{"id":"d2febb86-cbfe-48ea-ae8e-b5ab55cbc3fe","arxiv_id":"2607.29137","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"FIM shape optimization is shown in simulation to improve ISAC achievable rate and AoA sensing for OFDM, OTFS, and AFDM in doubly-dispersive MIMO channels.","lead":"Shape-changing intelligent surfaces are modeled as a new control knob for wireless links that both communicate and sense in fast-moving environments, with an optimization rule for tuning the surface geometry. The paper reports simulation gains of roughly 4.5 dB in achievable rate and clearer angle-of-arrival peaks for OFDM, OTFS, and AFDM.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"FIM gains claim lacks a non-morphing baseline: Eq. (13) fixes T=I_Nds, so the numerical comparison does not separate FIM-specific gains from the suboptimality of the no-FIM transmit covariance.","rationale":"I read the paper's central claim as the assertion that FIM geometry morphing provides ISAC gains that a non-morphing metasurface cannot provide. The mathematical core — the FPDD channel model, the waveform I/O matrices, and the closed-form gradients — appears internally coherent. The angle-dependent phase coupling y_b sinθ sinφ in Eq. (4) is a genuine distinction from electronic RIS/SIM phase coefficients, and the optimization over y is sensible. However, the load-bearing point is not the far-field constant-AoA approximation (Footnote 1), which is a standard and likely acceptable assumption; rather, it is the numerical evidence offered for the 'unattainable' claim. The paper fixes T=I_Nds in Eq. (13) for all cases, including the no-FIM baseline. Because the rate is concave in T, the no-FIM baseline is artificially handicapped. A non-morphing system with optimal transmit covariance could be much closer to, or even above, the FIM-optimized rate. The paper also does not compare against an electronic RIS/SIM with optimized angle-independent phase coefficients. Such a baseline has at least the same number of per-element degrees of freedom as the FIM (one complex phase vs. one real displacement), and its response is angle-independent; only an experiment can show whether the FIM's angle-dependent phase diversity yields a real advantage. Without these baselines, the central 'unattainable' claim is not demonstrated. This is consistent with the reader's CONDITIONAL verdict, though my identified weak point differs from the reader's named weakest assumption. I therefore recommend no change to the verdict: the paper remains conditionally viable pending a fair baseline comparison.","tokens_in":8794,"tokens_out":15751,"duration_ms":167930,"concrete_test":"Reproduce Fig. 1(a) (N=16, N_T=N_R=4, P=2, same channel realizations) with two added baselines: (i) non-morphing UPA with y=0 and transmit covariance T optimized via projected gradient ascent on (12a)-(12c) over T (or water-filling for the unconstrained case); (ii) non-morphing metasurface with per-element angle-independent phase coefficients w_T, w_R at TX/RX, optimized by gradient ascent over these coefficients, with the same sensing penalty. Plot the achievable rate vs SNR for optimized FIM (T=I) and these two baselines. If the optimized-FIM curve does not exceed both baselines by at least the margins reported in Fig. 1, then the 'unattainable with non-morphing metasurfaces' claim is not validated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that FIM shape optimization yields 'gains unattainable with non-morphing metasurfaces' is not supported by the numerical comparison. In Section IV, Eq. (13), the transmit covariance is fixed to T≈I_Nds for all cases, including the 'no FIM' baseline. Since the achievable rate in (12a) is concave in T, optimizing T for the non-morphing channel H0 can only increase the rate relative to the reported T=I curve. Thus Fig. 1 does not establish a gain over a non-morphing array that uses its own available DoFs (e.g., water-filling or digital precoding). Furthermore, the claim that the shape-gradients vanish under an electronic RIS/SIM model does not imply that an optimized RIS/SIM cannot achieve comparable or better rates; the gradient just has different parameters. The paper does not simulate a RIS/SIM with optimized angle-independent per-element phase coefficients, which is the natural equal-DoF non-morphing baseline. If the FIM-optimized rate does not exceed that baseline, the central claim's 'unattainable' qualification fails.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a doubly-dispersive MIMO channel model for flexible intelligent metasurfaces (FIMs), where each metasurface element can move along the normal direction. The element positions enter the array response phase nonlinearly, and the authors derive unified input-output relations for OFDM, OTFS, and AFDM. They formulate a rate-maximization problem with a sensing constraint and solve it via projected gradient ascent using closed-form gradients of the mutual information with respect to the FIM displacement vectors. Numerical results claim that optimized FIM shapes yield significant rate and sensing gains over non-morphing configurations.","tokens_in":9133,"tokens_out":5894,"duration_ms":59545,"significance":"If the model is physically valid and the optimization is meaningfully benchmarked, this work introduces a new degree of freedom for metasurface-aided ISAC, extending