CO rotational line emission in very red carbon stars in the Magellanic Clouds

Abstract: Stars of low and intermediate initial mass lose most of their stellar mass at the end of their lives during the asymptotic giant branch (AGB) phase. Determining their gas and dust mass-loss rates (MLRs) is crucial for quantifying the contribution of evolved stars to the life cycle of dust and gas in the Universe. The Atacama Large Millimeter/submillimeter Array was used to observe 38 carbon stars (C stars) in the large Magellanic cloud (LMC) and three C stars in the small Magellanic cloud (SMC) in the CO J= 2-1 line. Line profiles were fitted to derive stellar velocities and wind-expansion velocities (Vexp). CO emission is detected in two C stars in the SMC and 33 C stars in the LMC. This is the first detection of carbon monoxide around an AGB star in the SMC. One object in the LMC shows emission in $<sup>{13}$CO. The wind-expansion velocity ranges from \sim7.5 to \sim30 km/s. Archival data were used to determine the pulsation periods as well as construct and model the spectral energy distributions using two dust radiative transfer codes. Mass-loss rates were independently derived from these two codes as well as from the intensity of the CO line, using a simple formula. On average, the dust-based MLRs higher than the MLRs based on the CO line by a factor of 1.6. Additional CO data in other transitions, combined with proper modelling, is required to further investigate this possible discrepancy. Mass-loss rates, pulsation periods, and expansion velocities were compared to a sample of Galactic C stars. There is a strong bias, as the Magellanic Cloud targets sample the highest MLRs and luminosities, yet they represent only a minority of stars in a Galactic sample. Comparing this sample with a similarly extreme set of Galactic stars with periods longer than 500 days, we identify no correlation between metallicity and either the MLR or Vexp.

• The paper uses ALMA Band 6 observations of 38 LMC and three SMC carbon stars, detecting CO in 33 LMC and two SMC targets, including the first AGB-star CO detection in the SMC.
• The paper finds expansion velocities of about 7.5–30 km/s and no clear LMC–Milky Way metallicity trend in mass-loss rates or wind speeds, while the two SMC detections show notably slower winds.
• The paper identifies a factor-of-1.6 average difference between dust- and CO-derived mass-loss rates, a likely J-type carbon star in the LMC, and three sources whose millimetre continua indicate unusual wind histories or geometry.

Overview and motivation

This paper presents ALMA Band 6 observations of the $^{12}$CO(2–1) line (and $^{13}$CO(2–1)) toward 38 carbon stars (C stars) in the Large Magellanic Cloud (LMC) and three C stars in the Small Magellanic Cloud (SMC), selected as the reddest, highest mass-loss rate (MLR) objects from dust radiative-transfer (RT) grids built on Spitzer SAGE photometry. The work extends the earlier four-object LMC detection study of Groenewegen et al. (2016) and addresses a specific open question: whether carbon-star winds depend on metallicity. Because C stars self-enrich in carbon via dredge-up, their refractory content is largely decoupled from initial metallicity, so hydrodynamical wind models such as those of Bladh et al. (2019) predict only weak metallicity dependence of both MLR and expansion velocity ($V_{\rm exp}$) — solar/LMC MLR ratios of 0.87 ± 0.02 and solar/SMC of 0.83 ± 0.02, with $V_{\rm exp}$ ratios near unity. The observations test these predictions empirically.

The sample represents roughly the reddest 10% of known and likely C stars in the LMC ([3.6]–[4.5] > 0.37 mag among C-AGB, x-AGB, and FIR classes). Although it constitutes only ~0.5% of the total LMC C-star population, its combined gas return of $\sim 8 \times 10^{-4}\ M_\odot\,{\rm yr^{-1}}$ accounts for ~6% of the total gas and ~16% of the total dust returned by the entire population, underscoring that a small number of extreme objects dominate the dust budget.

