{"id":"9641115e-b049-40af-8bcc-935568c57d35","arxiv_id":"1908.03452","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Multicore waveguides written by ultrafast laser pulses in chalcogenide glass transmit 4.5 µm light with 0.20 ± 0.05 dB/cm loss and over 60% coupling efficiency.","lead":"Ultrafast laser writing created light-guiding structures in a chalcogenide glass that carry mid-infrared light with very low loss. The reported loss of about 0.2 decibels per centimetre is much lower than earlier mid-infrared waveguides, which could enable compact chemical sensors.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Back-reflection loss value sits near the resolution limit of the method; independent cutback verification is needed to anchor the 0.20 dB/cm central claim.","rationale":"The reader's weakest-assumption analysis identifies the same load-bearing concern: the back-reflection method is used without independent verification, and the measured one-way loss over 28 mm is only about 0.56 dB, near the practical resolution of the approach. This is a genuine internal-validity issue, not merely a departure from consensus. The paper's careful comparison of two mesh geometries, the use of the same host glass, and the observation of identical minimum loss across configurations are real strengths, but they do not calibrate the loss-measurement method itself. The coupling efficiency in Eq. (1) is also under-determined because TOpt is not specified, but the central claim of record-low propagation loss is the more consequential number. The proposed cutback test is a standard, decisive check: if it reproduces 0.20 ± 0.05 dB/cm, the conditional concerns are resolved; if it does not, the headline claim would need to be revised downward in significance. Since the reader's verdict already conditions acceptance on such additional verification, our stress-test does not change that verdict.","tokens_in":5593,"tokens_out":3843,"duration_ms":43906,"concrete_test":"Perform a destructive cutback measurement: using a stable free-space or fiber launch at 4.5 µm, record the transmitted power through the full 28 mm waveguide, then repolish the same sample to a shorter length (e.g., 14 mm) and remeasure under identical launch conditions. Compute the propagation loss as the slope of insertion loss versus length, and repeat on at least two independently written waveguides from the same glass. If the cutback-derived loss deviates from 0.20 dB/cm by more than the stated ±0.05 dB/cm, the back-reflection result is biased and the headline claim requires revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline claim of 0.20 ± 0.05 dB/cm propagation loss at 4.5 µm rests on a single back-reflection measurement [22] over a 28 mm waveguide. At that length the total measured round-trip loss is roughly 2 × 0.20 dB/cm × 2.8 cm ≈ 1.1 dB, and the one-way loss is only about 0.56 dB. Systematic effects in the back-reflection technique—imperfect spatial separation of input and output Fresnel reflections, back-coupling losses at the output facet, mode conversion in the multicore structure, and calibration of the Fresnel coefficient—can be comparable to this small number. The paper reports no independent cutback, length-dependent, or two-length measurement, so the accuracy of the loss value is not externally established. If the true loss were, for example, 0.35–0.40 dB/cm, the claim that this is 'far below' photowritten and silicon-photonic mid-IR waveguides would be substantially weakened. The coupling-efficiency values in Eq. (1) also inherit any error in α through the correction term 10^(0.1αL), but the propagation-loss claim is the more load-bearing one.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports ultrafast laser inscription of multicore waveguides in 72GeS2-18Ga2S3-10CsCl chalcogenide glass and characterizes them at 4.5 µm. Two hexagonal-mesh configurations and one circular-mesh configuration are written with varying burst duration τ. The authors report a minimum propagation loss of 0.20 ± 0.05 dB/cm, coupling efficiencies above 60%, and Gaussian-like near-field mode profiles, and they interpret the results as evidence of homogeneous refractive-index channels with no additional scattering from slice concatenation. They also discuss the trade-off between coupling efficiency and propagation loss as τ varies.","tokens_in":5823,"tokens_out":5940,"duration_ms":57658,"significance":"The reported value of 0.20 ± 0.05 dB/cm at 4.5 µm would be a substantial improvement over previously published photowritten chalcogenide and silicon-based mid-IR waveguides, and the ability to independently control Δn and structure diameter is useful for practical coupling. Strengths of the paper include the direct, non-circular measurement procedure: all quantities are measured externally, the Fresnel correction uses an independently known refractive index [23], and the back-reflection method has been established for waveguide loss measurement [22]. If the loss value is confirmed by an independent method, the result would be of clear interest to the mid-IR photonics community.","major_comments":[{"comment":"The central claim of 0.20 ± 0.05 dB/cm rests on a single back-reflection measurement on a 28 mm waveguide. The one-way loss over this length is only about 0.56 dB, and the stated uncertainty corresponds to ±0.14 dB over the same length; systematic