{"id":"81edaa94-9a9f-41b5-bf8c-7e58d9b5d346","arxiv_id":"2607.05351","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":4,"one_line_summary":"A prototype photonic ring resonator device was characterized in the upper H-band, with a quadratic model for suppressed-wavelength progression found to be more consistent with data than a linear model, though depth and FWHM trends remain unresolved.","lead":"This paper measures a prototype photonic ring resonator filter chip across a broader wavelength range (upper H-band, ~1620–1665 nm) than previously tested, and fits polynomial models to predict how the filter's suppression dips shift with wavelength order. A smart generalist might read it to gauge how close this technology is to enabling ground-based near-infrared astronomy free of atmospheric OH emission noise.","discovery_kind":"unclear","skeptic_critique":{"model":"glm-5.2","headline":"The m=6 validation relies on visually inspected dip centers from a blended, unflattened spectrum with no quantified uncertainty; the chi-squared model selection uses laser resolution (10 pm) rather than actual measurement errors, making the statistical significance of the quadratic-linear preference","rationale":"The reader correctly identified the core statistical fragility: 4-point polynomial extrapolation with poor chi-squared values and unquantified validation uncertainties. However, the reader slightly mischaracterized the validation as resting on 'a single experimental data point' — Table 3 shows all five dips have m=6 measurements, and the quadratic-linear model outperforms linear on all five (prediction errors 0.1–1.9 nm vs 7–11 nm for linear). This is qualitatively more compelling than a single-point validation. The real load-bearing question is not whether the quadratic-linear model is better than linear (it clearly is, by both chi-squared ratio and m=6 prediction errors), but whether the m=6 visual measurements are reliable enough to constitute meaningful validation. The paper does not quantify visual inspection uncertainty, and the chi-squared analysis uses instrumental resolution (10 pm) rather than actual measurement uncertainty. This prevents formal statistical assessment of the model's predictive power. The theoretical justification (Eq. 5–6) is qualitative — n₀ and n₁ are never independently measured, so the connection between the fitted polynomial and the physics is not parameter-free. The paper is transparent about these limitations (§5–6), which is appropriate for a prototype characterization. The CONDITIONAL verdict is correct: the qualitative finding (quadratic progression) is well-supported, but the quantitative claims cannot be formally validated without proper uncertainty quantification. No adjustment to the verdict is needed.","tokens_in":11779,"tokens_out":5072,"duration_ms":254819,"concrete_test":"Have 3–5 independent observers measure the dip centers in the m=6 unflattened spectrum (Fig. 5) without knowledge of the model predictions. Compute the standard deviation of their measurements as the visual inspection uncertainty for each dip. Then recompute χ² for the m=6 predictions (quadratic-linear and linear) using these empirical uncertainties. If the visual inspection uncertainty is ≥1 nm for the blended dips (1, 2, 4, 5), the 0.5–2 nm prediction errors become statistically indistinguishable from measurement noise, weakening the validation. If the uncertainty is ≤0.3 nm, the quadratic-linear model's predictive success is robust.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim — that suppressed wavelengths follow a quadratic progression — rests on two pillars: (1) chi-squared comparison favoring quadratic-linear over linear (Table 2), and (2) m=6 predictions matching experimental measurements (Table 3). Both pillars have a shared weakness: the uncertainty is not properly characterized. For pillar (1), σ = 10 pm is the laser resolution, not the actual uncertainty on the DF suppressed wavelengths extracted by the dip-fitting script. The script is documented to fail for depth and FWHM on most dips (§4), and even for the third dip, DF and VI suppressed wavelengths differ by up to ~0.3 nm across orders (Table 1: e.g., m=1: DF=1491.696 vs no VI comparison given, but m=3: DF=1543.763). If the true fitting uncertainty is ~0.1–0.5 nm rather than 0.01 nm, the chi-squared values drop by a factor of 100–2500, and while the ratio between linear and quadratic-linear χ² is preserved (so model selection direction is unchanged), the absolute goodness-of-fit assessment becomes impossible to evaluate. For pillar (2), the m=6 experimental measurements for all five dips come from visual inspection of an unflattened, blended spectrum (Fig. 5) with no uncertainty quoted. The quadratic-linear predictions agree with these measurements to within 