{"id":"ec66234e-9336-4347-9da7-814f8a9143eb","arxiv_id":"2506.18636","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"For BIC metasurface infrared sensors, silicon outperforms gold in low-loss environments, gold outperforms silicon above a solvent loss of about kenv = 2e-3, and the difference vanishes in highly lossy solvents.","lead":"This study compares gold and silicon metasurface sensors in liquids with different infrared losses. It finds that silicon works best in air or low-loss liquids, while gold wins once the liquid's infrared absorption is large enough, and both become equally poor in very lossy solvents.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The kenv≈2e-3 crossover is simulated only at fixed solvent real index n=1.33, while the experiments include DMSO with n≈1.48; the paper does not establish that the crossover is invariant to n, so the quantitative roadmap may be solvent-specific.","rationale":"I read the paper as a practical comparison of representative Si and Au BIC metasurfaces for SEIRAS in lossy solvents; the crossover claim is the key novel result. The design uses nearly identical planar geometry with different resonator heights, which is a defensible representative-device comparison, and the gradient metasurface plus θ optimization is a sensible way to handle resonance shifts and radiative loss. The weakest point is the univariate loss sweep: kenv is varied while n is held at 1.33, although the experiments include DMSO with n≈1.48. Since the claimed threshold is quantitative and the experimental data do not densely sample kenv, the threshold's dependence on n is load-bearing. This is the same concern the reader identified, so I agree with the CONDITIONAL verdict; a numerical n-sweep would settle whether the concern actually lands.","tokens_in":11992,"tokens_out":5689,"duration_ms":64463,"concrete_test":"Repeat the numerical sweep underlying Fig. 3d with solvent n = 1.33, 1.40, and 1.48 (or, better, with the wavelength-dependent n and k of H2O, D2O, and DMSO near 1730 cm-1 from the Max & Chapados data cited in Methods), re-optimizing θopt for Si and Au at each kenv, and record the zero crossing of max(AbsSi) - max(AbsAu). If the crossing leaves the range 1e-3 to 5e-3, the 2e-3 threshold is not robust to n and the central crossover claim must be weakened or explicitly recalibrated for each solvent.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim—Si is superior for kenv < 2e-3 and Au for kenv > 2e-3—is extracted from Fig. 3d, which is computed with the solvent real refractive index fixed at n = 1.33 (Section 2.1: 'the real part of the refractive index (n) is held fixed at 1.33'). The experimental confirmation includes DMSO, whose mid-IR refractive index is near 1.48, not 1.33, so Fig. 4d mixes solvents with different n values. The gradient metasurface compensates the resonance-frequency shift caused by n, but the absorbance amplitude and the optimal tilt angle θopt also depend on n: changing n alters the mode's overlap with the analyte layer, the radiative coupling, and the relative field confinement inside Si versus at the Au surface. Nothing in the paper establishes that the zero crossing of max(AbsSi,θopt) - max(AbsAu,θopt) is invariant to n. If the crossover shifts by a factor of 2-3 with n, the stated threshold and the claimed roadmap for solvent-based SEIRAS require recalibration for each solvent, and the experimental agreement in Fig. 4d could be specific to the particular n values of the solvents tested.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper compares the sensing performance of gold (plasmonic) and silicon (dielectric) quasi-BIC metasurfaces for surface-enhanced infrared absorbance spectroscopy (SEIRAS) in lossy liquid environments. Numerically, the authors fix the real refractive index of the environment at n = 1.33 and sweep the imaginary part kenv from 0 to 0.1, simulating a 5 nm PMMA analyte layer on both platforms and optimizing the tilt angle (asymmetry) per platform and environment. They report a crossover at kenv ≈ 2 × 10^-3: silicon yields higher absorbance for lower losses, gold for higher losses, and both converge to poor performance at very high loss. Experimentally, they use dual-gradient metasurfaces in a microfluidic setup with air, DMSO, D2O, and H2O, and find that the measured absorbance difference between Si and Au follows the simulated trend. The paper claims this provides a roadmap for choosing the metasurface material based on solvent losses.","tokens_in":12267,"tokens_out":5379,"duration_ms":54258,"significance":"If the crossover is robust, the result is practically valuable for SEIRAS optofluidics: it provides a quantitative rule for choosing between dielectric and plasmonic BIC metasurfaces as a function of solvent absorption. The dual-gradient metasurface design that compensates spectral shifts while allowing independent control of radiative loss is an elegant and useful experimental tool, and the combination of simulations with microfluidic measurements is a strength. The main limitation is that the quantitative threshold is established only at a single real refractive index (n = 1.33) and the experimental validation mixes solvents with different real indices, so the generality of the numerical roadmap is not yet fully