{"id":"9acb00b5-332b-4b23-bb0b-0bc399283abe","arxiv_id":"2607.16699","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A vanishing ninth Fourier peak in difference-of-Gaussians maps of raw STM images identifies the triple-step period of nominally Si(557) wafers as 18b = 5.99 nm, i.e. local Si(8 8 11) orientation.","lead":"A new image-analysis method reads the period of atomic-step arrays on angled silicon surfaces directly from raw microscope images, without fixing scanner distortions first. Applied to nominally Si(5 5 7) wafers, it reports a local step period of 5.99 nm — the signature of the Si(8 8 11) orientation, not the nominal one.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"DoG contrast is assumed, not derived, to reproduce the adatom-density zero at m=9; step-edge curvature with period L may fill the predicted suppression.","rationale":"The reader's weakest_assumption already identified the idealised profile in Eq. (4) and the unproven DoG-as-density mapping as the key vulnerability. Our stress test sharpens this to a specific, testable mechanism: the DoG filter's sensitivity to step-edge curvature, which is absent from the flat model images but present in the experimental D. This is not an internal inconsistency in the algebra (Eq. (6) is correct and parameter-free), but it is a missing link in the application to real surfaces. The paper's own Conclusion admits the approach cannot be applied directly to topography because of quantized height changes, yet the experimental pipeline computes D from raw topography, so the step-edge contribution must be quantified. The proposed simulation would settle whether the m=9 zero survives the DoG of a realistic triple-step profile. Given that the reader's verdict was already CONDITIONAL on this type of validation, our concern does not change the verdict; it reinforces the need for the stated conditions. Agreement is partial because we focus on the step-edge contamination rather than the exact adatom offsets, though both are facets of the same model-dependence.","tokens_in":14575,"tokens_out":3875,"duration_ms":38676,"concrete_test":"Simulate a one-dimensional profile combining the 7×7 adatom density of Eq. (4) (L=18b, weights 3:1:2:2:1:3, σ≈0.1 nm) with the triple-step height profile: a sawtooth of period 18b and step height 3d=0.941 nm (with rounding to mimic tip convolution). Apply Eq. (1)'s DoG with σ1=0.1 nm, σ2=0.18 nm to the sum, then compute the FFT amplitude at k_x=2π·9/(18b), k_y=0. Repeat without the step-height term. If the amplitude with the step term is not near zero while the density-only case is zero, the central inference fails; if both are zero, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central inference equates the zero in the Fourier amplitude of the averaged adatom density (Eq. (4), amplitudes A_m in Eq. (6)) with suppression of the ninth peak in the FFT of the experimental DoG image D(x,y). But D is not a direct measurement of \\bar{z}(x); it is a band-pass filtered version of raw topography (Eq. (1)), sensitive to local curvature. The theoretical model images in Fig. 4 are flat 7×7 stripe density maps with no triple-step height jumps; the real surface combines adatom corrugation with ~0.94 nm step edges repeating with the same period L. DoG responds to curvature, so step edges produce their own Fourier harmonics at all m, with amplitudes generally nonzero at m=9. The paper asserts without derivation that D 'provides simultaneous imaging of atomic lattices' and that the FFT of D can be compared with the density model. The Conclusion even states the approach cannot be directly applied to topography because of tilt and quantized height changes, yet the experimental D is computed from raw topography via Eq. (1). Unless the step-edge contribution to D is shown to have a node at k_x=2π·9/L, the observed suppression of the ninth peak may be a contrast artifact rather than evidence for L=18b.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes that the periodicity of multiatomic step arrays on vicinal Si(hhm) surfaces can be determined from raw STM topography images by computing difference-of-Gaussians (DoG) maps and examining the FFT of these maps. A one-dimensional model represent the averaged adatom density of Si(111)7x7 stripes, with the adatom rows at positions ±b/2, ±3b/2, ±5b/2 relative to the stripe center and weights 3:1:2:2:1:3, leads to an analytical expression (Eq. 6) for the relative amplitudes A_m of Fourier orders. The zeros of A_m depend on L/b: for L=16b, m=±8 vanish; for L=17b, m=±5 and ±12 vanish; for L=18b, m=±9 vanish. The authors then apply this to a nominally Si(557) wafer, observe strong suppression of the ninth Fourier peak in the FFT of the DoG map, and conclude that the local period is 18b = 5.99 nm and that the surface corresponds to Si(8811). A second example on Si(556) gives L=32b and suppression of m=±16.","tokens_in":14753,"tokens_out":5865,"duration_ms":60826,"significance":"The analytic result in Eq. (6) is clean, parameter-free, and gives a striking and verifiable interference effect: the period-dependent suppression