{"id":"271cc0b7-2d99-41eb-a548-161497083aef","arxiv_id":"2607.14590","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":4,"one_line_summary":"Angle-resolved SHG on MBE AlScN films extracts d31 up to 4.93 pm/V and unconstrained d33, but the tensor model assumes d15=d31 without justification.","lead":"The paper reports a 60x enhancement in the second-order nonlinear coefficient d31 of MBE-grown AlScN films with a suppressed d33, based on angle-resolved SHG measurements. The analysis appears to conflate the independent d15 coefficient with d31, and the d33 values are so uncertain that the suppression claim is not supported.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The SHG extraction sets d15=d31 in the 6mm tensor (Eqs. 1–2), so the claimed 60x d31 enhancement may be a mislabeled d15.","rationale":"Read in good faith, the paper attempts to resolve conflicting reports of AlScN second-order nonlinearity by measuring MBE films on nitrided sapphire and fitting angle-dependent SHG with a TMM model. The structural characterization (XRD, RHEED, AFM) supports wurtzite-phase films, and the direct-growth strategy is reasonable since c-plane sapphire is centrosymmetric. However, the quantitative claim that d31 reaches 4.92 pm/V rests on the tensor in Eq. (1). That tensor sets the shear coefficient d15 equal to d31. This is not a symmetry requirement of P6_3mc; it is a silent modeling assumption. For AlN the two coefficients differ by an order of magnitude, and the p-pump SHG—the stronger signal by more than 10×—is governed by d15. The fitting therefore cannot reliably separate d31 from d15; the extracted 'd31' is biased and may actually be a d15-dominated value. This is not a context-dependent disagreement with published d33 values; it is a mechanical flaw in the analysis. The d33 suppression conclusion is additionally weakened by confidence intervals that span zero and include AlN-like magnitudes, but the d15 issue alone is sufficient to invalidate the headline result as presented. A corrected refit with d15 free and an s-pump-only d31 extraction is the one check that would settle whether the enhancement is physical. Until that is done, the paper's core claim is unsupported, so the reader's REJECT verdict with moderate confidence should stand unchanged.","tokens_in":10685,"tokens_out":10242,"duration_ms":105306,"concrete_test":"Refit the raw angle-dependent SHG data from Fig. 3(b,c) using the correct 6mm tensor with d15 as an independent free parameter, fitting p-pump data for {d15,d31,d33} and, separately, fitting the s-pump data alone for {d31,d33}. Also re-run the AlN calibration with literature d15≈3.6 pm/V and d31=0.08 pm/V. If the s-pump-only d31 remains ≈4.9 pm/V at 20% Sc and p-pump d15 is comparable, the 60× claim is supported. If the p-pump fit is dominated by d15≈5 pm/V while the s-pump-only d31 returns toward ~0.1 pm/V, the headline d31 enhancement is an artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—d31 ≈ 4.9 pm/V at 20% Sc, a 60× enhancement over AlN—is not established because the fitting model incorrectly conflates d15 with d31. In wurtzite P6_3mc symmetry, d15 appears in the (x,z)-shear positions and is independent of d31. Eq. (1) places d31 in those shear positions, and Eq. (2) then writes P_x(2ω)=2d31 E_x E_z for a p-polarized pump. Symmetry does not force d15=d31; reported AlN values differ by more than an order of magnitude (d15≈3.6 pm/V vs d31≈0.08 pm/V). Since the measured p-pump SHG is more than an order of magnitude stronger than the s-pump SHG (Fig. 3), the p-pump data are dominated by the shear term. With d15 removed as a free parameter, the fit assigns that large shear response to d31. The extracted 15–20% Sc 'd31' values (3–5 pm/V) are quantitatively on the scale expected for d15, not for the AlN d31 used in calibration. If Kleinman symmetry was intended, it is not stated and is inconsistent with the cited AlN values. The d33 suppression claim is also unsupported: the 20% sample's 95% CI spans -8.85 to 2.66 pm/V, so AlN-like d33 ≈ 5 pm/V cannot be excluded. This is an internal modeling error, not a disagreement with prior literature.