Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

Rotational modulation and long-term evolution of the small-scale magnetic fields of M dwarfs observed with SPIRou

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Rotation periods of M dwarfs can be recovered from time series of the average small-scale magnetic field in unpolarized spectra, matching polarized-light periods in four of six stars, with month-to-year variations largely independent of the

desk verdict A careful multi-year SPIRou study that convincingly recovers rotation periods from small-scale field series for three M dwarfs; the '4 of 6' headline count overstates the independent detections. read the letter →

arxiv 2508.04569 v1 pith:ENSFUKCZ submitted 2025-08-06 astro-ph.SR

P. I. Cristofari , J.-F. Donati , S. Bellotti , É. Artigau , A. Carmona , C. Moutou , X. Delfosse , P. Petit
show 2 more authors
B. Finociety J. Dias do Nascimento
This is my paper · ORCID
classification astro-ph.SR
keywords Mdwarfssmall-scalemagneticfieldsZeemanbroadeningrotationperiodsGaussianprocessesSPIRouactivitycyclesspotcoverage
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to show that the rotation periods of M dwarfs can be recovered from the way the average small-scale magnetic field, measured by fitting Zeeman-broadened synthetic spectra to unpolarized near-infrared observations, varies over time. Monitoring six M dwarfs with SPIRou between 2019 and 2024 (two partially convective, three fully convective, and EV Lac near the fully convective boundary), the authors detect the rotation period in the small-scale-field series for four of them, with values consistent with periods from polarized longitudinal-field measurements: EV Lac at 4.362±0.001 d (vs 4.371 d), DS Leo at 13.953±0.088 d (vs 14.001 d), and the slow rotator Barnard's star at 144±6 d. This matters because the small-scale field carries most of the unsigned magnetic flux, so a demonstration that it rotates coherently means the large unpolarized surveys already in hand can deliver rotation periods and magnetic activity timescales without polarimetry. The paper also reports year-to-year field fluctuations of up to 0.3 kG (CN Leo), phase-to-phase swings of up to 1 kG (EV Lac), little correlation between small-scale and large-scale field series, and a persistent anti-correlation between field strength and temperature-variation measurements for three stars.

What carries the argument

The method rests on two pieces. First, synthetic spectra computed with ZeeTurbo from MARCS model atmospheres, with surface fields from 0 to 10 kG in 2 kG steps assumed radial, are fitted to each night's unpolarized spectrum as a linear combination $S=\sum_i f_i S_i$ of magnetic components with filling factors $f_i$ via an MCMC likelihood; each spectrum becomes one average field value $\langle B\rangle$. Second, a quasi-periodic Gaussian-process kernel $$\kappa(t_i,t_j)=\$alpha^{2}$\exp\!\left[-\frac{(t_i-t_j)^2}{$2l^{2}$}-\frac{1}{2\$beta^{2}$}\$sin^{2}$\!\left(\frac{\pi(t_i-t_j)}{P_\mathrm{rot}}\right)\right]+\$sigma^{2}$\delta_{ij}$$ recovers the recurrence period $P_\mathrm{rot}$ (the rotation period) from t

What would settle it

Feed the pipeline synthetic spectra with a constant, known magnetic field but sampled at the same nightly cadence with comparable noise: if the Gaussian-process fit returns a spurious rotation period, the detected modulation is a modeling artifact, not a stellar signal. As a cross-check, recompute $\langle B\rangle$ with a finer Zeeman grid (1 kG or 0.5 kG steps) and with non-radial field geometries and test whether recovered periods and phase-curve amplitudes shift beyond the quoted uncertainties.

Watch

Extended reading notes

Core claim

The central claim is that the average small-scale magnetic field $\langle B\rangle$ of an M dwarf, measured from unpolarized spectra via Zeeman-broadening fits, carries a coherent rotational modulation that a quasi-periodic Gaussian-process fit can extract as the star's rotation period. Across six SPIRou targets the paper recovers periods in the $\langle B\rangle$ series for EV Lac (4.362±0.001 d), DS Leo (13.953±0.088 d), Barnard's star (144±6 d), and, with a prior, PM J18482+0741 (2.762±0.009 d); CN Leo's period is not clearly detected, and pole-on AD Leo (~20°) shows no modulation. The paper also reports year-to-year variation of up to 0.3 kG (CN Leo) and up to 1 kG across rotation phases

Load-bearing premise

The load-bearing premise is that night-to-night changes in the fitted field strength are real variations of the star's small-scale magnetic field, not artifacts of the spectral model (radial fields, local thermodynamic equilibrium, fixed atmospheric parameters, discrete 2 kG field steps) or of the Gaussian-process rescaling of error bars (Secs. 3.1 and 4.1).