Observations and detections

Observations were carried out in early 2024 in three science goals (SMC; "LMC strong"; "LMC weak"), with on-source integration times of 7–24 minutes per target and typical beams of 0.5″–0.8″. Line spectra were extracted within a conservative 2.0″ radius aperture, sized to encompass the maximum photodissociation radius (~1.4″ at 50 kpc) predicted by the Saberi et al. (2019) grid for the most extreme plausible parameters. Dust continuum was extracted at 1330 μm from a 4.5″ radius aperture, with fluxes scaled from four spectral windows assuming a $\lambda^{-2}$ dependence.

CO emission is detected in 33 of 38 LMC targets and two of three SMC targets, the latter constituting the first CO detection around an AGB star in the SMC. The high detection rate validates the target-selection strategy based on predicted line intensities accurate to within a factor of two. One object, IRAS 05568, was also detected in $^{13}$CO with a $^{12}$CO/$^{13}$CO integrated intensity ratio of 1.8 ± 0.1. Since the $^{12}$CO profile is parabolic (optically thick), the true $^{13}$0C/$^{13}$1C ratio must exceed this value but likely not by a large factor; the authors note this ratio resembles that of J-type C stars (4–10, close to the CN-cycle equilibrium value of 3). This is the first indication of J-type chemistry in a high-MLR LMC carbon star, and only the second such object after the Galactic IRAS 15194−5115. Among non-detections, the strongest lower limits are $^{13}$2CO/$^{13}$3CO ≳ 15–19 (IRAS 05416, TRM 74, IRAS 05190).

Line profiles were fitted with parabolic shell profiles ($^{13}$4 fixed), yielding stellar velocities and full widths at zero intensity; $^{13}$5 ranges from ~7.5 to ~30 km s$^{13}$6. A few LMC objects reach 31–35 km s$^{13}$7, velocities with no counterpart in the Galactic comparison samples used here, though such stars do exist in the Galaxy (13 of 331 C stars in Groenewegen et al. 2002 have $^{13}$8–43 km s$^{13}$9).

Pulsation periods

Periods were derived from VISTA VMC DR6 $V_{\rm exp}$0-band light curves and WISE/NEOWISE W1/W2 data, with saturation corrections applied at the bright end. Periods were adopted as weighted means when the amplitude exceeded 0.2 mag and its error was below one third of its value. These periods supersede those reported in the 2016 study for IRAS 05506, IRAS 05125, ERO 0529379, and ERO 0518117. Most detected sources have periods between ~500 and ~1350 days, with several objects showing very long or poorly constrained periods (>2500 days).

SED modelling and mass-loss rates

Two independent dust RT approaches were employed: (i) MoD (More of DUSTY) fits to broad-band photometry and Spitzer/IRS spectra, using amorphous carbon, SiC, and MgS with distribution-of-hollow-spheres grain optics, where the gas-to-dust (GTD) ratio is tuned to match the observed $V_{\rm exp}$1; and (ii) the self-consistent dust-growth wind models of Nanni et al., coupled a posteriori to MoD spectra, with model parameters constrained to reproduce $V_{\rm exp}$2 within 3σ. The geometric mean of the two independent determinations was adopted.

A key diagnostic comes from the ALMA 1330 μm continuum. For most sources the MoD models fall below the observed upper limits or agree with detections, but three objects — TRM 74, IRAS 05315, and IRAS 05495 — show severe discrepancies: the standard MoD fit over-predicts the millimetre continuum while under-predicting the mid-infrared, with fitted inner dust temperatures near 300 K versus the ~950 K condensation temperatures implied by the Nanni models. Two explanations are considered. A steeper density profile ($V_{\rm exp}$3 instead of $V_{\rm exp}$4) improves the fit but implies current MLRs of order $V_{\rm exp}$5, i.e. an increase by factors of ~15 over the past 22–34 kyr. Alternatively, a recently halted wind would require flow timescales of only 110–150 yr, which is statistically implausible for three objects given post-AGB transition timescales of a few thousand years. Deviations from spherical symmetry or binarity remain possible but cannot be tested at the current spatial resolution; notably, the CO profiles of these objects show standard parabolic shapes with no evidence of companion-induced structure.

