{"id":"5018f33e-5f33-48f9-a228-b4839018dd6b","arxiv_id":"2502.02114","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A Fourier-patterned hBN waveguide with an embedded WS2 monolayer shows an anti-crossing with a 40 meV Rabi splitting, placing the system at the onset of strong coupling.","lead":"The authors carved a wavy pattern into a boron nitride light guide, embedded an ultra-thin tungsten disulfide layer inside, and used the pattern to couple light into the guide. The result is a step toward 2D-material optoelectronic devices where light-matter coupling can be designed by shaping the dielectric.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 40 meV Rabi splitting may be an artifact of under-constrained Lorentzian fits to unresolved, broad branches; the quantitative claim needs a global-fit or pole-retrieval validation.","rationale":"The paper is a careful experimental demonstration, and the qualitative observation of an anti-crossing when WS2 is present is supported by same-grating control without WS2 and by COMSOL simulations. I credit the authors for explicitly stating that the system does not meet the ERabi > γCav + γEx criterion and for presenting a second analysis route via energy conservation. However, my concern targets the step that converts raw reflectance into the headline numbers. The branch energies are obtained by fitting each spectrum to three interfering Lorentzians at a detuning where the two resonances are separated by only ~1.2 times the cavity HWHM; such fits are known to be poorly constrained, and the 95% confidence intervals (±9 meV) already reflect this fragility. The energy-conservation method is not an independent check of the branch energies: it uses the same E± values and merely tests consistency with a two-oscillator model. Since the central claim's quantitative content (ERabi = 40±9 meV, g = 22±4 meV, and the 'onset' classification) depends on these fits, a validation of the extraction procedure is the single most load-bearing open question. A global fit to the full 2D reflectance map or a pole-retrieval from an electromagnetic model would settle whether the fitted branches correspond to true polariton modes. Pending such a check, I would adjust the verdict to CONDITIONAL rather than full ACCEPT, because the experiment is otherwise convincing but the headline numbers are not yet independently secured.","tokens_in":17966,"tokens_out":9933,"duration_ms":96598,"concrete_test":"Re-analyze the raw R(E,kx) maps with a global fit: either (a) fit the full 2D map to a coupled-oscillator model (or a Fano model) with shared parameters, or (b) extract the complex reflection poles from a transfer-matrix/COMSOL model of the hBN/WS2/hBN grating and compare the pole splitting with the reported ERabi. Separately, generate synthetic reflectance maps with known g=22 meV, γCav, and γEx and run them through the paper's three-Lorentzian pipeline; if recovered g deviates by more than ±4 meV, the published uncertainty underestimates the systematic error.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The quantitative claim rests on extracting upper and lower branch energies by fitting reflectance spectra to a sum of three interfering Lorentzians (Methods, Fig. 5c). In the coupled regime, the two branch resonances are separated by ERabi ≈ 40 meV while the uncoupled cavity HWHM is γCav = 33.8 meV and the exciton HWHM is γEx = 15.1 meV; near zero detuning the branches are therefore not resolved as distinct peaks. The fit can trade peak positions, amplitudes, and phases, and the extracted branch energies may be biased or even spuriously anti-cross for an uncoupled crossing. The energy-conservation cross-check (Fig. S5, Table S2) uses the same fitted E± values, so it does not independently validate the branch positions. Thus ERabi = 40±9 meV and g = 22±4 meV are not secured against a fitting degeneracy.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports the fabrication and optical characterization of van der Waals heterostructures in which a monolayer WS2 is embedded inside an hBN waveguide and a two-sinusoid Fourier grating is patterned into the top hBN layer by thermal scanning-probe lithography. The authors demonstrate diffractive coupling of free-space light to TE0 and TE1 waveguide modes, show that the TE0 field maximum overlaps the WS2 layer, and observe an avoided crossing in angle-resolved reflectance when WS2 is present. From Lorentzian fits of the reflectance spectra they extract an uncoupled exciton energy of 1984.3 ± 0.6 meV, half-widths γEx = 15.1 ± 0.6 meV and γCav = 33.8 ± 0.2 meV, a Rabi splitting ERabi = 40 ± 9 meV, and a coupling strength g = 22 ± 4 meV. They conclude that the system is at the onset of the strong-coupling regime.","tokens_in":18103,"tokens_out":4615,"duration_ms":45593,"significance":"If the quantitative claim is secure, the paper presents a useful and transferable platform: placing the active TMD at the field maximum of a low-loss dielectric waveguide while using Fourier-surface gratings for far-field coupling is a practical route to exciton-polaritons in vdW