{"id":"ecde91ca-7833-473a-aa61-b23bc9978999","arxiv_id":"2508.12963","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Traditional Doppler Imaging of the Sun from HARPS-N spectra produces brightness maps matching SDO dopplergrams and enables detection of injected long-period planets with mass estimates comparable to Gaussian process regression.","lead":"Doppler Imaging, a technique that maps star spots from spectra, was applied to two years of HARPS-N solar spectra and recovered large-scale solar surface features that match SDO dopplergrams for active epochs. The paper shows this approach can filter stellar activity down to about 0.6 m/s and, if jointly fit with a planet model, recover low-mass long-period planet signals at accuracies comparable to Gaussian process methods.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The iterative intrinsic-profile self-calibration in Sec. 2.2.1 may absorb the chunk-mean of long-period planet signals, so the joint-fit mass recovery should be re-tested without using blind-DI-derived priors.","rationale":"The reader's weakest assumption identifies the same load-bearing element: the self-calibrated intrinsic CCF profile can absorb non-modulated signals. I agree with the conditional verdict and do not see a reason to move it. The concern is concrete and testable, but it has not yet been shown to actually break the central claim, because the joint fit may still constrain the within-chunk shape of the planet signal well enough for the reported 100-d recovery. The paper's independent support includes the SDO dopplergram comparison, the robustness simulations, and the injection-recovery tests; these are real evidence, but none of them directly isolates the chunk-mean absorption channel. I considered the lack of code release and the cited rather than recomputed GP residual comparison, but those are reproducibility issues rather than internal correctness risks. The proposed 500-d injection with uniform priors would settle whether the self-calibration biases planet mass estimates for the long-period regime that the paper emphasizes.","tokens_in":28655,"tokens_out":12113,"duration_ms":137202,"concrete_test":"Inject a synthetic planet with Porb=500 d and K=0.5 m/s into the same 439 solar CCFs, run the joint DI+planet fit of Section 5.2 with broad uniform priors on Porb and phase (no use of the blind-DI values from Table 3), and compare the recovered K. If K is biased low by more than about 20%, the self-calibration is absorbing the chunk-mean planet component and the long-period mass-recovery claim fails; if K is recovered within about 2 sigma, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 2.2.1 introduces a self-calibration loop because the intrinsic local CCF profile of a slow rotator is unknown: the median difference between observed and best-fit synthetic CCFs is subtracted iteratively until flat. This re-estimates a non-modulated profile component separately for each 27-day chunk (Section 2.2.2). A planet with Porb much larger than the chunk length produces a nearly constant RV offset within a chunk; that chunk-mean offset is exactly the kind of non-modulated signal the self-calibration can absorb into the intrinsic profile. The residual within-chunk variation is only a fraction of the full planet signal (about 0.4 times K for Porb=100 d and a 27-d chunk), so the recovered K is vulnerable to bias. The blind DI test in Section 5.1 indeed measures a roughly 40% underestimate of K, which the authors attribute to cross-talk between CCF derivatives. The joint fit in Section 5.2 is the proposed remedy, but its Porb and phase priors are taken from the same blind DI run (Section 5.2.2), so it is not a fully independent validation. If the chunk-mean absorption persists in the joint fit, the mass-recovery claim for long-period planets is weakened. A decisive test is therefore to run the joint fit on a much longer-period injected planet with wide uniform priors.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a proof-of-concept application of traditional maximum-entropy Doppler Imaging (DI) to Sun-as-a-star observations, using HARPS-N CCFs collected between December 2021 and January 2024. The data are split into 23 independent 27-day chunks, and for each chunk a relative brightness map is inverted from the CCF time series, after an iterative self-calibration of the unknown intrinsic line profile (Sec 2.2.1) and assuming non-evolving active regions (Sec 2.2.2). Three lines of evidence are presented. First, RVs extracted from the best-fit CCFs reproduce the observed activity-driven RVs with a residual RMS of 0.58 m/s, and a GLS periodogram of the residuals shows no remaining power at the rotation period or its harmonics. Second, the DI