{"id":"c1aaf5f6-74a8-4c9f-b65e-14887e05a5c6","arxiv_id":"1908.09815","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"Backward stimulated Brillouin scattering is observed for the first time in fully cladded silicon nitride waveguides, with a measured gain spectrum and the first reported |p12| = 0.047 for Si3N4.","lead":"Researchers report the first measurement of stimulated Brillouin scattering in silicon nitride waveguides, a process that limits optical power in photonic chips. The work also gives the material's photoelastic constant and explains why silicon nitride chips can handle high optical power, which matters for integrated nonlinear photonics.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Observation is credible, but extracted |p12| and intrinsic gain carry an unquantified model error: the FEM fit uses literature Si3N4 acoustic constants plus a uniform 220 MHz shift, and Supplementary S3 explicitly sets simulation error to zero.","rationale":"The reader's CONDITIONAL verdict is the right call. I agree with the weakest assumption: the FEM fit depends on literature Si3N4 acoustic parameters and on a 220 MHz shift that the authors themselves ascribe to parameter mismatch. My stress-test refines the concern in two ways. First, the quoted ±0.004 on p12 covers only calibration uncertainties; the simulation error is explicitly assumed negligible in Supplementary S3, even though p12 is defined by fitting that simulation to data. Second, a uniform frequency shift is not a valid substitute for parameter sensitivity, because the HBAR peaks (A–D in Fig. 3) have different spatial characters and would respond non-uniformly to changes in layer thicknesses or sound velocities; the multi-peak matching is thus less constrained than a single scalar shift suggests. The intrinsic gain also depends on a separate, unvalidated assumption that material damping in this LPCVD film is exactly 5× silica, with the rest of the 390 MHz linewidth assigned to leakage. I do not see an internal inconsistency in the TIM measurement, and the spectral separation from the silica patchcord peak supports the first-observation claim, so the verdict should not move to REJECT or ACCEPT-as-is. The paper should be published with the derived p12 and intrinsic gain explicitly labeled as model-dependent fits, and with the promised Zenodo data/code so others can rerun the sensitivity analysis. Hence the reader's CONDITIONAL verdict is unchanged.","tokens_in":15404,"tokens_out":10458,"duration_ms":111848,"concrete_test":"Re-run the COMSOL acoustic eigenmode sweep with Si3N4 Young's modulus and density varied over reported ranges (e.g., E = 250–300 GPa, rho = 3000–3200 kg/m3) and with top SiO2 thickness varied ±5%, then compare the simulated multi-peak spectrum to Fig. 3(b) without applying a global frequency shift. The HBAR assignment and p12 fit are robust only if the same eigenmode ordering and spacing reproduce all measured peaks for a contiguous parameter set; if the required shift becomes mode-dependent or the ordering changes, the fit is not unique. Additionally, refit p12 with the simulation-induced uncertainty included, and report a model-error term alongside the current ±0.004 experimental error.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central observation—backward SBS near 25 GHz in a fully cladded Si3N4 waveguide—is well supported: the TIM method separates the signal from the 11 GHz silica patchcord peak, and the authors quantify stray backgrounds. I do not dispute the observation itself. The load-bearing weakness is in the derived constants that appear in the abstract and the strongest claim. |p12| = 0.047 ± 0.004 is obtained by adjusting p12 until FEM-computed Brillouin peak heights match the measured spectrum. Supplementary S3 propagates only the experimental uncertainties (PS, PP, detector responsivity, waveguide length) and states: 'It is assumed that the additional error caused the simulation as well as the estimation of acoustic Q-factor ... is negligible.' That assumption is load-bearing because the same FEM model required a 220 MHz uniform frequency shift to match the measured eigenfrequencies, which Supplementary S4 attributes to a 'slight mismatch of parameters in the simulation, such as Young's modulus or density.' A single global shift does not test the mode assignment: if E or rho is off, the spacing and ordering of the hybrid HBAR modes (peaks A–D in Fig. 3) change non-uniformly, so the peak-to-eigenmode matching used for the p12 fit is not uniquely constrained. The intrinsic gain 7×10−13 m/W is even more model-dependent: it removes the phonon-leakage half of the linewidth by assuming ΓM,Si3N4 = 5 ΓM,SiO2 via a ν_B^2 scaling. If the material damping of this specific LPCVD film differs, the leakage correction and the derived intrinsic gain change. The paper discloses these limitations, but they are absent from the quoted error bars.