prior RIS/SIM frameworks. The unified waveform treatment and the closed-form gradient derivation are useful contributions; the algebra in Eqs. (16)-(24) is internally consistent and appears correct. The paper also provides a reproducible algorithmic recipe (Algorithm 1) for shape optimization. However, the numerical validation does not yet support the headline claim of gains 'unattainable with non-morphing metasurfaces,' and the physical premise of the channel model is in question. The paper is therefore promising but requires substantial additional evidence before the central claim can be accepted.","major_comments":[{"comment":"The transmit covariance is fixed to T≈I_Nds for all cases, including the no-FIM baseline. Since the achievable rate in (12a) is concave in T, optimizing T for the fixed non-morphing channel H0 would yield a rate at least as high as the reported T=I curve. The figure therefore does not establish that FIM morphing produces gains 'unattainable with non-morphing metasurfaces.' To support the claim, the authors should compare against a non-morphing baseline that uses its own available DoFs—e.g., water-filling over the singular values of H0, or a conventional RIS/SIM with optimized per-element phase coefficients under the same number of tunable parameters. The current comparison conflates the benefit of FIM geometry with the suboptimality of the fixed transmit covariance for the no-FIM case.","section":"Section IV-A, Eq. (13) and Fig. 1"},{"comment":"The model assumes that path parameters (AoAs/AoDs, delays, Dopplers) remain fixed while element positions vary over [-λ, λ]. A displacement of up to one wavelength is not 'slight'; it can change the local scattering geometry, alter path delays by up to a full carrier period, and affect the effective angles, especially for non-planar deformation or near-field conditions. The derived gradients in Eqs. (21)-(24) capture only the phase variation of a fixed far-field steering vector, not any geometry-induced changes in the channel's angles, delays, or Doppler. The authors should either restrict the morphing range to a small fraction of λ where the constant-angle assumption is quantitatively justified, or extend the model to include displacement-dependent path parameters and verify that the gradients still hold (or derive corrected gradients). As written, the central premise that the channel c","section":"Footnote 1 and Eq. (5)"},{"comment":"The problem is stated as joint optimization of T, y_T, and y_R, but in (13) T is fixed to identity. This is not an achievable-rate maximization over all resources; it is a shape optimization under a specific covariance. The abstract and introduction claim 'achievable rate maximization' without this caveat. Please either solve the joint problem (e.g., alternating optimization with water-filling for T) or explicitly rephrase the contribution as 'rate maximization over FIM shapes for a fixed i.i.d. transmit covariance.' This also affects the interpretation of Fig. 1, as noted above.","section":"Section IV-A, Eqs. (12)-(13)"},{"comment":"The sensing results in Fig. 2 are a single realization with no error bars, no quantitative estimation metrics (e.g., RMSE vs. SNR, probability of resolution), and no sensitivity analysis. The covariance approximation in Appendix A requires cross-terms to vanish via orthogonality conditions 'met in practice,' but the paper does not verify these conditions for the simulated waveform lengths, subcarrier spacings, and angular separations. If the cross-terms are non-negligible, the MUSIC spectrum may be degraded, and the claim that FIM optimization 'substantially impacts sensing quality' is not quantitatively supported. Please provide statistical trials, a quantitative sensing metric, and a verification of the orthogonality assumptions for the actual parameters.","section":"Section V-B and Appendix A"}],"minor_comments":[{"comment":"The symbol \\check H_p is defined as \\tilde h_p b_{R:p} b^H_{T:p}, but the notation is introduced abruptly in (7). A brief definition of \\check H_p before its use would improve readability.","section":"Eq. (7) and surrounding text"},{"comment":"The auxiliary variables \\gamma_{R:p} and \\gamma_{T:p} are defined with in/out angles; consider adding a small table or glossary of symbols to avoid confusion between transmit and receive angles.","section":"Eqs. (18a)-(18b)"},{"comment":"The statement that rates across waveforms are 'nearly identical ... consistent with Shannon's capacity formula' is imprecise because the effective channels \\bar H differ per waveform. The similarity likely stems from the specific simulated parameters; please clarify the reasoning.","section":"Section V-A, first paragraph"},{"comment":"The claim that OFDM exhibits slightly lower sidelobes is hard to discern from the plotted spectra; consider using a zoomed inset or a numerical sidelobe-level metric.","section":"Fig. 2"},{"comment":"Reference [15] is duplicated (it is the same as [9]). Please merge or renumber.","section":"References"},{"comment":"The sensing QoS threshold Ψ is used in the problem formulation but never appears in the simulations. Please specify how Ψ is set, or state that the penalty is used without an explicit threshold.","section":"Section IV-A, Eq. (12c)"},{"comment":"The algorithm stops after a fixed number of iterations i_GD; consider