0.1–1.9 nm (Table 3), which is clearly better than the linear model's 7–11 nm errors. However, without knowing the visual inspection uncertainty, one cannot assess whether 0.5–2 nm agreement is statistically significant or merely consistent with an unquantified measurement error. The reader slightly overstated the fragility by saying validation 'rests on a single data point' — all five dips have m=6 measurements, and the quadratic-linear model outperforms linear on all five. But four of those five measurements come from blended dips in an unflattened spectrum, making their reliability the actual load-bearing question.","agreement_with_reader":"partial"},"referee_report":{"model":"glm-5.2","summary":"This manuscript characterizes a prototype photonic ring resonator (PRR) filter (ANT-07) for OH suppression in near-infrared astronomy, extending prior lower H-band (1480–1620 nm) measurements into the upper H-band (1620–1665 nm). The authors measure suppressed wavelengths, FWHM, and depth for the first four orders of five dips, fit polynomial models to these data, and extrapolate to the sixth order. They conclude that suppressed wavelengths progress quadratically with order, consistent with a wavelength-dependent group refractive index, while depth and FWHM show no reliable trend. The sixth-order predictions are compared against visual-inspection measurements from an unflattened upper H-band spectrum.","tokens_in":12802,"tokens_out":1191,"duration_ms":193898,"significance":"The paper addresses a practical instrumentation problem—extending PRR-based OH suppression to the upper H-band—and provides a transparent, falsifiable comparison of linear versus quadratic-linear models for wavelength progression. The theoretical motivation via a linear wavelength-dependent group refractive index (Eqs. 5–6) is a reasonable physical extension of the standard FSR formula. The work is incremental but appropriate for a prototype characterization study. The target venue (JAAVSO) appears suitable for this type of instrumentation development paper.","major_comments":[{"comment":"§4, Eq. (4) and surrounding text: The chi-squared statistic uses σ = 10 pm for suppressed wavelengths, described as the laser resolution. However, the actual uncertainty on the DF suppressed wavelengths extracted by the dip-fitting script is not established. The script is documented to fail for depth and FWHM on most dips (§4), and DF vs. VI suppressed wavelengths differ by up to ~0.3 nm in some cases (Table 1 vs. Table 5). If the true fitting uncertainty is ~0.1–0.5 nm rather than 0.01 nm, the absolute chi-squared values in Table 2 (ranging from 3.8 to 14,184) cannot be interpreted for goodness-of-fit. The ratio between linear and quadratic-linear chi-squared is preserved, so the model selection direction is unaffected, but the claim that the quadratic-linear model provides an acceptable fit (as opposed to merely a better-than-linear fit) is not currently supported. The authors should (","section":null},{"comment":"§4, Table 3 and Fig. 5: The sixth-order experimental validation relies entirely on visually inspected dip centers from an unflattened, blended spectrum with no quantified uncertainty. The quadratic-linear predictions agree with these measurements to within 0.1–1.9 nm (Table 3), which is clearly better than the linear model's 7–11 nm errors. However, without any uncertainty estimate on the visual inspection, the statistical significance of this agreement cannot be assessed. The authors should provide at minimum a rough uncertainty estimate for the visual measurements (e.g., based on dip width and blending).","section":null},{"comment":"§5, Eqs. (5)–(6): The theoretical derivation introduces n0 and n1 as parameters in ng(λ) = n0 + n1λ, but these are never fitted or constrained from the data. The quadratic-linear fit to suppressed wavelengths is a purely empirical polynomial fit; it is not derived from or constrained by Eqs. (5)–(6). The claim that the quadratic progression is 'consistent with the theoretical dependence' (§7) would be strengthened if the authors could extract n0 and n1 from their fit and check whether the values are physically reasonable for silicon waveguides, or at least clarify that the theoretical derivation provides qualitative motivation rather than quantitative prediction.","section":null}],"minor_comments":[{"comment":"§1: The abstract and introduction motivate the work with supernova cosmology and OH suppression, but the connection between the specific prototype measurement (orders 1–6 of one device) and the broader astronomical application could be stated more directly.","section":null},{"comment":"§2, Eq. (2): The notation λres,m = mλres = mLng is slightly ambiguous; the relationship between λres and Lng should be stated more explicitly.","section":null},{"comment":"§4, Table 