{"id":"e3105969-b861-40c6-9f57-d9a5e90850d2","arxiv_id":"2602.15366","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Passive acoustic multiple scattering is inverse-designed to realize reflectionless unitary mode transformations, experimentally demonstrating non-Abelian braid generators on four waveguide modes.","lead":"Researchers designed a slab of passive acoustic scatterers that lets sound pass without reflection while performing a chosen unitary operation on the waveguide modes. They demonstrate non-commuting braid operations and logic gates in a tabletop acoustic waveguide, pointing toward programmable multi-mode acoustic devices.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Loss-removal normalization in Supp. Note 5 relies on an unjustified simultaneous diagonalization of S and the Wigner-Smith delay operator; the reported R=0 and exact T-matrices may be artifacts.","rationale":"The reader's weakest assumption is the loss normalization; I agree and sharpen it: the supplement's derivation is not just relying on weak loss but on a specific simultaneous-diagonalization claim that is false in general. This is a correctness risk rather than a consensus disagreement. The paper's theoretical framework (Eq. 2 and the optimization) is not threatened by this concern, and the field-map observations of mode conversion provide partial independent support, so I would not reject outright. But the quantitative claims—zero reflection, exact σ_i entries, and gate matrices—are supported only by normalized data, and the proof of the normalization is invalid as written. A conditional verdict, with the loss-normalization test above as the condition, matches the reader's assessment. I set verdict_should_be to UNCHANGED because the reader already chose CONDITIONAL and this concern reinforces rather than redirects that decision. The 'arbitrary unitary' overclaim is a secondary limitation (no proof of reachability), but the single most load-bearing issue is the unvalidated loss-removal step.","tokens_in":22040,"tokens_out":20790,"duration_ms":211025,"concrete_test":"Measure the raw (unnormalized) S-matrix of the Fig. 2(b1) σ1 sample at 3.18, 3.20, and 3.22 kHz. Compute the Wigner-Smith delay operator by central difference, Q_ω ≈ -i S(3.20 kHz)† [S(3.22)-S(3.18)]/(2Δω), and evaluate the off-diagonal norm of Q_ω in the singular-vector basis of S(3.20 kHz). If this norm is not small compared with δ/2 ≈ 0.0016ω0 (the assumed loss-induced singular-value spread), then loss rotates the singular vectors, Supp. Eq. S52 is invalid, and the normalized T-matrix reported in Fig. 2(b2) cannot be taken as the physical T-matrix.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The experimental core of the paper—Figs. 2(b2-d2), 4(a3,b3,c3,d3), 5(a,c)—consists of S-matrices that have been singular-value-normalized to remove loss. The justification in Supplementary Note 5 (Eqs. S47-S52) requires that the Wigner-Smith time-delay operator Q_ω = -i S^{-1} dS/dω and the lossless symmetric unitary S(ω0) are simultaneously diagonalizable in the same WS/Takagi basis. For a generic smooth family of symmetric unitary matrices this is false: [Q,S] need not vanish, so a weak loss term δ/2 Q does not simply rescale the singular values of S; it also rotates its singular vectors. Equation S49 ('In the same WS basis, S admits a Takagi decomposition') is precisely the unproved assumption. If it fails, replacing singular values by 1 yields a unitary matrix that is not the physical lossless S(ω0), and the claimed R≈0 and exact T-matrix entries (σ_i(π/6), BNOT, CNOT) become normalization artifacts. The added selection of configurations with τ/⟨τ⟩≤0.9 concentrates on cases where this assumption is least directly tested. This is load-bearing because the quantitative demonstration of non-Abelian braiding and gates rests on those normalized matrices.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an inverse-design method, based on generalized Wigner–Smith operators, for engineering reflectionless complex acoustic media whose transmission matrices are prescribed unitary operations. Concatenating such media multiplies their T-matrices, enabling wave-based unitary transformations. The authors optimize 80-cylinder sections in a four-mode waveguide to realize the braid-group generators σ1, σ2, σ3 with a nontrivial phase η=π/6, experimentally demonstrate the non-commutativity of σ1σ2 versus σ2σ1 and the commutativity of σ1σ3 versus σ3σ1, and implement BNOT and CNOT gates. A numerical 8×8 DFT example is also presented. The central experimental claims rely on measured S-matrices that are post-processed by singular-value normalization to remove loss.","tokens_in":22409,"tokens_out":10398,"duration_ms":98004,"significance":"If the experimental results are valid, the work would substantially extend the range of operations achievable with multiple-scattering media: instead of treating