brightness maps correlate with large-scale (ell_max = 5) SDO/HMI dopplergrams at epochs of strong line-profile distortion: the median absolute Pearson correlation is about 0.63 at the phase of maximum |RV| per chunk, with correlations above roughly 0.5 only when |RV| exceeds about 2 m/s. Third, in planet injection-recovery tests, blind DI preserves the orbital period and phase of injected planets but underestimates K by about 40%, whereas a joint DI+planet fit recovers K within 1 sigma of the injected values for periods of 6 and 100 d, with uncertainties comparable to those from 1D and 2D Gaussian process regression.","tokens_in":28959,"tokens_out":21887,"duration_ms":209361,"significance":"The potential significance of this work is genuine: it appears to be the first demonstration that classic DI, designed for rapidly rotating stars, can be applied to a Sun-like slow rotator observed with a stable spectrograph, and that the reconstructed maps carry spatially resolved information grounded against independent, disc-resolved SDO observations. Several aspects of the paper are strengths that should be credited. The planet injection-recovery tests are built on the actual solar CCFs and include multiple phase realisations; the SDO comparison provides an external validation that is rare in this literature; the authors explicitly report the blind-DI K underestimate (about 40%) rather than hiding it; and the robustness tests in Sec 4.3 characterise the regime (noise level, number of epochs) in which the inversions are trustworthy. The paper is also appropriately hedged about map realism.","major_comments":[{"comment":"The joint activity+planet fit carries the abstract's central claim that DI 'yields planetary mass estimates with an accuracy comparable to, for example, multi-dimensional Gaussian process regression', but its validation has two gaps. First, the fit adopts Gaussian priors on Porb and phi_p that were obtained from the blind DI run on the same data (Sec 5.2.2: 'We thus adopt Gaussian priors for these two parameters, using the best-fit values and uncertainties reported in Tab. 3'), so the K recovery in Table 4 is not a blind validation of the full pipeline. Second, the longest injected period tested is 100 d, while the paper advertises the method for Porb > 100 d. The blind DI run already underestimates K by about 40% (Table 3), which the paper attributes to cross-talk between CCF derivatives; an equally plausible mechanism, acknowledged in Sec 2.2.1 ('could affect axisymmetric structures generating non-modulated profile distortions'), is that the iterative intrinsic-profile self-calibration absorbs the per-chunk mean of a long-period planet signal. For a 27-d chunk (Sec 3.2), a 100-d planet contributes only about 0.4 K of within-chunk RMS; for longer periods that constraint weakens further. I therefore request an additional injection-recovery test in which the joint fit is applied to a planet with Porb of about 200-300 d using wide uniform (or fully marginalised) priors on Porb and phi_p, reporting the recovered K and its credible interval. If K remains unbiased, the claim is supported; if not, the 'accuracy comparable to multi-dimensional GP' statement should be restricted to Porb of about 100 d or less.","section":"Sec. 5.2.1-5.2.2, Tables 3-4"},{"comment":"The headline residual RMS of 0.58 m/s is presented as 'consistent with that obtained with a Gaussian process regression on the RV time series (see Klein et al. 2024)', but the GP residual is not computed on the same 23-chunk dataset in this paper; the comparison rests entirely on a citation. Since the DI-corrected RVs are extracted from the best-fit CCFs to which the DI model was itself fitted (Sec 4.1 reports reduced chi2 = 1), the residual is partly a fit-quality statistic, and the per-chunk self-calibrated intrinsic profile (Sec 2.2.1) gives the model freedom on non-modulated components. To support the abstract's 'comparable with existing state-of-the-art activity correction techniques', I suggest computing a 1D GP regression on the same chunks, with the same noise model and outlier selection, and quoting the resulting residual RMS alongside 0.58 m/s; alternatively, the claim should be reworded to cite the specific Klein et al. 2024 residual and note that it was obtained on the same or a neighbouring time span.","section":"Sec. 4.1, Fig. 3"}],"minor_comments":[{"comment":"The units of the orbital phase phi_p are inconsistent: Eq. (4) adds phi_p to 2*pi*(T0 - t)/P as if it were an angle in radians, while the tabulated values (0.0, 0.33, 0.7) read as fractions of an orbital cycle. Please state the units explicitly and write 2*pi*phi_p in the sine argument if cycles are intended.","section":"Eq. (4) and Tables 