{"id":"8f4c5fd9-f420-4bf1-b6b9-2f021e37a927","arxiv_id":"1908.02059","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":5,"one_line_summary":"Cepheid radial velocities vary by a few percent depending on the correlation template and fitting method, and a standardized catalogue of 64 stars shows centroid velocities give the most consistent results.","lead":"This astronomy paper measures how different choices in the cross-correlation technique change the radial velocities of 64 pulsating stars called Cepheids. It finds that the wavelength range, the chosen spectral lines, and the velocity-fitting method all shift the results, and recommends using centroid velocities and stronger lines for more consistent measurements.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The recommended preference for centroid v_rad rests on a 'significantly smaller scatter' claim that the manuscript does not actually quantify; if that scatter comparison is absent or non-significant, the paper's main practical recommendation is unsupported.","rationale":"The reader's template-mismatch concern is legitimate, but it mainly affects the astrophysical interpretation of the depth-dependent differences rather than the empirical demonstration that differently constructed templates change v_rad. The unsupported scatter claim is more directly load-bearing because the paper's practical recommendation to favour centroid v_rad is built on it. The paper has independent support for the broad central claim: a large homogeneous dataset, the external DRS check in Appendix B.4, and the direct line-width experiment in Section 4.5.1. My concern does not overturn those results; it identifies a missing quantitative analysis that conditions the headline recommendation. Since the reader already returned CONDITIONAL, the appropriate final verdict is UNCHANGED.","tokens_in":33860,"tokens_out":11492,"duration_ms":144153,"concrete_test":"Recompute Section 4.3 for the 47 selected targets: for each target, fit a common spline or Fourier model to the phased pulsation curve and compute the RMS residual of RVcc-c, RVcc-g, and RVcc-2g about that model; then compare the per-target residual distributions with a paired test. If centroid residuals are not significantly smaller (e.g., paired p > 0.05, or the effect disappears after matching by S/N and phase coverage), the abstract's 'should thus be favoured' should be downgraded.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's actionable conclusion, stated in the Abstract and Section 5, is that centroid v_rad (RVcc-c) should be favoured because they show 'significantly smaller scatter' than Gaussian or biGaussian v_rad. But the quantitative support for that claim is not presented in the manuscript. Section 4.3 reports only linear-regression slopes (RVcc-g ~1.01 and RVcc-2g ~1.03-1.04 relative to RVcc-c) and their dispersions; no RMS residual, scatter measure, or statistical test around a common pulsation model is given for the three methods. This matters because RVcc-c is a first moment integrated over the CCF core, so a smaller scatter could partly reflect the smoothing effect of the integration window rather than a genuine improvement in consistency. Without a quantified scatter comparison, the central recommendation 'centroid v_rad should thus be favoured' overreaches the evidence. The core empirical statement that v_rad values are method-dependent is otherwise well supported, but the paper's headline advice depends on this missing analysis.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a large homogeneous spectroscopic survey of 64 Classical Milky Way Cepheids, based on 3919 high-resolution spectra from seven echelle spectrographs. The authors standardise all spectra through a single processing chain, cross-correlate them with six tailored binary correlation templates that select unblended lines of different depths on three wavelength ranges, and derive from each CCF a set of line-profile observables plus three radial-velocity measurements: centroid (RVcc-c), Gaussian (RVcc-g), and biGaussian (RVcc-2g). They then compare the resulting v_rad time series as functions of the v_rad computation method, template wavelength range, template line width, and template line depth. The central claimed result, stated in the Abstract and in Section 5, is that each of these steps significantly impacts the derived v_rad, and that centroid v_rad, which show slightly smaller amplitudes but 'significantly smaller scatter' than Gaussian or biGaussian v_rad, should be favoured, together with stronger lines and somewhat broader template lines, to obtain more consistent Cepheid v_rad for Baade-Wesselink distance determinations.","tokens_in":34077,"tokens_out":4892,"duration_ms":47624,"significance":"If the claims hold, the paper makes a