Chemically Peculiar A and F Stars with Enhanced s-process and Iron-peak Elements: Stellar Radiative Acceleration at Work

We adopt the stellar abundance catalog of Xiang et al. (2019), which includes abundance estimates for 16 elements (C, N, O, Na, Mg, Al, Si, Ca, Ti, Cr, Mn, Fe, Co, Ni, Cu, and Ba) in 6 million stars derived from the LAMOST DR5 low-resolution (R ∼ 1800) spectra. Abundances are measured using The DD–Payne (Ting et al. 2017; Xiang et al. 2019), a hybrid method that combines the philosophy of data-driven spectroscopy with The Payne, a flexible spectral fitting tool based on neural network modeling (Ting et al. 2019). As the training set to build up the data-driven spectral model, the LAMOST DD–Payne used the set of LAMOST stars that overlap with the GALAH (de Silva et al. 2015) and APOGEE (Majewski et al. 2017) surveys (for which stellar abundance measurements exist). In the training process, priors are assigned according to the first-order derivative of Kurucz model spectra (Kurucz 2005) with respect to each elemental abundance, to ensure that abundances are deduced from physically sensible features in the LAMOST spectra. For a spectrum with signal-to-noise ratio (S/N) higher than 50, the statistical uncertainties on the abundance estimates can be as small as 0.03 dex for [C/Fe], [Fe/H], [Mg/Fe], [Ca/Fe], [Ti/Fe], [Cr/Fe], and [Ni/Fe], and 0.15 dex for [Ba/Fe]. Systematic errors are expected to be similar in magnitude. In the catalogs provided by Xiang et al. (2019), unreliable estimates are marked with a flag. In this work, we make use of the "recommended catalog" created by Xiang et al. (2019) by combining two individual abundance catalogs scaled to either GALAH or APOGEE.

In this paper, we also draw on the GALAH DR2 (Buder et al. 2018) catalog, both for validating the LAMOST results and for studying the correlation of chemical abundance peculiarities with projected stellar rotation velocity v sin i. GALAH is a high-resolution (R ∼ 28,000) spectroscopic survey using the Anglo-Australian Telescope (de Silva et al. 2015; Martell et al. 2017). Its second data release (DR2) provides stellar parameters (T_eff, $\mathrm{log}g$, V_mic, and $v\sin i$) and elemental abundances of 23 elements for 342,682 stars (Buder et al. 2018).

The most pertinent result revealed by the Xiang et al. (2019) catalog is shown in Figure 1, which highlights where in the [Ba/Fe] versus [Fe/H]¹⁵ diagram the subset of LAMOST DR5 stars with a g-band spectral S/N higher than 50 is located. This includes both dwarfs and giants, which we define as stars with T_eff < 5600 K and $\mathrm{log}g\lt 3.8$ throughout this paper. The vast majority of stars have [Ba/Fe] values around solar ([Ba/Fe] ∼ 0), but there are also a significant fraction of stars with high [Ba/Fe] ([Ba/Fe] ≳ 1 dex) or low [Ba/Fe] ([Ba/Fe] ≲ −1 dex) values. A prominent population of metal-rich ([Fe/H] > −0.2 dex), Ba-enhanced ([Ba/Fe] > 1 dex) dwarfs and subgiants, in particular, has no counterpart in the giant population.

Zoom In Zoom Out Reset image size

We select the metal-rich and Ba-enhanced stars of interest using the criteria of [Fe/H] > −0.2 dex and [Ba/Fe] > 1.0 dex, and require S/N > 50, qflag_chi2 = "good" to ensure that the abundance estimates are of good quality. We further discard stars classified as O or B type or a WD according to the LAMOST pipeline (Luo et al. 2015), as their DD–Payne abundance estimates are problematic. These criteria lead to 15,009 stars in our sample. The median value of the reported measurement uncertainties for this sample of stars is 0.15 dex in [Ba/Fe] and 0.04 dex for [Fe/H] and [X/Fe], where X = C, Mg, Si, Ca, Ti, Cr, Mn, and Ni. Our sample of Ba-enhanced stars occupies a specific region in the [Ba/Fe]–[Fe/H] plane that is populated by dwarfs and subgiants but contains almost no red giant stars. We will use the terms "Ba-enhanced stars" and "Ba-enhanced chemically peculiar stars" interchangeably in this study, as we have found that these stars also exhibit chemical peculiarity in other elements.

Figure 1 demonstrates that stars can also have "peculiar" [Ba/Fe] values when, for instance, they are Ba enhanced (both dwarfs and giants) with [Fe/H] < −0.2 dex, or when they are metal rich but Ba depleted. In this study, we focus only on the metal-rich and Ba-enriched stars as defined above, which may be related to the classical Am/Fm stars. The exploration of stars with other apparent chemical "peculiarities" is deferred to future work.

