{"id":"21105640-47c9-4707-9898-dfbda0396d8f","arxiv_id":"2607.27525","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"In viscoelastic truss networks driven near resonance, the mass distribution that maximizes energy dissipation decays from the driven joint with a length scale set by the material's intrinsic attenuation length.","lead":"The paper extends a graph-based spectral method for elastic truss networks to linear viscoelastic materials, enabling fast computation of energy dissipation in large disordered networks. Using gradient optimization, it finds that the best way to distribute material in a driven viscoelastic truss is to concentrate mass near the source, with the decay length set by the material's attenuation length.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central ℓρ–ℓ scaling is asserted from a single trial-averaged plot with no fitted exponent, no error bars, and no evidence that the gradient runs reached the claimed optimum despite a rugged landscape.","rationale":"The reader identified the axial-only, pin-jointed assumption as the weakest assumption and recommended CONDITIONAL. I agree the paper deserves CONDITIONAL, but I locate the more load-bearing weakness elsewhere: the central quantitative claim (ℓρ ∝ ℓ) is supported by a single trial-averaged curve with no error bars, no fitted exponent, and no demonstration that the gradient-based optimizer reached a meaningful optimum in a landscape the authors themselves describe as highly rugged. This is more directly tied to the central claim than the bending limitation, which the authors explicitly acknowledge and which can be framed as a scope restriction. If the scaling were to change with more trials, a global optimizer, or a larger network, the abstract's design principle (b) would be misleading. The concrete test I propose—per-trial error bars, a fitted exponent with confidence intervals, a larger lattice, and a check against BC variation—would settle whether the scaling is real. I keep the verdict UNCHANGED because the flaw is one of missing quantitative support, not of demonstrated incorrectness, and it is addressable with additional computation and analysis. Credit is due for the exact analytical benchmark (Fig. 3), the closed-form dissipation expression, and the robustness checks under geometric perturbation (Figs. S10–S11), but these do not substitute for the missing error analysis on the main scaling plot.","tokens_in":20283,"tokens_out":12332,"duration_ms":109686,"concrete_test":"For each of the 20 values of |ξr| in Fig. 8, run at least 50 independent optimizations (or a stochastic global optimizer such as CMA-ES on a subset of parameters) and compute ℓρ for every trial individually. Report mean ± standard deviation and fit log ℓρ = a + b log ℓ over the nonsaturated regime (e.g., ℓ < Lnet/5) with bootstrap confidence intervals. Also repeat on a larger lattice (e.g., 15×15) and with the two boundary conditions; if b differs from 1 by more than the fit uncertainty, or if trial-to-trial spread is comparable to the trend, or if the scaling shifts with system size, then the asserted ℓρ ∝ ℓ design principle is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central design principle (b) is the scaling ℓρ ∝ ℓ between the optimal mass-decay length and the material attenuation length (Fig. 8B). This claim is load-bearing because the abstract, Section III C, and the Conclusion all present it as the main quantitative result. However, the evidence is only a visually guided line through trial-averaged data from 10 optimization runs per ξr, with no standard deviations, no fitted power-law exponent, and no statement of the scaling regime or goodness of fit. The paper itself notes in the Conclusion that 'the optimization landscape becomes highly rugged with many resonant solutions requiring the use of stochastic optimization techniques,' which raises the possibility that the reported 'optimal' architectures are local optima whose mass distribution reflects the optimizer's trajectory, not a true design principle. In addition, the ℓρ definition uses bins of width Lnet/(3Nseg) ≈ 0.2 in the units of the lattice, comparable to the smallest ℓ values in the sweep, so resolution may contaminate the small-ℓ end of the scaling. Because the central claim is quantitative and the supporting figure lacks error analysis, the claim is not yet established to the precision implied by the abstract. This is addressable, but it should be fixed before the design principle is stated as a general result.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper extends a graph-Laplacian spectral method for elastic truss networks to linear viscoelastic networks, retaining the exact continuum rod dynamics while reducing the problem size to the number of joints. The authors derive the network Laplacian for a standard linear solid, obtain a closed-form expression for the energy dissipated per rod (Eq. S18), and benchmark the implementation against the analytical single-rod solution. They then use random sampling and gradient-based optimization (with a fixed material cost) of the rod cross-sectional areas in 1D and 2D triangular trusses driven harmonically near a global resonance. The main reported findings are that random area redistribution typically lowers dissipation relative to a uniform network, while optimized architectures are source-weighted, loopy, and have a mass distribution with a decay length ℓρ that is claimed to scale with the material attenuation length ℓ = 1/(2|ξr|); at small ℓ the optimal architecture is claimed to be insensitive to far-field boundary conditions.","tokens_in":20551,"tokens_out":7700,"duration_ms":68762,"significance":"The