2-3"},{"comment":"The clause 'values consistently larger than 0.5 for relative levels larger or equal to ~10^-4' appears to contradict the preceding statement that the correlation 'decreases roughly linearly with the noise level'; presumably 'smaller than or equal to ~10^-4' is intended.","section":"Sec. 4.3"},{"comment":"'gosts' in the list of instrumental contamination sources should read 'ghosts'.","section":"Sec. 3.1"},{"comment":"'intensitigrams' appears twice and should be 'intensitygrams'.","section":"Sec. 4.2"},{"comment":"The caption states that the color scale depicts the logarithm of the relative surface brightness, but the color bar is labelled 'Relative brightness [%]'; please make the labelling consistent.","section":"Fig. 2 caption"},{"comment":"The '~2 m/s' threshold for a good DI-SDO match is calibrated on only 23 chunks (Pearson rho_RV = 0.85 in Fig. 5) and should be explicitly presented as a tentative, small-sample calibration rather than a general detection criterion.","section":"Sec. 4.2, Fig. 5"},{"comment":"Minor wording issues: 'one of the most promising way' should be 'ways', and 'systemtically' in Sec 5.2.2 should be 'systematically'.","section":"Sec. 6.2"}],"recommendation":"major_revision","confidential_remarks":"This is a competent and honest proof-of-concept from an experienced group, and the HARPS-N solar testbed is well matched to the question. My main reservation is that the mass-recovery claim for the advertised long-period regime is validated only with period/phase priors derived from the same data and only up to Porb = 100 d; both gaps are fixable with additional injection-recovery tests, which I would regard as a standard requirement for the headline claim. An in-paper GP comparison for the residual RMS is also worth adding. The paper is well within the scope of MNRAS and, subject to these tests, would be a useful contribution to the RV-activity literature."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a solid proof-of-concept that Doppler Imaging can be applied to the Sun from disk-integrated HARPS-N spectra, with maps that correlate with SDO dopplergrams at active epochs, and a joint DI+planet fit that recovers injected planet semi-amplitudes about as well as 2D GP. The novelty is real: brightness DI of the Sun from CCFs validated against resolved images, plus the injection-recovery calibration for the solar case, are genuinely new.\n\nWhat's done well: the SDO comparison is the right independent check, and the phase dependence of the correlation (better at max RV) is physically sensible. The 1000-injection completeness test is well constructed and honestly shows DI's strengths (long periods) and weaknesses (near rotation harmonics). The blind DI test finding a ~40% underestimate of K is reported clearly, and the joint fit fixes it. That is a useful caution for anyone using DI as a blind filter.\n\nSoft spots: the residual comparison to GP is cited from Klein et al. 2024 rather than recomputed on the same data; probably fine, but it is not a like-for-like measure here. The joint fit uses period and phase priors taken from the same blind DI run, so the recovery is not fully independent; this is a mild circularity, not a fatal one, and the K prior is uniform. The stress-test worry that the per-chunk intrinsic-profile self-calibration absorbs the chunk-mean of a long-period planet is plausible for the blind DI, and that is exactly what the 40% deficit shows; but in the joint fit the planet signal is shifted out of the CCFs before DI, so the absorption mechanism should not persist. Still, a long-period injection with wide uniform priors would settle it. The SDO-DI correlation is moderate (mean absolute rho ~0.34 at phase 0), so the maps are meaningfully but not tightly tied to resolved solar structure. The DI code is not released, which limits full reproduction.\n\nBottom line: this is a serious proof of concept, not a finished tool. The central claims—DI can model solar activity and, when jointly fit, recover injected low-mass planet signals with accuracy comparable to GP—are supported by the tests. The paper deserves referee time; I would suggest the authors add the independent long-period injection test and release the DI code or provide more detailed reproducibility notes.","headline":"Brightness DI of the Sun from disk-integrated HARPS-N CCFs is a solid proof of concept, validated against SDO images, and the joint DI+planet fit recovers injected signals as well as 2D GP; the caveats are the cited GP baseline and mildly circular priors.","tokens_in":29555,"tokens_out":2870,"would_cite":true,"duration_ms":30429,"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":"Doppler Imaging, a technique built for rapidly rotating stars, can