Editorial extensions

If this is right

  • Unpolarized spectra alone can yield M dwarf rotation periods, so multi-year unpolarized surveys already in hand can be mined for rotation and activity timescales without polarimetry; the recovery works for slow rotators (Barnard's star, 144±6 d) when inclination is high and signal-to-noise sufficient.
  • The average small-scale field fluctuates by up to 0.3 kG within a year (CN Leo) and up to 1 kG across rotation phases (EV Lac), with modulation amplitudes that can double between seasons — larger than typical measurement errors, so single-epoch spectra carry epoch-dependent activity contamination.
  • Long-term small-scale field fluctuations, with GP amplitudes varying by about a factor of two and means drifting by up to 0.3 kG, point to magnetic cycles in the parent dynamo; the near-zero correlation between $\langle B\rangle$ and $|B_\ell|$ implies small- and large-scale fields evolve largely independently.
  • For three stars (DS Leo, EV Lac, Barnard's star) stronger small-scale fields go with cooler surface temperatures (Pearson -0.93, -0.86, -0.71), indicating that magnetic flux is concentrated in dark spots; the derived spot coverage is about 8%, 19%, and 2%, respectively.
  • Joint Gaussian-process fits to $\langle B\rangle$ and $B_\ell$ sharing one rotation period converge cleanly and tighten period estimates (e.g., DS Leo 13.980±0.059 d), making combined fits the reliable route for stars where either series alone is ambiguous.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • My inference: the technique should transfer to archival spectra from other near-infrared spectrographs, turning the small-scale field into a cheap rotation-period diagnostic for hundreds of M dwarfs already monitored for exoplanets — with particular value for fully convective stars, where period–age relations built on the tachocline are doubtful.
  • My inference: the AD Leo null result makes inclination a selection filter — pole-on stars will be invisible to this method, so in a volume-limited sample the detection fraction should trace the inclination distribution, and measured modulation amplitudes could, in principle, be inverted for inclination if spot contrast is known.
  • My inference: because the grid assumes radial fields in 2 kG steps, recomputing a subset of nights with non-radial geometries and finer field steps would test whether the recovered periods and amplitude changes survive; if they do, the radial assumption is a safe simplification rather than the source of the signal.
  • My inference: the dTemp–$\langle B\rangle$ slope steepens toward earlier-type M dwarfs (DS Leo and AU Mic versus later M dwarfs), suggesting the relation could be calibrated into a spectroscopic spot-coverage indicator for monitoring activity cycles from unpolarized spectra alone.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper analyzes near-infrared SPIRou spectra of six M dwarfs (EV Lac, DS Leo, AD Leo, CN Leo, PM J18482+0741, Barnard's star) observed over 2019-2024. Using the ZeeTurbo spectral synthesis grid, the authors fit each spectrum as a linear combination of six magnetic-field components (0-10 kG) and derive nightly mean small-scale magnetic fields ⟨B⟩. They then apply quasi-periodic Gaussian Process (GP) regression to the ⟨B⟩ time series to search for rotational modulation, and compare with longitudinal-field (Bℓ) and dTemp series from the same data. The paper reports rotation periods detected in the small-scale-field series for 4 of 6 stars, agreement between ⟨B⟩ and Bℓ periods, long-term amplitude/mean variations, and anti-correlations between ⟨B⟩ and dTemp for three stars (EV Lac, DS Leo, Barnard's star). The central methodological claim is that rotation periods can be recovered from unpolarized spectra via Zeeman-broadening-derived small-scale magnetic field variability.

Significance. If the detection claim holds, this is a novel and useful result: unpolarized spectra, which are far more abundant than polarimetric data, could be used to monitor rotation and magnetic activity in M dwarfs, including fully convective stars and slow rotators. The cross-validation against independent Bℓ measurements for the three cleanest detections (EV Lac, DS Leo, Barnard's star) is a genuine strength, as is the agreement with literature periods and the long-term monitoring baseline. The dTemp-⟨B⟩ anti-correlations for three stars provide an interesting spot-physics link. The paper makes its codes and reduced data available (star-activity-tools, Zenodo, CDS), which supports reproducibility. However, the headline '4 of 6' detection rate is overstated on the paper's own evidence, and the statistical significance of the GP detections is not quantified. These issues are fixable but currently affect the central claim.