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports the first observation of backward stimulated Brillouin scattering (SBS) in fully cladded Si3N4 waveguides. Using a triple-intensity-modulation scheme to suppress Fabry-Pérot cavity noise, Kerr-induced probe modulation, and Raman background from fiber patchcords, the authors measure a Brillouin gain spectrum near 25 GHz with a main peak of (8±1)×10^-14 m/W and a linewidth of 390 MHz. They attribute the multi-peak structure to hybridization of the waveguide acoustic mode with high-overtone bulk acoustic resonances (HBARs) of the silica cladding, supported by FEM simulations. From fits of the simulated gain to the measured spectrum, they estimate the Si3N4 photoelastic constant |p12| = 0.047±0.004, and from Eq. (3) they estimate an intrinsic Si3N4 Brillouin gain of 7×10^-13 m/W and an SBS threshold of 87 kW in the 5-mm waveguide.","tokens_in":15734,"tokens_out":3243,"duration_ms":37209,"significance":"If the central claims hold, this is a valuable first measurement of backward SBS in a technologically important platform. The measurement technique itself is a notable experimental contribution: the TIM scheme quantitatively addresses three distinct noise sources that would otherwise bury the small Brillouin signal, and the direct comparison with silica-patchcord SBS provides an internal calibration. The derived quantities, however, are more model-dependent than the abstract suggests: |p12| is obtained by fitting an FEM model whose acoustic parameters are taken from the literature and whose eigenfrequencies require a 220 MHz correction, and the 'intrinsic' gain relies on an assumed splitting of the measured linewidth into material damping and phonon leakage. These caveats do not undermine the observation, but they need to be propagated into the error bars and clearly stated in the abstract and conclusions.","major_comments":[{"comment":"The reported uncertainty |p12| = 0.047 ± 0.004 is propagated only from experimental parameters (PS, PP, ρpd, L), while Supplementary S3 explicitly states that 'the additional error caused by the simulation as well as the estimation of acoustic Q-factor ... is negligible.' This assumption is load-bearing because |p12| is extracted by adjusting the FEM-simulated Brillouin gain to the measured spectrum. Supplementary S4 further states that the simulated spectrum had to be shifted by 220 MHz because of 'a slight mismatch of parameters in the simulation, such as Young's modulus or density.' The authors should either quantify how uncertainties in the acoustic parameters propagate into |p12| and the intrinsic gain, or explicitly reframe these values as model-dependent estimates rather than measurement results.","section":"Supplementary S3 and S4"},{"comment":"The intrinsic Si3N4 Brillouin gain of 7×10^-13 m/W is not directly measured. It is obtained by inserting the fitted |p12| into Eq. (3) together with the assumed material damping ΓM,Si3N4 = 5 ΓM,SiO2 and an overlap of η = 1. The measured linewidth of 390 MHz only determines the total damping ΓM + ΓL; the partition into equal material and leakage contributions relies on the ν_B^2 scaling relation and is not independently verified for this LPCVD film. If the film's acoustic damping differs from silica-scaled values, the estimated intrinsic gain changes accordingly. Please state this dependence explicitly or provide a sensitivity analysis over plausible ΓM values.","section":"Discussion, Eq. (3)"},{"comment":"The mode assignment underlying the HBAR interpretation rests on a single uniform 220 MHz frequency shift applied to the simulated spectrum. If the true Young's modulus or density of the Si3N4 film differs from the literature values used in the simulation, the relative spacings and ordering of the hybridized peaks A–D would change non-uniformly, so the peak-to-eigenmode matching used to fit |p12| is not uniquely constrained. A concrete robustness test—repeating the eigenmode calculation and p12 fit for, e.g., ±5% variations in E and ρ—would materially strengthen the central claim.","section":"Fig. 3(b) and Supplementary S4"}],"minor_comments":[{"comment":"The