useful contribution: it provides a large, consistently processed catalogue of Cepheid CCFs and v_rad time series, demonstrates quantitatively that v_rad depends on template depth, width, wavelength range, and measurement method, and validates the s1d-based pipeline against the HARPS-North DRS with Pearson correlation coefficients between 0.99 and 1 (Appendix B.4). The decision to compute linear regressions over all measurements rather than comparing only extrema is a methodological improvement over earlier studies. The recommendation that authors specify their template and v_rad method is well motivated. However, the paper's most actionable claim, that centroid v_rad should be favoured because of significantly smaller scatter, is not quantitatively supported in the present text; this is a load-bearing gap that requires additional analysis.","major_comments":[{"comment":"The central recommendation, stated in the Abstract and reiterated in Section 5, is that centroid v_rad (RVcc-c) should be favoured because they exhibit 'significantly smaller scatter' than Gaussian or biGaussian v_rad. No quantitative scatter comparison is presented anywhere in the manuscript. Section 4.3 reports only linear-regression slopes (RVcc-g ~1.01, RVcc-2g ~1.03-1.04 relative to RVcc-c) and dispersions of those slopes across targets (Fig. 9). No per-target RMS residual, scatter around a common phased pulsation curve, or statistical test is given for the three methods. This is load-bearing: the practical recommendation rests on this claim. In addition, RVcc-c is a first-moment integral over the CCF core, so a smaller scatter could partly reflect the smoothing effect of the integration window rather than a genuine improvement in consistency. The authors should provide a direct scatter comparison (e.g., RMS of residuals around a spline fit to the phased v_rad curve for each method and target) and a statistical test (e.g., paired F-test or Wilcoxon signed-rank test on per-target scatter).","section":"Abstract, §4.3, §5"},{"comment":"Section 3.5 states: 'We did not try to definitively assess which method is to be preferred.' This is in direct tension with the Abstract and Section 5, which explicitly recommend centroid v_rad over Gaussian and biGaussian v_rad. This internal inconsistency suggests that the scatter-based recommendation may not have been fully analysed or validated. The authors should either remove the disclaimer in §3.5 or qualify the conclusions to match the available evidence, which currently supports method-dependence but not a definitive ranking of the three methods.","section":"§3.5 vs Abstract and §5"},{"comment":"The template line selection is based on a single synthetic PHOENIX Cepheid spectrum (Teff = 5250 K, log g = 1, solar metallicity), while the sample spans spectral types F8-G5 and periods from roughly 2 to 68 days. The manuscript does not quantify how template mismatch would affect the selection of 'un-blended' lines for stars with different Teff/log g, and hence how it would affect the depth-dependent comparisons in Section 4.6. If the reference spectrum is not representative, the weak/medium/deep template comparisons could partly reflect template mismatch (blended or absent lines in real spectra) rather than cleanly probing line-formation depth. The authors should test the robustness of their line selection and of the Section 4.6 conclusions by repeating the selection with one or two additional PHOENIX models spanning the sample's parameter range (e.g., Teff = 6000 K, log g = 2; Teff = 5000 K, log g = 1) and verifying that the trends in CCF quality, asymmetry, and v_rad persist.","section":"§3.3 and §4.6"}],"minor_comments":[{"comment":"In the last full paragraph of Section 4.1, 'which we will show latter' should read 'which we will show later'.","section":"§4.1"},{"comment":"The uncertainty formula for RVcc-c, epsilon_cc-c = W/SNRCCF, is described in the text as 'arbitrary'. Consequently, the comparison of uncertainty magnitudes among RVcc-c, RVcc-g, and RVcc-2g in Fig. B.1 is not informative for ranking the methods and should be explicitly labelled as such in the main text.","section":"Appendix B.3"},{"comment":"The comparison with the G2 HARPS DRS template is performed only for δ Cep (103 HARPS-North spectra); this should be stated explicitly in the main text rather than only in the caption of Fig. 12.","section":"§4.5.2"},{"comment":"The left panel of Fig. 9 would benefit from a direct statement in the text or the caption of the median and 1σ dispersion of the slope distributions for RVcc-g vs RVcc-c and RVcc-2g vs RVcc-c, since these values are quoted in the text.","section":"Fig. 9"},{"comment":"For the last row (G2/HARPS template), the values N_l = 1725 and sigma_l = 0.08 Å are valid over the green range only; this