of selected Fourier peaks is a direct consequence of the specific arrangement of adatom rows. If the experimental connection can be made rigorous, the method could indeed provide a distortion-robust way to calibrate STM images and identify step periodicities on vicinal surfaces. The paper also correctly emphasizes that the zeros of A_m are unaffected by scaling or affine distortions, which is a genuine practical advantage. However, the central experimental claim relies on an identification between the DoG signal and the adatom-density model that is not theoretically justified. The manuscript does not yet establish that the suppression of the ninth peak in the experimental DoG spectrum is caused by the model's density profile rather than by the step-edge contrast that is necessarily present in DoG images of a triple-step staircase.","major_comments":[{"comment":"The model of Eq. (4) and the theoretical images in Fig. 4 describe the averaged adatom density on flat 7x7 stripes. The experimental observable D(x,y) is a band-pass filtered version of the raw topography z(x,y) via Eq. (1), and the surface contains triple steps of height ~0.94 nm repeating with the same period L. The Fourier transform of D is the sum of the adatom contribution and the step-edge contribution. The zero in A_m for m=9 does not imply a zero in the total DoG spectrum unless the step-edge contribution also vanishes at that k_x and k_y=0, or is proven negligible. The paper neither derives the step-edge contribution nor performs a simulation of a realistic stepped surface. The Conclusion explicitly states that the approach 'cannot be directly applied to topography images' because of tilt and quantized height changes, yet the experimental D is computed from raw topography via Eq","section":"§ Experiment: nominal Si(5 5 7)...; Eq. (1); Eq. (4); Fig. 4 vs Fig. 7"},{"comment":"The central experimental evidence is taken from a single, hand-selected region ('the right part of panel b') where the 7x7 stripes 'display enhanced uniformity.' No quantitative selection criterion is given, and no independent repeated experiments or statistical analysis over multiple areas are presented. The later direct measurement of inter-terrace distances in Fig. 8 uses the FFT-determined L=18b as the reference to correct the image, so it is not an independent confirmation. The suppression of the ninth peak should be shown to be robust across multiple images and across different choices of the DoG parameters to rule out selection bias.","section":"§ Experiment: nominal Si(5 5 7)...; Fig. 7"},{"comment":"The exact zero of A_m at m=9 depends on the assumed atomic row positions (at ±b/2, ±3b/2, ±5b/2) and the 3:1:2:2:1:3 weighting. The real atomic structure of triple steps on vicinal Si surfaces is explicitly disputed in the literature (refs [12–23]) and the authors defer the structural model to an unpublished manuscript (ref [25], 'in preparation'). If the actual structure places the adatom rows at different offsets, or if the step edges introduce additional rows with different weights, the zeros shift or disappear. The paper notes that adding/removing rows alters the effect, but it does not provide a sensitivity analysis. A small displacement of the rows could destroy the exact zero, so the inference from a single observed suppression to L=18b is less robust than the abstract implies.","section":"§ Model: periodic array of 7×7 stripes; Eq. (4)"}],"minor_comments":[{"comment":"There is a typographical error in the second integral: the Gaussian exponent is written with σ1² instead of σ2² in the denominator. As written, the second term is not a normalized Gaussian blur with width σ2. Please correct to exp(−((x−x')²+(y−y')²)/(2σ2²)).","section":"Eq. (1)"},{"comment":"The caption states 'image size 50×60 nm^{−2}' for the Fourier transforms. This should presumably be 'nm^{−1}' (or the axes labeled correctly) since k-space dimensions are inverse nanometers.","section":"Fig. 4 caption"},{"comment":"The abstract says the surface 'may correspond locally to Si(8 8 11)', while the experimental section states the observations 'readily deduce' that the local orientation is (8 8 11). Given the model-experiment gap noted above, the more cautious wording is appropriate.","section":"Abstract and § Experiment: nominal Si(5 5 7)..."},{"comment":"The Conclusion's statement that the approach 'cannot be directly applied to topography images' is confusing because the experimental method is applied to topography images after the DoG transformation. Clarify that the FFT analysis is applied to D(x,y), not to the raw z(x,y), and explain why the step-height information in z does not dominate the relevant part of the spectrum.","section":"Conclusion"}],"recommendation":"major_revision","confidential_remarks":"The paper's central mathematical derivation is sound and the method is potentially valuable, but the experimental conclusion hinges on an unproven equivalence between the DoG signal and the adatom-density model. This is fixable within the manuscript's scope by adding a simulation of a realistic stepped surface