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports angle-resolved transmission SHG measurements on MBE-grown AlScN films (10–25% Sc) deposited directly on nitrided c-plane sapphire, with ellipsometric and XRD/RHEED characterization. Using a transfer-matrix model for a P6_3mc (wurtzite) film on sapphire, the authors extract d31 and d33 as functions of Sc content, reporting d31 ≈ 4.9 pm/V at 20% Sc — about 60× the AlN value — while d33 appears suppressed. The paper argues that omitting a thick AlN buffer isolates the AlScN response and contrasts its findings with prior sputtered-film results reporting large d33.","tokens_in":10978,"tokens_out":1438,"duration_ms":15526,"significance":"If the extracted d31 enhancement is correct, the result would be significant for integrated nonlinear photonics on nitride platforms, and the direct-growth methodology would offer a cleaner route to measuring the intrinsic response of AlScN. The paper is also useful in that it provides a full tensor extraction and quantifies confidence intervals for d33, a step beyond many prior reports. The manuscript is clearly written and the experimental dataset appears carefully acquired; the transfer-matrix treatment of the layered SHG problem is a standard and appropriate framework.","major_comments":[{"comment":"The SHG model in Eq. (1) sets the (x,z) and (z,x) shear tensor elements equal to d31. In 6mm symmetry these elements are d15, which is independent of d31. The manuscript offers no justification for d15 = d31, and the cited AlN literature values for d15 (≈3.6 pm/V) and d31 (≈0.08 pm/V) differ by orders of magnitude. Because the p-polarized SHG is more than ten times stronger than the s-polarized signal (Fig. 3), the p-pump data are dominated by the shear coefficient. Setting d15 = d31 in the fit therefore misattributes the large shear response to d31. The headline claim 'd31 as high as 4.92 pm/V' is thus not established by this analysis. The authors should either (i) include d15 as a separate free parameter and report its confidence interval, or (ii) explicitly invoke and justify a Kleinman-symmetry condition with a stated uncertainty, or (iii) reframe the claim as an effective shear-rela","section":"Eqs. (1)–(2) and Fig. 3"},{"comment":"The AlN calibration uses literature d31 = 0.08 pm/V and d33 = 5.1 pm/V, but then says the 'd31/d33 ratio slightly adjusted to optimize the fit to our baseline data.' This adjusted ratio is not reported, nor is the resulting set of AlN coefficients. If the ratio change is large (e.g., if d31 is adjusted upward relative to d33), the calibration itself would partly determine the magnitude of the extracted d31 values. The dependence of the final d31 values on this calibration choice must be stated, and the actual AlN fit parameters should be listed.","section":"Sec. 'To calibrate the experimental setup...' and Table I"},{"comment":"The claim that d33 'appears suppressed' is not supported by the fit. For the 20% sample, the 95% CI spans -8.85 to 2.66 pm/V, which includes the AlN value of 5.1 pm/V (and certainly includes d33 ≈ 0). The statement in the text that 'a significant enhancement in d33 would still have been captured' is not quantitatively backed; given the elongated MSE contour in Fig. 4(c), a large positive d33 cannot be excluded at 95% confidence. The paper should downgrade the d33 suppression claim to 'not measurable in this geometry with the present precision.'","section":"Sec. 'Fig. 4' and Table I, 20% Sc row"}],"minor_comments":[{"comment":"The sentence in the Results section describing the 10% film's refractive index appears incomplete: 'For the 10% film, no increases from 2.16 at 1600 nm to 2.4 at 300 nm.' This should be rewritten, e.g., 'For the 10% film, n_o increases from 2.16 at 1600 nm to 2.4 at 300 nm.'","section":"Abstract and Fig. 2(b)"},{"comment":"The notation SE for the electric-field discontinuity could be confused with the time-averaged Poynting vector; suggest using ΔE_x or similar. Also, the sign convention for kx should be defined.","section":"Eq. (4)"},{"comment":"The inset axes labels ('p' and 's') are not defined clearly in the caption; specify whether the inset shows the detected polarization angle or the analyzer angle.","section":"Fig. 3 inset"},{"comment":"Ref. [21] is cited as 'V. Yoshioka et al.' but the author list includes 'J. Jin, R. H. O. III, and B. Zhen'; please verify the author names are complete.","section":"References"},{"comment":"The text states 'The lowest RMS roughness of 0.353 nm' and later Fig. 1(b) caption says '3.53 nm RMS' — there is a factor-of-ten inconsistency. Check whether the RMS is 0.353 nm or 3.53 nm.","section":"Sec. 'The AlScN samples...'"