framework is a genuine methodological contribution: the graph-Laplacian formulation with exact rod solutions and the closed-form dissipation expression (Eq. S18) are elegant, and the 1D benchmark reproduces the analytic solution exactly. The complexity reduction from element-level to joint-level unknowns is potentially valuable for large-network topology optimization. If the ℓρ–ℓ scaling were established quantitatively, it would provide a concrete, falsifiable design principle for additively manufactured viscoelastic metamaterials. However, as presented, the central quantitative claim rests on a single trial-averaged plot without error bars or a fitted power law, so the significance is currently prospective rather than demonstrated.","major_comments":[{"comment":"The central claim that the optimal mass-decay length scales with the material attenuation length is not quantified. The solid line in Fig. 8B is a guide to the eye; there is no fitted exponent, no confidence interval, and no measure of trial-to-trial variation, even though each point is an average over only Ntrials = 10 optimizations. Because design principle (b) and the abstract state this as a quantitative result, the authors should fit ℓρ = A ℓ^β in the resolved regime, report β with uncertainty and goodness of fit, and overlay error bars or a shaded region representing the spread over independent optimization runs.","section":"Section III C, Fig. 8B"},{"comment":"The paper's own Conclusion states that 'the optimization landscape becomes highly rugged with many resonant solutions requiring the use of stochastic optimization techniques,' yet the reported optima are the maximum of only 10 gradient-descent runs from random initializations (Fig. 7 caption). Without evidence that these runs approach the global maximum, the phrase 'optimal mass distribution' may describe the optimizer's trajectory rather than a true design principle. The authors should report the distribution of Qtotal and ℓρ over the 10 runs and validate the scaling on a subset of ξr values with a more global search method (e.g., simulated annealing or a substantially larger number of restarts).","section":"Section III B, Fig. 7; Conclusion"},{"comment":"The definition of ℓρ as the position where the cumulative mass c(z) reaches 1−e^(−1) presumes that the mass profile decays exponentially, but the paper does not test this assumption. Moreover, the binning used to build ρnet(z) (3Nseg = 30 bins over the network length, with Nseg = 10) gives a bin width comparable to the smallest ℓ values in the sweep (ℓ ≈ 0.16 for |ξr| = 3.14), so the small-ℓ end of the scaling may be resolution-limited. The authors should fit an exponential (or other functional form) to ρnet(z) and report the goodness of fit, and demonstrate that ℓρ is converged with respect to bin resolution or restrict the scaling claim to ℓ values well above the bin width.","section":"Section III C, Fig. 8A"},{"comment":"The design principle is derived under the pin-jointed axial-rod assumption, but the optimized architectures are loopy and the paper presents the principle without that qualification. Since real additively manufactured trusses often have rigid joints and non-negligible bending stiffness, the authors should either demonstrate (even in a simple test case) that the ℓρ–ℓ scaling is unchanged when bending elasticity is included, or explicitly restrict the design principle to axial-dominated structures. This would clarify the scope without requiring a new framework.","section":"Section IV (limitations); design principle (b)"}],"minor_comments":[{"comment":"The supplementary information title contains a typo: 'met amaterials' should be 'metamaterials'.","section":"SI title"},{"comment":"The statement that each rod is 'discretized into Nseg = 500' conflicts with the elsewhere-claimed exactness of the rod dynamics; clarify that Nseg is used only for post-processing the dissipation profile along the rod, not for solving the dynamics.","section":"Section II C, Fig. 3 caption"},{"comment":"The axes labels do not state whether Qtotal is normalized or what units it is reported in; the text cites values such as 5×10^(−10) but no normalization or dimensional basis is given.","section":"Fig. 4C and Fig. 5B"},{"comment":"The sentence about the energy dissipated by two rods under identical displacement boundary conditions equaling that of a single rod with the sum of their areas is confusing and should be rewritten for clarity.","section":"Section III A"},{"comment":"The paper sets γ = 1 without discussing the sensitivity of the results to the cost exponent; a brief statement on how γ affects the optimization or whether the scaling persists for γ ≠ 1 would be useful.","section":"Section III B, Eq. (17)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is potentially publishable, but the central quantitative claim (ℓρ–ℓ scaling) is not yet supported with error bars, a fitted exponent, or a demonstration that the reported optima are reliable. The three issues identified in the major comments are addressable within the manuscript's scope, so I recommend major revision rather than rejection. I would also encourage the authors to make the code and data available to allow independent reproduction of Fig. 8B."