Figure 2 shows the LAMOST spectra of a Ba-enhanced A star and a Ba-normal star with similar stellar parameters. The Ba-enhanced star shows a stronger Ba ii λ4554 Å line that cannot be fitted by a normal Ba abundance, suggesting that the Ba enhancement is a genuine feature recognizable in LAMOST low-resolution spectra.

Zoom In Zoom Out Reset image size

Figure 3 plots [Ba/Fe] as a function of effective temperature for two subpopulations of stars, those with [Fe/H] > −0.2 dex and [Fe/H] < −0.2 dex. For both metallicity populations, there is an increasing trend in [Ba/Fe] with T_eff, reaching a [Ba/Fe] value of 0.5 dex at T_eff ∼ 6500 K. This trend has also been revealed by measurements from high-resolution spectra and has been argued to originate from NLTE effects (Bensby et al. 2014). Beyond 6500 K, however, these two populations exhibit markedly different behavior. In the low-metallicity, [Fe/H] < −0.2 dex, population, the [Ba/Fe] ratio turns over and begins decreasing with T_eff, dropping back to the solar value at T_eff ∼ 7000 K for. The [Fe/H] > −0.2 dex subset, on the other hand, exhibits a bifurcation in [Ba/Fe] at T_eff ≳ 7000 K, with one branch reaching the low-[Ba/Fe] values typical of the [Fe/H] < −0.2 dex population and a second branch of high-[Ba/Fe] values. This latter branch constitutes our sample of Ba-enhanced stars. Such a bifurcation structure suggests that the high-[Ba/Fe] stars are not simply due to a systematic bias in the measurements, but emerges as a result of a genuine enhancement in Ba. We verify this using high-resolution spectra in Section 2.3.

Zoom In Zoom Out Reset image size

Figure 4 highlights the distribution of Galactocentric rotational velocities for our set of Ba-enhanced stars. The rotational velocity for each star is derived by adopting the Gaia distance from Bailer-Jones et al. (2018), the Gaia DR2 proper motion (Gaia Collaboration et al. 2018), and the LAMOST DR5 radial velocity (RV; for details, see Coronado et al. 2020). The Ba-enhanced stars are indistinguishable from chemically normal metal-rich stars and exhibit the same disk-like kinematics typical of metal-rich stars.

Zoom In Zoom Out Reset image size

The spectral signals of Ba enhancement in the LAMOST spectra are significant but appear subtle at face value (Figure 2). To verify them, we have carried out high-resolution (R ∼ 48,000) follow-up observations for a few of these Ba-enhanced stars with the FEROS spectrograph on the ESO La Silla 2.2 m telescope (Kaufer et al. 1999). Figure 5 shows some of the key spectral lines from the FEROS spectrum for one example star, together with the predictions from 1D LTE Kurucz models for a range of abundances. In all panels, we assign v sin i = 3.0 km s⁻¹ based on the method presented in Strassmeier et al. (1990) using the λ6432 Å line. The Kurucz model spectra are computed adopting the T_eff and $\mathrm{log}g$ from the LAMOST DD–Payne catalog. The red line in each panel highlights the high-resolution spectrum predicted specifically for the abundance ratios derived from the low-resolution LAMOST spectra. In all models, the microturbulence velocity V_micro is set to 2.0 km s⁻¹ to match the line profile, consistent with expectations from the V_micro–T_eff relation measured in the literature (Bruntt et al. 2010; Gebran et al. 2014). The models also adopt a macroturbulence velocity V_macro = 8.0 km s⁻¹, which is roughly consistent with an extrapolation of the V_macro–T_eff relation in the literature (Bruntt et al. 2010; Doyle et al. 2014). It should be noted, though, that this extrapolation is suited to much hotter stars than those where the V_macro–T_eff relation more strictly applies (i.e., stars cooler than 6400 K). The good agreement of line profiles between the Kurucz synthetic spectra and the high-resolution FEROS observations lends credence to this extrapolation.

Zoom In Zoom Out Reset image size

Figure 5 illustrates the remarkable power of low-resolution spectra for determining even "peculiar" element abundances: for Ba, Ni, Fe, Mn, Cr, Ti, Si, and Mg, the observed high-resolution FEROS spectral lines are in good, though not perfect, agreement with the Kurucz model spectra (red line) adopting the abundance values derived with The DD–Payne applied to the LAMOST data. Our focus here is on isolated single lines with good S/N but weak features to avoid saturation. For this set of lines, NLTE effects should be mostly minimal. For Ba, NLTE effects are expected to be negligible for the 5854 Å line (Korotin et al. 2015). For the Mg 5529 Å line, the impact of NLTE effects for late A-type stars is smaller than 0.1 dex (Alexeeva et al. 2018). In the case of Cr, Mn, Fe, and Ni, the spectral lines are even stronger features than in the Kurucz model that adopts the best-fitted DD–Payne abundance, which is opposite to the expected impact of NLTE effects and possibly implying that the star is iron-peak enhanced. The Ca and C i lines are glaring exceptions that we will follow up with a more thorough analysis in the future. We emphasize, though, that the chemical peculiarity in these cases is unlikely due to NLTE, given that this should be ubiquitous for all stars at this temperature and $\mathrm{log}g$.

