{"id":"e4f12f53-02ff-4939-b304-5fd215ca33bf","arxiv_id":"2411.13035","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":4,"one_line_summary":"A simulation study finds that GaAs, InP, and GaSb rib waveguides on sapphire could have propagation losses of 0.32, 0.67, and 0.70 dB/cm at 1330, 1550, and 2000 nm.","lead":"We simulated rib and strip waveguides made of GaAs, InP, and GaSb on sapphire and estimated propagation losses between 0.32 and 0.70 dB/cm at telecom wavelengths. The work supports a proposed all-on-sapphire photonic integrated circuit platform that would combine III-V active devices with silicon-on-sapphire electronics.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline loss numbers are not reproducible from §4 as written: the conversion k = 4.3(λα/4π) with α in dB/cm is a factor ~18.7 larger than the standard k = αλ/(4π×4.343), so every simulated loss value used to support 'acceptable for PICs' is called into question.","rationale":"The reader's weakest-assumption statement already flagged both the suspicious 10 dB/cm interface loss and the dimensionally inconsistent k formula, so there is substantial overlap. However, the reader emphasized the 10 dB/cm typo as the main issue, whereas the more load-bearing concern is the k-conversion formula: if the formula is wrong as printed, then even after correcting 10 dB/cm to 0.10 dB/cm, the simulated losses are scaled by a factor of about 18.7 and the reported numbers cannot be reproduced. The mode maps, single-mode width analysis, and PCF data in Figs. 2–4 are useful and are not affected by this concern, and the paper does give credit for a systematic parametric study. The loss model in §4, however, is the sole quantitative basis for the abstract's headline loss figures and the conclusion that III-V waveguides on sapphire are 'acceptable for PICs.' The correction is straightforward to test with a re-run of the Lumerical simulation using the standard conversion, so the appropriate verdict remains CONDITIONAL rather than outright rejection: the paper should not be accepted until the conversion formula is corrected and the headline losses are recomputed and re-reported.","tokens_in":7895,"tokens_out":10272,"duration_ms":102277,"concrete_test":"Recompute the imaginary refractive index at λ=1550 nm for α=0.05 dB/cm using the standard relation k=αλ/(4π×4.343), then rerun the Lumerical Mode simulation for the InP rib waveguide with this corrected k for sidewall roughness and with the dislocation loss set to 0.10 dB/cm (instead of 10 dB/cm). If the simulated loss changes by more than ±10% from the reported 0.67 dB/cm, the headline values depend on the erroneous conversion; if it does not, the k formula was not used in the way the text suggests. Repeat the same check for GaAs at 1330 nm and GaSb at 2000 nm.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In §4 (Loss analysis), the paper states that 'the attenuation index (k) for the losses due to sidewall roughness and dislocations ... are calculated by using k = 4.3 (λα/4π), where λ and α are the operating wavelength and attenuation constant in dB/cm.' This relation is dimensionally inconsistent. For a loss α expressed in dB/cm, the power attenuation coefficient is α = 4.343 × (4πk/λ), hence k = αλ/(4π × 4.343) ≈ αλ/54.6. The printed expression gives k ≈ 0.342 αλ, which is about 18.7 times too large. Because the imaginary index k is then used in the Lumerical Mode simulation to represent scattering at the sidewalls and the wetting-layer/sapphire interface, the resulting propagation losses scale with k; the claimed 0.32, 0.67, and 0.70 dB/cm values cannot be trusted. Separately, the paper lists the dislocation/interface loss as '10 dB/cm' while reporting total losses below 1 dB/cm, which points to a decimal typo (0.10 dB/cm); even after that correction, the k-conversion error remains. Since these are the only quantitative inputs for the headline loss figures and no measured III-V-on-sapphire data are provided, the central claim that the waveguides are 'acceptable for PICs' is presently unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a simulation study of group III-V (GaAs, InP, GaSb) rib and strip waveguides on sapphire substrates, with AlAs, GaP, and AlSb wetting layers. Using Lumerical Mode simulations, the authors map guided TE/TM modes, identify single-mode, multimode, and cutoff regions, and compute power confinement factors for various aspect ratios. The paper's central quantitative claim is that propagation losses for the GaAs, InP, and GaSb rib waveguides at 1330 nm, 1550 