errors in the back-reflection method (spatial separation of input and output reflections, back-coupling at the exit facet, mode conversion in the multicore structure, and Fresnel-coefficient calibration) can easily be of this magnitude. No cutback, length-dependent, or independently repeated measurement is reported, so the accuracy of the headline loss value is not externally established. I ask the authors to add a cutback or two-length verification, or at minimum to report the repeatability and the dominant systematic-error budget.","section":"Section III, Fig. 2"},{"comment":"The coupling-efficiency values rely on the optics transmission TOpt, but the manuscript does not state how TOpt was calibrated or measured, nor are uncertainties in η, R, or the power measurements given. Although the propagation-loss correction 10^(0.1αL) is only about 1.14 for α = 0.2 dB/cm and L = 28 mm, an uncalibrated system-level factor in TOpt could shift η by a larger, unknown amount. Please report the TOpt calibration procedure and propagate uncertainties into η, including error bars in Fig. 4; without this information, the 'higher than 60%' coupling claim is not quantitatively supported.","section":"Eq. (1), Fig. 4"}],"minor_comments":[{"comment":"The first sentence of the conclusions states 'propagation loss below 0.2 dB/cm,' which is not consistent with the abstract's measured value of 0.20 ± 0.05 dB/cm; use the same number with its uncertainty throughout.","section":"Section IV"},{"comment":"The final sentence of the conclusions says that the authors 'do not take into account Fresnel loss' after reporting efficiencies obtained with Eq. (1), which includes the (1 − R)^2 factor; clarify whether η is Fresnel-corrected and whether the 'carried power' statement refers to η or to a different quantity.","section":"Section III and Section IV"},{"comment":"The text contains 'channelx' and 'channely' that should read 'channel x' and 'channel y', and 'in the plan of the transverse section' should be 'in the plane of the transverse section'.","section":"Section II"},{"comment":"The coupling efficiencies are presented as point measurements without error bars, so it is unclear whether the ranking of the three structures at a given τ is significant; add repeat-measurement statistics.","section":"Fig. 4"},{"comment":"The single-mode claim is based solely on a near-Gaussian near-field image; a mode-cutoff study or an M² measurement would make this claim more robust.","section":"Section III, Fig. 5"}],"recommendation":"major_revision","confidential_remarks":"The single-length, single-sample loss measurement is the main substantive obstacle to accepting the record-class claim; a cutback or two-length verification would resolve my primary concern. The paper is otherwise concise and well suited to a rapid-communication format."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe thing to know: this paper extends the authors' multicore femtosecond-laser writing method from 1.55 µm to 4.5 µm in an arsenic-free chalcogenide glass and reports propagation losses of 0.20 ± 0.05 dB/cm at 4.5 µm. If that number is right, it is genuinely better than the suspended silicon (0.82 dB/cm below 4 µm) and germanium (>1 dB/cm) guides they compare against. But the claim is only as good as the measurement, and the measurement is only a single back-reflection on a 28 mm waveguide, with one-way loss of about 0.56 dB. That is close to the resolution limit of the method, and no cutback or length-dependent check is provided.\n\nWhat the paper does well: the writing procedure is clearly described, with a useful parameter study of burst duration τ and its effect on Δn and loss. The observation that the minimum loss is the same for two different hexagonal meshes, and similar for a circular mesh, is a nice piece of evidence that the method is robust and the index structure is homogeneous. The coupling-efficiency study, while simple, gives a practical sense of how to choose τ depending on device length. The near-field mode images support (but don't prove) single-mode behavior.\n\nSoft spots, in order: (1) The loss measurement. The back-reflection technique of Ramponi et al. is legitimate, and the authors took care to angle the input face and account for Fresnel reflection. But with such a small total loss, systematic errors from mode conversion or imperfect coupling can be comparable to the signal. A cutback measurement is standard and would settle it. (2) The coupling-efficiency formula depends on TOpt, the transmission of the focusing/collimating optics, but its calibration is not described. No error bars are given for η, so the 'higher than 60%' claim is not yet quantified. (3) The single-mode claim is inferred only from a Gaussian-like near-field profile; that is suggestive, not conclusive.\n\nNone of these are disqualifying, but they are exactly the details that matter for a record-class loss value. This is a solid, honest letter from a group that knows the material and the method; the self-citations are appropriate. It deserves a serious referee. If I were the editor, I'd send it to review and ask for a cutback measurement and TOpt details before publication.