major comments (3)
  1. [Abstract/Sec. 4.6/Sec. 6] The abstract's claim of detecting rotation in the small-scale magnetic field series for 4 of 6 stars conflates an independent detection with a prior-assisted confirmation. For PM J18482+0741 (Sec. 4.6, Table A.5), the authors state that with uninformed priors the GP posterior for Prot shows multiple peaks, and that they 'fix the decay time and smoothing to typical values of 300 and 1.50' and 'adopt a Gaussian prior centered on the expected rotation period' (2.76 d, σ=0.6 d) to obtain Prot=2.762±0.009 d. Because the prior is centered on the literature period and is much wider than the quoted error, the data do narrow the period within the prior, but the prior selects which peak is explored. The paper's own Sec. 6 acknowledges that only EV Lac, DS Leo, and Barnard's star show clear modulation 'with little prior assumptions', while PM J18482+0741 is obtained only 'with more drastic priors'
  2. [Sec. 4.1] The error rescaling procedure is a potential source of over-confidence. The paper re-scales the ⟨B⟩ error bars using the dispersion of residuals from the best GP fit so that the minimum reduced χ² is near 1. Because the same data are used both to fit the GP and to calibrate the noise, this can absorb genuine quasi-periodic variability into the white-noise term and inflate the apparent significance of the periodic component. For the three clean detections the agreement with independent Bℓ periods is strong evidence, but for PM J18482+0741 and for the joint fits (e.g., AD Leo in Sec. 4.3, where the authors themselves note the constraint 'primarily arises from Bℓ') the contribution of the ⟨B⟩ series is less clear. I request a quantitative significance assessment: a likelihood-ratio test of the quasi-periodic GP against a non-periodic GP (or an equivalent false-alarm probability) for the ⟨B⟩
  3. [Sec. 4.3/Sec. 6] For AD Leo, CN Leo, and to a large extent PM J18482+0741, the rotation periods quoted from the joint ⟨B⟩+Bℓ GP fits are essentially inherited from the Bℓ series. In Sec. 4.3 the authors state that for AD Leo the simultaneous fit yields Prot=2.230±0.001 d and that this 'suggest[s] that the constraint on Prot primarily arises from Bℓ'. Section 6 similarly notes that for PM J18482+0741, CN Leo, and AD Leo the joint-fit periods are 'very close to those obtained from the Bℓ data set alone'. This means the small-scale-field series alone does not provide independent detections for these three stars. The paper should clearly separate (a) independent detections from ⟨B⟩ alone, (b) detections confirmed by ⟨B⟩+Bℓ joint fits, and (c) targets where only Bℓ constrains the period. This distinction is essential for the abstract and conclusions.
minor comments (5)
  1. [Fig. 6] The caption reports 'Pearson = 0.93' for EV Lac, while the text (Sec. 5) reports a correlation coefficient of -0.86 for EV Lac (and Fig. 6 shows a negative slope). The sign convention should be made consistent, e.g., by using the signed coefficient in both places or explicitly noting that the figure shows |r|.
  2. [Table A.6] For CN Leo, the Prot entry is reported as '2.694*' with a note that the posterior is multi-peaked and no reliable error bars can be extracted. This should be stated in the main text's Sec. 4.7 as well; currently the table note is the only place where the lack of a meaningful uncertainty is explicit.
  3. [References] The emcee package is referenced as 'Mackey et al. 2008', which appears to be an erroneous citation (the emcee software is Foreman-Mackey et al. 2013, ApJ, 795, 64). Please correct.
  4. [Fig. B.4] Caption typo: 'PMJ J18482+0741' should be 'PM J18482+0741'.
  5. [Throughout] 'dT emp' is frequently written with an irregular space; use consistent notation (e.g., dTemp or dTemp). Also, the reduced χ² notation 'χ2 r' is difficult to read; use χ²_r consistently.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: rotation periods are free GP hyperparameters on ZeeTurbo-derived ⟨B⟩ series and are cross-checked against independent Bℓ/dTemp measurements; the prior-assisted PM J18482+0741 detection is disclosed in Sec 6.

