WIYN OPEN CLUSTER STUDY. LVII. OXYGEN ABUNDANCES OF SOLAR-TYPE DWARFS IN THE HYADES AND NGC 752

The expectation of an age–metallicity relation in the Galatic disk is an important component of our understanding of Galactic chemical evolution. It has often been convenient to assume that Fe exhibits a clear correlation with age and thus can be used as a chronometer for chemical enrichment. However, the arguments against Fe being a suitable chronometer have been acknowledged for over two decades, and were first notably raised by Wheeler et al. (1989). Fe has two production sources, Type Ia supernovae (SNe) which arise via white dwarfs from low-mass stars with lifetimes ∼10⁹ yr, and Type II SNe which are the end result of high-mass stars with lifetimes ∼10⁷ yr. While these two sources thus contribute Fe on very different timescales, comparison of theoretical SN elemental yields with observed SN rates indicates that each SN type has contributed roughly equally to Galactic Fe over the history of the disk (Nomoto et al. 1997) suggesting that Fe should not be expected to exhibit a robust correlation with age. In addition the yield of Fe is sensitive to the details and parameters of the explosion for both SN types. The Fe ejected by Type II SNe is produced both by Si burning prior to the explosion and by O burning during the explosion; models indicate that the amount of Fe ejected is sensitive both to the amount of fallback onto the core and to the explosion energy (Woosley & Weaver 1995; Nomoto et al. 2006). Fe production by Type Ia SNe results from radioactive decay of ⁵⁶Ni synthesized during the explosion (e.g., Kuchner et al. 1994), and thus is sensitive to details of the explosion as well as to the progenitor system (Woosley 2001), the latter not yet being fully understood (e.g., Kobayashi & Nomoto 2009). Indeed, since Type Ia luminosity is driven by the decay of ⁵⁶Ni, Howell et al. (2009) argue that the observed scatter in Type Ia luminosities is difficult to explain if most Type Ia SNe explode near the Chandrasekhar mass.

In addition to the theoretical arguments against Fe as a robust age indicator, the lack of a strong Fe versus age relationship in the Galactic disk during the past ∼10 Gyr has been made clear by observational studies for over 30 years. The seminal work of Twarog (1980) investigated the age–metallicity relationship in the solar neighborhood via derivation of [Fe/H] from Hβ and uvby photometry for over 900 field F dwarfs. The results, when binned by age, clearly demonstrate a substantial rise, ∼1 dex, in mean metallicity over the history of the Galaxy, although with a rise of only ∼0.1 dex occurring in the past ∼5 Gyr. However, the high-resolution spectroscopic study of 189 field F and G dwarfs by Edvardsson et al. (1993) casts serious doubt on the reality of a well-defined age–metallicity relationship. Edvardsson et al. (1993) found that while the oldest and youngest groups of stars exhibited the "expected" low and high average [Fe/H], respectively, the intermediate group, roughly 2–10 Gyr, had large scatter and populated the [Fe/H] range equally across all ages. These field star results are corroborated by the low-resolution spectroscopic study of 39 open clusters by Friel et al. (2002). Open clusters (OC) can provide well-constrained abundances and are also more reliably dated than field stars; they are an ideal tool for examining Galactic chemical evolution. Friel et al. (2002) derived [Fe/H] from Fe and Fe-peak element spectral features by using the ratio of the average flux in a bandpass centered on each feature to the average flux of the local spectral continuum. The Friel et al. (2002) sample exhibits a uniform spread in [Fe/H] across the entire age range, and this result holds even when their cluster averages are adjusted for the [Fe/H] versus R_G gradient exhibited by their cluster sample; regarding their use of such a gradient, Friel et al. (2002) acknowledge that, while the gradient exhibited by their sample is consistent with previous determinations in the literature, it is possible that the assumption of a simple gradient is not correct, as argued by Twarog et al. (1997). More recently, Friel et al. (2010) again found no Fe versus age correlation using high-resolution spectroscopic abundances of 11 open clusters.

It is important to note that the roughly flat [Fe/H] versus age relationship found by these studies is actually predicted by chemical evolution models which incorporate the in-fall of unenriched gas onto the Galactic disk (e.g., Pilyugin & Edmunds 1996; Chiappini et al. 1997; Kobayashi et al. 2006). Such in-fall effectively "resets" the chemical clock, from which Fe would not easily recover due to its long production timescale, undermining its utility as an age indicator.