nm, and 2000 nm are 0.32, 0.67, and 0.70 dB/cm, respectively, and that these values are acceptable for photonic integrated circuits on a sapphire platform.","tokens_in":8187,"tokens_out":3312,"duration_ms":34566,"significance":"If the loss estimates were reliable, the study would provide useful design rules for a monolithic III-V-on-sapphire PIC platform, complementing the authors' broader program on sapphire-based photonics. The systematic mode-confinement and single-mode-condition analysis is a genuine contribution and appears internally consistent with the stated refractive-index data. However, the headline loss numbers are not independently derived: they are dominated by assumed sidewall and interface scattering constants taken from other material systems, and the conversion formula used to turn those constants into simulation inputs contains a large numerical error. The lack of any measured III-V-on-sapphire propagation-loss data further limits the strength of the 'acceptable for PICs' conclusion. The paper's value would be much higher if the loss model were corrected, the assumed constants were clearly separated from computed results, and a sensitivity analysis were provided.","major_comments":[{"comment":"The conversion relation k = 4.3(λα/4π), where α is given in dB/cm, is dimensionally inconsistent and numerically wrong by a large factor. For an attenuation constant α_dB in dB/cm, the power attenuation coefficient is α_dB = 4.343 × (4πk/λ), so the correct conversion is k = α_dB λ / (4π × 4.343) ≈ α_dB λ / 54.6. The printed expression gives k ≈ 0.342 α_dB λ, which is about 18.7 times too large. Because the imaginary index k is used as the input to the Lumerical Mode simulation for sidewall roughness and interface defects, the resulting propagation-loss values scale with this erroneous k. The headline numbers 0.32, 0.67, and 0.70 dB/cm therefore cannot be trusted as computed predictions.","section":"§4, Loss analysis"},{"comment":"The paper states that 'loss due to dislocations at nanoscale wet layer/substrate interface was 10 dB/cm' while simultaneously reporting total losses of 0.32 to 0.70 dB/cm. A 10 dB/cm interface-loss contribution cannot be consistent with sub-dB/cm totals unless it is either a typo (likely 0.10 dB/cm) or the loss mechanism is not actually included in the reported totals. This inconsistency must be resolved and the corrected value re-simulated, since this assumed constant directly sets the final loss numbers.","section":"§4, Loss analysis"},{"comment":"The three loss constants used as simulation inputs — 0.05 dB/cm sidewall roughness for rib waveguides, 0.25 dB/cm for strip waveguides, and the 10 dB/cm interface-defect loss — are taken from references on other waveguide material systems and are not measured for III-V-on-sapphire. Since material absorption is stated to be negligible (≤0.01 dB/cm), the reported propagation losses are essentially the assumed scattering constants propagated through the mode solver. The conclusion that the waveguides are 'acceptable for PICs' is therefore circular unless these constants are validated or the results are presented as a scenario-dependent estimate with a sensitivity analysis. The authors should either supply measured loss data or explicitly reframe the claim as conditional on the assumed defect densities.","section":"§4 and Conclusions"},{"comment":"The paper states that 'the single-mode condition in the plot corresponds to a power confinement factor close to or above 80%.' This conflates a threshold on power confinement with the existence of a single guided mode. The single-mode boundary should be determined by counting guided modes, not by a confinement-factor criterion. If the maps in Figures 2 and 3 are based on actual mode counting, the text should say so; if the 80% PCF value is an additional design preference, it should be presented as such.","section":"§3, Figures 2-4"}],"minor_comments":[{"comment":"The word 'systemically' should be 'systematically', and the phrase 'we proposed to use sapphire a s a high-performance PIC platform' contains a spacing typo that should be corrected.","section":"Abstract"},{"comment":"The enumerated rationale in the introduction uses 'i)' twice; the list should be re-lettered or renumbered.","section":"Introduction, Table 1"},{"comment":"The text says the effective index method was used with 'Ansys Lumerical Mode's tool', but Lumerical Mode is a full-vectorial mode solver, not an effective-index-method implementation. The wording should be corrected to avoid an inaccurate description of the simulation method.","section":"§2, Simulation Methods"},{"comment":"Figure 5 plots propagation losses over a broad wavelength range, but the text does not specify whether the plotted quantity is the total loss or a sum of the modeled components, nor does it define the symbols or line styles used for the different polarizations and waveguide types. The figure caption and text should be expanded for reproducibility.","section":"§4"},{"comment":"Reference [28] is listed as IEEE Trans. Electron. Dev. 2020, 32, 1–13; the volume and year appear inconsistent with a 1990s-era GaAs waveguide paper. The bibliographic details should be checked and corrected.