should be stated explicitly in the table caption or column header to avoid ambiguity.","section":"Table 2"},{"comment":"In the discussion of the red versus blue v_rad comparison, the sentence 'Such studies would need to be extended to infrared (IR) wavelengths in order to be confirmed' is vague; the authors could indicate which specific infrared wavelengths or instruments would be relevant.","section":"§4.4"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reports a valuable and carefully processed dataset, and the core empirical finding that Cepheid v_rad depend on template and method is well supported. The main obstacle is the unsupported 'significantly smaller scatter' claim for centroid v_rad, which underpins the headline recommendation; this requires a dedicated quantitative analysis. The template-mismatch concern is also worth addressing, but it is less severe and can be handled with robustness tests. With the missing scatter analysis added, the paper would likely be acceptable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Colleague],\n\nRead Borgniet et al. on consistent Cepheid RVs from cross-correlation. The paper is worth a serious look. It's a large, homogeneous study—3900+ spectra, 64 Cepheids, seven spectrographs—that quantifies how template wavelength range, line depth, line width, and RV fitting method shift the derived velocities. The central empirical result, that Cepheid RVs depend strongly on these choices, is solid and extends earlier work by Nardetto and Anderson to a much larger sample with a consistent framework.\n\nWhat the paper does well: the processing pipeline is transparent; the validation against the HARPS-North DRS in Appendix B.4 gives Pearson correlations of 0.99–1.0, which is real evidence of reproducibility; and the regression approach uses all measurements rather than extrema. If the catalogue and templates are actually released, this becomes a useful community resource. The authors also honestly flag the red-blue wavelength trend as provisional.\n\nThe soft spots are addressable. First, the paper's headline recommendation—that centroid RVs should be favoured because they show 'significantly smaller scatter' than Gaussian or biGaussian RVs—is not quantified anywhere in the text. Section 4.3 reports only regression slopes and their dispersions; there is no scatter comparison around a common pulsation model for the three methods. So the practical advice rests on a statistic the reader never sees. The authors need to provide that comparison and, ideally, check whether the smaller scatter is partly a smoothing effect of the centroid integration window. Second, the single PHOENIX template (Teff=5250 K, log g=1, solar metallicity) is used to select unblended lines for all 64 Cepheids across F8–G5. The paper does not quantify how the line selection degrades for the hotter or cooler stars. If lines are blended or missing at those temperatures, the depth-dependent RV differences could be partly a template-mismatch artifact. That needs a robustness test, e.g., repeating the template construction with a different Teff. Third, the catalogue and templates are promised but not provided with the preprint—a practical condition for final acceptance.\n\nWho this is for: anyone using Cepheid RVs for Baade-Wesselink distances or p-factor work, and anyone cross-correlating spectra of pulsating stars. The paper deserves a serious referee. The main claim holds up; the recommendation on centroid RVs needs the missing scatter analysis, and the template robustness needs a test. I'd send it to review with those conditions, not reject it.","headline":"A methodologically solid, large-sample study confirming that Cepheid RVs are strongly method-dependent; the practical recommendation to favor centroid RVs is under-supported by a missing scatter analysis.","tokens_in":34668,"tokens_out":3852,"would_cite":true,"duration_ms":40562,"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":"Cepheid radial velocities shift with every cross-correlation choice, and centroid velocities from deep, broad-line templates are the most consistent.","keywords":["Cepheids","radial velocities","cross-correlation function","binary correlation templates","Baade-Wesselink technique","line asymmetry","pulsating stars"],"falsifier":"Rebuild the weak, medium, and deep templates from synthetic spectra at the hot and cool ends of the sample, for example Teff near 6000 K and 4500 K, and re-derive the depth-dependent v_rad offsets on the same spectra. If the offsets change sign or disappear for the extreme stars, the single-reference template, not a physical velocity gradient, produced the reported trend; a check of whether the selected unblended lines are actually unblended in observed spectra of the hottest and coolest Cepheids would settle