disorder as a scrambler, the authors show that tailored complex media can perform targeted unitary operations, including non-commuting braid-group operations, without relying on adiabatic evolution. The acoustic GWSO derivation in Supplementary Note 4 is a useful contribution in its own right. The method is, in principle, transferable to other wave platforms. However, the paper does not provide a rigorous argument for the 'arbitrary unitary T-matrix' claim, and the experimental demonstration depends on a loss-removal normalization whose mathematical justification is questionable. The paper ships no public code or data, so the reproducibility of the optimization and measurements rests on the 'available upon request' statement.","major_comments":[{"comment":"The loss-removal normalization is not justified. Eq. (S49) assumes that in the basis that diagonalizes the Wigner–Smith operator Qω, the symmetric unitary S(ω0) also admits a Takagi decomposition with the same basis. This is equivalent to [Qω,S]=0, which does not hold for a generic smooth family of symmetric unitary matrices. For example, S(ω)=R(ω) diag(e^{iφ1},e^{iφ2}) R(ω)^T with a real rotation R and φ1≠φ2 gives [Q,S]≠0. Then the weak-loss term (δ/2)Q does not simply rescale the singular values of S; it rotates its singular vectors. Replacing the measured singular values by unity then yields a unitary matrix that is not the physical lossless S. Since the experimental T-matrices in Figs. 2(b2–d2), 4(a3–d3), and 5(a,c) are all shown after this normalization, the reported R≈0 and exact T-matrix entries could be artifacts. The authors should present raw, unnormalized S-matrices or provide","section":"Supplementary Note 5, Eqs. (S47)–(S52)"},{"comment":"The headline claim of arbitrary unitary transmission matrices is not demonstrated. The manuscript shows inverse-design optimization for three braid generators, BNOT, CNOT, and one numerical 8×8 DFT. There is no mathematical or scaling argument that any N×N unitary T can be realized by a sufficiently large complex medium, nor a study of how optimization success depends on N or target structure. The word 'arbitrary' in the abstract should be relaxed to 'a broad class of target unitary operations' unless a constructive proof or systematic scalability evidence is added.","section":"Abstract; Results, 'Realization of braiding'; Supp. Note 3"},{"comment":"The experimental configurations are post-selected from 40 optimizations by the dwell-time criterion τ/⟨τ⟩≤0.9, but the manuscript does not report how many of the 40 realizations satisfied this criterion or the distribution of τ. This selection makes it impossible to assess the success rate of the design method, and it concentrates the demonstration on cases in which the loss correction is smallest. The criterion should be reported with the number of candidates, and ideally raw S-matrix data for unselected configurations should be shown to establish that the method is not cherry-picked.","section":"Methods, 'The effect of dissipation'; Supp. Note 5"}],"minor_comments":[{"comment":"The term 'braiding' may be misleading because no physical exchange of scatterers occurs; the operations are static matrices satisfying braid relations. Consider clarifying in the introduction or discussion that this is braiding in the sense of matrix representations of the braid group, not topological braiding of scatterers.","section":"Eq. (4) and Fig. 3"},{"comment":"The experimental S-matrix figures show magnitude/color only. Including numerical values or a quantitative error metric (e.g., deviation from the target T-matrix) would help support claims such as 'exactly σ1(π/6)'.","section":"Figs. 2(b2–d2), 4(a3–d3), 5(a,c)"},{"comment":"The Yang–Baxter relation is only validated numerically, and the text states 'Experimental validation is currently beyond our capability'. This limitation should be clearly stated in the main text as well.","section":"Supp. Note 2"},{"comment":"The global phases φ1, φ2, φ3 are stated but the extraction procedure is not described. Please indicate how these phases are determined from the measured S-matrices.","section":"Methods, 'The effect of dissipation'"},{"comment":"The 'available upon request' statements are weak. Depositing the optimization code and the raw experimental S-matrix data would materially improve reproducibility, especially given the loss-normalization step.","section":"Data/Code availability"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern about the loss-removal normalization lands: the simultaneous diagonalization of S and the Wigner–Smith operator is not generally valid, and the experimental demonstration leans heavily on this step. The paper is otherwise interesting and the inverse-design methodology is worth publishing, but the authors need to either re-analyze the raw data with a defensible loss model or present unnormalized S-matrices. Also, the 'arbitrary unitary T-matrix' claim should be tempered unless a general construction is supplied."