2: The caption refers to 'quad-lin. fit' while the text uses 'quadratic-linear' and §5 uses 'linear-quadratic'. Consistent naming would help.","section":null},{"comment":"§4, Table 5 footnote: The footnote text is garbled ('1585 a · · ·3.3· · ·1.10'). This should be corrected.","section":null},{"comment":"§5: The phrase 'linear-quadratic fit' is used where 'quadratic-linear' is used elsewhere. Consistent naming would help.","section":null},{"comment":"Fig. 5: The green lines indicating dip centers are helpful but the dip numbering in blue could be larger or placed more clearly to improve readability.","section":null},{"comment":"§6: The future work items are reasonable but somewhat lengthy; the licensing/construction paragraph could be shortened or removed as it is not directly relevant to the scientific findings.","section":null},{"comment":"Table 4: Predicted depths for dips 4 and 5 are negative (-0.72 and -0.39 dB), which is unphysical. This should be noted and discussed, or the predictions should be presented without comment.","section":null}],"recommendation":"major_revision","confidential_remarks":"The reader's concerns about the statistical fragility (4 data points, 2 parameters) and the use of laser resolution rather than actual fitting uncertainty are well-founded and constitute the primary basis for the major revision recommendation. The m=6 validation is a genuine independent check and not circular, but its value is limited by the lack of quantified uncertainty on the visual inspection measurements. The paper is a reasonable prototype characterization study but needs to be more honest about the limitations of its statistical claims."},"author_rebuttal":{"model":"glm-5.2","summary":"We thank the referee for a careful and constructive report. All three major comments identify legitimate weaknesses that we can and should address in revision. We agree fully with Comments 1 and 2 and will revise accordingly. For Comment 3, we agree that the theoretical derivation should be clarified as qualitative motivation rather than quantitative prediction, and will revise the text; extracting physically constrained values of n0 and n1 from our data is not straightforward with the current four-point fits, and we explain why below.","responses":[{"response":"The referee is correct on all points. The 10 pm value reflects the laser step size and spectrum granularity, not the uncertainty on the dip-fitting script's extracted suppressed wavelength. The discrepancies between DF and VI suppressed wavelengths (up to ~0.3 nm in Table 1 vs. Table 5) make clear that the true fitting uncertainty is larger than 10 pm, likely in the 0.1–0.5 nm range as the referee suggests. We agree that the absolute chi-squared values in Table 2 therefore cannot be interpreted for goodness-of-fit; only the ratio between linear and quadratic-linear chi-squared is meaningful for model selection. We will revise the manuscript to (1) explicitly state that σ = 10 pm represents the measurement granularity rather than the fitting uncertainty, (2) acknowledge that the true fitting uncertainty is not yet established and is likely orders of magnitude larger, (3) remove or qualify the claim that the quadratic-linear model provides an acceptable fit, replacing it with the more limited claim that the quadratic-linear model is strongly preferred over the linear model, and (4) note that proper goodness-of-fit assessment awaits improved characterization of the dip-fitting script's uncertainties.","revision_made":"yes","referee_comment":"§4, Eq. (4): The chi-squared statistic uses σ = 10 pm (laser resolution), but the actual uncertainty on DF suppressed wavelengths extracted by the dip-fitting script is not established. The script fails for depth and FWHM on most dips, and DF vs. VI suppressed wavelengths differ by up to ~0.3 nm. If the true fitting uncertainty is ~0.1–0.5 nm, absolute chi-squared values cannot be interpreted for goodness-of-fit. The model selection direction is preserved, but the claim that the quadratic-linear model provides an acceptable fit is not supported."