full rationale

The derivation chain is: ZeeTurbo synthetic spectra (Cristofari et al. 2023a) are fit to SPIRou spectra to obtain a time series of ⟨B⟩; a quasi-periodic GP (Eq. 1) is then fit with Prot as a free hyperparameter. Nothing in this chain defines Prot in terms of ⟨B⟩ or vice versa, so there is no self-definition. The three cleanest detections (EV Lac, DS Leo, Barnard's star) are obtained under wide priors and agree with independent Bℓ/literature periods (e.g., EV Lac 4.362±0.001 d vs. 4.3715±0.0006 d), so the central claim has independent content. The abstract's fourth detection, PM J18482+0741, is obtained only after fixing l and β and adopting a Gaussian prior centered on the expected 2.76 d period (Sec. 4.6, Table A.5). This is a confirmation rather than an independent detection, but the paper itself states this in Sec. 6 ('with more drastic priors... unable to unambiguously constrain...'), so it is a disclosed counting choice, not a hidden reduction by construction. The simultaneous two-GP fits share a common Prot, and the paper explicitly notes for AD Leo/PM/CN Leo that the constraint primarily arises from Bℓ; those stars are not counted as independent small-scale detections. The dTemp anti-correlation is an external observable (line-profile temperature index) compared with Zeeman-broadening field estimates, so it is not circular. The main self-citations (ZeeTurbo, previous ⟨B⟩ analyses) are methodological and are not invoked as evidence for the new periods. Overall, the paper is largely self-contained against external period measurements; any overstatement in counting is modest and disclosed.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claims (rotation periods and long-term variations) depend on the ZeeTurbo spectral model, the filling-factor decomposition, and the Gaussian process model. These are standard tools in the field, but they carry several domain assumptions, and the paper's own convergence problems for some stars indicate sensitivity to priors. No new physical entities are introduced.

free parameters (5)
  • Magnetic filling factors f0, f2, f4, f6, f8, f10 = fractions per spectrum summing to 1
    MCMC fit to each nightly spectrum (Sec. 3.1); the average field is derived from these factors.
  • Radial-tangential broadening zeta_RT = 2.19 to 5.51 km/s (Table 2)
    Fitted per star from template spectra; absorbs extra line broadening not accounted for by v sin i and the instrumental profile.
  • Atmospheric parameters Teff, log g, [M/H] = Table 2
    Derived from template spectra, then fixed when fitting nightly spectra (Sec. 3.2).
  • Gaussian process hyper-parameters mu, sigma, alpha, l, beta, Prot = Tables A.1 to A.6
    MCMC fit to each time series; some values fixed (e.g., l=300 d, beta=1.5 for AD Leo) to achieve convergence.
  • Error rescaling factors for field and temperature measurements = not explicitly quoted
    Rescaled in Sec. 4.1 so that the minimum reduced chi-square is close to 1; affects all reported error bars.
assumptions (5)
  • domain assumption LTE and MARCS model atmospheres are valid for synthesizing M dwarf spectra.
    Invoked in Sec. 3.1 for the ZeeTurbo grid; departures affect absolute field values but likely less the relative time variability.
  • domain assumption The Zeeman broadening grid assumes a purely radial magnetic field geometry.
    Sec. 3.1 states 'assuming that the magnetic field is radial in all points of the photosphere'; mixed-polarity or transverse fields would change filling factors and the average field.
  • ad hoc to paper The observed spectrum is a linear combination of six synthetic spectra with fields 0, 2, ..., 10 kG.
    Sec. 3.1: this discretization is a modeling choice from Cristofari et al. (2023a,b); a different field distribution within each component cannot be diagnosed.
  • domain assumption The quasi-periodic Gaussian process kernel (Eq. 1) adequately models the temporal modulation.
    Sec. 4.1; the kernel is flexible but degenerate (Angus et al. 2018), and several fits require fixing hyper-parameters or adding priors.
  • ad hoc to paper Error bars are re-scaled so that the best fit yields reduced chi-square near 1.
    Sec. 4.1; this converts absolute errors into relative errors and does not provide an independent estimate of systematic uncertainty.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Rotational modulation and long-term evolution of the small-scale magnetic fields of M dwarfs observed with SPIRou." pith.science (2026). https://pith.science/paper/ENSFUKCZ

@misc{pith2026250804569,
  author       = {Pith},
  title        = {Pith review of: Rotational modulation and long-term evolution of the small-scale magnetic fields of M dwarfs observed with SPIRou},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ENSFUKCZ}},
  note         = {Machine review of arXiv:2508.04569}
}
abstract