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Let me cut to it. The real contribution is the viscoelastic graph Laplacian: they extend their earlier elastic framework [36] to linear viscoelastic rods with exact continuum rod dynamics, and the per-rod dissipation formula (Eq. 14) is closed-form and benchmarks exactly against the single-rod analytical solution (Fig 3). That part is solid, and the fact that problem size scales with joints rather than discretization points genuinely matters for optimizing large disordered trusses. The new physical insight — that optimal area distributions taper from the driven joint with length scale ℓρ that tracks the material attenuation length ℓ = 1/(2|ξr|) — is believable and nicely connected to the physics of stress-wave penetration.\n\nBut the scaling claim is the soft spot. Fig 8B is a trial-averaged curve with a guide-to-the-eye line: no standard deviations, no fitted power-law exponent, no goodness-of-fit, and no statement of the fitted range or slope. Given the optimizer runs on a landscape the paper itself calls 'highly rugged,' the reported maxima may be local optima whose mass distribution reflects the optimizer's trajectory. That doesn't kill the claim — the robustness tests under geometric perturbation (Figs S10-S11) support it qualitatively — but it means the abstract's 'decays with the attenuation length scale' is ahead of the evidence. The ℓρ definition via 1−e−1 also presumes an exponential cumulative mass profile; if the true profile is not exponential, the metric could be measuring something else. I'd want to see per-trial scatter, a fitted exponent with confidence interval, and ideally the same scaling extracted from a different threshold or from fits of ρnet(z) rather than a single cumulative-mass crossing.\n\nThe axial-only restriction is flagged honestly in the conclusion; bending could alter the loopy architectures but the paper doesn't pretend to cover it. The BC-independence claim is milder in the body ('less sensitive') than in the abstract ('independent'), and the supporting figure is representative, not quantitative. Those are minor.\n\nRandom-network results (negative skew, thick-at-source for high dissipation) are nice but not the main event. No code/data, which is a pity for a computational methods paper; availability would materially strengthen a revision.\n\nWho's it for: people working on architected/vibration-damping trusses, and anyone who wants an efficient spectral solver for viscoelastic networks. It deserves a serious referee — the framework is solid and the scaling is plausible but needs quantitative hardening. I'd send it out, with a request for error bars and a fitted exponent on the central figure.","headline":"A clean spectral framework for viscoelastic truss networks with a plausible design principle, but the headline scaling needs error bars and a fitted exponent before it becomes a stated law.","tokens_in":21076,"tokens_out":1744,"would_cite":true,"duration_ms":16590,"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":"For resonant driving, the best damping trusses taper at the material's attenuation length.","keywords":["viscoelastic network metamaterials","energy dissipation","truss network","graph Laplacian","standard linear solid","attenuation length","topology optimization","gradient-based design"],"falsifier":"Recompute the optimal area distributions for the same triangular network using rods or beams that can bend, keeping the same resonant drive and cost constraint; if the optimal mass decay length $\\ell_\\rho$ no longer tracks the axial attenuation length $\\ell=1/(2|\\xi_r|)$, the central design principle is limited to axially loaded trusses.","tokens_in":20071,"feed_emoji":"🛠️","tokens_out":6354,"duration_ms":53071,"temperature":0.7,"pith_summary":"This paper asks how to arrange material in a network of viscoelastic rods so that it absorbs the most mechanical energy when driven at a steady frequency, and it proposes a design rule. Treating each rod's cross-sectional area as a design variable under a fixed total material cost, the authors show that the optimal arrangement concentrates mass near the driven joint and tapers away, with a decay length set by the material's intrinsic attenuation length. The work also introduces a graph-Laplacian spectral solver that retains full continuum rod dynamics while scaling with the number of joints rather than element discretization points, making large-network optimization feasible. A sympathetic reader would care because the result turns a costly finite-element search into a length-scale design principle for vibration damping and impact protection.","feed_headline":"Best damping trusses taper at the material's attenuation length","feed_subtitle":"The design rule: thick rods near the source, then thin at a rate set by the material's absorption depth.","key_machinery":"The central object is the frequency-dependent, complex-valued network Laplacian $\\Leftrightarrow\\mathbf{D}$ that relates joint forces to joint displacements in Fourier space; it is assembled by solving the uniaxial continuum wave equation exactly within each rod and enforcing force balance and velocity compatibility at pin joints. Its entries involve the dispersion function $\\xi(\\omega) = \\sqrt{2\\pi(i\\omega)^3\\rho \\tilde{c}(\\omega)}$, which for a standard linear solid becomes $\\xi(\\omega) = i\\omega\\sqrt{(\\rho/E)\\,(1+i\\omega\\tau_\\epsilon)/(1+i\\omega\\tau_\\sigma)}$. The real part of $\\xi$ sets the attenuation length $\\ell=1/(2|\\xi_r|)$ and the imaginary part sets the spatial oscillation wavelength, and the dissipation of each rod is computed from a closed-form integral of $|\\tilde{\\sigma}_n(z)|^2$. Using this Laplacian, the optimization gradient with respect to rod areas is evaluated in a problem whose matrix size scales with the number of joints rather than the number of element discretization points, which is what makes gradient-descent optimization of cross-sectional areas feasible.","core_discovery":"For a