WIYN Open Cluster Study 89. M48 (NGC 2548) 2: Lithium Abundances in the 420 Myr Open Cluster M48 from Giants through K Dwarfs^*†

Detecting Li in stars that are expected to be rotating rapidly and where the Li line is expected to be very weak can be quite challenging. We have nevertheless observed as many stars as was feasible in case there were discoverable surprises. Unfortunately, nearly all stars have upper limits near or above the meteoritic A(Li) (Figure 2), so it is not possible to discern whether these stars are depleted in Li or even slightly enriched. Although star 3010 has A(Li) < 2.9 dex, we caution against concluding that it is depleted. Note that the Hyades and Praesepe Am stars of Burkhart & Coupry (1989, 1998) were not included in the prime sample of C17.

Several Pleiads show a plateau near 3.3 dex, consistent with the meteoritic A(Li) and an assumption of being undepleted. Most Praesepe stars and possibly one Pleiad are depleted relative to the Pleiades trend and to M48. Of the five M48 stars in this T_eff range, at least two stars (2080, 2091) are extraordinary in exhibiting significant Li enrichment above 3.3 dex at levels 3.65–3.75 dex. We cannot discern whether star 2083 (A(Li) < 3.6 dex) is also extraordinary. It may be interesting that whatever physical processes that have enriched Li in the M48 stars have not affected the Pleiades or Praesepe.

In the region that is just slightly hotter than the Li Dip (6850–7200 K), three Hyades/Praesepe stars have detections near 3.3 dex, and two are slightly depleted (∼3.0 and 2.7 dex). Interestingly, one or maybe two Pleiads may also be slightly depleted. Four of the five M48 stars comprise two interesting pairs where each pair has nearly identical T_eff but different A(Li). Figure 5 shows spectra of and syntheses for the two pairs. In each pair the Fe i lines near λ6705 line up well, but the Li lines are clearly different. In both pairs, the higher A(Li) is consistent with meteoritic, but the other star is lower by 0.4 dex, illustrating clear Li depletion. The upper limit of star 2119 (A(Li < 2.5 dex) also suggests depletion. Possibly so does star 2113, A(Li) < 2.9 dex. While Li depletion in this mass range has been observed in substantially older clusters (e.g., in NGC 7789, NGC 3680, and NGC 6819 with ages 1.5–2.2 Gyr; Deliyannis et al. 2019), these examples show that Li depletion begins at least as early as 420 Myr.

Zoom In Zoom Out Reset image size

Figure 2 clearly shows the formation and evolution of the Li Dip in going from the Pleiades to M48 to the Hyades/Praesepe. In the Pleiades, a few stars may already be showing significant depletion (A(Li) = 2.5–2.9 dex), but most stars show at most a small depletion (if any) relative to the two stars near 6700 K that straddle the meteoritic A(Li).

The Hyades/Praesepe prime sample of C17 defines a very precise Li Dip. In the T_eff range 6200–6635 K (the "cool" or "red" side of the Li Dip), there is increasing Li depletion with little scatter. Then, remarkably, in a range of less than 100 K (from 6635 to 6735 K) there is an apparent very steep rise ("the Wall," or the "hot" or "blue" side of the Li Dip) defined by six Hyades/Praesepe detections from A(Li) = 1.70 to 3.33 dex and enhanced by three upper limits between 1.85 and 2.01 dex. Note that the two Pleiads mentioned above lie at the top of the Wall. A similar vertical structure is seen in the older clusters NGC 752 (1.45 Gyr; Boesgaard et al. 2022), NGC 3680 (1.75 Gyr; Anthony-Twarog et al. 2009), and the much richer samples of NGC 7789 (1.5 Gyr; Twarog et al. 2020) and NGC 6819 (2.2 Gyr; Deliyannis et al. 2019) that separates hotter stars that show a large variety of A(Li) (from 3.3 to < 1.6 dex) from Li Dip stars whose A(Li) are uniformly <2.4 dex across the hotter 400 K in a portion of the Li Dip. Unfortunately, only four M48 stars fall in this T_eff range, so the situation in M48 is unclear. Two are detections at A(Li) = 2.2 and 2.6 dex, while the other two are upper limits at 2.5 and 2.6 dex.

All M48 stars on the cool side of the Li Dip have detections and generally lie between the Pleiades and the Hyades/Praesepe. Like the Hyades/Praesepe, the cool side of the M48 Li Dip shows a clear pattern of increasing Li depletion with increasing T_eff, although with more scatter than in the Hyades/Praesepe. Figure 6(a) shows a pair near 6400 K with A(Li) that differ by 0.55 dex. Either such a difference becomes smaller by the age of the Hyades/Praesepe, or the Hyades/Praesepe is unusual, or the samples are too small to fully describe how Li Dip stars act.