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The core problem is not the mode-analysis part, which is plausible, but the loss-prediction section, where a numerical error and a consistency issue undermine the paper's central quantitative claim. The errors appear fixable within the scope of a revised manuscript: the authors can correct the k-conversion formula, re-run the loss simulations, clarify the interface-loss constant, and add a sensitivity analysis over the assumed defect densities. If the corrected losses still support sub-dB/cm performance, the revised paper could be acceptable; if not, the conclusion must be softened. I see no citation or novelty-disclosure concerns beyond the usual expectation that the authors acknowledge that the loss constants are transferred from other platforms."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nRead this one if you care about design studies for alternative PIC platforms. The genuinely new content is the systematic single-mode and power-confinement maps for GaAs, InP, and GaSb rib and strip waveguides on sapphire, including the straight-line single-mode width approximations. That part is a legitimate and useful numerical contribution. The mode-solving work is standard, but it is done carefully and the parametric sweeps give a designer something to use. I would cite that portion in a platform-comparison paper.\n\nThe soft spots are in the loss analysis, and they are load-bearing. The paper claims 0.32, 0.67, and 0.70 dB/cm for the three rib waveguides, but those totals reduce almost entirely to assumed sidewall and interface scattering constants carried over from other material systems, not measured III-V-on-sapphire values. The text lists the dislocation/interface loss as 10 dB/cm while reporting totals below 1 dB/cm; that has to be a typo for something like 0.10 dB/cm, but as written it is a direct internal contradiction. Worse, the conversion k = 4.3(λα/4π) with α in dB/cm is off by roughly a factor of 18.7 relative to the standard relation k = αλ/(4π × 4.343). Since the imaginary index k is what seeds the scattering loss in the simulation, the headline loss numbers cannot be reproduced from the paper as written. This is not a minor formatting issue; it directly undermines the 'acceptable for PICs' conclusion.\n\nAlso note the overstatement in the conclusions: they say low losses were 'proved,' but everything here is simulation with assumed loss parameters. That language should be toned down to 'estimated' or 'modeled.'\n\nMy verdict: conditional, not reject. The single-mode and confinement analysis is independent of the loss-conversion error and stands on its own. The paper is worth serious referee time because the platform concept is plausible and the mode maps are a solid first systematic pass. But the loss section needs a corrected conversion formula, a sensitivity analysis over the assumed scattering constants, and an explicit statement that these are projections pending experimental validation. If the authors fix that, the manuscript becomes a reasonable design-route paper for a niche but real audience.\n\nRecommendation: send it to peer review, but flag the loss-formula error and the 10 dB/cm inconsistency in the review request. A referee should check the corrected loss numbers before publication.