Given the poor theoretical and observational prospects for using Fe as a reliable age indicator, other elements should be more thoroughly explored. Wheeler et al. (1989) suggested the potentially excellent alternative that oxygen may be a superior chronometer for Galactic chemical evolution. In contrast to Fe, O has the advantage of having primarily one dominant source, namely, production by He burning in the cores of massive stars and ejection by the subsequent Type II SNe. Since O is produced by hydrostatic burning prior to the explosion, its yield is well-described by nucleosynthesis models alone, the results of which depend primarily on the progenitor properties rather than the hydrodynamical details of the explosion (Kobayashi et al. 2006). In addition to having a single production source, the yield of O from Type II SNe is independent of metallicity (Woosley & Weaver 1995), and the 10⁷ yr timescale of Type II SNe in principle allows stellar O abundances in the Galaxy to respond quickly to enrichment, and thus O has the potential to exhibit a robust correlation with stellar age. A possible caveat is that Type II progenitors do not fully disburse into the Galaxy before exploding, possibly leading to self-enrichment of star-forming regions and creating scatter in the [O/H] versus age relationship (Pilyugin & Edmunds 1996). Nonetheless, there is compelling observational evidence for a strong O versus age correlation, as demonstrated by King (1993, hereafter K93), using low-resolution 7774 Å O i triplet data of mostly solar-type dwarfs in 8 open clusters. One of the principal aims of our current study is to further test this encouraging endorsement of O as an age indicator.

In addition to its potential as an age indicator, O is frequently used as a tracer of Galactic chemical evolution via the [O/Fe] versus [Fe/H] relationship; this relationship reflects the relative contributions of Type Ia and Type II SNe to chemical enrichment. For [Fe/H] > −1 there is universal consensus for an approximately linear trend where [O/Fe] decreases with increasing metallicity with slope ∼  − 0.5, as shown by both field star data and chemical evolution models. A possible break in slope near [Fe/H] ∼ −1, the existence of which is debated (see below), would indicate the point at which Type Ia production of Fe reaches its maximum rate (Chiappini et al. 2003). However, the form of this relationship for [Fe/H] < −1 remains uncertain despite being the focus of many investigations; at least two possible slopes have been demonstrated, each having distinct implications for enrichment in the early Galaxy. Some studies (e.g., Carretta et al. 2000) suggest that, at low [Fe/H], the [O/Fe] ratio plateaus at values near ∼0.5, indicating dominant Type II enrichment, in agreement with recent chemical evolution models (Kobayashi et al. 2006). In contrast, a number of investigations (e.g., Boesgaard et al. 1999; Israelian et al. 2001) have found linear trends with slope ∼  − 0.3 (i.e., [O/Fe] increasing with decreasing metallicity), as predicted by chemical evolution models that incorporate a possible early contribution of Type Ia SNe to Fe production at Galactic age ⩽1 Gyr (Chiappini et al. 1999). Alternatively, values of [O/Fe] ∼ 1 at [Fe/H] ∼ −3 (e.g., Boesgaard et al. 1999; Israelian et al. 2001), higher than the pure Type II value of [O/Fe] ∼ 0.4, may suggest that O production was higher in the early galaxy (Israelian et al. 2001) and/or at very low metallicity. This finding is supported by the chemical evolution models of Kobayashi et al. (1998) and Kobayashi et al. (2006), which include metallicity-dependent SN yields, and show a roughly flat relationship for −3 ≲ [Fe/H] ≲ −1 that subsequently changes slope for [Fe/H] ≲ −3 [O/Fe], rising to [Fe/H] ≳ 0.7 near [Fe/H] ∼ −4.

The debate about the constancy of [O/Fe] at low metallicity arises from apparent discrepancies between O abundance indicators in different stellar types. Studies utilizing the 6300 Å [O i] line in giants frequently find an [O/Fe] plateau below [Fe/H] ∼ −1 (though see Figures 1 and 8 of King 2000, and associated discussion), while studies based upon the 7774 Å O i indicate a non-zero slope. Studies using the temperature sensitive near-UV OH lines (located between 3100 Å and 3300 Å) have variously yielded both results depending on the atmospheric parameters used (see Boesgaard et al. 1999; Israelian et al. 2001; Shchukina et al. 2005, for detailed discussions of these results; issues affecting the abundance indicators will be discussed in detail below). There have been various attempts to reanalyze and reconcile these findings using revised atmospheric parameters, non-LTE (NLTE) effects, and three-dimensional model atmospheres (Boesgaard et al. 1999; King 2000; Israelian et al. 2001; Shchukina et al. 2005; Gonzalez Hernandez et al. 2010). These studies appear to have achieved agreement in favor of a non-zero slope for [Fe/H] < −1, or even a single, constant slope over the entire metallicity range. Nonetheless, the overall picture remains unclear, as do the implications for Galactic chemical evolution.