The distribution of the metal-rich, Ba-enhanced stars in the T_eff–$\mathrm{log}g$ (Kiel) diagram is plotted in Figure 6. From their location in the plot, we can see that these stars are almost exclusively MS (turnoff) and subgiant stars with relatively high temperature (T_eff ≳ 6300 K). Their distribution exhibits a sharp border toward low T_eff (aside from the imposed temperature cut at T_eff < 7500 K), with few, if any, stars at the lowest temperatures. The lack of cooler stars immediately implies that the origin of these Ba-enhanced stars cannot lie in normal Galactic chemical evolution, which would predict stars across a broad temperature range, including at the low-T_eff side of the diagram.

Zoom In Zoom Out Reset image size

The shape of the low-temperature border—which traces a decrease in T_eff with decreasing $\mathrm{log}g$—can yield insight into the nature of these stars. We study this further in Figure 6, which incorporates the PARSEC stellar evolution tracks (Bressan et al. 2012). The distribution of Ba-enhanced stars in the Kiel diagram is consistent with the MS (turnoff) and subgiant regimes of the stellar evolutionary tracks for (initial) stellar mass between 1.4 and 1.8 M_⊙. The lower border is consistent with mass in the range 1.4–1.5 M_⊙, depending on $\mathrm{log}g$.

In Figure 7, we compare the distribution of our sample of Ba-enhanced stars with the stellar envelope models from Ludwig et al. (1999) for a range of convective envelope mass ratios in the T_eff–$\mathrm{log}g$ plane. The low-T_eff border of these Ba-enhanced stars corresponds to a convective envelope mass ratio of about ≃10⁻⁴, and this convective envelope mass fraction drops by four orders of magnitude toward the hottest stars. This offers a satisfying stellar evolutionary explanation for the distribution of Ba-enhanced stars and provides a valuable diagnostic of external accretion events (see Section 4).

Zoom In Zoom Out Reset image size

Finally, we note that ∼3% (482) of our sample are giants, with T_eff < 5600 K and $\mathrm{log}g\lt 3.6$. Among these giants, only 12 of them (or 2.5%) have [Ba/Fe] > 1.5 dex, a much lower fraction than in the overall sample (∼1/3). Throughout this work, we will focus on the relatively hot A/F stars. The nature of the cooler Ba-enhanced stars will be explored in an upcoming study (M. Zhang et al. 2020, in preparation; see also Norfolk et al. 2019).

Although The DD–Payne is designed to provide abundances for 16 elements from the LAMOST spectra, for the relatively hot stars of interest in this work, some abundances cannot be determined (i.e., the given spectral features are too weak to allow for reliable abundance determination; Xiang et al. 2019) and the fraction of stars with useful abundance estimates varies.

Figure 8 highlights the Ba-enhanced sample in elemental abundance space [X/H]–[Fe/H] for X = Mg, Si, Ca, Ti, Cr, Mn, Fe, Ni, and Ba for stars with 6700 < T_eff < 7500 K. Ba enhancement coincides with an enhancement of iron-peak elements (Cr, Mn, Fe, Ni), as well as with Si and Ti. Mg and Ca, on the other hand, show no significant overabundance. Mg even exhibits a slight underabundance. For the most Ba-enhanced ([Ba/Fe] > 1.5 dex) stars, [Fe/H] is 0.3 dex higher than that of Ba-normal stars on average. For this set, the overabundance of Cr and Ni are even more prominent, reaching about 0.5 dex and 0.8 dex, respectively. Ba can be enhanced by up to 1000 times (3 dex), making this the most overabundant element.

Zoom In Zoom Out Reset image size

Wide binaries are ideal sources to verify this pattern of abundances. The member stars in a wide binary system should have identical initial elemental abundances, given that they are likely to have formed in the same birth cloud. Wide binary pair stars have the added advantage that they can be targeted by LAMOST due to their wide separation. El-Badry & Rix (2018) identified more than 50,000 wide binary systems using parallax and proper motions in Gaia DR2. Among them, 4714 systems have LAMOST DR5 spectra for both of the pair stars. About 80% of these systems have a physical separation smaller than 0.08 pc, and the maximal separation is 0.23 pc. Such small separations suggest they are indeed conatural (e.g., Kamdar et al. 2019). From this set of stars, we select systems with S/N > 40 for both stars in the pair, and additionally require that the pair star with the lower temperature has T_eff < 6500 K, in an effort to ensure that these stars have chemically normal abundances. This yields a total of 1200 wide binary pairs that can be analyzed in the manner described above. Of these, 15 pairs hold a Ba-enhanced member with [Ba/Fe] > 1 dex in the temperature range of 6700 < T_eff < 7500 K.