and/or an analytic treatment of the step-edge contribution. The reliance on an unpublished companion paper (ref [25]) for the structural model is also a concern for reproducibility; I would encourage the editor to insist that the necessary structural information be included in this manuscript or that the companion be made available. The paper is within the scope of the journal, and the level of interest is reasonably high given the longstanding controversy over the Si(557) surface orientation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Good to see the FFT suppression paper finally. The math part is genuinely nice: Eq. (6) is a closed-form, parameter-free expression for the Fourier amplitudes of a 7x7-stripe array, and the predicted zeros at m=9 for L=18b (and m=8, 5, 12 for other periods) check out exactly. That is a useful fingerprint for calibration-free metrology on narrow-terrace vicinal surfaces. The DoG trick on raw STM data is a sensible way to bring out atomic lattices, and the paper is honest that the approach can't be applied to plain topography because of tilt and step-height jumps.\n\nThe soft spot is the bridge between the model and the experiment. The model computes the Fourier transform of the averaged adatom density. The experimental signal D(x,y) is a difference of Gaussians of the raw topography, which is a curvature-sensitive filter. Step edges produce their own curvature signal, periodic with the same period L, and that signal will in general put weight at m=9. The paper simply asserts that the DoG map mirrors the adatom density; it doesn't derive it. Given they state they can't use the raw topography directly, they need to show that the step-edge contribution to D is also zero at the predicted positions, or that it's negligible at those k-vectors. Without that, the suppressed ninth peak might be a contrast artifact, not a proof of L=18b. This is the load-bearing chink.\n\nThe experimental evidence is also thin: one hand-selected subregion of a single image, qualitative 'strongly suppressed' / 'notably intense' statements, no noise analysis or multiple images, and the raw data aren't released. The rescaling uses the predicted period to calibrate the image, then finds 18b, 19b, 20b terraces, which is partly circular; the 8.9° slope and the 7x7 adatom registration give some independent support, but not enough. The companion paper [25] may have the missing statistical detail, but it's 'in preparation.'\n\nOn its own terms, the math is clean and the paper is serious about its limitations, so it deserves a proper referee—conditional, but a real candidate for publication if the authors can justify the DoG-density relation and provide quantitative FFT intensities. If I got this for review, I'd send it back for those two things before acceptance.","headline":"The zero-prediction in Eq. (6) is real and useful; the Si(8811) claim rests on an unproven link between DoG contrast and the adatom-density model.","tokens_in":15457,"tokens_out":5654,"would_cite":true,"duration_ms":51187,"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 period of triple-step arrays on silicon vicinal surfaces can be read directly from raw scanning tunneling microscopy images: when the ninth Fourier peak vanishes, the array period is 18 atomic rows, identifying the local Si(8 8 11) plan","keywords":["difference-of-Gaussians","Fourier analysis","scanning tunneling microscopy","vicinal silicon surfaces","triple steps","Si(557)","Si(8 8 11)","period determination"],"falsifier":"Acquire an atomically resolved STM image of an individual terrace on a Si(557) triple-step staircase and measure the actual lateral offsets of the 7×7 adatom rows relative to the 1×1 lattice. If those offsets are not ±b/2, ±3b/2, ±5b/2, then Eq. (6) predicts a nonzero ninth peak, and the claimed 18b period inference fails. Alternatively, if a similar wafer prepared by the same recipe shows no suppression of the ninth peak in the DoG-FFT analysis, the identification of Si(8 8 11) is contradicted.","tokens_in":14331,"feed_emoji":"🔬","tokens_out":4869,"duration_ms":44203,"temperature":0.7,"pith_summary":"This paper argues that the periodicity of multiatomic step arrays on high-index silicon surfaces can be determined from raw STM images, without tilt or drift corrections, with precision close to one atomic row. The key is to Fourier-analyze difference-of-Gaussians maps of the topography; the relative intensities of the Fourier peaks at k_y=0 are controlled by the array period, and specific peaks are completely suppressed for particular periods. For the triple-step staircase on nominally Si(557) wafers, the suppression of the ninth peak implies a period of 18b = 5.99 nm in projection onto the Si(111) terrace plane, meaning the local surface is actually Si(8 8 11). The same method is shown to work for wider terraces on Si(556), and should generalize to other vicinal surfaces with terraces containing an integer number of reconstruction unit cells.","feed_headline":"One vanished Fourier peak exposes Si(557)'s true step period","feed_subtitle":"Raw STM maps without corrections