},{"response":"We agree. The visual inspection measurements were made on an unflattened spectrum with significant blending, and we did not assign uncertainties to these measurements. We will provide a rough uncertainty estimate in the revised manuscript. Based on the dip widths (FWHM ~2.5 nm for the third dip at sixth order, and broader for other dips) and the degree of blending visible in Fig. 5, we estimate that the visual inspection uncertainty on suppressed wavelength is approximately ±1–2 nm for the third dip (which is relatively isolated) and approximately ±2–4 nm for the more blended dips. We will add these estimates to Table 3 and discuss their implications: the quadratic-linear predictions for the third dip (0.5 nm agreement) are within this uncertainty, while for the more blended dips the agreement (0.1–1.9 nm) is comparable to the estimated uncertainty. We will also note explicitly that the sixth-order validation is qualitative rather than statistically rigorous, and that a proper test requires an improved spectrum (flattened, higher SNR) and automated dip-fitting.","revision_made":"yes","referee_comment":"§4, Table 3 and Fig. 5: The sixth-order experimental validation relies entirely on visually inspected dip centers from an unflattened, blended spectrum with no quantified uncertainty. The quadratic-linear predictions agree to within 0.1–1.9 nm, but without any uncertainty estimate on the visual inspection, the statistical significance of this agreement cannot be assessed. The authors should provide at minimum a rough uncertainty estimate."},{"response":"The referee is correct that the quadratic-linear fit is purely empirical and is not derived from or constrained by Eqs. (5)–(6). We will revise the manuscript to clarify this distinction explicitly: the theoretical derivation in §5 provides qualitative motivation for expecting a quadratic component in the wavelength progression, but the fit itself is an empirical polynomial. We agree that extracting n0 and n1 from the fit and checking their physical reasonableness would strengthen the paper. However, doing so rigorously is not straightforward with our current data. The empirical quadratic-linear fit has three free parameters fitted to four data points per dip, leaving only one degree of freedom. Mapping these empirical coefficients back to n0 and n1 requires knowledge of the ring circumference L (which is set during fabrication but has manufacturing tolerances) and the relationship between the empirical fit parameters and the physical model parameters, which involves the Taylor expansion in Eq. (6). With only four data points and uncertain fitting errors, the resulting constraints on n0 and n1 would be too weak to be physically meaningful. We will state this limitation transparently and frame the theoretical derivation as qualitative motivation. We will also soften the claim in §7 from 'consistent with the theoretical dependence' to 'qualitatively consistent with the expectation from a wavelength-dependent group refractive index.'","revision_made":"partial","referee_comment":"§5, Eqs. (5)–(6): n0 and n1 are introduced as parameters in ng(λ) = n0 + n1λ but are never fitted or constrained from the data. The quadratic-linear fit is purely empirical. The claim that the quadratic progression is 'consistent with the theoretical dependence' would be strengthened if the authors could extract n0 and n1 and check whether they are physically reasonable, or at least clarify that the theoretical derivation provides qualitative motivation rather than quantitative prediction."}],"tokens_in":11612,"tokens_out":1213,"duration_ms":160063,"standing_objections":[]},"desk_editor":{"model":"glm-5.2","letter":"Here's my read on the PRR filter paper (Hermann et al.). The bottom line: this is an honest instrumentation characterization that extends the collaboration's prior lower-H-band work on ANT-07 into the upper H-band (1620–1665 nm), and proposes a quadratic model for how suppressed wavelengths progress with order. The qualitative finding is probably right. The quantitative case is weak in ways the authors are mostly transparent about, but a referee should still push on them. What's genuinely new: the m=6 upper-H-band measurements on ANT-07, and the quadratic-linear model motivated by including a linear wavelength-dependent group refractive index (Eq. 5–6). The theoretical motivation is sound — if ng depends on λ, the FSR picks up higher-order terms, and a quadratic correction to the suppressed-wavelength progression follows naturally. The m=6 validation for the third dip is a real independent check: the quadratic-linear prediction (1633.48 nm) lands within 0.5 nm of the measured value (1634.00 nm), while the linear model is off by ~9 nm. That's a meaningful difference. The paper is also commendably honest about where the dip-fitter failed and where they fell back to visual inspection. Now the soft spots, in rough order of importance. The biggest issue is the uncertainty treatment. The chi-squared analysis uses σ = 10 pm, which is the laser resolution — not the actual uncertainty on extracted dip centers. The dip-fitting script failed for depth and FWHM on most dips, and the m=6 measurements for four of five dips come from visual inspection of an unflattened, blended spectrum with no uncertainty quoted. If the real uncertainty on those visually inspected centers is ~0.1–1 nm (which seems likely given the blending visible in Fig. 5), the absolute chi-squared values in Table 2 are