heterostructures. The work has clear strengths: the fabrication is carefully characterized by AFM with reported fit RMSEs; the patterned hBN-only control and the WS2-free part of the same grating provide a direct comparison; the COMSOL simulations use literature refractive indices and reproduce the qualitative anti-crossing; and the authors report confidence intervals and an alternative energy-conservation analysis. The qualitative observation of an anti-crossing is well supported. The main weakness is that the central quantitative result, ERabi = 40 ± 9 meV, rests on a three-Lorentzian decomposition of spectra in which the two branches are not spectrally resolved, and the cross-check uses the same fitted branch energies.","major_comments":[{"comment":"The central Rabi-splitting claim is not yet secured against fitting degeneracy. Near zero detuning the expected branch separation ERabi ≈ 40 meV is smaller than or comparable to the sum of the uncoupled half-widths (γCav + γEx ≈ 49 meV), so the two branch resonances are not resolved as distinct spectral features. The spectra in Fig. 5c are fitted independently at each kx with a sum of three interfering Lorentzians, and the 95% confidence intervals shown in Fig. 5c reflect the precision of each individual fit, not the uniqueness of the three-peak decomposition. I ask for a global fit of all spectra with shared parameters, and/or a validation on simulated reflectance spectra: the COMSOL model already reproduces the anti-crossing in Fig. 4d, so applying the same extraction pipeline to simulated spectra with a known input coupling would directly test whether the procedure recovers the input coupling without bias.","section":"Fig. 5 and Methods, 'Analysis of coupling strength using the coupled oscillator model'"},{"comment":"The energy-conservation cross-check does not independently validate the fitted branch energies E±. It uses the same fitted E+ and E− values and only replaces the estimate of the uncoupled cavity energy, so it tests the zero-detuning assumption, not the Lorentzian decomposition. The consistency of the two ERabi estimates (40 ± 9 meV vs 38 ± 4 meV) is reassuring for the former, but it leaves the central fitting-degeneracy concern unaddressed.","section":"Fig. S5 and Table S2"},{"comment":"The Rabi splitting is extracted as the energy difference between the upper and lower branches at the data points whose kx values are closest to the zero-detuning intersection. This nearest-neighbor selection is not a fit at zero detuning and can introduce a systematic error that is not included in the 95% confidence intervals. The authors should report the actual detuning of the selected points and, preferably, interpolate E± to δ = 0 with the coupled-oscillator model, or use a global fit that includes the zero-detuning position as a fitted parameter.","section":"Methods, 'Analysis of coupling strength using the coupled oscillator model'"}],"minor_comments":[{"comment":"In Table S1, the equation for the double-sinusoid profile is written with A1 in both terms; it should read A2 cos(q2 x − π/2) to be consistent with the text and with the reported A2 value.","section":"Table S1"},{"comment":"The main text states that the dispersion in Fig. 1 is computed with an isotropic refractive index n = 2.1, while the simulations are said to use the anisotropic refractive index of hBN; please clarify which approximation is used in the waveguide-dispersion plots and in the COMSOL simulations.","section":"Main text, Fig. 1 and Methods, 'Simulations'"},{"comment":"The uncoupled hBN mode is represented by a linear fit over a limited kx range, and the extrapolation to the exciton intersection is used to define zero detuning; reporting the linear fit parameters and residuals would make the uncertainty in the intersection position more transparent.","section":"Fig. 5d"},{"comment":"The two-Lorentzian fit expression places the phase factor e^{iφ} on only one of the two Lorentzian terms; as written, the relative phase is not displayed symmetrically, and the text should clarify whether the phase multiplies the amplitude of one oscillator or the entire interference term.","section":"Methods, 'Analysis of coupling strength using the coupled oscillator model'"}],"recommendation":"major_revision","confidential_remarks":"The qualitative demonstration is solid and within the scope of the journal, but the quantitative strong-coupling claim needs stronger validation before acceptance. The reader's accept is somewhat optimistic given that the main cross-check uses the same fitted branch energies; a global fit or a simulation-based extraction would resolve the central concern."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take. What's new: this is the first combination of Fourier-surface gratings with a TMD monolayer embedded inside an hBN waveguide, placed at the TE0 field antinode. That placement is the right physical idea, and the same-grating control (WS2 present vs absent) makes the anti-crossing comparison clean. The COMSOL simulations using literature refractive indices line up with experiment, including the modified TE0 response above the exciton and the largely unaffected TE1 mode. Credit where due: the authors are explicit that they are at the onset of strong coupling, not fully in it, and they provide two extraction methods.