supported.","major_comments":[{"comment":"The central quantitative claim—the crossover at kenv ≈ 2 × 10^-3, with Si superior below and Au superior above—is computed with the real part of the solvent refractive index held fixed at n = 1.33. The text gives no sweep over n and no argument that the crossover is invariant to n. Since the experiments include DMSO with n ≈ 1.48, and since changing n alters the mode overlap with the analyte, the radiative coupling, and the field confinement in Si versus Au, the threshold may shift with n. Without an n-dependence study (or a perturbation argument), the stated roadmap for solvent-based SEIRAS is only established for n = 1.33 and may require recalibration for solvents with different real indices.","section":"Section 2.1 and Figure 3d"},{"comment":"The experimental confirmation mixes solvents with different real refractive indices: air (n = 1), DMSO (n ≈ 1.48), and D2O/H2O (n ≈ 1.33). The observed negative (Au-better) difference for DMSO could in principle be caused by its higher real index rather than by its loss kenv. To uniquely attribute the crossover to kenv, the simulations should be repeated with the actual complex refractive index dispersions of the measured solvents, or the experiments should include solvents with matched n but differing kenv. As presented, the experimental agreement is qualitative and does not independently validate the quantitative kenv threshold.","section":"Section 2.4 (labeled 2.3) and Figure 4d"}],"minor_comments":[{"comment":"The section titled \"2.3. Experimental comparison of sensing performance\" is numbered the same as the preceding section \"2.3. Numerical comparison of sensing performance for analyte coated metasurfaces\"; the experimental section should be renumbered (e.g., 2.4).","section":"Section numbering"},{"comment":"The text states that the field is \"completely quenched for kenv = 0.3\" and later mentions \"kenv = 0.3\" again, but the relevant figure and the range discussed elsewhere use kenv = 0.03; this appears to be a typo and should be corrected.","section":"Section 2.2, text near Figure 2f"},{"comment":"The sentence \"At kenv = 0.005 and, θ = 100, the maximum reflectance...\" contains a typo: θ should be 10° (the text later refers to \"θ = 20°\" and \"θ = 30°\"), and the comma after \"and\" should be removed.","section":"Section 2.2, paragraph on asymmetry"},{"comment":"The text gives the lattice periods as \"Px = 4000 μm and Py = 2400 μm\" for both metasurfaces; for mid-infrared resonances at ~1730 cm^-1 and a 2 × 2 mm^2 field of view, these values seem to be in nanometers or otherwise misstated, and should be corrected.","section":"Section 2.1, lattice parameters"},{"comment":"The field-enhancement crossover in Figure 2g is reported at kenv > 1.2 × 10^-2, while the absorbance crossover in Figure 3d is at kenv ≈ 2 × 10^-3; the paper should clarify that these are different metrics (mean electric field versus absorbance) to avoid confusion about the crossover location.","section":"Figures 2g and 3d"},{"comment":"The choice to average the 50 highest measured Abs values is not justified; a brief rationale or a sensitivity test showing that the result does not depend on this threshold would strengthen the analysis.","section":"Section 2.4, Figure 4c"},{"comment":"Several references are duplicated: reference 24 is the same as reference 17, reference 25 is the same as reference 31, and reference 35 is the same as reference 37; these should be merged or cross-referenced consistently.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reports a useful and timely comparison, and the dual-gradient BIC metasurface platform is well executed. The main concern is the lack of any test of the dependence of the crossover on the real refractive index of the solvent; this is central to the quantitative roadmap claim. The experimental section also mixes solvents with different real indices, so the confirmation is weaker than it appears. I recommend major revision with a request for an n-sweep in the simulations or a matched-index experimental set, and a re-plotting of the experimental data against the simulation curve for a more quantitative comparison."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper gives a clean, head-to-head comparison of gold and silicon BIC metasurfaces for SEIRAS in lossy liquids, and the central result holds up: silicon wins in low-loss environments, gold once the solvent absorption kenv passes about 2e-3, and they converge at high loss. What's new is that the comparison is done at fixed resonance frequency and optimized radiative coupling via the tilt angle, so the material difference is isolated, and they back it with experimental measurements in air, DMSO, D2O, and H2O. That is a useful dataset for anyone choosing a platform for aqueous or DMSO-based assays.\n\nThe strongest part is the methodology: the dual-gradient design tunes spectral position and asymmetry independently, making the comparison fairer than prior air-only studies. The field-enhancement analysis and the crossover in max(Abs) are internally consistent, and the experimental confirmation, though limited to four environments, follows the simulation trend.