meaningless, and one cannot assess goodness of fit. The model selection direction (quadratic-linear beats linear) is probably robust since it depends on the ratio, but the authors should say this explicitly rather than reporting chi-squared values that imply a precision they don't have. Second, fitting polynomials to 4 data points and extrapolating to m=6 is inherently fragile. The quadratic-linear model has 2 effective parameters on 4 points (2 dof), and the cubic fit over-constrains to near-zero chi-squared, which the authors correctly flag as overfitting. This is a small-data problem and there's no way around it without more measurements, but the paper should frame its claims as suggestive rather than confirmatory. Third, the theoretical justification for the linear dispersion model (Eq. 5) rests on private communication from a PhD student in the collaboration. The functional form is plausible, but n0 and n1 are never independently measured or constrained — they enter as free parameters absorbed into the polynomial fit. The depth predictions for dips 4 and 5 are negative, which is unphysical, and the authors don't discuss this. The stress-test note correctly identifies the uncertainty quantification as the load-bearing weakness. I'll note one place where I think the reader's report slightly overstates the fragility: it says validation 'rests on a single data point,' but all five dips have m=6 measurements and the quadratic-linear model outperforms linear on all five. The real concern is that four of those five measurements come from blended, unflattened spectra — their reliability is the question, not their existence. This paper is for instrumentation specialists working on OH suppression or integrated photonics for astronomy. It's a modest step within an established program, not a breakthrough. The device is early-stage and not yet telescope-tested. It deserves a serious referee who can demand proper uncertainty quantification, framing of statistical claims, and at minimum a discussion of the unphysical depth predictions. I'd recommend conditional acceptance pending revisions on the uncertainty treatment and claim calibration.","headline":"Honest prototype characterization with a reasonable qualitative finding but statistically fragile quantitative claims. The quadratic model for suppressed-wavelength progression is probably directionally correct, but the evidence is thin.","tokens_in":12721,"tokens_out":1496,"would_cite":false,"duration_ms":145360,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"glm-5.2","headline":"Quadratic law predicts where ring resonator filters suppress light","keywords":["photonic ring resonator","OH suppression","near-infrared astronomy","free spectral range","group refractive index","H-band","supernova cosmology","atmospheric emission"],"falsifier":"If the suppressed wavelength at order 6 for dips other than the third were measured with a calibrated dip-fitter and deviated from the quadratic-linear prediction by more than the ~1 nm agreement seen for the third dip, the quadratic model would be undermined.","tokens_in":11810,"feed_emoji":"🔵","tokens_out":1423,"duration_ms":128016,"temperature":0.7,"pith_summary":"Photonic ring resonators (PRRs) are tiny circular waveguides that can filter out specific wavelengths of light, making them attractive for suppressing bright atmospheric OH emission lines that plague ground-based near-infrared astronomy. This paper tests a prototype PRR device (ANT-07) across the lower H-band (1480–1620 nm) and attempts to predict its performance in the upper H-band (1620–1800 nm) by fitting polynomial models to the suppressed wavelengths, depths, and linewidths (FWHM) measured at four lower orders. The central finding is that the suppressed wavelength of a PRR progresses with resonance order according to a quadratic function, rather than a linear one. This quadratic behavior is consistent with the theoretical free spectral range formula when the group refractive index of the waveguide material is allowed to vary linearly with wavelength, rather than being held constant as in the simplest theory. The quadratic-linear model yields lower chi-squared values than a pure linear fit across all five dips, and the one experimentally measured upper H-band data point (the third dip at order 6, measured at 1634.00 nm) matches the quadratic-linear prediction (1633.48 nm) more closely than the linear prediction (1624.93 nm). No reliable trends were established for how dip depth or FWHM evolve with order.","feed_headline":"Quadratic law predicts where ring resonator filters suppress light","feed_subtitle":"Silicon photonic filters for removing sky glow in near-infrared astronomy follow a quadratic wavelength rule tied to material dispersion, a ","key_machinery":"The free spectral range (FSR) of a