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my read. The genuinely new thing is that you can inverse-design a passive multiple-scattering medium in a waveguide so its transmission matrix is a specific unitary, and concatenate such sections so the T-matrices multiply. The braid generators with a complex phase eta=pi/6, the non-commutation of sigma1 sigma2, and the BNOT/CNOT gates all appear in both simulation and experiment. That's a real advance over the anti-reflection work, which gave perfect transmission but no control over the output wavefront.\n\nThe GWSO-based optimization is the strongest part. The acoustic version is derived carefully, including the boundary-integral form with tangential gradients, and the optimization seems robust. The field-map measurements in Figs. 3 and 4 independently show the mode conversion and the 30/150 degree phases—those aren't just reconstructed S-matrix entries—so I trust the qualitative demonstration.\n\nThe soft spots are real but not fatal. 'Arbitrary unitary' is an overclaim; they show a few targets, not a general construction. The experimental S-matrices are all shown after singular-value normalization, and the justification in Supp Note 5 rests on an unproved simultaneous diagonalizability of S and the Wigner-Smith Q. For a generic symmetric unitary S, [Q,S] need not vanish, so the loss correction can rotate singular vectors. The paper doesn't bound this. The tau/<tau> <= 0.9 selection and the lack of error bars add to the concern. And the Faugno-Ozawa work is only a supplementary reference; if it overlaps, the authors should say so.\n\nNone of this sinks the central result. The agreement with lossless simulation and the field measurements make the normalized S-matrices credible. But the abstract should say 'targeted' instead of 'arbitrary,' and the loss normalization needs either a proof or a numerical check of the commutator. I'd send this to peer review; the referee load is justified.","headline":"Inverse-designed acoustic T-matrices work in experiment, but the 'arbitrary' claim is overbroad and the loss normalization needs scrutiny.","tokens_in":22830,"tokens_out":3351,"would_cite":true,"duration_ms":34104,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that reflectionless complex media can be given an arbitrary unitary transmission matrix, so cascading them multiplies their T-matrices and performs targeted unitary operations—demonstrated as non-Abelian braiding of four ac","keywords":["multiple scattering","reflectionless media","unitary transmission matrix","non-Abelian braiding","Wigner-Smith operator","waveguide modes","wave-based computing","acoustic waveguide"],"falsifier":"Measure one optimized section's raw scattered power with a calibrated absolute-energy probe at 3.2 kHz, then repeat the measurement after adding a known absorbing liner to the waveguide walls; if the normalized T-matrix changes substantially or the recovered zero reflection fails to hold under increased loss, the singular-vector-invariance assumption is disproved.","tokens_in":21932,"feed_emoji":"🔊","tokens_out":5294,"duration_ms":51099,"temperature":0.7,"pith_summary":"The authors set out to show that the messy, multidirectional scattering of sound inside a dense arrangement of cylinders can be precisely engineered rather than merely endured. They use a generalized Wigner-Smith operator to move each cylinder along a gradient that drives the medium's scattering matrix toward a target with zero reflection and a specified unitary transmission matrix. With this design tool they experimentally build acoustic sections that behave as the three generators of the four-strand braid group, and they demonstrate that concatenating sections multiplies their transmission matrices, making the order of operations matter for neighboring generators and not matter for distant ones. They further realize binary NOT and CNOT gates and numerically a discrete Fourier transform, arguing that complex media can serve as compact, passive wave-based computing platforms.","feed_headline":"Sound braids waveguide modes via reflectionless scattering media","feed_subtitle":"Concatenating reflectionless acoustic sections multiplies transmission matrices, enabling braid and logic gates in sound.","key_machinery":"The load-bearing identity is the multiplication rule for reflectionless sections: if the reflection blocks of each section vanish, the total transmission matrix of stacked media is the ordered product of the individual T-matrices, turning propagation into a matrix product. To realize each T, the optimization uses the generalized Wigner-Smith operator (GWSO): for every cylinder, the derivative of the scattering matrix under a small displacement equals i S Q_r, and Q_r is a boundary integral of pressure and tangential gradient