Figure 9 shows that the abundances of iron-peak elements (Cr, Fe, Mn, Ni) in the set of Ba-enhanced pair stars are enhanced compared to their low-T_eff companions, whereas stars with [Ba/Fe] < 1 dex in the same temperature range exhibit little to no enhancement with respect to their low-T_eff companions. On the other hand, the [Mg/H] for the majority of the Ba-enhanced stars are slightly underabundant with respect to their low-T_eff companions. This is similar to our finding for the overall sample. Meanwhile, Si and Ti abundances for the Ba-enhanced stars are moderately enhanced compared to their low-T_eff companions, again consistent with findings for the overall sample.

Zoom In Zoom Out Reset image size

The above metal-rich, Ba-enhanced, chemically peculiar stars constitute 16% of the entire stellar population within the temperature range 6700 < T_eff < 7500 K. Figure 10 illustrates an increase in the fraction of Ba-enhanced stars with both effective temperature and stellar mass. The fraction could be as high as 40% at temperatures above 7200 K or stellar masses higher than 1.5 M_⊙. Note that we see a slightly decreasing trend at the high-T_eff and high-mass ends. Given the imposed hard cut at 7500 K in the abundance estimates derived with DD–Payne and the small number of high-mass (>2 M_⊙) stars, it is difficult to unambiguously relate such a trend to either a genuine phenomenon or to data artifacts. The trend can be also sensitive to the mass estimates used here, which is difficult to assign for chemically peculiar stars, given that stellar models are not precise enough to account for the atomic diffusion effects experienced by these stars. Our mass estimates are derived by matching stellar isochrones from the Dartmouth Stellar Evolution Database (Dotter et al. 2008) using the Bayesian method presented in Xiang et al. (2019), assuming that all stars have an initial metallicity of [Fe/H] = −0.1 dex and [α/Fe] = 0.0 dex.

Zoom In Zoom Out Reset image size

Figure 10 also shows the fraction of peculiar stars defined with respect to other elements, selected according to the abundance pattern revealed in Figure 8. About 10%–20% of the stars with T_eff > 7000 K are found to have enhanced nickel abundances with respect to iron ([Ni/Fe] > 0), and a similar fraction of stars have depleted calcium abundances with respect to iron ([Ca/Fe] < 0).

The high incidence of these Ba-enhanced stars, alongside their distribution in T_eff–$\mathrm{log}g$ space and elemental abundance patterns, suggests that these stars are plausibly related to (but not necessarily identical with) previously known Am stars (Conti 1970; Smith 1971; Preston 1974; Abt 1981; Gray et al. 2016; Qin et al. 2019). In particular, Gray et al. (2016) identified 1067 Am stars from the LAMOST spectral database in the Kepler field utilizing the MK classification method, corresponding to an Am frequency of 34.6% for stars with spectral type between A4 and F1. In a dedicated search of metal-line stars in LAMOST DR5, Qin et al. (2019) identified 10,503 Am/Fm stars with the random forest method and reported an incidence of about 22% for stars with (J − H) > 0.1. These numbers are qualitatively consistent with ours, with slight differences possibly due to differences in the approach used for peculiar star characterization. On closer inspection, however, we find that the methods do not always select identical sets of stars. The cross-match of our sample with Gray et al. (2016) using the LAMOST spectra ID yields only 258 stars in common, 175 of which are identified as either Am stars or stars with strong Sr lines. A cross-match with Qin et al. (2019) yields only 1881 stars in common. This relatively low overlap is in part due to the conservative design of The DD–Payne, which recommends Ba-abundance determinations only when they can be deemed physically sensible (Xiang et al. 2019). On the other hand, it is known that some Am/Fm stars do not exhibit Ba enhancement (Ghazaryan et al. 2018). Thus, our selection of Ba-enhanced stars omits a fraction of the Am/Fm population. We therefore emphasize that, although we conclude from this study that our sample is mostly related to Am/Fm stars, we do not necessarily expect it to be a strict subset of the Am/Fm stars identified by Qin et al. (2019).

Am/Fm stars have been suggested to be mostly slow rotators compared to chemically normal stars (Abt & Morrell 1995; Abt 2000). Considering that the resolution of LAMOST spectra is too low to yield robust measurements of rotation velocity, here we make use of the GALAH DR2 catalog (Buder et al. 2018) to examine the rotation of Ba-enhanced A/F stars. Figure 11 shows [Ba/Fe] versus [Fe/H] as well as the T_eff–log g diagram from GALAH DR2. We have selected stars with good S/N in the GALAH spectra by requiring S/N > 20 for both the c2 and c3 spectral segments. No cut based on abundance quality flags in GALAH DR2 is adopted because this would discard the peculiar stars that are of interest here; we find that all stars with [Ba/Fe] > 1.0 dex are flagged (i.e., marked as questionable estimates) in GALAH DR2. This flagging indicates that the abundance measurements are extrapolations outside the range of the training set. But as we have demonstrated above, the spectral features are strong, suggesting that the measurements still have value in a relative sense. Indeed, in the distribution of selected stars across the T_eff–$\mathrm{log}g$ diagram in Figure 11, we see the same behavior exhibited by the LAMOST sample in Figure 6. Note that here we show stars with [Ba/Fe] > 1.0 dex and [Fe/H] > −0.1 dex, rather than [Fe/H] > −0.2 dex as adopted for the LAMOST data, given that the [Fe/H] of GALAH DR2 was found to be 0.05–0.1 dex higher than the recommended value in the LAMOST catalog based on the APOGEE (Payne) scale (Xiang et al. 2019).