}],"recommendation":"major_revision","confidential_remarks":"The central issue is the d15=d31 assumption, which directly affects the headline number. This is a correctable modeling issue, not a fundamental flaw in the measurement approach. However, unless the authors can show that d15 and d31 are indeed equal in their AlScN films (e.g., by independent measurement or by a valid symmetry argument), the paper's main claim will need to be substantially reframed. I would be willing to review a revised version. Also note that the d33 suppression claim is statistically weak; the paper should be more careful in its wording."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know up front. The experiment is clean and the write-up is honest about its weak spots: the authors grow MBE AlScN directly on nitrided sapphire, measure angular SHG with a proper transfer-matrix treatment, and show their d33 confidence intervals plainly. But the headline number — d31 = 4.93 pm/V at 20% Sc, a 60x enhancement over AlN — rests on a tensor model that is wrong for the stated symmetry. Eqs. (1) and (2) put d31 in the (x,z) shear positions that belong to d15 in P6_3mc. The two coefficients are independent, and the paper's own AlN calibration values (d31 = 0.08, d33 = 5.1 pm/V) rule out any Kleinman assumption that would make them equal.\n\nThe p-polarized pump response, which is an order of magnitude stronger than the s-pump response, is dominated by the shear term, so the fit is effectively extracting something on the d15 scale (3–5 pm/V for AlN) and calling it d31. That is why the numbers land where they do. This is a mechanical modeling error, not a quibble. The reader's stress test lands squarely on this.\n\nWhat survives is real. The direct-on-sapphire template is a legitimate improvement over the buffered films in Ref. [22]; the 25% Sc point is new; and the s-pump channel (P_z = d31 E_y^2) genuinely constrains d31. The s-pump data show a clear increase with Sc, so the qualitative trend of d31 enhancement is probably right — just not the specific value, and not the \"60x\" slogan. The d33 \"suppression\" claim is softer: the paper itself reports the 20% sample's 95% CI as -8.85 to 2.66 pm/V and admits the transmission geometry is far less sensitive to d33 than to d31. That is honest, but \"suppressed\" overstates what the data can say. The AlN calibration step with a d31/d33 ratio \"slightly adjusted\" to optimize the fit also needs to be described rather than tucked away.\n\nThe fix is straightforward: add d15 as a free parameter in the fit, or report the s-pump-only d31 values. This paper deserves peer review, not a desk reject. Send it back for major revision with that request; the growth work and the measurement set are valuable to the AlScN nonlinear optics community even if the corrected numbers land lower than 4.93.","headline":"A clean SHG experiment whose headline 60x d31 claim rides on a tensor that silently sets d15=d31 — the fix is straightforward and the data set is worth keeping.","tokens_in":11637,"tokens_out":6490,"would_cite":false,"duration_ms":70571,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["42.65.Ky"],"model":"deepseek-v4-flash","headline":"MBE-grown AlScN films show a d31 of up to 4.92 pm/V, about 60 times AlN, with d33 suppressed.","keywords":["AlScN","second harmonic generation","d31 coefficient","molecular beam epitaxy","wurtzite","transfer matrix method","nonlinear optics","nitride photonics"],"falsifier":"Measure the SHG for a p-polarized pump while detecting only the x-polarized harmonic component, and separately extract d15 and d31 from geometries that isolate each coefficient. If d15 and d31 differ by more than the reported fit uncertainties, the paper's tensor assignment and its d31 value are incorrect.","tokens_in":10480,"feed_emoji":"🔬","tokens_out":4027,"duration_ms":44932,"temperature":0.7,"pith_summary":"This paper measures the second-order optical nonlinearity of AlScN films grown directly on nitrided sapphire by molecular beam epitaxy, using angle-resolved transmission second-harmonic generation. It reports that the in-plane d31 coefficient increases sharply with scandium content, reaching about 5 pm/V at 20% Sc, while the longitudinal d33 coefficient stays near zero. The direct-growth method is designed to isolate the AlScN response from any buffer-layer contributions. If correct, this large d31 would make AlScN a viable material for on-chip nonlinear photonics and would help resolve conflicting reports between sputtered and MBE-grown films.","feed_headline":"AlScN d31 hits 4.9 pm/V, 60 times AlN","feed_subtitle":"Angle-resolved SHG on films