Gas MLRs were also estimated directly from the CO(2–1) integrated intensity using the Ramstedt et al. (2008) scaling relation, adopting updated CO abundances relative to H$V_{\rm exp}$6 of $V_{\rm exp}$7 (SMC) and $V_{\rm exp}$8 (LMC) from new nucleosynthesis calculations. The dominant error source is the calibration coefficient, giving uniform relative errors of 54–59%. Comparing the 26 objects with CO detections and reliable RT models, dust-based MLRs exceed CO-based MLRs by a factor of 1.6 on average (0.8 in the median), with very large scatter — a discrepancy the authors explicitly flag as requiring additional CO transitions and proper radiative-transfer modelling of the lines before it can be considered established.

Comparison with Galactic samples and metallicity dependence

Comparisons with Galactic C-star samples reveal a strong selection bias: the Magellanic Cloud targets sample the longest periods (≳500 d), highest luminosities ($V_{\rm exp}$9–3.85), and highest MLRs, properties held by only a minority of Galactic C stars. Restricting the Galactic comparison to stars with periods longer than 500 days improves the overlap considerably. Within this matched comparison, no correlation between metallicity and either MLR or $V_{\rm exp}$0 is identified for the LMC versus Milky Way. The two SMC detections, however, have among the lowest expansion velocities measured (6.7 and 9.9 km s$V_{\rm exp}$1, against an LMC median of 15.7 km s$V_{\rm exp}$2).

This SMC result is broadly consistent with radiation-driven wind theory, $V_{\rm exp}$3: the derived median GTD ratios are 1500 (SMC, three stars), 400 (LMC), and 190 (Galactic reddest stars), predicting an SMC-to-LMC velocity difference of a factor of ~1.9 — close to what is observed, though resting on only two detections. By contrast, the GTD-implied LMC-to-Galactic velocity difference of ~1.4 is not supported by the data. The authors also highlight a population of intermediate-luminosity LMC C stars with large MLRs that has no counterpart in the Galactic sample; they suggest this may reflect the larger carbon excess available to an LMC C star at fixed luminosity, since less oxygen must be overcome.

The paper further notes practical difficulties in constructing volume-complete Galactic comparison samples via the NESS survey: several highly red stars (AFGL 3068 among them) should fall in NESS's 'high' or 'extreme' tiers but were not included, illustrating how distance uncertainties propagate into luminosity and MLR classification. Estimated tier abundances imply the 'high' tier comprises ~1% of all AGB stars and the 'extreme' tier roughly 20 times fewer.

Limitations and open questions

Several caveats bear directly on the results. The metallicity-independence conclusion for $V_{\rm exp}$4 rests on only two SMC detections, and the claimed factor-of-two agreement with GTD scaling is correspondingly fragile. The dust-versus-CO MLR discrepancy (factor 1.6 in the mean) is based on a single CO transition and a simple scaling relation whose calibration uncertainty dominates; multi-transition CO observations with proper line RT modelling are required to confirm or refute it. The three objects with anomalous millimetre continua admit multiple interpretations — evolving density structure, recent MLR cessation, or asymmetry/binarity — none of which can be discriminated with current data. Finally, ISM contamination affected one non-detection outright (MSX LMC 527) and required channel excision in MSX LMC 1780, while deeper observations or velocity priors could recover the four remaining ambiguous non-detections.

Conclusion

This study expands the census of spectrally resolved CO emission in extragalactic AGB stars from 4 to 37 LMC carbon stars and provides the first CO detection of an AGB star in the SMC, together with the first $V_{\rm exp}$5CO detection in an LMC carbon star pointing to J-type chemistry. Expansion velocities span ~7.5–30 km s$V_{\rm exp}$6, and the combined evidence — particularly the low SMC velocities and the absence of an LMC–Galaxy difference at matched period — offers tentative support for weak metallicity dependence of carbon-star winds, consistent with self-enrichment arguments and hydrodynamical predictions. The principal quantitative uncertainties — the dust/CO MLR offset and the nature of the three continuum-anomalous sources — define the immediate observational agenda: multi-transition CO follow-up of the brightest targets, feasible with roughly 100 hours of ALMA time for a dozen objects.