the same question directly.","tokens_in":33631,"feed_emoji":"🔭","tokens_out":7323,"duration_ms":128172,"temperature":0.7,"pith_summary":"Using 3919 high-resolution spectra of 64 Milky Way Cepheids, the paper argues that every step of the cross-correlation pipeline changes the measured radial velocity: the wavelength range, the template's mean line depth and line width, and the method used to read a velocity off the cross-correlation function. A sympathetic reading of the results is that published Cepheid v_rad values are strongly method-dependent, and that the same star can yield different pulsation amplitudes, and therefore different Baade-Wesselink distances, depending on those choices. The paper's recommendation follows from its own observables: centroid (first-moment) velocities show smaller scatter than Gaussian or biGaussian fits, and templates built from stronger and broader lines reduce the CCF asymmetry that biases velocity measurements. This matters because Cepheid v_rad feed directly into projection factors and distance estimates.","feed_headline":"Cepheid velocities shift with every cross-correlation choice","feed_subtitle":"A 64-star survey shows centroid velocities and strong, broad-line templates give the most consistent results.","key_machinery":"The load-bearing object is the cross-correlation function (CCF), a single average line profile obtained by sliding a binary correlation template across the spectrum. The templates are built from a synthetic PHOENIX Cepheid spectrum with Teff = 5250 K, log g = 1, and solar metallicity, by selecting un-blended lines in three depth bins (weak, medium, deep), plus an all-line depth-weighted template on the green range and medium-depth templates on blue and red ranges. The CCF is characterized by depth, width, equivalent width, bisector inverse span (BIS) for asymmetry, a quality factor Q, and a signal-to-noise proxy; velocities are then extracted three ways: centroid, Gaussian, and biGaussian fits. This machinery isolates the effect of each choice because only one ingredient changes at a time across otherwise identical spectra.","core_discovery":"The central discovery is that the Cepheid radial velocity is not a single number but a function of the measurement recipe. Cross-correlating the same spectra with six tailored templates, weak, medium, and deep lines on a green range, an all-line depth-weighted template, and medium templates on blue and red ranges, and then extracting velocities by centroid, Gaussian, and biGaussian fits, the paper finds significant offsets and amplitude changes in every comparison. Gaussian velocities run about 1% larger than centroid velocities, biGaussian velocities about 3-4% larger, and the differences grow for shallower lines and shorter periods. Deeper lines produce less asymmetric CCFs and more robust velocities; wider template lines reduce wing noise and asymmetry at the cost of a shallower core. The paper concludes that consistent Cepheid v_rad time series should favor centroid velocities and templates made of stronger, broader lines, and that any Baade-Wesselink study should specify these choices because each implies a different projection factor and distance.","pith_inferences":["Beyond the paper: the same line-selection logic could be applied to other pulsating stars, such as RR Lyrae, whose data-reduction templates are similarly built from dwarf spectra and may carry the same method-dependent bias.","Beyond the paper: building the weak, medium, and deep templates from synthetic spectra spanning the sample's full effective-temperature range would directly test whether the depth-dependent v_rad offsets persist or are partly an artifact of the single 5250 K reference spectrum.","Beyond the paper: the released CCFs could be used to check whether centroid-v_rad based p-factors actually reduce the scatter of Baade-Wesselink distances across the 64-star sample, a test the paper does not run.","Beyond the paper: the observed decrease in v_rad amplitude from blue to red wavelengths, if confirmed on more targets, could be used as a spectroscopic probe of the Cepheid atmospheric velocity gradient rather than treated as noise."],"forward_implications":["Published Cepheid v_rad values are method-dependent: the same star can show different pulsation amplitudes, and hence different projection factors and Baade-Wesselink distances, depending on template and estimator.","Centroid v_rad should be favored for distance work: despite slightly smaller amplitudes, their scatter is significantly smaller than Gaussian or biGaussian v_rad.","Templates built from deeper and broader lines are more robust: they reduce CCF asymmetry and yield more consistent v_rad time series.","Any v_rad publication for Cepheids should report the wavelength range, template line