grown directly on sapphire isolates the true nonlinear response, settling a sputtered-vs-MBE conflict.","key_machinery":"The wurtzite (P63mc) second-order susceptibility tensor, with its contracted d-coefficients d31 and d33, combined with a transfer-matrix model that computes the pump field distribution inside the film, treats the nonlinear polarization as local sheet sources, and propagates the generated harmonic field to the substrate and detector. A central element is the screening factor 1/εzz in the source discontinuity for the transverse electric field, which suppresses out-of-plane (Pz) radiation and makes the measurement far more sensitive to the in-plane coefficient than to d33.","core_discovery":"The central claim is that in wurtzite AlScN grown by MBE, the coefficient d31 is enhanced roughly 60-fold relative to AlN, peaking at about 4.9 pm/V for 20% scandium, whereas d33 remains suppressed within a large uncertainty. This is inferred from the angular dependence of the p- and s-polarized pump SHG signals, matched by transfer-matrix simulations that account for multilayer reflectance and the strong longitudinal dielectric screening in the high-index film. The authors argue that because the films are grown directly on nitrided sapphire, the measured response is intrinsic to AlScN rather than an artifact of a thick AlN buffer.","pith_inferences":["In strict wurtzite 6mm symmetry, the p-polarized pump response in the x-z plane is governed by d15, not d31; the paper's tensor model implicitly assumes d15 = d31, so the reported 'd31' enhancement may actually be a d15 enhancement, which would change which waveguide geometries benefit.","The apparent suppression of d33 could be partly a measurement-geometry effect; a normal-incidence or reflection-geometry SHG experiment with greater sensitivity to the out-of-plane polarization could either confirm the suppression or reveal a residual d33.","If the drop in d31 at 25% Sc is caused by strain relaxation through crack formation, then strain-engineering or lattice-matched substrates could extend the enhancement to higher scandium fractions than the paper observed.","The stark contrast with sputtered-film results reporting d33 up to 62 pm/V may stem from the analysis method: the TMM explicitly accounts for longitudinal screening, which simpler ray-optic fits could miss, potentially overestimating d33."],"forward_implications":["AlScN films could provide a strong effective nonlinearity for integrated photonics even if d33 is small, since the in-plane coefficient is large.","The d31 peak near 20% scandium suggests a composition sweet spot for device design before crystalline quality degrades.","Direct growth on nitrided sapphire offers a clean platform for characterizing the intrinsic nonlinearity of other nitride alloys.","The transfer-matrix analysis can serve as a rapid metrology method for optically characterizing transparent epitaxial nonlinear films."],"fun_headline_variants":["AlScN d31 60x AlN at 4.9 pm/V","MBE AlScN: d31 soars 60x, d33 suppressed","Direct-grown AlScN reveals d31 boost to 4.9 pm/V","AlScN SHG: d31 60x AlN, d33 lower","No buffer: AlScN's true d31 is 60x AlN"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The tensor model in Eq. (1) assumes the shear coefficient d15 equals d31; if they are independent, the extracted 'd31' is a biased mixture of the two and the claimed 60-fold enhancement is not established.","fun_headline_variants_meta":{"raw":{"variants":["AlScN d31 60x AlN at 4.9 pm/V","MBE AlScN: d31 soars 60x, d33 suppressed","Direct-grown AlScN reveals d31 boost to 4.9 pm/V","AlScN SHG: d31 60x AlN, d33 lower","No buffer: AlScN's true d31 is 60x AlN"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000272,"raw_usage":{"total_tokens":1443,"prompt_tokens":689,"completion_tokens":754,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":433,"completion_tokens_details":{"reasoning_tokens":655}},"tokens_in":433,"tokens_out":754,"duration_ms":7173,"temperature":1.0,"reasoning_tokens":655,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T01:38:46.979699+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the SHG for a p-polarized pump while detecting only the x-polarized harmonic component, and separately extract d15 and d31 from geometries that isolate each coefficient. If d15 and d31 differ by more than the reported fit uncertainties, the paper's tensor assignment and its d31 value are incorrect.","supporting_citations":[],"review_version":1}