over the cylinder's rigid surface, proportional to the acoustic radiation force. The gradient-descent objective g(r) = 1 - |Tr(S_obj^dagger S)|^2/(4N^2) drives the actua","core_discovery":"On the paper's own terms, the central discovery is that the transmission matrix of a reflectionless complex medium is not a nuisance to be averaged over but a designable object: when R = 0, cascading K sections yields T_total = T_K ... T_2 T_1, so each section is a linear operator on the four waveguide modes. The authors implement this by minimizing the distance g(r) between the actual and target S-matrices, with gradients supplied by the generalized Wigner-Smith operator, a boundary integral involving the acoustic radiation force on each cylinder. The optimized sections realize the braid generators sigma1, sigma2, sigma3 with a complex phase eta = pi/6 in a 2x2 block Y(eta), experimentally","pith_inferences":["I would read the paper's broader bet as: if the T-matrix is fully programmable, then a stack of such sections is a programmable multimode linear processor; a natural next test is whether reconfiguring cylinder positions in real time can switch the unitary operation while preserving zero reflection, which the static demonstrations do not prove.","The loss-compensation step is the part most worth probing: the claimed exactness of the T-matrix entries after singular-value normalization assumes losses only scale the singular values. A stronger test would compare normalized S-matrices against full-wave simulations that include thermoviscous losses in the air and boundary layers rather than a uniform complex frequency shift.","The optimization demonstrably works for four modes and eighty cylinders; whether the scheme scales to larger modal counts is open, since the S-matrix dimension grows and the gradient landscape may develop more local minima. One could test convergence statistics for N = 8 and N = 16 targets."],"forward_implications":["Because reflectionless sections multiply their transmission matrices, any sequence of unitary operations on the N modes can be composed by stacking sections, making a stack of such media a linear wave-based processor.","Non-Abelian braiding of acoustic modes can be produced by static passive scatterers, without requiring the adiabatic evolution or parity-time-symmetric Hamiltonians used in earlier waveguide braiding demonstrations.","The same optimization can encode discrete Fourier transforms and two-qubit-style gates such as CNOT into the transmission matrix, offering a route to multimode acoustic signal processing and wave-based logic.","Since the scheme relies only on linear interference and rigid scatterers, it should transfer to microwaves, elastic waves, and photonics wherever the corresponding Wigner-Smith operator can be computed."],"fun_headline_variants":["Acoustic non-Abelian braiding via designed reflectionless media","Tailored reflectionless sections braid sound modes in a waveguide","Sound waves braid non-Abelian style with custom scattering","Experimentally realized non-Abelian braiding in acoustic waves","Designer reflectionless media enable braided acoustic logic"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The experimental demonstration assumes that weak dissipation only reduces the singular values of the scattering matrix while leaving its singular vectors unchanged, and that configurations with dwell time below 0.9 times the ensemble average are therefore representative; if that loss-compensation assumption fails, the reported zero reflection and exact T-matrix entries could be artifacts of the normalization rather than properties of the physical medium.","fun_headline_variants_meta":{"raw":{"variants":["Acoustic non-Abelian braiding via designed reflectionless media","Tailored reflectionless sections braid sound modes in a waveguide","Sound waves braid non-Abelian style with custom scattering","Experimentally realized non-Abelian braiding in acoustic waves","Designer reflectionless media enable braided acoustic logic"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000716,"raw_usage":{"total_tokens":3059,"prompt_tokens":752,"completion_tokens":2307,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":496,"completion_tokens_details":{"reasoning_tokens":2232}},"tokens_in":496,"tokens_out":2307,"duration_ms":16164,"temperature":1.0,"reasoning_tokens":2232,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T22:52:01.796135+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure one optimized section's raw scattered power with a calibrated absolute-energy probe at 3.2 kHz, then repeat the measurement after adding a known absorbing liner to the waveguide walls; if the normalized T-matrix changes substantially or the recovered zero reflection fails to hold under increased loss, the singular-vector-invariance assumption is disproved.","supporting_citations":[],"review_version":1}