 5) is a nice addition. If the scope is properly qualified, the paper should be of interest to the optics and microwave communities. The derivation is explicit and the experiments support the claim; no fitted parameters enter the central rank-one argument.","major_comments":[{"comment":"The universal phrasing 'for a localized perturbation, the differential scattering matrix of any complex medium has rank one' is stronger than what is derived. The derivation assumes N_S=1 and a lumped (point-like) scalar load ρ (Sec. 2.2.1). If the perturbed object has finite electrical size or multiple independent internal degrees of freedom, it corresponds to several ports/pixels and ΔS is a sum of rank-one terms, generically of higher rank. Please qualify the claim throughout (abstract, intro, Sec. 2.2.2) as 'a perturbation that changes a single scalar degree of freedom' (e.g., a single lumped load or a single MPLC pixel) and explicitly state the multi-rank case.","section":"Abstract / Sec. 2.2.2"},{"comment":"The MPLC derivation in the SI needs its 'sub-resolution' condition spelled out precisely. A phase-mask pixel is an abstract single control degree of freedom; a physical modification spanning several resolution cells would correspond to multiple pixels and destroy exact rank one. Please state the condition as 'the modified area must correspond to exactly one independent pixel in the discretized model' and note that the rank-one result does not extend to multi-pixel modifications.","section":"SI: MPLC"},{"comment":"The STNR derivation relies on the independence assumption ⟨n n̂†⟩=0. This is physically reasonable for measurements taken at well-separated times, but it should be stated explicitly, since for simultaneous measurement with common internal noise sources the cross-correlation terms would not vanish and would modify the formula. Please add one sentence justifying the assumption and indicating the regime in which it applies.","section":"Sec. 4.2"}],"minor_comments":[{"comment":"Please define p, q, and g explicitly before the formula for ΔS, and note that reciprocity gives p=q^T. Currently the notation is introduced somewhat abruptly.","section":"Sec. 2.2.2"},{"comment":"The coherence result 𝚪out,Δ = (ΔS)𝚪in(ΔS)† assumes the same input realization x is used before and after the perturbation. If independent input realizations are used in the two measurements, the coherence of the difference is not rank one. Please clarify this condition in the text.","section":"Sec. 4.2"},{"comment":"The aligned phase (blue solid) and raw phase (blue dashed) are visually very different; a sentence explaining why the raw phase is not meaningful for the comparison would improve readability.","section":"Fig. 3"},{"comment":"The phrase 'perfectly coherent' could be misread as a statement about temporal coherence. Consider saying 'spatially single-mode' or 'rank-one spatial coherence' to avoid ambiguity.","section":"Abstract"},{"comment":"In the one-hot optimization, two examples are shown; it would be helpful to indicate which output index is targeted in each case, and whether the observation agrees for the other tested targets.","section":"Sec. 5.2"}],"recommendation":"major_revision","confidential_remarks":"The overstatement of the theorem's scope in the abstract and intro is the main issue; it is fixable with careful qualification. The experimental work and the customization demonstration are convincing. I see no grounds for rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: the central identity — that changing one scalar, point-like load updates the scattering matrix by a rank-one term — is correct and already appears in the cited literature. What this paper adds is the physical reading: the change in the output wavefront is \"frozen,\" independent of the input wavefront, and it works for reflection as well as transmission. That's a useful way to think about differential scattering, and the paper develops some clean consequences for coherence and thermal noise.\n\nThe experimental work is genuinely good. Across 250 random configurations of the other meta-atoms, the differential scattering matrix shows effective rank 1.3 for S and 1.1 for T, with collinearity 0.99–1.00. That is strong evidence that the lumped-element approximation holds for their PIN-diode meta-atoms. The customized freezing section is more suggestive than conclusive — four optimizations, no error bars — but the idea is sensible.\n\nWhere the paper overstates is the claim that \"for a localized perturbation, the differential scattering matrix of any complex medium has rank one.\" The proof only covers a perturbation modeled as a single scalar load at a point-like port. A spatially extended or multi-mode localized perturbation will generally produce multiple significant singular values. The authors hedge later, acknowledging finite size, but the abstract and Sec. 2.2.2 phrase it as universal. They should qualify it.\n\nTwo other soft spots. First, the thermal noise argument assumes the noise realizations before and after the perturbation are independent. That is plausible but not justified, and the STNR formula depends on it. Second, no code or data is released, so the experimental claims can't be independently checked. That's a bigger deal for the customized freezing part than for the rank-one demonstration, which is almost a mathematical tautology once you accept the multiport model.\n\nOverall: the core result is a simple, correct algebraic fact given the model, and the experiments support it. The paper deserves a serious referee, with the expectation of revision to tone down the \"any complex medium\" wording and address the independence assumption. I'd cite it for the interpretation, but not for the identity itself.","headline":"Rank-one differential scattering is a clean identity with a useful physical interpretation and solid supporting experiments, but the 'any complex medium' claim outruns the proof, and the paper still deserves a serious referee.","tokens_in":17353,"tokens_out":2383,"would_cite":true,"duration_ms":29667,"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 localized perturbation makes the differential scattering matrix of any complex medium rank one, so the change in output wavefront is frozen.","keywords":["frozen differential scattering","rank-one perturbation","scattering matrix","complex media","reconfigurable metasurface","wavefront shaping","coherence","thermal noise"],"falsifier":"Take a complex medium and a tunable element that is not point-like—e.g., a phase-mask pixel larger than one diffraction-limited resolution cell or a multi-parameter tunable resonator with two independent controls—and measure the differential output for two very different inputs. If the two Δy vectors are not collinear, or if ΔS has two comparable singular