\n\nWho is this for: the mid-IR photonics community, especially people building on ULI in chalcogenides. I'd cite it with a caveat until the loss is independently verified.","headline":"The 0.20 dB/cm mid-IR loss claim is a plausible record, but it rests on one back-reflection measurement with no cutback check; worth reviewing, needs experimental corroboration.","tokens_in":6381,"tokens_out":3430,"would_cite":true,"duration_ms":32927,"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":"Multicore waveguides written by ultrafast laser inscription in chalcogenide glass guide 4.5 µm light with 0.20 ± 0.05 dB/cm loss and over 60% coupling efficiency.","keywords":["mid-infrared photonics","ultrafast laser inscription","chalcogenide glass","multicore waveguide","propagation loss","coupling efficiency","femtosecond laser writing","single-mode waveguide"],"falsifier":"A cutback test with identical guides of several lengths (for example 10, 20, and 40 mm) would settle the loss claim: the slope of measured insertion loss versus length must equal $0.20$ dB/cm and the intercept must match an independently calibrated coupling loss. Separately calibrating $T_{\\mathrm{Opt}}$ with a reference optic of known transmission would check the reported coupling efficiencies above 60%.","tokens_in":5419,"feed_emoji":"🔬","tokens_out":10327,"duration_ms":88247,"temperature":0.7,"pith_summary":"This paper reports that multicore waveguides written by ultrafast laser inscription in the arsenic-free chalcogenide glass 72GeS$_2$--18Ga$_2$S$_3$--10CsCl guide mid-infrared light at 4.5 µm with propagation loss $0.20 \\pm 0.05$ dB/cm and coupling efficiency above 60%. The authors argue these figures are far below those of previously photowritten waveguides or silicon-photonics-derived mid-IR waveguides. The result matters because mid-infrared photonics is the practical route to compact chemical and biological sensors, and loss and coupling have been the main barriers for photowritten glass devices. The paper also shows the measured loss minimum is the same for different mesh geometries, which it interprets as evidence that the writing process is homogeneous, and that the guided mode is near-Gaussian, i.e. single-mode.","feed_headline":"Femtosecond-written waveguides reach 0.20 dB/cm at 4.5 µm","feed_subtitle":"Multicore chalcogenide waveguides beat silicon-photonics losses and couple over 60 percent of mid-IR light.","key_machinery":"The central object is the multicore waveguide: an array of parallel positive-refractive-index channels, each produced by a femtosecond-laser filament whose diameter is fixed by the glass, with the channels arranged on a hexagonal or concentric-ring mesh. The index contrast $\\Delta n$ of each channel is controlled by the burst duration $\\tau$ of the pulse train, which sets the compromise between poor confinement (small $\\tau$) and field localization inside single channels (large $\\tau$); the optimal $\\tau$ minimizes propagation loss. The overall transverse size is set separately by the number and spacing of channels, giving independent control of $\\Delta n$ and guide diameter. Losses are obtained with the back-reflection method; coupling efficiency is derived from Eq. (1), which corrects the power ratio for Fresnel reflection, optics transmission, and propagation loss.","core_discovery":"On its own terms, the paper's discovery is that a multicore geometry—parallel positive-index channels written on a hexagonal or concentric-ring mesh—lets ultrafast laser inscription produce a single-mode mid-infrared waveguide in an arsenic-free chalcogenide glass with propagation loss $0.20 \\pm 0.05$ dB/cm at 4.5 µm and coupling efficiency higher than 60%. The authors find the minimal loss is identical for 4-row and 5-row hexagonal meshes and for a 5-ring circular mesh, and take this as evidence that the index-contrast channels are homogeneous and that stitching successive transverse slices introduces no extra scattering. They further show the index contrast $\\Delta n$ can be tuned through the burst duration $\\tau$ while the overall guide diameter is set independently by the channel spacing and count, so mode size and confinement can be engineered for efficient light collection. Combining these results with the earlier $0.11 \\pm 0.03$ dB/cm at 1.55 µm, the authors conclude the method can cover the 1.5–4.5 µm range in one glass platform.","pith_inferences":["If an independent cutback measurement on several guide lengths confirms the 0.20 dB/cm figure, femtosecond-laser-written chalcogenide waveguides would move into the loss class needed for practical mid-IR evanescent-wave gas sensors, not just laboratory demonstrations.","The equality of the loss minimum across different meshes suggests the residual loss is set by the glass matrix or the channel material rather than by mesh geometry; writing the same multicore structure in glasses with different germanium/gallium ratios would test this.","Eq. (1) implies a design rule the paper states only qualitatively: for a fixed input power, the best $\\tau$ maximizes transmitted power, which for a given device length is a trade-off between $\\alpha$ and $\\eta$ that could be plotted explicitly.","The paper projects but does not demonstrate operation beyond 4.5 µm; measuring the same guides at 7–10 µm would check whether the method's advantage persists up to the glass's transmission edge."],"forward_implications":["A laser-written chalcogenide waveguide can carry 4.5 µm light at propagation loss below 0.2 dB/cm, a level the paper argues is better than current photowritten and