selection, line width, and v_rad estimator, because each materially changes the result.","The published catalogue of templates, CCFs, observables, and v_rad time series enables homogeneous studies of Cepheid binarity, period-luminosity relations, and p-factor calibration."],"supporting_citations":[{"why":"Introduces the CCF technique and the template line-width effects this paper exploits; predicts that broader template lines broaden the CCF core and reduce wing noise.","marker":"Queloz 1995"},{"why":"Defines the binary-correlation-template transparency and the CCF method for precise radial velocities.","marker":"Baranne et al. 1979"},{"why":"Establishes single-line centroid versus Gaussian and biGaussian v_rad conventions and the p-factor decomposition that motivates the three v_rad estimators.","marker":"Nardetto et al. 2006"},{"why":"Shows Cepheid CCF asymmetry depends on template line strength; supplies the weak-line versus strong-line comparison this paper extends.","marker":"Anderson 2016"},{"why":"Reports decreasing Cepheid v_rad amplitude with wavelength, the trend the paper checks across green, red, and blue ranges.","marker":"Nardetto et al. 2009"},{"why":"Provides the PHOENIX model atmosphere code used to generate the reference synthetic Cepheid spectrum for line selection.","marker":"Hauschildt & Baron 1999"},{"why":"Supplies the line-selection approach of depth-bin selection of unblended lines that is adapted to build the templates.","marker":"Hindsley & Bell 1986"},{"why":"Describes HARPS DRS binary-mask cross-correlation and Gaussian fitting against which the paper compares its own template choice.","marker":"Pepe et al. 2002"},{"why":"Supplies the public G2 template data used for the default-template comparison.","marker":"Brahm et al. 2017"}],"fun_headline_variants":["Cepheid velocities shift with each cross-correlation method","Centroid fits yield most consistent Cepheid velocities","Template choice alters Cepheid radial velocity readings","Strong-line templates improve Cepheid velocity precision","Cepheid radial velocities depend on measurement recipe"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The single synthetic spectrum used to choose template lines is assumed to represent all 64 Cepheids; if real spectra of hotter or cooler stars differ, the depth-dependent velocity differences could be an artifact of that template choice.","fun_headline_variants_meta":{"raw":{"variants":["Cepheid velocities shift with each cross-correlation method","Centroid fits yield most consistent Cepheid velocities","Template choice alters Cepheid radial velocity readings","Strong-line templates improve Cepheid velocity precision","Cepheid radial velocities depend on measurement recipe"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000228,"raw_usage":{"total_tokens":1560,"prompt_tokens":1115,"completion_tokens":445,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":731,"completion_tokens_details":{"reasoning_tokens":371}},"tokens_in":731,"tokens_out":445,"duration_ms":4701,"temperature":1.0,"reasoning_tokens":371,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:55:00.241668+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Rebuild the weak, medium, and deep templates from synthetic spectra at the hot and cool ends of the sample, for example Teff near 6000 K and 4500 K, and re-derive the depth-dependent v_rad offsets on the same spectra. If the offsets change sign or disappear for the extreme stars, the single-reference template, not a physical velocity gradient, produced the reported trend; a check of whether the selected unblended lines are actually unblended in observed spectra of the hottest and coolest Cepheids would settle the same question directly.","supporting_citations":[{"cited_title":"1995, in IAU Symposium, Vol","cited_arxiv_id":null,"evidence_quote":"Introduces the CCF technique and the template line-width effects this paper exploits; predicts that broader template lines broaden the CCF core and reduce wing noise."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the binary-correlation-template transparency and the CCF method for precise radial velocities."},{"cited_title":"2006, , 453, 309","cited_arxiv_id":null,"evidence_quote":"Establishes single-line centroid versus Gaussian and biGaussian v_rad conventions and the p-factor decomposition that motivates the three v_rad estimators."},{"cited_title":"2009, , 502, 951","cited_arxiv_id":null,"evidence_quote":"Reports decreasing Cepheid v_rad amplitude with wavelength, the trend the paper checks across green, red, and blue ranges."},{"cited_title":"& Bell , R","cited_arxiv_id":null,"evidence_quote":"Supplies the line-selection approach of depth-bin selection of unblended lines that is adapted to build the templates."}],"review_version":1}