Further progress in understanding the chemical evolution of O has been impeded by debate over the reliability of the principle O abundance indicators, namely: (1) the 6300 Å and 6363 Å [O i] lines, the latter being less commonly used, (2) the near-IR 7774 Å O i triplet, (3) the near-UV, electronic OH lines in the 3100 Å to 3300 Å region, and (4) the near-IR, vibrational-rotational OH lines in the 1.5μ to 1.7μ region, which are used only rarely, despite being regarded as potentially very reliable (e.g., Balachandran & Carney 1996; Melendez & Barbuy 2002). Both sets of OH lines are temperature sensitive, although it has been argued that the near-UV set can produce abundances in agreement with other indicators when appropriate atmospheric parameters are used (e.g., Boesgaard et al. 1999). In addition, the near-UV OH lines suffer from blends, thus requiring high-resolution spectra for accurate analysis, and furthermore have been found to be sensitive to temperature inhomogeneities (Gonzalez Hernandez et al. 2010, and references therein), thus indicating that abundance analyses using three-dimensional model atmospheres may be required.

The most widely trusted feature is the [O i] line at 6300 Å which is free of NLTE effects (Kiselman 1991; Takeda 2003). However, this line is very weak even in metal-rich dwarfs, and is blended with a Ni i feature, thus requiring synthesis and an accurate Ni abundance. As a result, the forbidden line is most reliably used in giants, where the gravity-sensitive feature is enhanced. However, possible modification of surface CNO abundances by evolutionary effects in evolved stars raises possible concerns about using such stars to trace Galactic chemical evolution. It is established that convective dredge-up on the red giant branch should have a negligible effect on the surface oxygen abundance, but effects by other forms of mixing (such as rotationally induced mixing) and processing have not been ruled out (Boesgaard et al. 1999; Schuler et al. 2006a). In addition, recent asymptotic giant branch models suggest that rotationally induced mixing could indeed dilute the surface O abundance with CNO-processed material (Stasin¨ska et al. 2010). Even in the seemingly simpler case of dwarfs, Garcia Lopez et al. (1993) found that Hyades and Ursa Majoris group F dwarfs exhibit a possible surface depletion of ∼0.1 dex in [O/H] for 6500 K < T_eff < 6900 K, perhaps related to microscopic diffusion. Beyond these concerns over possible evolutionary effects, concerns regarding abundance analyses of the [O i] lines have also been raised. Takeda (2003) and Israelian et al. (2004) each found substantial discrepancies (as much as ∼1 dex in the latter work) in metal-poor giants between the abundances derived from the forbidden lines and those derived from the near-IR O i triplet, and both studies also found that these discrepancies are not resolved using NLTE corrections. Israelian et al. (2004) argued that these discrepancies could be due to unreliable model atmospheres. In addition, Shchukina et al. (2005) have reported that the 6300 Å [O i] shows granulation inhomogeneity effects as large as 0.2 dex, and thus requires three-dimensional model atmospheres for proper analysis.

Unlike the forbidden lines, the near-IR O i triplet at 7772, 7774, and 7775 Å (hereafter simply "the triplet") has easily observed line strengths of tens of mÅ in solar-type dwarfs; in such stars, the feature is also free of any known significant blending effects. Furthermore, the O abundances of dwarfs should be free of evolutionary effects. These high excitation potential (9 eV) lines arise from a transition from the 3s to 3p orbitals with three possible values of the total angular momentum in the final state. Models of the triplet indicate that it suffers from NLTE effects (e.g., Kiselman 1991; Takeda 2003; Shchukina et al. 2005), which result from a sub-Plankian source function thus leading to over-excitation and overpopulation of the lower state, ultimately causing an increasing trend in LTE abundances with increasing temperature for T_eff ≳ 6200 K (Takeda 2003; Schuler et al. 2006b). Fortunately, the work of Shchukina et al. (2005), who used using three-dimensional model atmospheres, has indicated that the triplet is not significantly affected by granulation inhomogeneities.