Zoom In Zoom Out Reset image size

Considering the significant variation of [Ba/Fe] from GALAH DR2 with T_eff, we introduce a further T_eff-dependent criterion to differentiate between Ba-normal and Ba-enhanced stars, thereby yielding a clean sample that can be used to characterize the shape of the Ba-enhanced v sin i distribution. Specifically, we compute the median and dispersion of [Ba/Fe] in 100 K wide T_eff bins and then take stars with [Ba/Fe] values deviating by more than 3σ from the median as the sample of Ba-enhanced stars. In total, there are 1188 Ba-enhanced stars with 6700 < T_eff < 7500 K, constituting 18% of the total number of stars in this temperature range, as is consistent with the LAMOST results (16%). The v sin i distribution of these Ba-enhanced stars from GALAH DR2 is shown in Figure 12. The majority of the Ba-enhanced stars rotate slower than the normal stars, consistent with the Am stars studied by Abt (2000).

Zoom In Zoom Out Reset image size

However, a considerable fraction of Ba-enhanced stars have large v sin i; around 37% have v sin i > 25 km s⁻¹, which is comparable to the bulk of normal stars. These stars may have slightly different levels of abundance enhancement than the remainder of the Ba-enhanced population. As we found in Section 3.2 with the LAMOST data, Ba-enhanced stars exhibit enhanced [Fe/H] but depleted [Mg/Fe] with respect to normal stars on the whole. However, the Ba-enhanced stars with larger v sin i values tend to have slightly larger [Mg/Fe] than the slower rotators. On the other hand, 18% of the Ba-normal stars have a v sin i smaller than 20 km s⁻¹. These stars tend to have higher [Fe/H] than the stars with larger v sin i, indicating they are probably also peculiar to some extent. It should be noted, however, that their [Mg/Fe] values are comparable to those of the faster rotators and systematically higher than those of the Ba-enhanced stars.

As a final note, we highlight that the GALAH DR2 v sin i distribution for the Ba-normal stars exhibits a peak near ∼35 km s⁻¹. This is systematically lower than found in the literature (e.g., Abt 2000); previous studies have suggested that a large fraction of A stars can be expected to have v sin i larger than 100 km s⁻¹. This discrepancy can mainly be attributed to the tendency by GALAH DR2 to underestimate v sin i at the high-v sin i end, which will be improved on by GALAH DR3 (S. Buder 2020, private communication). Given that GALAH DR2 v sin i estimates are robust in the relative sense, however, this does not affect the main conclusion of this paper.

The properties presented in Section 3 imply that the Ba-enhanced, chemically peculiar A/F stars are related to the Am/Fm stars that have been widely studied in the literature. Extensive effort has been made to describe these stars in the context of atomic diffusion due to stellar radiative acceleration (Michaud 1970; Turcotte et al. 1998; Richer et al. 2000; Richard et al. 2001; Talon et al. 2006; Vick et al. 2010; Michaud et al. 2011; Deal et al. 2020), and the abundance patterns of elements from C to Ni have indeed been reproduced quite well by introducing mixing mechanisms such as turbulence (Turcotte et al. 1998; Richer et al. 2000; Richard et al. 2001) and mass loss/stellar winds (Vick et al. 2010; Michaud et al. 2011).

Here we compare our observed abundance patterns with the stellar radiative acceleration models from the literature, making two main assumptions. Our first main assumption is that, on average, the Ba-enhanced, chemically peculiar stars share the same initial (birth) abundances as chemically normal stars of similar mass. This is supported by several direct and indirect pieces of evidence. First, the similarity in their spatial distributions and kinematics, and the fact that all MS and subgiant stars with masses larger than 1.4 M_⊙ are younger than ∼2.5 Gyr, suggests that the birth metallicities of Ba-enhanced chemical peculiars should not be very different from those of normal stars. Indeed, in the temperature range 6700 < T_eff < 7500 K, stars more massive than 1.5 M_⊙ are subgiants (Figure 6), for which there is a tight relation between mass and age (where age mostly indicates MS lifetime), implying that stars with similar mass also have similar age. Meanwhile, stars with similar ages can be expected to have similar birth metallicities, given that the dispersion in the metallicity of the interstellar medium at a given radius in disk galaxies is small (<0.1 dex, e.g., Kreckel et al. 2019). Second, and more directly, the low-T_eff star in Gaia wide binary pairs have identical abundances when the pair star is either a Ba-enhanced chemically peculiar star or a normal star (Figure 13). The abundance of the low-T_eff pair star in a Gaia wide binary system serves as an indicator of the birth abundance of the binary system (and thus the high-T_eff star) because neither stellar evolution processes nor external accretion events will have a large impact given the thick convective envelopes and young ages of these stars.