\n\nThe soft spot is the quantitative claim. The Rabi splitting is extracted by fitting reflectance spectra to three interfering Lorentzians in a regime where the two branches are barely resolved. With an uncoupled cavity HWHM of 33.8 meV and a claimed splitting of 40 meV, the fit has genuine freedom in trading peak positions, amplitudes, and phases. The energy-conservation cross-check in Fig. S5 uses the same fitted E± values, so it is not an independent validation of the branch positions. This does not overturn the qualitative result — you can see the anti-crossing by eye in the reflectance maps, and the simulations reproduce it — but the specific numbers (ERabi = 40 ± 9 meV, g = 22 ± 4 meV) are softer than the abstract implies. A global fit across all kx values or a pole-retrieval analysis would secure them. Raw data and code are not provided, which limits reproducibility but is common for this type of experimental paper.\n\nThe citation pattern is appropriate: the self-cited Fourier-surface and tSPL work is the direct foundation, and the prior hBN waveguide/TMD coupling papers are properly acknowledged. No red flags there.\n\nWho is this for? People working on polaritonics in van der Waals stacks, especially anyone interested in all-dielectric platforms for exciton-polaritons. It deserves a serious referee. My recommendation would be to send it to review, with the expectation that the quantitative extraction is either strengthened or the splitting is presented more cautiously as an estimate from overlapping resonances. I would bring it to a reading group and would cite it for the platform and the careful same-grating comparison.","headline":"A genuinely useful all-dielectric platform with a clean same-grating control; the qualitative anti-crossing holds, but the quantitative Rabi splitting is softer than the abstract implies because the branch fits are under-constrained.","tokens_in":18773,"tokens_out":1764,"would_cite":true,"duration_ms":18427,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A WS2 monolayer placed at the field maximum inside an hBN waveguide shows a 40 meV anti-crossing.","keywords":["van der Waals heterostructures","hexagonal boron nitride","monolayer WS2","exciton-polaritons","Fourier surfaces","Rabi splitting","thermal scanning probe lithography","angle-resolved reflectance spectroscopy"],"falsifier":"Fabricate an identical hBN waveguide and grating but with the WS$_2$ monolayer placed at the vertical node of the TE$_0$ mode; if angle-resolved reflectance still shows an avoided crossing near $E_{\\mathrm{Rabi}} \\approx 40$ meV, the claimed field-overlap mechanism is not what produces the splitting.","tokens_in":17712,"feed_emoji":"⚛️","tokens_out":10565,"duration_ms":90957,"temperature":0.7,"pith_summary":"The paper sets out to show that a monolayer semiconductor embedded inside a dielectric waveguide can couple strongly enough to the waveguide mode to form exciton-polaritons, if it is placed where the mode's electric field is largest and a Fourier-patterned surface couples light in from free space. The demonstration embeds a WS$_2$ monolayer between two hBN flakes, with a two-sinusoid grating written into the top flake, and probes the device with angle-resolved reflectance. The payoff would be a low-loss, metal-free, integrated platform for polaritonic and quantum devices in which the active material is protected inside the heterostructure. The paper reports an anti-crossing with $E_{\\mathrm{Rabi}} = 40 \\pm 9$ meV and $g = 22 \\pm 4$ meV, and classifies the system as being at the onset of strong coupling.","feed_headline":"WS2 inside hBN shows a 40 meV light-matter splitting","feed_subtitle":"With the monolayer at the waveguide's field maximum, a dielectric hBN cavity reaches the edge of strong coupling.","key_machinery":"The central object is the Fourier surface grating, a topographic profile $f(x)=\\sum_j A_j\\cos(q_j x + \\phi_j)$ whose spatial frequencies $q_j$ supply the extra wavevectors that shift free-space light into the waveguide mode through $k_{i,x}+q_j=k_{\\mathrm{WG}}$. A second sinusoid at $q_2=2q_1$ folds bandgaps into the light cone. The design rule that carries the light-matter coupling argument is the placement of the WS$_2$ monolayer at the vertical position where the TE$_0$ electric field is maximal, since $g \\sim \\boldsymbol{\\mu}\\cdot\\mathbf{E}$. The quantitative extraction uses a coupled-oscillator model with $E_{\\mathrm{Rabi}}=\\sqrt{4g^2-(\\gamma_{\\mathrm{Cav}}-\\gamma_{\\mathrm{Ex}})^2}$, fed by resonance energies obtained from two- and three-Lorentzian fits of the Fano-like reflectance lineshapes.","core_discovery":"By encapsulating a monolayer of WS$_2$ between two hBN flakes at a depth where the TE$_0$ waveguide mode has its electric-field maximum, and patterning the top hBN with a two-sinusoid Fourier grating, the paper reports