sentence 'the Brillouin gain spectrum exhibits an unusual multi-peak structure resulting from hybridization with with high-overtone bulk acoustic resonances' contains a duplicated 'with' that should be removed.","section":"Abstract"},{"comment":"The symbol L is used both for the waveguide length and for the lineshape function L(ν∆), which is confusing; consider using L_wg for the length or a different symbol for the lineshape.","section":"Eq. (2)"},{"comment":"The caption states that simulation results show 'eigenfrequency and gain for each eigenmode,' but the simulated spectrum is shifted by 220 MHz relative to the measurement; mentioning this shift in the main text and caption would improve transparency.","section":"Fig. 3 caption"},{"comment":"In the uncertainty table, the entry for L is described as 'waveguide effective length, including inverse nanotapers' with value '5 +0/−0.3 mm'; defining this quantity as L_eff in Eq. (12) would avoid ambiguity with the waveguide physical length.","section":"Supplementary S3"}],"recommendation":"major_revision","confidential_remarks":"The experimental observation is credible and the measurement methodology is a real strength. The main issue is that the abstract's headline numbers—|p12| = 0.047 ± 0.004 and the intrinsic gain 7×10^-13 m/W—are presented with uncertainties that exclude the dominant model error, while the supplementary text explicitly assumes that simulation error is negligible. This is fixable by adding a sensitivity analysis for the acoustic parameters and by cautiously rewording the abstract. I do not see grounds for rejection, but the revision should be substantive rather than cosmetic."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take. The core observation is solid: backward SBS near 25 GHz in fully cladded Si3N4, cleanly separated from the silica patchcord peak, with a useful new technique (triple intensity modulation) that kills the Fabry-Perot Kerr background and SRS. The multi-peak HBAR structure is genuinely new and the FEM mode assignment is plausible. Credit where due: the measurement is careful, the noise analysis is quantitative, and the paper discloses its own limitations.\n\nThe soft spot is exactly where the stress test lands. The |p12| = 0.047 ± 0.004 is not a direct measurement: it is adjusted until FEM peak heights match the spectrum, and the error bar only propagates experimental uncertainties in the detection chain, explicitly assuming simulation error is negligible. That assumption is load-bearing because the same simulation required a 220 MHz global shift to align eigenfrequencies, which the authors attribute to a mismatch in Young's modulus or density. A single global shift doesn't test the mode assignment; if E or rho are off, the spacing and ordering of hybrid HBAR modes change non-uniformly. So the fitted p12 carries an unquantified model error that is not in the ±0.004. The intrinsic gain 7×10−13 m/W is even more model-dependent: it removes phonon leakage by assuming a ν_B^2 scaling for material damping and assigning half the linewidth to leakage. That's a reasonable estimate, but it is an estimate, not a measurement.\n\nI don't think this sinks the paper. The observation—the actual first observation of SBS in Si3N4—is credible and important for the platform. The derived constants should be framed as estimates consistent with the data under a specific model, not as measured values with those error bars. The paper already says much of this, but the abstract presents |p12| with error bars without the caveat. Referees should ask for that reframing and for the promised Zenodo data/code, since the fit is the crux.\n\nThis deserves peer review. The measurement technique and the observation are real advances, and the flaws are in the interpretation of fitting outputs, not in the core experiment. My verdict would be: accept after revision, with changed framing of p12 and intrinsic gain.","headline":"First credible backward SBS measurement in Si3N4 waveguides, but the headline photoelastic constant and intrinsic gain are fitting outputs with unquantified model error.","tokens_in":16334,"tokens_out":1594,"would_cite":true,"duration_ms":15912,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"First measurement of backward stimulated Brillouin scattering in silicon nitride waveguides reports a 25 GHz gain peak, a photoelastic constant of |p12| = 0.047 ± 0.004, and an SBS threshold near 87 kW.","keywords":["stimulated Brillouin scattering","silicon nitride","integrated