Zoom In Zoom Out Reset image size

The Li Plateau of the Pleiades shows very little slope, if any, but may be slightly depleted, as compared to the hotter stars discussed above and to the meteoritic A(Li). The increasing slope from the hottest dwarfs to G (and K) dwarfs is consistent with SSET, as discussed earlier. However, the older clusters show a slope in the Li Plateau, which steepens further for T_eff < 6000 K, so we limit our definition of the Li Plateau from about 6200 to 6000 K. SPTLBs provide strong evidence that the Li Plateaus in the Hyades and in M67 are depleted (Section 1), but any differences in the Plateaus of the clusters discussed here are subtle. M48 and the Hyades/Praesepe appear to be lower than the Pleiades, but not necessarily distinct from each other. However, if Galactic Li production affected the initial abundances of these clusters (end of Section 4), then to study the relative Li depletion of these clusters we must place them on a common scale for initial A(Li). In particular, relative to the Sun, the A(Li) of all Pleiads must be lowered by 0.03 dex, all Hyads and Praesepids must be lowered by 0.15 dex, and all M48 members must be increased by 0.06 dex. Figure 4 shows a close-up of a portion of the bottom panel of Figure 2 using this adjusted scale for A(Li), for the regions where this issue is most relevant. Under these assumptions, the Hyades/Praesepe Li Plateau is clearly lower than the one in M48, and the Pleiades Li Plateau remains slightly above M48. So here, too, we see a progression of Li depletion from the Pleiades to M48 to the Hyades/Praesepe.

Just hotter than our defined boundary for the Li Plateau (near 6257 K) there is evidence for some scatter, as illustrated in Figure 6(b). (Figures 2 and 4 show that the Pleiades might also show scatter at this same T_eff.) However, the M48 Li Plateau itself is very tight with remarkably little scatter.

The Li–T_eff trend in M48 lies significantly below that of the Pleiades and slightly above that of the Hyades/Praesepe. We argue below that the dominant factor is increasing depletion with age; however, metallicity may also play a role. In addition, whereas the early G stars show remarkably tight Li–T_eff relations, cooler stars show large scatter, as exemplified by direct comparisons of spectra of three pairs of stars in Figures 6(c), (d), and (e). Each pair has nearly identical T_eff (near 5300, 4960, and 4485 K) but significantly different A(Li) (0.6, 0.75, and >0.8 dex, respectively).

Standard Li depletion of G/K dwarfs increases with increasing metallicity (e.g., P97). Since M48 is metal-poor compared to the Pleiades, age may be the dominant parameter affecting the difference in the Li–T_eff trends of these clusters.

In the Pleiades (Butler et al. 1987; Soderblom et al. 1993; Bouvier et al. 2018) and in the slightly older M35 (Anthony-Twarog et al. 2018a; Jeffries et al. 2021), ultrafast rotators (UFRs) exhibit significantly higher A(Li) than slower rotators, at least in part creating the large Li dispersions seen in these clusters. If M48 formed with its share of UFRs, they would appear to have all spun down (Barnes et al. 2015); all of our G and K dwarfs have v sin i < 20 km s⁻¹. It would be nice to know what happens to the A(Li) of UFRs as they spin down, but clusters with ages closer to those of the Pleiades and M35 may be required to address this. M48 itself provides at best a mild hint: we have detected Li in all M48 stars in the T_eff range 5400–5050 K, which is the upper part of the T_eff range where high-Li UFRs appear in the Pleiades and M35, and these M48 stars show a Li dispersion of no more than 0.6–0.7 dex. By contrast, the dispersions in the Pleiades and M35 near 5300–5200 K are at least 1.3 dex, so it is possible that the Li dispersions in this T_eff range have shrunk by the age of M48. Cooler stars in the Pleiades and M35 show even larger Li dispersions, at least 1.8 dex in both clusters, but because our M48 data include upper limits for T_eff < 5050 K, we cannot evaluate the full dispersion range for cooler M48 stars. All we can say is that it is at least 0.7–0.8 dex near 5000 and 4500 K.