\n\nSoft spots, in proportion. The main one is the fixed real refractive index n=1.33 in the numerical sweep. The experiments include DMSO with n≈1.48, and the paper does not check whether the crossover at kenv≈2e-3 moves with n. The gradient metasurface compensates the resonance shift, but absorbance amplitude and optimal tilt angle can also depend on n, so the 2e-3 threshold is best treated as a rule of thumb for water-like solvents rather than a universal calibration. That does not sink the qualitative conclusion, but it limits the strength of the 'roadmap' claim.\n\nTwo smaller issues: the absorbance metric is the mean of the 50 highest measured values per solvent, which is defensible as a best-region estimator but is not justified in the text, and the raw data and analysis code are not deposited despite the data availability statement. The height difference between Au (100 nm) and Si (750 nm) is physically necessary but is a confounder; the authors acknowledge it indirectly but do not discuss how much it affects the comparison. For a fully quantitative design rule, I would want an n-sweep and a sensitivity analysis.\n\nOverall, this is a solid, honest paper. The central crossover is a real design guideline, and the experimental confirmation is a plus. It deserves serious peer review, with the n-invariance question as the main thing to ask the authors to address.\n\nI would bring it to the reading group and would cite it in work on solvent-based SEIRAS.","headline":"A careful material comparison for SEIRAS in lossy solvents with a plausible crossover, but the fixed n=1.33 sweep leaves the quantitative threshold solvent-specific.","tokens_in":12796,"tokens_out":2411,"would_cite":true,"duration_ms":25306,"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":"Neither gold nor silicon is the best metasurface material in every environment: the paper demonstrates a loss-dependent crossover at which silicon BIC sensors hand the lead to gold ones.","keywords":["Bound states in the continuum","SEIRAS","Plasmonic metasurfaces","Dielectric metasurfaces","Infrared sensing","Lossy environment","Mid-infrared spectroscopy","Optical losses"],"falsifier":"Measure absorbance for optimally tilted Si and Au BIC metasurfaces in two solvents with nearly equal $k_{\\mathrm{env}}$ but clearly different real refractive index, e.g. one near $n=1.33$ and one near $n=1.48$. If the sign of $\\mathrm{Abs}(\\mathrm{Si})-\\mathrm{Abs}(\\mathrm{Au})$ at $k_{\\mathrm{env}}\\approx 2\\times10^{-3}$ differs between the two solvents, the crossover is not set by loss alone and the roadmap must be redrawn per solvent; if it is the same, the rule holds.","tokens_in":11795,"feed_emoji":"🔬","tokens_out":7707,"duration_ms":75281,"temperature":0.7,"pith_summary":"This paper asks which metasurface material—gold or silicon—gives stronger molecular sensing signals in realistic infrared solvents. Existing comparisons were done in air, where silicon's high-quality resonances win. The authors show that once the solvent absorbs infrared light, the ranking flips: gold's strong surface fields survive losses, while silicon's fields leak away. There is a crossover near $k_{\\mathrm{env}}\\approx 2\\times10^{-3}$, and in highly lossy solvents both platforms do equally poorly. The practical payoff is a solvent-aware material selection rule for surface-enhanced infrared absorbance spectroscopy (SEIRAS).","feed_headline":"Gold beats silicon for infrared sensing in lossy solvents","feed_subtitle":"Silicon leads in air; gold leads once solvent absorption passes kenv ≈ 0.002; near 0.1 they tie.","key_machinery":"The central object is the quasi-bound-state-in-the-continuum (quasi-BIC) metasurface with a dual gradient: a lateral scaling factor $S$ tunes the resonance frequency continuously across the unit cells, while an elliptical tilt angle $\\theta$ breaks the in-plane symmetry that turns a dark BIC into a radiative mode, setting the radiative loss. Because resonance frequency and radiative loss can be matched between silicon and gold designs, the comparison isolates the effect of intrinsic material losses. The sensing metric is the logarithmic absorbance ratio $\\mathrm{Abs}=-\\log(R_{\\mathrm{analyte}}/R_{\\mathrm{ref}})$ evaluated at the analyte's 1730 cm$^{-1}$ vibrational band.","core_discovery":"On its own terms, the paper claims that no universally superior metasurface platform exists for SEIRAS: the best material is determined by the optical loss of the solvent. Using quasi-BIC metasurfaces where radiative loss can be tuned independently via the resonator tilt angle, the authors find numerically that dielectric silicon metasurfaces give higher analyte absorbance when the environment's imaginary refractive index is below about $2\\times10^{-3}$, gold plasmonic metasurfaces give higher absorbance above that, and the two converge at high losses. Experiments with PMMA-coated metasurfaces in air, D2O, DMSO, and water, delivered through a microfluidic cell, reproduce this ordering with silicon best in air and gold better in the lossy solvents.","pith_inferences":["If the same sweep is repeated with solvents of different real refractive index while keeping $k_{\\mathrm{env}}$ fixed, the crossover threshold may move; DMSO with $n\\approx1.48$ is the natural first test.","The loss-leakage mechanism should be generic: any dielectric resonator whose mode spreads through its volume will quench faster than a metal's