Zoom In Zoom Out Reset image size

Our second main assumption is that chemically normal stars do not experience the radiative acceleration process or, if they do, that the effects are negligible compared to the impact on Ba-enhanced, chemically peculiar stars. In this way, differential abundances between peculiar stars and normal stars can be taken as indicators of the impact of the radiative acceleration process on elemental abundance enhancements. We emphasize that the validity of this assumption is in need of verification.

Figure 14 shows the mean abundance and dispersion for both peculiar and normal stars with a range of masses and compares the differential abundances of these stars with the stellar radiative acceleration models of Vick et al. (2010) and Talon et al. (2006). There we see that the mean abundances of normal stars are similar (within ∼0.1 dex; e.g., panel a) among different masses. This suggests that the abundances generally do not vary strongly with age for these relatively young stellar populations. The normal stars with similar masses exhibit a dispersion of ∼0.2 dex, significantly larger than the reported measurement errors (≲0.05 dex). This could be due to spatial variation stemming from local abundance gradients or possibly nonnegligible atomic diffusion. Interestingly, the Ba-enhanced, chemically peculiar stars of different masses also exhibit consistent mean abundances (e.g., panel b). This might be a useful constraint on the stellar physical processes causing the chemical peculiarity, such as variations in mass-loss rate with stellar mass. All masses exhibit large scatter in abundances (0.1–0.2 dex for Mg, Si, Ca, Ti, Fe; 0.3–0.5 dex for Cr, Mn, Ni, Ba). For a few elements, such as Cr, Mn, and Ni, the scatter is apparently larger than that of the chemically normal stars. This is probably due to the fact the abundances of these Ba-enhanced, chemically peculiar stars have been altered by a variety of factors that depend, for instance, on age, mass-loss rate, and rotation speed.

Zoom In Zoom Out Reset image size

The differential abundances between the peculiar and normal stars exhibit a similar pattern to predictions from the model of Vick et al. (2010), in which surface abundances are the consequence of competition between atomic diffusion due to gravitational settling and radiative acceleration, modulated by stellar mass loss (panel c). However, there are significant differences between the model and the observations for a few elements. Whereas stars with M = 1.5 M_⊙ and M = 1.7 M_⊙ exhibit iron-peak elemental abundances (Cr, Mn, Fe, Ni) in good agreement with the models for a mass-loss rate of 5 × 10⁻¹⁴M_⊙ yr⁻¹, the models predict stronger depletion of Mg, Si, Ca, and Ti. The pattern in abundances predicted for these elements, however, is similar to the observations; in both cases, Si and Ti abundances are higher than Ca. For stars with M = 1.9 M_⊙, the model with a mass-loss rate of 1 × 10⁻¹³M_⊙ yr⁻¹ provides a much better match to the observations than the model with a mass-loss rate of 5 × 10⁻¹⁴M_⊙ yr⁻¹ (panel d).

The observed differential abundance patterns are also qualitatively consistent with the model of Talon et al. (2006), in which the surface abundances are modulated by turbulence induced by stellar rotation (panel e). The values of the abundance alteration depend sensitively on the rotation velocity V_rot. The observed differential abundances are comparable to the prediction of an isotropic diffusion model with a V_rot of about 15 km s⁻¹ or an anisotropic diffusion model with V_rot between 15 and 50 km s⁻¹. These rotation velocities are qualitatively consistent with the observed v sin i from GALAH DR2 (Figure 12). For Mg, Si, Ca, and Ti, the turbulence models seem to match the data better than the mass-loss models. This is consistent with the finding of Michaud et al. (2011).

These results suggest the Ba-enhanced, chemically peculiar stars are inevitably the consequence of stellar internal atomic transport processes. We cannot, however, strongly prefer one model over the other. Indeed, there are a number of reasons why we might expect the models and the observations to differ. First, as mentioned above, the "normal" stars may not actually reflect true initial abundances. Second, the mass-loss rate adopted by the model may not be the optimal one, and it is possible that stars with the same mass exhibit different mass-loss rates, depending on a number of factors, such as age and rotation speed. Furthermore, it is possible that both processes, turbulence and mass loss, are present in stars. Understanding how these two processes are related and act together is key from both observational and theoretical points of view.

Stellar photospheric abundances can also be altered by external accretion events, such as supernova (SN) pollution, mass transfer from an AGB companion, and planet engulfment. In this section, we briefly discuss the potential impact (if any) of these processes on the origin of these A- and F-type Ba-enhanced, chemically peculiar stars.