M dwarfs are known to host magnetic fields, impacting exoplanet studies and playing a key role in stellar and planetary formation and evolution. Observations revealed the long-term evolution of the large-scale magnetic field reconstructed with Zeeman-Doppler imaging, and a diversity of their topologies. These large-scale magnetic fields only account for a small amount of the unsigned magnetic flux that can be probed by directly modeling the Zeeman broadening of spectral lines in unpolarized spectra. We aim at investigating the long-term behavior of the average small-scale magnetic field of M dwarfs with time, and assess our ability to detect rotational modulation from time series of field measurements derived from unpolarized spectra. We perform fits of synthetic spectra computed with ZeeTurbo to near-infrared high-resolution spectra recorded with SPIRou between 2019 and 2024 in the context of the SLS and SPICE large programs. The analysis is performed on the spectra of 2 partially convective (AD Leo, DS Leo) and 3 fully convective (PM J18482+0741, CN Leo, Barnard star) M dwarfs, along with EV Lac whose mass is close to the fully-convective limit. Our analysis provides measurements of the average small-scale magnetic field, which are compared to longitudinal magnetic field and temperature variation measurements (d$Temp$) obtained from the same data. We were able to detect the rotation period in the small-scale magnetic field series for 4 of the 6 stars in our sample. We find that the average magnetic field can vary by up to 0.3 kG throughout the year (e.g., CN Leo), or of up to 1 kG across rotation phases. The rotation periods retrieved from longitudinal and small-scale magnetic fields are found in agreement within error bars. d$Temp$ measurements are found to anti-correlate with small-scale magnetic field measurements for three stars (EV Lac, DS Leo and Barnard's star).

Figures

Figures reproduced from arXiv: 2508.04569 by the authors.

Figure 1
Figure 1. Simultaneous fit of two GPs on our small-scale and large-scale magnetic fields measurements. The rotational velocity Prot is the same for the two GPs. The pink shaded area shows the uncertainty on the GP fit. to decrease and then increase, although the minimum is reached in early-2022, and a maximum is reached around July 2021 (see Fig. D.1, available on Zenodo). The series of ⟨B⟩ and |Bℓ | ap￾pear uncorrelated, yie… view at source ↗
Figure 2
Figure 2. Posterior distribution of the hyper-parameters obtained for EV Lac. Red lines mark the median of the 1% of walkers which have the highest likelihood. We perform a fit of a GP on the dT emp measurements. Re￾lying on wide uniform priors yields Prot = 1.81 ± 0.10 d. We note that two peaks are visible in the posterior distribution with a secondary peak around the expected rotation period. To help convergence, we repeate… view at source ↗
Figure 3
Figure 3. Best fit GP fit (green) obtained on our small-scale magnetic fields measurements (black circles). The top panel shows the results obtained over our entire data set, while the middle and bottom panels are zoomed on different observation periods. measurements appear uncorrelated, with a Pearson correlation coefficient of 0.12. Here again, we run our process on the dT emp measurements secured for Barnard’s star, and ob… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Evolution of the filling factors distribution with respect to phase for EV Lac for observations recorded between MJD 59385 and 59515. Solid lines show a fit of a sinusoidal with a first harmonic to the data, assuming the rotation period obtained from our GP fit (Prot =…
Figure 5
Figure 5. Figure 5: Same as [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Correlation between ⟨B⟩ and dT emp for EV Lac. The red line shows a linear with slope −19.8±0.5 K kG−1 and intercept 90.0±2.3 K. and Barnard’s star being the least magnetic star of the three, al￾though the reported fractions should be considered with caution given the …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Analyzing and Mitigating Object Hallucination: A Training Bias Perspective

    cs.CV 2025-08 unverdicted novelty 6.0 of 10

    LVLMs hallucinate most on images from their own training data; the bias lives in the LM head, and unlearning just that head with label-output discrepancy signals cuts object hallucination across 2B to 72B models.

  2. Unstable magnetospheric accretion on the T Tauri star TW Hya

    astro-ph.SR 2026-07 accept novelty 4.5 of 10

    TW Hya’s large-scale field is a ~0.83 kG tilted dipole that varies yearly; accretion is unstable (rmag/rcor ≈ 0.33–0.40) and no close-in planet is detected above ~0.3–1 Mjup.

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages · cited by 2 Pith papers

  1. [1]

    W., & O’Neil, M

    Ambikasaran, S., Foreman-Mackey, D., Greengard, L., Hogg, D. W., & O’Neil, M. 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence, 38, 252 Angus, R., Morton, T., Aigrain, S., Foreman-Mackey, D., & Rajpaul, V . 2018, MNRAS, 474, 2094 Artigau, É., Cadieux, C., Cook, N. J., et al. 2024, AJ, 168, 252 Barnes, S. A. 2003, ApJ, 586, 464 Bellotti...

    work page arXiv 2015

  2. [500]

    The posterior distribution shows multiple peaks that does not allow us to extract statisfactory error bars

    βdT emp 1.00 Fixed Prot,dT emp (d) 3 .003+0.031 −0.536 G(2.70, 0.30) χ2 r,dT emp 1.02 RMSdT emp 1.78 (K) Notes.∗ Value corresponding to the maximum of likelihood in the pos- terior distribution. The posterior distribution shows multiple peaks that does not allow us to extract statisfactory error bars. Article number, page 13 of 16 A&A proofs: manuscript n...

    work page arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.