surface-localized field, so the ranking likely extends beyond silicon to other high-index dielectrics.","A testable extension is to replace the reflectance-ratio absorbance with a phase-sensitive readout; the material ordering could change because phase responses depend differently on radiative and absorptive losses."],"forward_implications":["For dry or low-loss use, silicon BIC metasurfaces give the stronger PMMA absorbance, so they are the natural default when air or dry-gas measurements dominate.","For solvents like D2O, DMSO, or water, gold BIC metasurfaces outperform silicon once $k_{\\mathrm{env}}$ exceeds about $2\\times10^{-3}$, making gold the default for in-situ bioassays.","At very high solvent loss ($k_{\\mathrm{env}}\\approx0.1$), the two platforms converge to similarly weak signals, so material choice becomes secondary to optimizing geometry and readout.","Tuning the tilt angle $\\theta$ shifts the crossover point, giving sensor designers an explicit knob to keep a preferred material competitive in a given solvent."],"supporting_citations":[{"why":"Dual-gradient continuous spectral and coupling-strength encoding; this is the method used to keep resonances on the PMMA band while varying the solvent.","marker":"[42]"},{"why":"Gradient high-Q dielectric metasurfaces for broadband sensing; supplies the gradient design for the silicon metasurface.","marker":"[43]"},{"why":"Plasmonic bound states in the continuum; establishes the tilted-ellipse qBIC geometry for gold and radiative-loss control via asymmetry.","marker":"[25]"},{"why":"Pixelated high-Q metasurfaces for in situ biospectroscopy in lossy media; provides the earlier lossy-environment result this work extends to a material comparison.","marker":"[32]"},{"why":"Dielectric versus plasmonic sensing comparison; supplies the earlier air-based comparison that this paper extends to lossy solvents.","marker":"[38]"},{"why":"Optical dielectric function of gold; provides the gold material data used in the simulations.","marker":"[45]"},{"why":"Infrared spectra of H2O and D2O; provides the solvent loss constants used to set $k_{\\mathrm{env}}$ values.","marker":"[46]"},{"why":"Complex refractive indices of polymers in the infrared; provides the PMMA analyte model.","marker":"[47]"}],"fun_headline_variants":["Silicon metasurfaces win in air, gold in lossy solvents","Solvent loss decides: dielectric BICs in air, plasmonic in liquids","Gold vs silicon metasurfaces: crossover set by solvent absorption","Best BIC metasurface platform? It depends on solvent losses","Plasmonic BICs outperform dielectric in lossy solvents, silicon in dry"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The numerical sweep that defines the crossover fixes the solvent's real refractive index at $n=1.33$ while varying only its imaginary part, so real solvents with different real indices could in principle shift the threshold.","fun_headline_variants_meta":{"raw":{"variants":["Silicon metasurfaces win in air, gold in lossy solvents","Solvent loss decides: dielectric BICs in air, plasmonic in liquids","Gold vs silicon metasurfaces: crossover set by solvent absorption","Best BIC metasurface platform? It depends on solvent losses","Plasmonic BICs outperform dielectric in lossy solvents, silicon in dry"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000213,"raw_usage":{"total_tokens":1416,"prompt_tokens":936,"completion_tokens":480,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":552,"completion_tokens_details":{"reasoning_tokens":385}},"tokens_in":552,"tokens_out":480,"duration_ms":4998,"temperature":1.0,"reasoning_tokens":385,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:45:34.807588+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure absorbance for optimally tilted Si and Au BIC metasurfaces in two solvents with nearly equal $k_{\\mathrm{env}}$ but clearly different real refractive index, e.g. one near $n=1.33$ and one near $n=1.48$. If the sign of $\\mathrm{Abs}(\\mathrm{Si})-\\mathrm{Abs}(\\mathrm{Au})$ at $k_{\\mathrm{env}}\\approx 2\\times10^{-3}$ differs between the two solvents, the crossover is not set by loss alone and the roadmap must be redrawn per solvent; if it is the same, the rule holds.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gradient high-Q dielectric metasurfaces for broadband sensing; supplies the gradient design for the silicon metasurface."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Pixelated high-Q metasurfaces for in situ biospectroscopy in lossy media; provides the earlier lossy-environment result this work extends to a material comparison."},{"cited_title":"& Cai, H","cited_arxiv_id":null,"evidence_quote":"Dielectric versus plasmonic sensing comparison; supplies the earlier air-based comparison that this paper extends to lossy solvents."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Optical dielectric function of gold; provides the gold material data used in the simulations."},{"cited_title":"& Chapados, C","cited_arxiv_id":null,"evidence_quote":"Infrared spectra of H2O and D2O; provides the solvent loss constants used to set $k_{\\mathrm{env}}$ values."},{"cited_title":"& Liu, L","cited_arxiv_id":null,"evidence_quote":"Complex refractive indices of polymers in the infrared; provides the PMMA analyte model."}],"review_version":2}