4.2.1. SN Pollution

4.2.2. Mass Transfer from AGB Companion

The overabundances could be due to the accretion of material from AGB companions, which produce a large amount of Ba (e.g., Busso et al. 1999; Karakas & Lattanzio 2014). In this case, we might expect the enhanced stars to hold unseen WD companions that have evolved from progenitor AGB stars. To further consider this possibility, here we examine the observed kinematics and ultraviolet (UV) properties of the Ba-enhanced stars for signals of WD companions.

First, we examine the scatter RVs between multiepoch observations from both Gaia DR2 and LAMOST for signals of binarity. For Gaia DR2, the RV scatter is derived by multiplying the reported RV errors by the number of measurements, given that the former is defined as the standard deviation of multiepoch RV measurements (see Equation (1) of Katz et al. 2019). For LAMOST, about one-third of the entire LAMOST spectral data set are repeat visits of common targets, and for each visit, there are two or three consecutive exposures (for the exposure strategy of LAMOST, see, e.g., Yuan et al. 2015). Recently, RVs have been derived from these single-exposure spectra (Y. Yang et al. 2020, in preparation) using the LAMOST stellar parameter pipeline at Peking University (LSP3; Xiang et al. 2015a). Uncertainties in the RV measurements depend on both the apparent magnitude and the spectral type of each star. To ensure sufficient RV precision, we select only Gaia DR2 sample stars with G-band magnitude brighter than 12 mag and LAMOST sample stars with S/N > 50. For these relatively hot A/F-type stars, the typical RV error for a single-epoch measurement is expected to be 4–5 km s⁻¹ for both Gaia DR2 and LAMOST. Note that this is ∼$\sqrt{\pi /2N}$ times larger than the reported Gaia RV uncertainty, which is for the combined (mean) velocity, where N represents the number of observation epochs. We further require that each target is observed in at least five epochs, leading to a total of 4002 stars from Gaia DR2 and 5276 stars from LAMOST with 6700 < T_eff < 7500 K and [Fe/H] > −0.2 dex. The RV variations measured with LAMOST are quantified as the ratio between the scatter of the RV measurements and the measurement error, $R=\sqrt{{{\rm{\Sigma }}}_{i=1}^{N}{({v}_{i}-\bar{v})}^{2}/{\sigma }_{v,i}^{2}}/\sqrt{N}$, where v_i is the velocity of the i_th epoch and σ_{v, i} is the velocity error.

For a binary system in a Keplerian orbit, the velocity amplitude of the A/F star's companion would be

$$\begin{eqnarray*}&&{K}_{1}=\sqrt{{{GM}}_{2}^{2}/[a({M}_{1}+{M}_{2})(1-{e}^{2})]}\times \sin i,\end{eqnarray*}$$

where M₁ and M₂ are the masses of the observed A/F star and the unseen WD, respectively; a is the total separation; e is the orbital eccentricity; and i is the inclination angle of the system. Here we consider M₁ = 1.5 M_⊙ and adopt M₂ = 0.6 M_⊙ typical of DA WDs (e.g., Rebassa-Mansergas et al. 2015). We also assume that e = 0, which is plausible for a close binary, leading to

$$\begin{eqnarray*}&&{K}_{1}=\displaystyle \frac{12.3}{\sqrt{a(\mathrm{au})}}\sin i\,(\mathrm{km}\,{{\rm{s}}}^{-1}).\end{eqnarray*}$$

Given the precision of RVs in Gaia DR2 and LAMOST discussed above, this implies that unseen WD companions should be identifiable within about 10 au of the target stars. About half of the discovered binary systems match this criterion (see Figure 2 of Duchêne & Kraus 2013).

Figure 15 compares the distribution in RV scatter between multiepoch measurements for Ba-enhanced, chemically peculiar stars and for chemically normal stars, measured with both Gaia DR2 and LAMOST. Chemically peculiar stars do not exhibit significantly greater RV scatter than chemically normal stars, implying that binary evolution is unlikely an important driver of chemical peculiarity. This finding is consistent with previous results for Ba-rich MS (turnoff) stars (Milliman et al. 2015), but is notably different than appears to be the case for Ba-rich giant stars, which are widely believed to be binary products (McClure 1983).