Comparison of Li depletion in M48 to that in the Hyades/Praesepe is complicated by the difference in [Fe/H] of ∼0.21 dex. As discussed above, Galactic Li production suggests that the Hyades/Praesepe A(Li) should be moved down relative to M48 by about 0.21 dex, in which case the entire G/K dwarf Li–T_eff trend in the Hyades/Praesepe lies clearly below that of M48 (Figure 4). But which is more important, age or metallicity (or something else)? By way of example, SSET models of P97 show a ∼0.4 dex difference in Li depletion between [Fe/H] of −0.06 dex (M48) and +0.15 dex (Hyades/Praesepe) at 5500 K. The Galactic-production-adjusted difference between the Li–T_eff trends of these clusters at 5500 K is 0.66 dex, suggesting that the difference between the two is partly due to metallicity and partly due to age. However, Somers & Pinsonneault (2014) highlight a number of uncertainties that affect standard depletion, some of which (like opacity) are themselves metallicity dependent. So the relative effects of metallicity and age on Li depletion could be a bit different than indicated here. But what does seem clear is that age-related Li depletion is far greater between the Pleiades and M48 than it is between M48 and the Hyades/Praesepe. Since the age differences are similar, where M48 is about 300 Myr older than the Pleiades and about 250 Myr younger than the Hyades/Praesepe, MS G/K dwarf Li depletion slows down over time.

SSET predicts that as stars evolve past the turnoff to lower T_eff and become subgiants, the SCZ deepens past the boundary of the Li preservation region, thereby diluting the surface A(Li) (Iben 1967; Deliyannis et al. 1990). The SCZ reaches a maximum depth by mass fraction, resulting in maximum dilution, after which it retreats, leaving behind a diluted Li–T_eff plateau. Standard dilution alone can explain the difference between the turnoff A(Li) and the diluted plateau in metal-poor stars (e.g., Ryan & Deliyannis 1995). However, open clusters show a more complex pattern (Anthony-Twarog et al. 2009; Cummings et al. 2012; Boesgaard et al. 2015; Anthony-Twarog et al. 2018b; Deliyannis et al. 2019; Anthony-Twarog et al. 2021). Near solar metallicity, decreasing giant mass (older cluster age) shows increasing departure from SSET, perhaps largely due to Li depletion during earlier phases. Decreasing metallicity suggests closer conformity to SSET.

Since M48 has very few giants, we make an exception and consider the "slm" giant (star 2002) in addition to the two "sm" giants (stars 2003–4). Using a 420 Myr, [Fe/H] = −0.063 dex Y² isochrone (Yi et al. 2001), we estimate our giants' masses to be ∼2.9 M_⊙. Of particular relevance is the right panel of Figure 14 from Charbonnel & Lagarde (2010), which shows post-MS Li evolution tracks for 2.5–2.7 M_⊙ giants with zero-age MS (ZAMS) equatorial velocities (V_ZAMS) of 0 (SSET), 110, and 250 km s⁻¹ (these models are also shown in our Figure 7). The standard track leaves the MS with no Li depletion, whereas the other two leave with depletions of 1.15 and 2.55 dex, respectively, due to rotational mixing. All three begin dilution near 5600–5500 K (a similar T_eff to the metal-poor models of Ryan & Deliyannis 1995) and dilute by 1.85, 1.55, and 1.55 dex, respectively (factors of 71, 35, and 35) down to 4800–4600 K.

Zoom In Zoom Out Reset image size

In Figure 7, star 2002 is shown with a filled green disk (slm), to be distinguished from 2003–4 shown with filled blue disks (sm). Star 2002 is remarkable in that its T_eff of 5520 K places it in the middle of the Hertzsprung gap, right at the boundary of where model dilution begins. The depleted A(Li) of 1.9 dex is consistent with MS Li depletion of the model with V_ZAMS = 110 km s⁻¹, or just slightly larger. Star 2004 (T_eff = 5071 K) is consistent with being in the middle of dilution, needing also MS depletion from the same 110 km s⁻¹ track as 2002. Star 2003 (T_eff = 4752 K) is consistent with being near the end of dilution and would require a slower rotator than the previous two, perhaps near 50 km s⁻¹. It is possible that stars 2004 and 2003 are more evolved red clump stars that have suffered an additional 0.1–0.2 dex depletion between the RGB and the red clump, according to the models of Charbonnel & Lagarde (2010). In that case, a model with V_ZAMS = 0–20 km s⁻¹ could match star 2004, whereas 30–40 km s⁻¹ could match star 2003. Except for this one possibility with star 2004, the other scenarios discussed above require approximately 0.5–1.0 dex Li depletion during the MS. Note that additional internal effects such as thermohaline mixing become important only for significantly lower model masses.

It is possible to envision more complex possibilities (see Section 5.2.2). For example, faster rotators may have depleted more Li, and then enrichment may have occurred through accretion/engulfment. Such a scenario may be discernible through abundance signatures other than Li.