\n\nBest,\n[You]","headline":"Useful mode-map design study for III-V-on-sapphire waveguides, but the headline loss numbers are not trustworthy as written because of a bad k-conversion formula and an internal 10 dB/cm inconsistency.","tokens_in":8769,"tokens_out":642,"would_cite":true,"duration_ms":7464,"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":"The paper predicts GaAs, InP, and GaSb rib waveguides on sapphire can reach 0.32–0.70 dB/cm propagation loss, comparable to III-V on silicon.","keywords":["waveguide","sapphire","III-V semiconductors","rib waveguide","propagation loss","photonic integrated circuit","numerical simulation","power confinement factor"],"falsifier":"Fabricate a GaAs/AlAs rib waveguide on sapphire with the simulated dimensions, measure propagation loss at 1550 nm by the cut-back method or from a ring-resonator Q factor, and compare with the predicted 0.32 dB/cm; a measured loss above about 1 dB/cm would show the assumed scattering constants do not transfer to this platform.","tokens_in":7682,"feed_emoji":"💡","tokens_out":7329,"duration_ms":76529,"temperature":0.7,"pith_summary":"The paper proposes and simulates straight rib and strip waveguides made of GaAs, InP, and GaSb on a sapphire substrate, with thin wetting layers (AlAs, GaP, AlSb) to aid epitaxial growth. It finds that single-mode rib waveguides can confine light with power confinement factors near or above 80 percent across broad near- and mid-infrared ranges. Estimated propagation losses at 1330 nm, 1550 nm, and 2000 nm are 0.32 dB/cm, 0.67 dB/cm, and 0.70 dB/cm for GaAs, InP, and GaSb rib waveguides respectively, which the authors say is comparable to reported III-V waveguide losses on silicon. The authors conclude that III-V-on-sapphire waveguides perform well enough to serve as passive building blocks in a monolithic sapphire photonic integrated circuit platform.","feed_headline":"III-V on sapphire: waveguides lose under 1 dB/cm","feed_subtitle":"Simulations put GaAs, InP, and GaSb rib losses at 0.32–0.70 dB/cm, matching III-V on silicon.","key_machinery":"The load-bearing design is the rib waveguide: a III-V core layer (GaAs, InP, or GaSb) on a 40 nm wetting layer (AlAs, GaP, or AlSb) deposited on sapphire, surrounded by air. The paper uses a commercial mode solver based on the effective-index method to map fundamental TE and TM modes as functions of width and wavelength, extract power confinement factors, and identify single-mode, multi-mode, and cut-off regions. Loss is modeled as the sum of material absorption (negligible for free-carrier densities below 1e15 $cm^{-3}$), sidewall scattering (assumed 0.05 dB/cm for ribs and 0.25 dB/cm for strips), and scattering from dislocations at the wet-layer/sapphire interface, converted to an attenuation index through $k = 4.3(\\lambda\\alpha/4\\pi)$.","core_discovery":"The central claim is that group III-V straight rib waveguides on sapphire are low-loss passive elements for photonic integrated circuits, with estimated losses of 0.32 dB/cm (GaAs at 1330 nm), 0.67 dB/cm (InP at 1550 nm), and 0.70 dB/cm (GaSb at 2000 nm). Strip waveguides show larger losses because their stronger horizontal confinement pushes optical field into the substrate. Over the simulated range from 800 nm to 3500 nm, TE polarization loses less than TM, and rib waveguides lose less than strips. These values are obtained by adding material absorption, sidewall-roughness scattering, and scattering from dislocations at the wetting-layer/sapphire interface, and they are comparable to reported III-V waveguide losses on silicon wafers.","pith_inferences":["If the assumed sidewall and interface scattering values are too optimistic for actual epitaxial III-V on sapphire, measured losses could be several dB/cm; a fabricated ring-resonator Q measurement at 1550 nm would settle this.","The same simulation approach could be extended to bends and directional couplers by adding radius-of-curvature and gap parameters, giving a more complete passive-component library for the sapphire platform.","Because GaSb has the highest index contrast with sapphire and the smallest single-mode widths, it may be the best choice for compact mid-infrared circuits, an implication the paper does not draw explicitly."],"forward_implications":["Straight-waveguide results give a starting point for designing bends, splitters, couplers, and ring resonators on the sapphire platform.","Active III-V devices such as lasers, modulators, and detectors grown on the same sapphire substrate can be monolithically integrated with the low-loss passive waveguides.","Operating near 1550 nm avoids the strong two-photon absorption that limits silicon photonics, extending usable power levels.","Because silicon-on-sapphire wafers are available, the photonic layer can be combined with CMOS control electronics on one substrate.","The broad single-mode wavelength range, from visible to shortwave infrared depending on material, supports multi-wavelength photonic integrated circuits."],"supporting_citations":[{"why":"Drives the inclusion of a wetting layer between the III-V core and sapphire to improve growth and reduce