silicon-based mid-IR waveguides.","The same inscription procedure yields low loss at 1.55 µm and at 4.5 µm, so one material and one writing method can serve devices across the 1.5–4.5 µm band.","Because the guide diameter and index contrast are independently adjustable, the transverse mode can be matched to fibers or free-space beams, reducing coupling losses in practical systems.","The burst duration can be chosen according to device length: short devices should prefer higher $\\tau$ to maximize coupled power, while long devices should prefer the $\\tau$ that minimizes loss.","The loss minimum being independent of channel density suggests the writing method is stable across mesh designs, which supports extension to curved guides and more complex circuits."],"supporting_citations":[{"why":"Earlier report of the same multicore writing procedure in this glass, giving the 1.55 µm loss baseline and the method this paper extends.","marker":"[18]"},{"why":"Supplies the back-reflection method used to measure the propagation losses reported at 4.5 µm.","marker":"[22]"},{"why":"Provides the quantitative phase microscopy and Abel inversion used to measure the refractive-index contrast of written channels.","marker":"[19]"},{"why":"Explains the filament-formation and charge-accumulation mechanism behind the index contrast and channel diameter.","marker":"[20]"},{"why":"Recent germanium-based photowritten waveguide with 1–1.5 dB/cm loss at 7.8 µm, the benchmark the paper's loss is compared with.","marker":"[17]"},{"why":"Suspended silicon waveguide with 0.82 dB/cm loss below 4 µm, one silicon-photonics baseline the paper claims to beat.","marker":"[6]"},{"why":"Silicon-on-sapphire nanowire with 2 dB/cm loss at 5.18 µm, another silicon-photonics baseline for the comparison.","marker":"[8]"},{"why":"Gives the refractive index of the glass used to check the measured 13% Fresnel reflection coefficient entering Eq. (1).","marker":"[23]"}],"fun_headline_variants":["Multicore chalcogenide waveguides hit 0.20 dB/cm at 4.5 µm","Femtosecond-written guides in chalcogenide achieve >60% coupling at 4.5 µm","Tunable index contrast enables efficient mid-IR waveguides","Hexagonal and circular meshes yield same loss in chalcogenide waveguides","Mesh-based chalcogenide guides couple >60% at 4.5 µm"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the back-reflection measurement on a single 28 mm guide, whose total loss is only about 0.56 dB, gives the true propagation loss without an independent cutback check, and that the optics transmission $T_{\\mathrm{Opt}}$ in Eq. (1) is correctly calibrated.","fun_headline_variants_meta":{"raw":{"variants":["Multicore chalcogenide waveguides hit 0.20 dB/cm at 4.5 µm","Femtosecond-written guides in chalcogenide achieve >60% coupling at 4.5 µm","Tunable index contrast enables efficient mid-IR waveguides","Hexagonal and circular meshes yield same loss in chalcogenide waveguides","Mesh-based chalcogenide guides couple >60% at 4.5 µm"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001685,"raw_usage":{"total_tokens":6640,"prompt_tokens":871,"completion_tokens":5769,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":487,"completion_tokens_details":{"reasoning_tokens":5656}},"tokens_in":487,"tokens_out":5769,"duration_ms":40183,"temperature":1.0,"reasoning_tokens":5656,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:12:12.503086+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A cutback test with identical guides of several lengths (for example 10, 20, and 40 mm) would settle the loss claim: the slope of measured insertion loss versus length must equal $0.20$ dB/cm and the intercept must match an independently calibrated coupling loss. Separately calibrating $T_{\\mathrm{Opt}}$ with a reference optic of known transmission would check the reported coupling efficiencies above 60%.","supporting_citations":[{"cited_title":"Masselin, E","cited_arxiv_id":null,"evidence_quote":"Earlier report of the same multicore writing procedure in this glass, giving the 1.55 µm loss baseline and the method this paper extends."},{"cited_title":"Ramponi, R","cited_arxiv_id":null,"evidence_quote":"Supplies the back-reflection method used to measure the propagation losses reported at 4.5 µm."},{"cited_title":"Ampem-Lassen, S","cited_arxiv_id":null,"evidence_quote":"Provides the quantitative phase microscopy and Abel inversion used to measure the refractive-index contrast of written channels."},{"cited_title":"Caulier, D","cited_arxiv_id":null,"evidence_quote":"Explains the filament-formation and charge-accumulation mechanism behind the index contrast and channel diameter."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Recent germanium-based photowritten waveguide with 1–1.5 dB/cm loss at 7.8 µm, the benchmark the paper's loss is compared with."},{"cited_title":"Soler Penades, A","cited_arxiv_id":null,"evidence_quote":"Suspended silicon waveguide with 0.82 dB/cm loss below 4 µm, one silicon-photonics baseline the paper claims to beat."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Silicon-on-sapphire nanowire with 2 dB/cm loss at 5.18 µm, another silicon-photonics baseline for the comparison."},{"cited_title":"Masselin, D","cited_arxiv_id":null,"evidence_quote":"Gives the refractive index of the glass used to check the measured 13% Fresnel reflection coefficient entering Eq. (1)."}],"review_version":1}