an avoided crossing in angle-resolved reflectance between the TE$_0$ mode and the WS$_2$ A exciton. The extracted Rabi splitting is $E_{\\mathrm{Rabi}} = 40 \\pm 9$ meV and the coupling strength is $g = 22 \\pm 4$ meV. Because $E_{\\mathrm{Rabi}}$ does not quite exceed $\\gamma_{\\mathrm{Cav}} + \\gamma_{\\mathrm{Ex}} = 48.9 \\pm 0.6$ meV, the system is described as being at the onset of strong coupling rather than fully inside it. The TE$_1$ mode, whose field is nearly zero at the WS$_2$ plane, remains largely uncoupled, which the paper uses to confirm that the coupling is governed by field overlap.","pith_inferences":["Extension: because $g \\sim \\boldsymbol{\\mu}\\cdot\\mathbf{E}$, moving the monolayer across the hBN thickness should trace the TE$_0$ field profile directly; the paper does not report such a thickness sweep.","Extension: a twin control stack with the identical grating but no WS$_2$ should show no anti-crossing, providing a clean experimental falsifier that the paper's data already approximate by comparing regions with and without WS$_2$.","Extension: the same fabrication chain should work for other excitonic 2D semiconductors by choosing the hBN thickness and grating period so the chosen mode crosses the exciton at zero detuning.","Extension: aligning an existing two-sinusoid band edge to the exciton could exploit slow light and raise the coupling beyond the loss threshold, turning the onset system into a more clearly strong-coupled one."],"forward_implications":["The vertical-overlap rule becomes a general design principle: active 2D layers should be placed at the antinode of the desired waveguide mode inside the dielectric, not on its surface.","hBN-encapsulated TMD monolayers can be probed by far-field reflectance through Fourier gratings while remaining protected from environmental degradation.","The two-sinusoid Fourier surface folds photonic bandgaps into the light cone, allowing band-structure engineering and coupling measurements in the same device.","Following the paper's stated next steps, adding more WS2 layers or aligning a band edge with the exciton should push the splitting past the loss sum and move the system from onset to full strong coupling."],"supporting_citations":[{"why":"introduces optical Fourier surfaces and the momentum-matching grating concept this paper extends to hBN waveguides.","marker":"[34]"},{"why":"demonstrates thermal scanning-probe lithography of freeform landscapes in hBN, the fabrication route used for the gratings.","marker":"[40]"},{"why":"supplies the optical dielectric function of monolayer WS2 and the excitonic response used in simulations and line-shape analysis.","marker":"[8]"},{"why":"provides the slab-waveguide dispersion theory used to model the TE0 and TE1 modes.","marker":"[41]"},{"why":"provides the anisotropic refractive index of hBN used in the waveguide dispersion and finite-element simulations.","marker":"[42]"},{"why":"explains the Fano-like lineshapes of resonant waveguide gratings used to fit the reflectance spectra.","marker":"[43]"},{"why":"establishes the energy of the WS2 A exciton used as the uncoupled exciton resonance.","marker":"[45]"},{"why":"states the strong-coupling criteria used to classify the system as at the onset of strong coupling.","marker":"[46]"}],"fun_headline_variants":["Fourier-patterned vdW stack approaches strong-coupling onset","WS2 in hBN cavity reaches 40 meV Rabi splitting","Grating-tuned heterostructure edges toward strong coupling","40 meV splitting from field-overlap-optimized vdW stack","Nanotextured hBN nudges WS2 toward strong coupling"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The Rabi splitting is obtained by treating the waveguide mode measured on the hBN-only part of the same grating as the uncoupled cavity, even though adding the WS$_2$ monolayer changes the waveguide's dielectric environment, so the zero-detuning point is inferred rather than directly measured.","fun_headline_variants_meta":{"raw":{"variants":["Fourier-patterned vdW stack approaches strong-coupling onset","WS2 in hBN cavity reaches 40 meV Rabi splitting","Grating-tuned heterostructure edges toward strong coupling","40 meV splitting from field-overlap-optimized vdW stack","Nanotextured hBN nudges WS2 toward strong coupling"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000826,"raw_usage":{"total_tokens":3618,"prompt_tokens":963,"completion_tokens":2655,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":579,"completion_tokens_details":{"reasoning_tokens":2571}},"tokens_in":579,"tokens_out":2655,"duration_ms":15888,"temperature":1.0,"reasoning_tokens":2571,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T13:17:43.345513+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fabricate an identical hBN waveguide and grating but with the WS$_2$ monolayer placed at the vertical node of the TE$_0$ mode; if angle-resolved reflectance still shows an avoided crossing near $E_{\\mathrm{Rabi}} \\approx 40$ meV, the claimed field-overlap mechanism is not what produces the splitting.","supporting_citations":[],"review_version":1}