pin the triple-step spacing to 5.99 nm and reveal an unlisted Si(8 8 11) plane.","key_machinery":"The machinery is the difference-of-Gaussians (DoG) filter — subtracting two Gaussian-blurred copies of the topography image — applied to raw STM data, followed by a 2D FFT. DoG highlights local curvature where adatoms sit, flattening tilt and background so that atomic lattices appear across all terraces simultaneously. The Fourier amplitudes along k_y=0 are then interpreted with the closed-form envelope A_m(L) of Eq. (6), derived from a 1D model of the averaged adatom concentration with a 3:1:2:2:1:3 row-weight pattern. Commensurability between the 7×7 reconstruction wave number k0/7 and the array period wave number k0/L causes the zeros of this envelope to fall exactly on certain integer or","core_discovery":"On the paper's own terms, the central discovery is an interference effect in the Fourier transform of a periodic array of atomically flat terraces: for a stripe of width 5b with the Si(111)7×7 reconstruction repeating with period L, the amplitude of the m-th Fourier peak at k_y=0 is A_m = 3cos(5πm b/L) + cos(3πm b/L) + 2cos(πm b/L). This expression vanishes for specific (m, L) pairs — m=±8 for L=16b, m=±5 and ±12 for L=17b, m=±9 for L=18b — because the adatom rows at offsets ±b/2, ±3b/2, ±5b/2 with weights 3:1:2:2:1:3 produce exact cancellation at the corresponding wave vectors. In the experimental DoG map of a nominally Si(557) wafer, the ninth Fourier peak is strongly suppressed while the","pith_inferences":["If the same envelope logic is applied to other reconstruction widths, one can predict the full set of suppressed orders for any L/b ratio; this could be turned into a general lookup rule for step-period metrology.","The zeros of A_m are sensitive to the assumed row offsets; comparing predicted and observed suppressions for different step geometries could discriminate between competing atomic models of triple steps.","The paper leaves open whether LEED patterns show analogous suppressed spots; modeling the electron-beam tilt could test the idea in reciprocal space, where no STM drift artifacts exist.","Using the crystal's own Fourier zeros as a ruler, one could calibrate piezo scanners in situ by measuring which peaks disappear, without any external length standard."],"forward_implications":["The period of triple-step arrays on Si(557) can be measured to within one atomic row directly from raw STM images, without prior plane subtraction or drift correction.","The nominal Si(557) wafers studied are locally Si(8 8 11) with a 5.99-nm period, not Si(557) or Si(7 7 10) as earlier works claimed.","Applying the same reasoning to Si(556) yields a 32b-period array with a fully suppressed 16th Fourier peak, confirming the rule for wider terraces.","Because scaling and affine distortions do not change which orders vanish, the pattern serves as an in-image calibration reference for rescaling raw data.","The method extends to any vicinal surface whose terraces contain an integer number of reconstruction unit cells (e.g., 5×5, 7×7, 9×9)."],"fun_headline_variants":["Vanished Fourier peak nails Si(557) step period at 5.99 nm","Raw STM maps reveal hidden Si(8 8 11) plane via peak suppression","Missing ninth peak exposes true spacing of triple steps on Si","Fourier peak silence uncovers Si(557)'s real terrace period","Step spacing decoded from STM without calibration corrections"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The whole period assignment rests on the model that each 7×7 terrace's adatom rows sit exactly at ±b/2, ±3b/2, ±5b/2 within a 5b-wide cell with fixed weights; if the real atomic offsets differ, the predicted vanishing of the ninth peak shifts or disappears.","fun_headline_variants_meta":{"raw":{"variants":["Vanished Fourier peak nails Si(557) step period at 5.99 nm","Raw STM maps reveal hidden Si(8 8 11) plane via peak suppression","Missing ninth peak exposes true spacing of triple steps on Si","Fourier peak silence uncovers Si(557)'s real terrace period","Step spacing decoded from STM without calibration corrections"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000252,"raw_usage":{"total_tokens":1484,"prompt_tokens":919,"completion_tokens":565,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":663,"completion_tokens_details":{"reasoning_tokens":469}},"tokens_in":663,"tokens_out":565,"duration_ms":6217,"temperature":1.0,"reasoning_tokens":469,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T20:13:02.029348+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Acquire an atomically resolved STM image of an individual terrace on a Si(557) triple-step staircase and measure the actual lateral offsets of the 7×7 adatom rows relative to the 1×1 lattice. If those offsets are not ±b/2, ±3b/2, ±5b/2, then Eq. (6) predicts a nonzero ninth peak, and the claimed 18b period inference fails. Alternatively, if a similar wafer prepared by the same recipe shows no suppression of the ninth peak in the DoG-FFT analysis, the identification of Si(8 8 11) is contradicted.","supporting_citations":[],"review_version":1}