values, the rank-one freezing claim fails in that regime.","tokens_in":16419,"feed_emoji":"📡","tokens_out":5836,"duration_ms":67905,"temperature":0.7,"pith_summary":"This paper establishes a universal property of any wave-scattering system: when the medium is changed at a single point-like spot, the change in the scattered field is always proportional to one fixed output pattern, no matter what wavefront is sent in. The author calls this 'frozen differential scattering' and proves it by showing that a localized perturbation is a rank-one update of the scattering matrix, so the differential scattering matrix has exactly one significant singular value. Because the differential output lies in a single mode, it is perfectly coherent even for incoherent inputs, and the way thermal noise enters the measurement has a low-rank structure that can be managed. The claim is tested in a programmable-metasurface-controlled radio chamber, and then the frozen pattern is 'customized' by optimizing other metasurface elements, including maximizing the signal-to-thermal-noise ratio. If right, this makes differential wavefront measurements robust and offers new ways to filter, stabilize, and communicate through complex media.","feed_headline":"A localized tweak freezes the wavefront change in any complex medium","feed_subtitle":"Even scrambled, incoherent inputs leave one fixed differential output mode—cleaning up sensing and 6G links.","key_machinery":"The load-tunable-port description of a programmable metasurface: each meta-atom is a lumped, point-like port terminated by a tunable reflection coefficient, so flipping it is a rank-one update of the full scattering matrix via a matrix-inversion lemma. Singular value decomposition then leaves one dominant left singular vector s1 as the frozen output shape; the scalar coupling encoded in the adjacent matrix entries r1 governs how strongly a given input excites that mode. The optimization adds a second layer: tuning the other meta-atoms changes the effective background scattering matrix, thereby shaping s1 or the signal-to-thermal-noise ratio.","core_discovery":"The central claim is that for a sufficiently localized perturbation of any linear, passive, matched complex medium, the differential scattering matrix ΔS has rank one, regardless of the structure of the full scattering matrix S. Consequently the differential output wavefront Δy = ΔS x is always collinear with a single fixed vector s1, so its shape is frozen: changing the input wavefront only rescales it by a complex scalar. The same holds for the transmission block ΔT, so frozen differential behavior occurs in transmission and reflection alike. The paper proves this algebraically from the multiport-network representation, verifies near-rank-one behavior in experiments with a PIN-diode-contro","pith_inferences":["Beyond the paper: the linear-algebraic mechanism—a rank-one matrix update—is independent of wave type, so acoustical, elastic, or matter-wave systems with a point-like perturbation should exhibit the same frozen differential response.","Beyond the paper: the same reasoning suggests that in large programmable arrays, one can precompute each element's frozen response and use it as a basis for fast, model-agnostic wavefront control without full channel estimation.","Beyond the paper: because the frozen output mode is set by the background medium, 'customized freezing' could be used to steer a differential signature toward a specific receiver or away from interceptors, which matters for covert communication.","Beyond the paper: quantifying how the second singular value grows with the perturbation's electrical size would give a practical design rule for when the rank-one approximation is good enough for a given sensing or communication task."],"forward_implications":["If the central claim holds, any differential scattering measurement—optical label-free sensing, coherent change-detection radar, differential detection—inherits a single-mode structure whose output pattern is independent of the input.","The differential signal is perfectly coherent even when the input wavefront is fully incoherent, so freezing-aware receivers do not need coherent illumination.","Thermal-noise coherence changes caused by the perturbation lie in a low-dimensional span (rank at most two for ΔΓ_th), enabling principled denoising and covert nearly-passive signaling.","The frozen mode can be engineered: optimizing other programmable meta-atoms imposes a desired output shape (e.g., one-hot or uniform) or raises the signal-to-thermal-noise ratio (6.3× in the reported experiment).","Differential freezing applies to scattering, reflection included, unlike previously known frozen-transmission cases, so the same principle covers radar and backscatter settings."],"fun_headline_variants":["Frozen differential output: any localized change fixes wavefront shape","Rank-one scattering: localized tweaks freeze differential output in any medium","Incoherent input? Differential output wavefront still freezes","Localized perturbation freezes differential output mode in any complex medium","Differential scattering rank one: perturbation shapes output regardless of input"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The whole proof relies on the changed element being a single tiny, point-like knob with one scalar setting; if the change covers an extended region or has several independent controls, the frozen pattern is lost.","fun_headline_variants_meta":{"raw":{"variants":["Frozen differential output: any localized change fixes wavefront shape","Rank-one scattering: localized tweaks freeze differential output in any medium","Incoherent input? Differential output wavefront still freezes","Localized perturbation freezes differential output mode in any complex medium","Differential scattering rank one: perturbation shapes output regardless of input"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001518,"raw_usage":{"total_tokens":5925,"prompt_tokens":756,"completion_tokens":5169,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":500,"completion_tokens_details":{"reasoning_tokens":5082}},"tokens_in":500,"tokens_out":5169,"duration_ms":32051,"temperature":1.0,"reasoning_tokens":5082,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T18:58:46.121685+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a complex medium and a tunable element that is not point-like—e.g., a phase-mask pixel larger than one diffraction-limited resolution cell or a multi-parameter tunable resonator with two independent controls—and measure the differential output for two very different inputs. If the two Δy vectors are not collinear, or if ΔS has two comparable singular values, the rank-one freezing claim fails in that regime.","supporting_citations":[],"review_version":1}