defects.","marker":"[19]"},{"why":"Supplies the sapphire refractive index used in the mode simulations.","marker":"[20]"},{"why":"Supplies the refractive indices and attenuation indices of GaAs, InP, and GaSb cores.","marker":"[21]"},{"why":"Provides optical constants for the AlAs wetting layer on GaAs waveguides.","marker":"[22]"},{"why":"Establishes the low-loss III-V waveguide design rules and rib versus strip comparison used here.","marker":"[25]"},{"why":"One of the reported III-V-on-silicon loss values the authors compare their estimates to.","marker":"[26]"},{"why":"Source of the assumed sidewall-roughness and interface scattering loss constants.","marker":"[31]"},{"why":"Source for modeling scattering loss from dislocations at the wet-layer/substrate interface.","marker":"[32]"}],"fun_headline_variants":["III-V rib waveguides on sapphire: losses under 1 dB/cm","Sapphire-based III-V waveguides achieve sub-1 dB/cm loss","Low-loss III-V rib waveguides on sapphire for photonics","GaAs, InP, GaSb on sapphire: rib loss below 1 dB/cm","Simulated III-V rib waveguides on sapphire lose 0.32–0.70 dB/cm"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The headline loss figures depend on assumed values for sidewall roughness scattering (0.05 dB/cm for ribs) and for scattering at dislocations in the wetting-layer/sapphire interface that are taken from other waveguide systems rather than measured in III-V layers on sapphire; if the real epitaxial interface is more defective, measured propagation loss will be far higher than the estimated 0.3–0.7 dB/cm.","fun_headline_variants_meta":{"raw":{"variants":["III-V rib waveguides on sapphire: losses under 1 dB/cm","Sapphire-based III-V waveguides achieve sub-1 dB/cm loss","Low-loss III-V rib waveguides on sapphire for photonics","GaAs, InP, GaSb on sapphire: rib loss below 1 dB/cm","Simulated III-V rib waveguides on sapphire lose 0.32–0.70 dB/cm"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000852,"raw_usage":{"total_tokens":3729,"prompt_tokens":996,"completion_tokens":2733,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":612,"completion_tokens_details":{"reasoning_tokens":2624}},"tokens_in":612,"tokens_out":2733,"duration_ms":18861,"temperature":1.0,"reasoning_tokens":2624,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T16:54:00.282201+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fabricate a GaAs/AlAs rib waveguide on sapphire with the simulated dimensions, measure propagation loss at 1550 nm by the cut-back method or from a ring-resonator Q factor, and compare with the predicted 0.32 dB/cm; a measured loss above about 1 dB/cm would show the assumed scattering constants do not transfer to this platform.","supporting_citations":[{"cited_title":"Crystalline GaAs Thin Film Growth on a c -Plane Sapphire Substrate","cited_arxiv_id":null,"evidence_quote":"Drives the inclusion of a wetting layer between the III-V core and sapphire to improve growth and reduce defects."},{"cited_title":"Refractive Index and Birefringence of Synthetic Sapphire,” J","cited_arxiv_id":null,"evidence_quote":"Supplies the sapphire refractive index used in the mode simulations."},{"cited_title":"Optical dispersion relations for GaP, GaAs, GaSb, InP, InAs, InSb, AlxGa1−xAs, and In1−xGaxAsyP1−y","cited_arxiv_id":null,"evidence_quote":"Supplies the refractive indices and attenuation indices of GaAs, InP, and GaSb cores."},{"cited_title":"Modeling the optical dielectric function of GaAs and AlAs: Extension of Adachi’s model","cited_arxiv_id":null,"evidence_quote":"Provides optical constants for the AlAs wetting layer on GaAs waveguides."},{"cited_title":"Low-Loss III-V Semiconductor Optical Waveguides","cited_arxiv_id":null,"evidence_quote":"Establishes the low-loss III-V waveguide design rules and rib versus strip comparison used here."},{"cited_title":"Single -mode GaAs/AlGaAs W waveguides with a low propagation loss","cited_arxiv_id":null,"evidence_quote":"One of the reported III-V-on-silicon loss values the authors compare their estimates to."},{"cited_title":"Estimation of Propagation Losses for Narrow Strip and Rib Waveguides","cited_arxiv_id":null,"evidence_quote":"Source of the assumed sidewall-roughness and interface scattering loss constants."},{"cited_title":"Study of crystalline defect induced optical scattering loss inside photonic waveguides in UV – visible spectral wavelengths using volume current method","cited_arxiv_id":null,"evidence_quote":"Source for modeling scattering loss from dislocations at the wet-layer/substrate interface."}],"review_version":1}