photonics","photoelastic constant","high-overtone bulk acoustic resonances","Brillouin gain","nonlinear optics","acousto-optic interaction"],"falsifier":"Measure the Brillouin gain spectrum of a Si3N4 waveguide in which the top silica-air boundary is removed (e.g., a suspended or differently capped waveguide) or the cladding thickness is varied; if the multi-peak structure persists without the reflecting boundaries, the HBAR interpretation is wrong, and if the fitted |p12| changes with cladding, the photoelastic constant estimate is not intrinsic. Alternatively, measure the Si3N4 acoustic velocity independently (e.g., by picosecond ultrasonics) and check whether the 220 MHz simulation shift vanishes when using the measured sound velocity.","tokens_in":15216,"feed_emoji":"🔬","tokens_out":4569,"duration_ms":38572,"temperature":0.7,"pith_summary":"This paper reports the first observation of backward stimulated Brillouin scattering (SBS) in silicon nitride (Si3N4) waveguides fully cladded in silica. It measures the Brillouin gain spectrum and finds an unusual multi-peak structure caused by the hybridization of the acoustic mode with high-overtone bulk acoustic resonances of the silica cladding. It estimates the intrinsic Si3N4 Brillouin gain at 25 GHz as 7×$10^{-13}$ m/W and the photoelastic constant magnitude |p12| = 0.047 ± 0.004. Because SBS normally limits optical power, the high SBS threshold of about 87 kW explains why Si3N4 waveguides can handle high optical power, which is central for integrated nonlinear photonics such as soliton microcombs.","feed_headline":"First Brillouin scattering seen in silicon nitride waveguides","feed_subtitle":"The 25-GHz gain peak explains why silicon nitride handles high optical power without Brillouin breakdown.","key_machinery":"The load-bearing machinery is a triple intensity modulation (TIM) measurement technique combined with 3D finite-element simulations of the acoustic eigenmodes. TIM cancels the Fabry-Perot cavity noise, Kerr-effect-induced background, and Raman background by using two counter-propagating, π-phase-shifted pump beams locked to a multiple of the cavity free spectral range, so that only the pump that is resonant with the probe contributes Brillouin gain. The simulations include the full waveguide cross-section with silica-air and silica-silicon boundaries and a perfectly matched layer, and they reproduce the multi-peak spectrum as hybridization with high-overtone bulk acoustic resonances (HBARs) of the silica cladding; the simulated eigenfrequencies are shifted by 220 MHz to match the measured spectrum.","core_discovery":"The central claim is that backward SBS occurs in the Si3N4 material itself, not in the silica cladding, and that its gain spectrum carries fingerprints of the cladding's finite thickness. The measured main Brillouin gain peak of (8±1)×$10^{-14}$ m/W at 25 GHz corresponds to an acoustic velocity of 10.5 km/s, consistent with literature values for Si3N4. The multi-peak structure is reproduced by finite-element simulations that treat the entire chip cross-section, showing that each peak is an acoustic supermode formed by hybridization with high-overtone bulk acoustic resonances reflecting from the top silica-air and bottom silica-silicon boundaries. Fitting the simulated peak heights to the measured spectrum yields |p12| = 0.047 ± 0.004, and the derived intrinsic gain of 7×$10^{-13}$ m/W is about 30 times smaller than silica's.","pith_inferences":["The cladding-dependent HBAR hybridization suggests a design lever: by changing the SiO2 cladding thickness or the substrate boundary, one could tune the Brillouin spectrum to a single dominant line or suppress SBS entirely, which is not explicitly proposed by the paper.","The triple intensity modulation technique could be applied to other low-gain integrated platforms where Fabry-Perot and Kerr backgrounds mask weak Brillouin signals.","If the photoelastic constant is confirmed by an independent method, the 33-fold reduction relative to silica implies that materials engineering of p12, rather than just acoustic confinement, dominates SBS gain suppression in Si3N4."],"forward_implications":["Because the SBS threshold in Si3N4 is estimated at 87 kW, silicon nitride waveguides can handle much higher optical powers than silica fibers, supporting applications like soliton microcombs and supercontinuum generation.","The measured 25 GHz Brillouin shift is the largest reported on integrated platforms, offering a distinctive frequency scale for on-chip