map the Sun's surface from high-stability spectra and filter its radial-velocity activity down to 0.58 m/s, opening a wavelength-domain route to long-period low-mass…","keywords":["Doppler imaging","stellar activity","radial velocities","Sun as a star","low-mass planet search","Gaussian process regression","injection-recovery","activity filtering"],"falsifier":"Repeat the joint DI-plus-planet injection-recovery on a star with a known transiting planet and semi-amplitude near 0.4 m/s, or on the Sun with the true local line profile measured from disc-resolved spectroscopy; if the joint fit is biased beyond its 1-sigma uncertainties, or if using the true profile instead of the iteratively flattened one changes the maps or residuals, the slow-rotator self-calibration is absorbing signal and the central claim fails.","tokens_in":28448,"feed_emoji":"☀️","tokens_out":13160,"duration_ms":127063,"temperature":0.7,"pith_summary":"The paper sets out to prove that Doppler Imaging can be moved from fast rotators to Sun-like stars and act as a stellar-activity filter for planet hunting. It reconstructs the Sun's relative surface brightness every 27 days from HARPS-N spectra, finds that the maps agree with large-scale SDO Dopplergrams when activity is strong enough, and reports RV residuals of 0.58 m/s after subtracting the DI model. In planet injection-recovery tests, DI recovers long-period, low-mass planet signals as accurately as Gaussian-process regression when the planet fit is run jointly with the imaging. If true, this gives planet hunters a physically motivated, wavelength-domain way to remove stellar noise that does not rely on time-domain covariance assumptions.","feed_headline":"Doppler imaging clears Sun-like starlight down to 0.58 m/s","feed_subtitle":"The same maps recover injected low-mass planet signals as accurately as 2D Gaussian-process regression.","key_machinery":"The central object is the Doppler Imaging inversion itself: a maximum-entropy reconstruction of relative surface brightness from a time series of cross-correlation functions. The surface is divided into roughly 10,000 cells, each with a brightness and a local line profile computed from an assumed radiative-transfer model; synthetic disc-integrated profiles are matched to the observed CCFs by conjugate-gradient iteration. Two adaptations make this work for the slow-rotating Sun: an iterative self-calibration of the unknown intrinsic line profile (subtract the median observed-minus-model CCF, re-fit, repeat until flat), and division of the two-year dataset into 27-day chunks over which active regions are assumed static. The planet search is an extension of the same code: shift the CCFs by a trial Keplerian signal, re-run the inversion, and map $\\chi^2$ over the planet grid, following the joint-fit method.","core_discovery":"The paper's central claim is that a maximum-entropy Doppler Imaging inversion of observed cross-correlation functions can serve as an activity model for Sun-like stars. Applied chunk-by-chunk to two years of HARPS-N solar spectra, DI reconstructs brightness maps with an angular resolution of about 36 degrees, matching large-scale SDO Dopplergrams (modelled to spherical-harmonic degree $\\ell_{\\max}=5$) with a median absolute correlation of 0.63 at the phase of maximum profile distortion, provided the disc-integrated RV exceeds about 2 m/s. The RVs read off the best-fit profiles reproduce the observed solar RV curve with 0.58 m/s residuals, and no power remains at the rotation period or its harmonics. The same inversion, run blindly, preserves the period and phase of an injected planet but underestimates its semi-amplitude by roughly 40 percent; when the planet fit is folded directly into the inversion by Doppler-shifting the CCFs, the recovered semi-amplitudes match the injected values within $1\\sigma$, with uncertainties comparable to a two-dimensional Gaussian-process fit. The paper concludes that DI is not yet an operational planet-search tool, but it is a credible physical, wavelength-domain alternative for suppressing activity around slowly rotating solar twins.","pith_inferences":["Beyond the paper, a multi-cycle Sun-like dataset could test whether DI's wavelength-domain filtering stays stable while Gaussian-process hyperparameters drift over a magnetic cycle; the two-year baseline here cannot address that.","Beyond the paper, using CCF masks built from lines with different formation temperatures should separate convective-bluesshift inhibition from brightness contrast, potentially yielding simultaneous intensity and velocity maps from one spectral time series.","Beyond the paper, a hybrid pipeline suggests itself: run blind DI to supply period and phase priors, then switch