adding a convergence criterion based on the gradient norm or objective change, and state whether the penalty weight β is fixed or adaptively increased.","section":"Algorithm 1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript relies heavily on same-group prior work (refs. [1,5,6,13,14,19]), and the note that an extended version was published in IEEE TWC raises a novelty-disclosure question: the present submission must clearly delineate its incremental contribution over [1]. The technical core (gradients and unified waveforms) appears sound, but the numerical benchmark is inequitable and the physical assumption in Footnote 1 is dubious. If the authors can supply a fair non-morphing baseline and a defensible morphing-range justification, the paper may be publishable after major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Let me give you my quick take. This is a workmanlike paper that has already appeared in longer form in TWC [1], per its own footnote. So as a standalone submission it's largely a duplicate. If you need the full story, read [1].\n\nWhat's actually good here: the FIM parameterization of the array manifold is a genuine departure from RIS/SIM phase models, and the unified I/O treatment for OFDM, OTFS, and AFDM under one Kronecker structure is neatly done. I checked the shape-gradient derivation in Sec. IV-B; the calculus is consistent, and the gradients do vanish when y is fixed. That part is solid.\n\nThe soft spots are in the evidence, not the algebra. First, the rate \"maximization\" fixes transmit covariance to the identity for every case, including the no-FIM baseline. Since the achievable rate is concave in T, a non-morphing system with water-filling could close much of the reported gap. The claim of gains \"unattainable with non-morphing metasurfaces\" is therefore not established by the simulations. They also never run a RIS/SIM with optimized phase coefficients as an equal-DoF control, so the comparison is tilted. Second, the sensing evidence is a single MUSIC spectrum with no error bars or repeated trials; it's illustrative, not convincing. Third, the physical assumption in footnote 1 — that FIM displacements up to λ do not alter path parameters — is hand-waved. Moving elements by a wavelength might change far-field scattering geometry; the model keeps angles, delays, and Dopplers fixed. That could be acceptable in a narrowband far-field setting, but it needs a real justification, not a footnote.\n\nThe citation pattern is heavily self-referential, which is common in this group but doesn't help the perception of incremental novelty. The circularity concern is real but not fatal: the channel model is built so that y affects the array response, so optimizing y must help. That's a modeling choice, not an error.\n\nBottom line: if this were the first and only version of the FIM-DD channel work, I'd send it to a knowledgeable referee — the math is careful and the concept is worth a serious look. But with [1] already published, the preprint adds little. I would not spend referee time on this duplicate. For a reading group, it might serve as a digestible entry point, but cite [1] for the actual result.","headline":"Competent but redundant: the paper's own front matter says an extended version already appeared in TWC, and the headline gains rest on a fixed-identity transmit covariance and no equal-degree-of-freedom RIS/SIM baseline.","tokens_in":9609,"tokens_out":6542,"would_cite":false,"duration_ms":65629,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Optimizing the physical shape of flexible metasurfaces is a new degree of freedom for joint sensing and communications, delivering rate and sensing gains that phase-only reconfigurable surfaces cannot achieve.","keywords":["flexible intelligent metasurface","doubly-dispersive channel","integrated sensing and communications","MIMO","shape optimization","OFDM","OTFS","AFDM"],"falsifier":"Measure the angle of arrival of a known source from a flexible array whose elements are displaced by ±λ relative to their nominal planar positions; if the measured steering-vector phases deviate from the constant-angle prediction beyond noise, the derived shape-gradients describe a different system than the physical one.","tokens_in":8712,"feed_emoji":"📡","tokens_out":4920,"duration_ms":46300,"temperature":0.7,"pith_summary":"The paper introduces a channel model for doubly-dispersive MIMO systems in which flexible intelligent metasurfaces (FIMs) at the transmitter and receiver can bend their elements along the normal axis. The key idea is that element displacement enters the array response nonlinearly, coupling surface geometry to every path's angles, so the shape itself becomes an optimization variable. The authors extend the model to OFDM, OTFS, and AFDM waveforms, derive closed-form gradients of the achievable rate and sensing quality with respect to element positions, and optimize the rate under a sensing constraint via projected gradient ascent. Numerical results show that optimizing the surface shape yields about 2.5 dB gain over no FIMs and another 2 dB over randomly shaped FIMs, and that optimized shapes give clean MUSIC peaks for angle-of-arrival estimation. The shape gradients vanish identically for fixed or phase-only metasurfaces, indicating the optimization is genuinely new to geometry morphing.","feed_headline":"Morphing