Zoom In Zoom Out Reset image size

WDs can also reveal themselves through differences in the UV–optical colors of chemically peculiar stars versus chemically normal stars. To construct color measurements we use the far-UV (FUV) and near-UV (NUV) photometry from the Galaxy Evolution Explorer (Galex) DR5 (Bianchi et al. 2011), for which a photometric error of better than 0.1 mag is required for both filters, and use the G magnitude from the Gaia DR2 (Gaia Collaboration et al. 2018). The photometry is dereddened using E(B − V) derived from the star pair method (Yuan et al. 2013), which takes as input the LAMOST DD–Payne stellar parameters and multiband photometry (a brief introduction can be found in Xiang et al. 2019). The typical precision of the E(B − V) estimates is 0.01–0.02 mag. To convert E(B − V) to reddening in the Galex and Gaia passbands, we use an extinction coefficient depending on T_eff and E(B − V) that we have computed by convolving the Kurucz synthetic spectra (Castelli & Kurucz 2003) with the Fitzpatrick extinction law (Fitzpatrick 1999). Figure 16 shows the distributions of chemically peculiar and chemically normal stars in the FUV − G versus NUV − G color–color diagram.

Zoom In Zoom Out Reset image size

For reference, we have examined Gaia WDs within 300 pc of the solar neighborhood, finding an average FUV absolute magnitude of 11.0 mag, which is only 0.6 mag dimmer than the FUV absolute magnitude of our A/F sample stars (6700 < T_eff < 7500 K) within 300 pc. In the Gaia G band, these WDs are much fainter (by 8.5 mag) than A/F stars. This suggests that, if there is a WD companion around a Ba-enhanced, chemically peculiar star, we should observe an FUV − G color excess of 0.5 mag. However, Figure 16 demonstrates that the chemically peculiar stars have similar colors to normal stars, implying that they likely do not have WD companions.

The significant enhancement in iron-peak elements but depletion in Mg characteristic of Ba-enhanced, chemically peculiar stars also strongly argues against mass transfer from AGB companions as the responsible mechanism, given that AGB stars do not produce such an abundance pattern.

4.2.3. Planet Engulfment

It has been suggested that Earth-like planets could be engulfed onto the surface of a MS (turnoff) star, altering the observed chemical stellar surface abundances as the engulfed planet gets dissolved and mixed in the convective envelope over short timescales (a few years; Church et al. 2020). To investigate if this mechanism can provide a reasonable explanation for Ba-enhanced, chemically peculiar stars, we carried out rough estimates of the abundance pattern of an A/F star assuming it has engulfed a terrestrial planet with chemical composition similar to either Earth or Mercury into its convective envelope (because the chemically peculiar stars are iron-peak enhanced, if they formed via planet engulfment, the engulfed planets would seem most plausibly terrestrial ones). For element X, the minimum mass that must have been accreted from the planet in order to enrich the stellar envelope from an initial abundance [X/H]₀ to a current abundance [X/H] is

$$\begin{eqnarray}\begin{array}{rcl}{M}_{X,\mathrm{planet}} & = & {M}_{\mathrm{cenv}}\times {X}_{\odot }/{H}_{\odot }\\ & & \times \ \left({10}^{\left[X/{\rm{H}}\right]}-{10}^{{\left[X/{\rm{H}}\right]}_{0}}\right)\times {A}_{X},\end{array}\end{eqnarray} \tag{ 1 }$$

where M_cenv is the mass of the stellar convective envelope, X_⊙/H_⊙ is the solar abundance of X in absolute value, and A_X is the atomic mass number of X. This assumes that all accreted material is reserved in the convective envelope, although part of this material in reality is likely to enter below the base of the convective envelope, in which case planets with larger masses than considered here would be required. Our fiducial case adopts a star with a convective envelope mass of 1.5 × 10⁻⁴ M_⊙, typical for stars at the border of the T_eff–$\mathrm{log}g$ diagram (Figure 7), and an initial abundance (before accreting the planet) set to the median values of the abundances characteristic of chemically normal stars. We adopt the estimates of Morgan & Anders (1980) for the abundances of Mercury and adopt abundances from McDonough & Sun (1995) for Earth.

Figure 17 shows the alteration in stellar abundances for the fiducial star after accreting a Mercury or an Earth. The observed abundance enhancement/depletion patterns of the Ba-enhanced, chemically peculiar stars are shown for comparison, adopting the differential abundances between chemically peculiar and normal stars, as in Section 4.1. Ba-enhanced, chemically peculiar stars exhibit much stronger Ba enhancement than that caused by the engulfment of a Mercury- or an Earth-like planet. The depletion of Mg and Ca characteristic of Ba-enhanced, chemically peculiar stars is also inconsistent with the engulfment of terrestrial planets like Earth and Mercury, given that they contain large amounts of these elements and will cause a significant abundance enhancement to the star after the engulfment events. At the same time, these planets cannot be responsible for the enhancements in Cr, Mn, and Ni observed for the Ba-enhanced stars. These patterns are more consistent with enhancement via a stellar evolution mechanism, as discussed in the last section. We therefore conclude that planet engulfment is quite an unlikely explanation, with the caveat that the details of how mixing takes place in the convective envelope may be important. It has been suggested that the material accreted onto the surface of a star could be significantly reduced in a short timescale (1000 yr) due to thermohaline mixing (Vauclair 2004; Théado et al. 2009; Théado & Vauclair 2012).

Zoom In Zoom Out Reset image size