microwave photonics.","The reported photoelastic constant |p12| = 0.047 ± 0.004 provides the first reference for Si3N4 photoelasticity, enabling quantitative simulation of acousto-optic interactions in future Si3N4 devices.","The multi-peak gain structure arising from cladding boundaries implies that the Brillouin response of Si3N4 waveguides is not an intrinsic material property alone but is engineered by the surrounding cladding geometry."],"supporting_citations":[{"why":"Supplies the literature acoustic velocity of Si3N4 used to compare with the measured 10.5 km/s from the Brillouin shift.","marker":"[58]"},{"why":"Provides the physical basis for high-overtone bulk acoustic resonances (HBARs) that the paper invokes to explain the multi-peak gain spectrum.","marker":"[59]"},{"why":"Establishes the finite-element simulation framework for computing SBS gain from acoustic eigenmodes, which the paper adapts to the full chip cross-section.","marker":"[50]"},{"why":"Earlier demonstration of a Brillouin laser in a Si3N4 waveguide where SBS occurred in the silica cladding, providing the contrast that the present work observes SBS in the Si3N4 core itself.","marker":"[40]"},{"why":"Classic reference for SBS as an optical power limitation in waveguides and fibers, which motivates the paper's discussion of Si3N4 high-power handling.","marker":"[45]"},{"why":"Supplies the threshold formula used to estimate the 87 kW SBS threshold from the measured gain.","marker":"[61]"}],"fun_headline_variants":["First Brillouin scattering measured in silicon nitride waveguides","Backward SBS observed in fully cladded Si3N4 waveguides","Silicon nitride's 25 GHz Brillouin gain peak measured for first time","Intrinsic SBS in Si3N4 waveguides explains high optical power handling"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The finite-element simulation relies on literature values for Si3N4's Young's modulus (280 GPa) and density (3100 kg/m3) and is shifted by 220 MHz to match the measured spectrum; if those acoustic parameters are wrong for this film, the mode assignment, the fitted |p12|, and the derived gain would be wrong.","fun_headline_variants_meta":{"raw":{"variants":["First Brillouin scattering measured in silicon nitride waveguides","Backward SBS observed in fully cladded Si3N4 waveguides","Silicon nitride's 25 GHz Brillouin gain peak measured for first time","Intrinsic SBS in Si3N4 waveguides explains high optical power handling"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000627,"raw_usage":{"total_tokens":2885,"prompt_tokens":918,"completion_tokens":1967,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":534,"completion_tokens_details":{"reasoning_tokens":1888}},"tokens_in":534,"tokens_out":1967,"duration_ms":13175,"temperature":1.0,"reasoning_tokens":1888,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:00:06.693589+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the Brillouin gain spectrum of a Si3N4 waveguide in which the top silica-air boundary is removed (e.g., a suspended or differently capped waveguide) or the cladding thickness is varied; if the multi-peak structure persists without the reflecting boundaries, the HBAR interpretation is wrong, and if the fitted |p12| changes with cladding, the photoelastic constant estimate is not intrinsic. Alternatively, measure the Si3N4 acoustic velocity independently (e.g., by picosecond ultrasonics) and check whether the 220 MHz simulation shift vanishes when using the measured sound velocity.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the literature acoustic velocity of Si3N4 used to compare with the measured 10.5 km/s from the Brillouin shift."},{"cited_title":"Marowsky \\ and\\ author G","cited_arxiv_id":null,"evidence_quote":"Provides the physical basis for high-overtone bulk acoustic resonances (HBARs) that the paper invokes to explain the multi-peak gain spectrum."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the finite-element simulation framework for computing SBS gain from acoustic eigenmodes, which the paper adapts to the full chip cross-section."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier demonstration of a Brillouin laser in a Si3N4 waveguide where SBS occurred in the silica cladding, providing the contrast that the present work observes SBS in the Si3N4 core itself."},{"cited_title":"Gyger , author Z","cited_arxiv_id":null,"evidence_quote":"Supplies the threshold formula used to estimate the 87 kW SBS threshold from the measured gain."}],"review_version":1}