surface shapes add ~4.5 dB to ISAC rate","feed_subtitle":"Optimizing the physical shape of a flexible metasurface yields rate and sensing gains that phase-only reconfigurable surfaces cannot match.","key_machinery":"The FIM array response b(y, φ, θ) — the steering vector whose b-th entry contains the phase term exp(j(2π/λ)(x_b sinθcosφ + y_b sinθsinφ + z_b cosθ)) — is the load-bearing object: it makes element displacement y_b an optimization variable nonlinearly coupled to the elevation and azimuth angles of every path. The closed-form shape-gradients (21)–(24), derived via Kronecker-product differentials, turn that model into a projected-gradient-ascent algorithm. The constraint y_b ∈ [y_min, y_max] with morphing range ±λ fixes the feasible deformation space.","core_discovery":"The central claim is that a FIM-parameterized doubly-dispersive (FPDD) MIMO channel model, in which the array response b(y, φ, θ) depends nonlinearly on each element's normal displacement y_b through the product y_b sinθ sinφ, correctly captures the effect of surface morphing on both communication and sensing. From this model the paper derives a unified I/O structure for OFDM, OTFS, and AFDM, and closed-form shape-gradients (Eqs. 21–24) that describe how mutual information and a sensing-quality penalty change when a single element moves. These gradients vanish under a purely electronic RIS/SIM model, so the optimization freedom is claimed to arise exclusively from FIM geometry. Simulations t","pith_inferences":["Since the constant-angle approximation is the fragile premise (see weakest_assumption_plain), a natural extension is an iterative model that re-estimates angles, delays, and Dopplers from the deformed geometry; that model would likely show smaller gains at morphing ranges approaching a wavelength and at wider angular spreads.","The shape-gradient machinery is not limited to metasurfaces: any antenna whose element positions are mechanically reconfigurable — piezoelectric, liquid-metal, or origami arrays — could use the same optimization, though actuation speed may limit high-mobility use.","The reported gains come from a point-to-point system with P = 2 or P = 5 scatterers and NT = NR = 4; scaling to larger arrays or richer multipath may shift the communication–sensing trade-off, so the ~4.5 dB figure is an existence proof rather than a universal bound.","The sensing evaluation relies on a simplified covariance model that assumes orthogonal transmit steering or per-snapshot waveform orthogonality; if correlated scatterers violate those conditions, MUSIC performance would degrade, offering a concrete stress test."],"forward_implications":["Shaping the FIM geometry adds about 2.5 dB to achievable rate compared with a rigid (no-FIM) surface, and another ~2 dB when going from random to optimized shapes, across OFDM, OTFS, and AFDM.","Because the shape gradients vanish identically for fixed or phase-only metasurfaces, the reported rate and sensing gains are specific to geometry morphing and cannot be replicated by electronic RIS/SIM tuning.","Optimized FIM geometry yields well-isolated MUSIC peaks for angle-of-arrival estimation, whereas randomly shaped FIMs produce merged or spurious peaks and no-FIM surfaces fail to isolate scatterers.","Although achievable rates for OFDM, OTFS, and AFDM are nearly identical under the same physical channel, waveform choice still matters for bit error rate in high mobility, with OTFS/AFDM retaining inter-carrier interference resilience.","The unified I/O formulation means the same optimization framework applies to all three waveforms, so the geometry gains are not tied to a particular modulation scheme."],"fun_headline_variants":["Shape-shifting metasurfaces boost ISAC rate by 4.5 dB","Flexible metasurface shapes unlock 4.5 dB in ISAC","Morphing surface geometry adds 4.5 dB for ISAC","Why flexible metasurfaces beat phase-only in ISAC","ISAC gains 4.5 dB from morphing surface shapes"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The model assumes that moving FIM elements by up to one wavelength does not change the angles, delays, or Doppler shifts of the propagation paths, so the entire morphing effect is captured by the array response term alone.","fun_headline_variants_meta":{"raw":{"variants":["Shape-shifting metasurfaces boost ISAC rate by 4.5 dB","Flexible metasurface shapes unlock 4.5 dB in ISAC","Morphing surface geometry adds 4.5 dB for ISAC","Why flexible metasurfaces beat phase-only in ISAC","ISAC gains 4.5 dB from morphing surface shapes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000199,"raw_usage":{"total_tokens":1180,"prompt_tokens":691,"completion_tokens":489,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":435,"completion_tokens_details":{"reasoning_tokens":395}},"tokens_in":435,"tokens_out":489,"duration_ms":4497,"temperature":1.0,"reasoning_tokens":395,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T13:00:48.104990+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the angle of arrival of a known source from a flexible array whose elements are displaced by ±λ relative to their nominal planar positions; if the measured steering-vector phases deviate from the constant-angle prediction beyond noise, the derived shape-gradients describe a different system than the physical one.","supporting_citations":[],"review_version":1}