Possible explanations for the two Li-rich stars with A(Li) = 3.65 and 3.73 dex include diffusion (radiative levitation), planetesimal accretion, and engulfment of a planet or Li-preserving brown dwarf. Each scenario has advantages and disadvantages.

Diffusion is more efficient in slower rotators, and our two stars indeed have v sin i of 33 and 60 km s⁻¹, which is slower than most of our hotter stars and a few stars slightly cooler. Two other stars in the same T_eff range rotate much faster and have upper limits in A(Li) of 3.6 dex, which is uninformative, and 3.3 dex, which is more convincingly below the two Li-rich stars. Star 2095 has a detection of A(Li) at 3.1 dex, which is near though perhaps marginally below meteoritic, and a v sin i of 61 km s⁻¹. However, Li richness in the models of Richer & Michaud (1993) at this age is constrained to the narrow T_eff range of 6950–7100 K, the "Li Peak," which is clearly cooler than our stars. On the other hand, the super-Li-rich dwarf J37 of NGC 6633 may lie in this T_eff range (Deliyannis et al. 2002). It was suggested that J37 might be better explained by planetesimal accretion (Laws & Gonzalez 2003), but the overall abundance pattern of a number of elements fit neither scenario conclusively (Ashwell et al. 2005).

A dwarfs have extremely thin SCZs, so it only takes accretion of just a small amount of hydrogen-poor material to enhance the surface abundances of the accreted elements. Accretion of planetesimals and planet engulfment are two possibilities for such enhancement. However, given how easy it is to produce such enhancements, one wonders whether Li-rich A dwarfs are observed sufficiently often. In M48, two out of five stars in this T_eff range are Li-rich, but we would expect to see Li-rich stars from accretion in hotter stars as well, yet most of our hotter M48 stars have A(Li) <3.5 dex. Another issue is that rotational mixing may deplete Li, as suggested for the M48 giants (Section 5.2.1) and for stars a bit more massive than the Li Dip (Deliyannis et al. 2019). So accretion may enrich the surface Li and subsequent depletion may diminish it. Finally, some A dwarfs show a number of abundance anomalies, at least some of which might be explained by diffusion. Fortunately, the signatures of diffusion and accretion differ, so it may be possible to distinguish between these mechanisms by studying the abundances of appropriate elements; for example, accretion may show a pattern of more refractory elements than the volatiles (Nagar et al. 2020).

As stars evolve, differential contraction, together with loss of angular momentum from the surface, can trigger a secular shear instability (and others) that causes mixing and depletion of the surface Li, Be, and B (Pinsonneault et al. 1990; Chaboyer et al. 1995; Deliyannis & Pinsonneault 1997; Boesgaard et al. 2016). Much evidence from a variety of angles favors such rotational mixing as the dominant Li depletion mechanism in the Li Dip, as discussed in Section 1. In addition, since stars spin down (e.g., Barnes 2007) as the Li Dip develops (e.g., from the Pleiades to M48 to the Hyades/Praesepe), there is a correlation between stellar spin-down and Li depletion. Furthermore, the existence of Li dispersions at a given T_eff separates proposed mechanisms that are able to create such dispersions, such as the above rotational mixing, from those that cannot, such as diffusion and gravity waves. Our M48 data help define the evolution of the Li Dip and constrain the degree of Li dispersion in the Li Dip, which can help guide future models.

In principle, diffusion can occur in sufficiently stable stellar layers. It is thus possible that diffusion contributes to Li depletion in the very small T_eff range (of about 100 K near 6700 K, depending also slightly on age; Richer & Michaud 1993) where it is predicted to be significant. Diffusion is predicted to act differently on Be than rotational mixing. While both elements are fully ionized at the base of the SCZ (T_eff < 6650 K), they diffuse downward via gravitational settling at similar rates; the patterns become more complex at higher T_eff, as one or both elements are able to retain at least one electron at the base of the SCZ (e.g., Figure 12 of Richer & Michaud 1993). The preponderance of Li and Be data in the Li Dip (Stephens et al. 1997; Deliyannis et al. 1998; Boesgaard et al. 2001, 2004) is more consistent with the signature of rotational mixing. However, HR 6052 is an interesting exception (Stephens et al. 1997): at T_eff = 6712 K, it has A(Li) = 2.82 dex and A(Be) = 0.48 dex, which are 0.49 and 0.94 dex below meteoritic, respectively. Such a depletion pattern is inconsistent with rotational mixing, diffusion, or accretion, or combinations thereof. The diffusion models of Richer & Michaud (1993) have significant Be depletion at a T_eff slightly higher than 6850 K, but the A(Li) is then higher than meteoritic owing to radiative levitation. Interestingly, another star may be closer to this diffusion pattern: at T_eff = 6953 K, HR 1287 has A(Li) = 3.42 dex (near or slightly supermeteoritic) and A(Be) = 0.89 dex (submeteoritic; Boesgaard et al. 2001). It would be worthwhile to greatly expand the sample of stars in the T_eff range 7000–6600 K that have measurements of both A(Li) and A(Be).