viscoelastic truss network driven harmonically near a global standing-wave resonance, the cross-sectional area distribution that maximizes total dissipated energy under a fixed material budget is source-weighted and loopy: thick rods near the driven joint give way to progressively thinner rods, with cycles retained to keep the network rigid. The key quantitative claim is that the characteristic decay length of this optimal mass distribution, $\\ell_\\rho$, is proportional to the material's attenuation length $\\ell = 1/(2|\\xi_r|)$, where $\\xi_r$ is the real part of the complex wavenumber of the standard-linear-solid material. When $\\ell$ is small compared with the system size, the optimal architecture becomes essentially independent of boundary conditions at walls far from the source, because stress waves die out before reaching them. The authors also find that random redistribution of areas generally lowers dissipation relative to a uniform baseline, and that good dissipative structures resemble transport-optimized flow networks in their tapering but differ by remaining loopy.","pith_inferences":["If bending stiffness is added to the rods, the optimal architecture may shift, but the length-scale principle could survive in a modified form; this is testable with beam-based solvers.","The similarity to flow-network optimization suggests a broader analogy: transport-like objectives with a fixed source and fixed cost may generically produce source-tapered architectures, while mechanical rigidity forces loops where flow networks branch.","A local adaptation rule, such as thickening rods in proportion to local strain amplitude, might converge to these optimized architectures without global computation, which would connect the design principle to biological remodeling.","For applications, the result implies that graded-strut lattices, rather than uniform lattices, are the better damping topology near resonances, and that the grading should be set by material attenuation rather than by structural geometry alone."],"forward_implications":["Because dissipation depends on the area distribution, a designer can tune a viscoelastic structure's damping purely by rearranging material volume, keeping the same material composition.","Near a resonant drive, the optimal structure is a source-weighted, loopy gradient, so the design rule is to place thick elements near the excitation and taper away at the attenuation length scale.","For short attenuation lengths, far-wall boundary conditions barely matter, so a single optimized architecture can serve across different support configurations.","The $\\ell_\\rho \\propto \\ell$ scaling gives a predictive, material-only input, the attenuation length, for deciding where mass should be placed in a damping network.","Because the graph-Laplacian solver scales with the number of joints, the same optimization approach can be applied to larger and more disordered networks than conventional finite-element approaches allow."],"supporting_citations":[{"why":"Supplies the elastic graph-Laplacian framework that this paper extends to viscoelastic rods.","marker":"[36]"},{"why":"Supplies the Boltzmann superposition formulation of linear viscoelasticity used in the constitutive equation.","marker":"[38–40]"},{"why":"Supplies the Lagrange multiplier update used to enforce the fixed material-cost constraint during gradient optimization.","marker":"[41]"},{"why":"Provides the flow-network dissipation-minimization context that motivates comparing optimized architectures with tapering in transport networks.","marker":"[30]"},{"why":"Supplies the tree-like gradient architecture of dissipation-minimizing flow networks that the paper contrasts with its loopy mechanical optimum.","marker":"[33]"}],"fun_headline_variants":["Truss damping peaks when rods taper at material's attenuation length","Optimal damping trusses thicken near source, taper with absorption depth","Design rule: looped rods taper at material's attenuation length","Efficient dissipative trusses taper from source at absorption depth","Loopy, source-thick rods: optimal damping follows attenuation length"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The framework assumes every rod deforms only along its own axis at pin joints, with no bending stiffness; if bending contributes significantly in real trusses, the computed dissipation and the claimed length-scale scaling could change.","fun_headline_variants_meta":{"raw":{"variants":["Truss damping peaks when rods taper at material's attenuation length","Optimal damping trusses thicken near source, taper with absorption depth","Design rule: looped rods taper at material's attenuation length","Efficient dissipative trusses taper from source at absorption depth","Loopy, source-thick rods: optimal damping follows attenuation length"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000553,"raw_usage":{"total_tokens":2637,"prompt_tokens":946,"completion_tokens":1691,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":562,"completion_tokens_details":{"reasoning_tokens":1602}},"tokens_in":562,"tokens_out":1691,"duration_ms":11461,"temperature":1.0,"reasoning_tokens":1602,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:23:42.623361+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the optimal area distributions for the same triangular network using rods or beams that can bend, keeping the same resonant drive and cost constraint; if the optimal mass decay length $\\ell_\\rho$ no longer tracks the axial attenuation length $\\ell=1/(2|\\xi_r|)$, the central design principle is limited to axially loaded trusses.","supporting_citations":[],"review_version":2}