ring resonator, which governs the spacing between consecutive resonance wavelengths; the group refractive index n_g, which determines phase delay in the waveguide and is shown to vary linearly with wavelength for this silicon device; and the resonance condition (φ = 2πm), which selects which wavelengths are suppressed at each integer order m.","core_discovery":"The suppressed wavelengths of photonic ring resonator filters follow a quadratic dependence on resonance order, arising from the wavelength-dependence of the group refractive index in the waveguide material. By replacing the constant group refractive index assumption with a linear wavelength-dependent term (n_g = n_0 + n_1·λ), the free spectral range acquires quadratic corrections that explain the observed curvature in the progression of suppressed wavelengths across orders. This provides a predictive model for placing suppression dips at target OH emission wavelengths across a broader spectral range than the device was originally fabricated for.","pith_inferences":["The quadratic model is fitted to only four data points per dip and validated against a single upper H-band measurement; a denser sampling of orders (e.g., m = 5 and m = 7) would more rigorously test whether the quadratic term suffices or a cubic correction is needed, as the Taylor expansion of the FSR with linear n_g(λ) actually contains cubic terms.","If the linear wavelength-dependence of n_g can be independently measured (e.g., via ellipsometry or interferometry on the same silicon wafer), the quadratic-linear model could become fully predictive rather than fitted, reducing the number of free parameters and strengthening extrapolation confidence.","The dip broadening problem at higher orders suggests that future PRR designs for upper H-band suppression may need smaller ring circumferences or different waveguide materials with lower dispersion to keep FWHM within the ~0.2 nm target for effective OH line suppression."],"forward_implications":["If the quadratic model holds across more orders and devices, PRR filters fabricated for one wavelength band could be predictably characterized and tuned for OH suppression across adjacent bands without fabricating entirely new chips.","The wavelength-dependence of the group refractive index is a critical parameter that must be measured or modeled for each PRR material and geometry; assuming it constant leads to large prediction errors (chi-squared values of 10,000+).","The broadening and blending of dips at higher orders (FWHM grows roughly as λ²) sets a practical upper limit on how many OH lines a single PRR can suppress before adjacent dips become unresolvable.","Reliable extraction of dip depth and FWHM at higher orders remains an unsolved problem; without accurate predictions of these metrics, total OH suppression cannot be guaranteed."],"fun_headline_variants":["Quadratic model predicts OH suppression across photonic filter bands","Photonic ring resonators follow quadratic rule for OH suppression","Curvature in photonic filter suppression tied to material dispersion","Material dispersion shapes quadratic OH suppression in ring resonators","Predicting OH suppression in ring resonators via quadratic wavelength law"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The quadratic model for suppressed wavelength is established from only four lower-order data points per dip and corroborated by a single experimental measurement at order 6 for one dip; the remaining four upper H-band dip locations were read by eye from an unflattened spectrum where dips overlap, so the model's validity across all dips and orders rests on limited evidence.","fun_headline_variants_meta":{"raw":{"variants":["Quadratic model predicts OH suppression across photonic filter bands","Photonic ring resonators follow quadratic rule for OH suppression","Curvature in photonic filter suppression tied to material dispersion","Material dispersion shapes quadratic OH suppression in ring resonators","Predicting OH suppression in ring resonators via quadratic wavelength law","Photonic ring filters obey quadratic law for OH emission suppression"]},"model":"glm-5.2","effort":"high","cost_usd":0.0,"raw_usage":{"total_tokens":930,"prompt_tokens":459,"completion_tokens":471,"prompt_tokens_details":null},"tokens_in":459,"tokens_out":471,"duration_ms":14366,"temperature":1.0,"reasoning_tokens":479,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-07T15:22:24.537350+00:00","model_set":{"reader":"glm-5.2"},"falsifier":"If the suppressed wavelength at order 6 for dips other than the third were measured with a calibrated dip-fitter and deviated from the quadratic-linear prediction by more than the ~1 nm agreement seen for the third dip, the quadratic model would be undermined.","supporting_citations":[],"review_version":1}