Higher A(Li) in SPTLBs as compared to normal, single stars provide direct support for models with rotational mixing related to angular momentum loss. Here, too, stellar spin-down correlates with Li depletion, and Li dispersions distinguish models that can create such dispersions (rotational mixing) from those that cannot (SSET, including with convective overshoot, diffusion, and gravity waves). Convective overshoot for ZAMS-defined G dwarfs occurs during the pre-MS only, in contradiction to the degree of Li depletion seen in the Pleiades and to the MS Li depletion inferred from other clusters (including M48), while mass loss leads to various absurdities (see discussion in C17). Some diffusion may be possible, though helioseismology limits it to less than 0.1 dex in the Sun and it would be even smaller for cooler dwarfs. It is possible that mixing by gravity waves may play a role in addition to rotational mixing; however, the timing of the Li depletion places constraints. Li depletion slows down over time (Section 5.1.6), as does stellar spin-down, but mixing by gravity waves is more closely related to action of the base of the SCZ, whose depth remains steadier with age. Perhaps mixing by gravity waves becomes more important at older ages.

Given the preponderance of evidence suggesting that rotationally induced mixing through angular momentum loss (hereafter J-loss) is the dominant Li-, Be-, and B- depleting mechanism in dwarfs spanning a variety of spectral types, including G/K (C17, Section 1), we examine our data for possible connections between J-loss and Li depletion in G/K dwarfs.

Figure 8 shows rotational periods for the four clusters under consideration and clearly illustrates the spin-down of stars as a function of age and T_eff. Furthermore, the period–T_eff relations for the three older clusters have less scatter than the relation for the Pleiades. The Pleiades does show a more populous group of slower rotators with relatively little scatter, but there are also a number of UFRs. The fits for M48 and Hyades/Praesepe are for all stars, but for the Pleiades the fit is restricted to the more populous group of slower rotators. The fits stop at 4500 K because for the Pleiades and for M48 the typical scatter increases dramatically at lower T_eff. Figure 9 (panel (a)) shows the differences between these fits, ΔP, to help investigate how J-loss might be related to Li depletion.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Before making this comparison, we must take into account any (standard) pre-MS Li depletion at the base of the SCZ, which depends strongly on metallicity (P97; we ignore the possible complication that for T_eff < ∼5000 K additional Li depletion at the base of the SCZ may occur after 100 Myr). First, Figure 9(b) shows fits to the A'(Li)–T_eff relations for our clusters. For M48 and Hyades/Praesepe the fits go through the mean trends, but for the Pleiades the fit goes through the more populous group of lower A'(Li); below we will show that these tend to be slower rotators. Second, and keeping in mind the various uncertainties discussed by Somers & Pinsonneault (2014) about the absolute standard depletion of Li, Figure 9(c) shows the predicted standard pre-MS Li depletion (at 100 Myr) for the metallicities of our clusters, interpolated from the models of P97. Figure 9(d) shows the difference A'(Li)–pre-MS Li depletion, ΔLi, which might be related to Li depletion due to J-loss.

We can now compare ΔP to ΔLi. For the Pleiades and M48, ΔP is fairly constant in going from 6000 to 5300 K, and then it decreases with lower T_eff (panel (a)). Since mass decreases with lower T_eff, we might thus also expect stellar J content and perhaps J-loss to decrease from 6000 to 5300 K and further still below 5300 K. In striking contrast to this trend, ΔLi increases with decreasing T_eff throughout this T_eff range (panel (d)). This strongly suggests the importance of at least one additional parameter. We suspect that the key is how the Li preservation region divides into the SCZ and the radiative region below where Li remains preserved. The depth of the SCZ increases with decreasing T_eff, and the radiative Li preservation region correspondingly decreases (Deliyannis et al. 1990; Pinsonneault et al. 1990). Thus, to achieve a certain amount of ΔLi, less J-loss is needed with decreasing T_eff. The combined result is that greater Li depletion due to J-loss occurs for decreasing T_eff in spite of the diminishing J-loss with decreasing T_eff. We call upon new models to investigate these ideas.