to a joint DI-plus-planet fit for unbiased masses, avoiding the 40 percent blind bias while keeping DI's long-period sensitivity.","Beyond the paper, the 2 m/s threshold implies DI will be noise-dominated for very quiet stars, so a sensitivity-floor test on synthetic quiet-star spectra would show where DI should yield to time-domain methods."],"forward_implications":["DI removes the rotationally modulated activity signal from Sun-like-star RVs, leaving 0.58 m/s RMS residuals with no periodicity at the rotation period or its harmonics.","DI brightness maps agree with disc-resolved SDO Dopplergrams on scales of about 36 degrees only when the activity-induced RV is above about 2 m/s; quieter epochs yield unreliable maps.","Blind DI gives unbiased orbital periods and phases but underestimates injected low-mass planet semi-amplitudes by about 40 percent; the paper's joint DI-plus-planet fit is required for accurate masses.","With the joint fit, DI recovers injected planet semi-amplitudes as accurately as a two-dimensional Gaussian-process model, with comparable 1-sigma uncertainties.","Injection-recovery completeness reaches 100 percent for injected planets with orbital period above 100 days and semi-amplitude above 0.4 m/s, and 67 percent below 0.4 m/s, while planets with periods near the rotation period and its harmonics are lost."],"supporting_citations":[{"why":"Supplies the HARPS-N solar data reduction, the GP rotation-period value, and the DCPCA and GP baselines that the DI results are compared against.","marker":"Klein et al. 2024"},{"why":"Supplies the maximum-entropy DI inversion framework that maps brightness from CCF time series.","marker":"Donati & Brown 1997"},{"why":"Establishes the resolution and intrinsic-profile limits of DI that motivate the slow-rotator adaptation.","marker":"Kochukhov 2016"},{"why":"Supplies the multidimensional Gaussian-process activity model whose semi-amplitude recovery accuracy is the comparison benchmark for DI.","marker":"Rajpaul et al. 2015"},{"why":"Supplies the joint activity-plus-planet fitting method that the paper uses to recover injected planet signals.","marker":"Petit et al. 2015"},{"why":"Introduces the iterative intrinsic-profile self-calibration that makes DI applicable to slow rotators.","marker":"Klein et al. 2021b"},{"why":"Extends that self-calibration framework and validates it for joint planet searches.","marker":"Klein et al. 2022"},{"why":"Provides the SOLASTER pipeline used to obtain SDO/HMI Dopplergrams, intensitygrams, and magnetograms for map comparison.","marker":"Ervin et al. 2022"},{"why":"Defines the data-selection criteria (cloud probability and zenith velocity corrections) for the HARPS-N solar spectra.","marker":"Collier Cameron et al. 2019"},{"why":"Supplies the DCPCA activity-filtering baseline against which DI completeness is measured.","marker":"Jones et al. 2017"}],"fun_headline_variants":["Solar Doppler mapping strips activity noise to 0.58 m/s","Doppler imaging sees planets behind Sun-like star jitter","DI clears stellar noise, reveals injected planets blind"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the Sun's local line profile can be replaced by an assumed model and then fixed by iteratively subtracting the median data-minus-model difference until flat; if that step absorbs axisymmetric activity or part of a planet's Doppler shift, the reconstructed maps and the 0.58 m/s correction are biased.","fun_headline_variants_meta":{"raw":{"variants":["Solar Doppler mapping strips activity noise to 0.58 m/s","Doppler imaging sees planets behind Sun-like star jitter","DI clears stellar noise, reveals injected planets blind"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000231,"raw_usage":{"total_tokens":1554,"prompt_tokens":1083,"completion_tokens":471,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":699,"completion_tokens_details":{"reasoning_tokens":418}},"tokens_in":699,"tokens_out":471,"duration_ms":6564,"temperature":1.0,"reasoning_tokens":418,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T17:17:04.830294+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the joint DI-plus-planet injection-recovery on a star with a known transiting planet and semi-amplitude near 0.4 m/s, or on the Sun with the true local line profile measured from disc-resolved spectroscopy; if the joint fit is biased beyond its 1-sigma uncertainties, or if using the true profile instead of the iteratively flattened one changes the maps or residuals, the slow-rotator self-calibration is absorbing signal and the central claim fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the joint activity-plus-planet fitting method that the paper uses to recover injected planet signals."}],"review_version":2}