In addition to the predicted NLTE effects, Schuler et al. (2004) discovered an increasing rise in LTE abundances with decreasing temperature for T_eff ≲ 5500 K in Pleiades dwarfs, with analogous trends subsequently found by King & Schuler (2005) and Schuler et al. (2006b) for Ursa Major Group dwarfs and Hyades dwarfs, respectively; this behavior is not predicted by NLTE calculations. Schuler et al. (2006b) found no strong correlation between [O/H] and Ca ii HK emission indicators, but noted that while this perhaps pointed to photospheric magneto-rotational activity (starspots, etc.), the chromospheric emission data was not contemporaneous with the abundance data, leaving the matter unclear. Schuler et al. (2006b) also showed that the low temperature trend could indeed be emulated using synthetic spectra generated from a linear combination of model atmospheres representing a base photospheric temperature with contributions from hot and cool spots, but this encouraging result could not firmly establish stellar spots as the source of the trend. The triplet abundance trend at low T_eff was further investigated by Shen et al. (2007) who found some correlation between triplet abundances and contemporaneous Hα and Ca ii data for IC 4665, as well as finding a strong correlation (albeit based on only three stars) with coronal X-ray emission. Although they suggested these results favor age-related chromospheric activity as a possible source of anomalous triplet-based O abundances at low T_eff, they could not rule out contributions from starspots; at present this puzzle remains unsolved.

While the above works have established the complications present in using the triplet to derive reliable abundances, they have also provided guidance in how to derive abundances effectively. The NLTE calculations by Takeda (2003) indicate that the triplet NLTE corrections in dwarfs are approximately constant over the near-solar temperature range (5500–6000 K), with no significant dependence on metallicity. This has been verified by Schuler et al. (2006b), who found constant Hyades dwarf triplet abundances in the range 6200 K ≳ T_eff ≳ 5500 K. These results imply that triplet NLTE departures can be nulled by a purely differential analysis of solar-type dwarfs in this T_eff with respect to the Sun. In addition, such a differential analysis is arguably more robust than applying any particular set of triplet NLTE corrections directly, given that the NLTE corrections from various sources agree only tolerably well at best, as discussed by Takeda (2003); Figure 4 of Takeda (2003), which compares their NLTE corrections to those from other groups, indicates that the discrepancy is typically >0.05 dex. Furthermore, a differential analysis eliminates other complications, e.g., the discrepancy between the traditional value (∼8.90) of the absolute solar oxygen abundance supported by models of the solar interior derived from helioseismology, and the lower value (∼8.70) derived using recent three-dimensional model atmospheres (see the discussion in Pereira et al. 2009). In short, this differential approach should produce trustworthy relative abundances, thus allowing reliable study of the chemical evolution of O.

A promising means to explore O evolution is by using open clusters, which can provide accurate ages and large sample sizes capable of yielding precise abundances. Two important, unresolved questions exist for open clusters (hereafter OCs), namely, (1) do they exhibit the same linear trend in [O/Fe] versus [Fe/H] evinced by field star data and chemical evolution models, and (2) can the [O/H] versus age relation found by K93 be verified? Answering these questions requires large, uniform samples of OCs spanning a wide range in both age and metallicity. K93's sample was limited by a small range in [Fe/H] (≲ 0.25 dex) and showed no trend in [O/Fe] versus [Fe/H].

More recently, Friel et al. (2010) produced O abundances for giants in 11 OCs, spanning a range of ∼0.4 dex in [Fe/H]. This sample exhibits a linear trend in [O/Fe] versus [Fe/H], but seemingly offset ∼0.15 dex below the trend typically seen in field star samples, as seen in their Figure 6, which also includes the field star data of Bensby et al. (2005). Friel et al. acknowledge the possibility of a systematic offset between their and other studies, but we reiterate the concerns over the use of the 6300 Å [O i] line discussed above, and note their limited sample statistics of ⩽4 stars per cluster.

Uniform studies of OC dwarf triplet abundances, which would allow testing of the [O/H] versus age relationship, have been particularly sparse. K93 remains the largest such survey to date (eight clusters), and while his stars are primarily in the required temperature range (6200 K ≳ T_eff ≳ 5500 K) his data are low resolution (1.2–2.3 Å). As noted above, Schuler et al. (2004), King & Schuler (2005), and Schuler et al. (2006b) have produced dwarf triplet abundances for four OCs from high-resolution data, but all having ages ≲ 600 Myr, and these works do not focus on the "prime" T_eff range. Additional OC dwarf triplet-based abundances were determined by Shen et al. (2007) in IC 4665, by Ford et al. (2005) in Blanco 1, and by Garcia Lopez et al. (1993) in the Hyades and the Ursa Major Group. Each of these studies included some prime T_eff stars, though Garcia Lopez et al. (1993) focused mostly on hotter stars in vicinity of the Li Gap (Boesgaard & Tripco 1986), with stars near the warm end of the prime T_eff range included. To our knowledge, these represent all OC dwarf triplet abundances reported in the literature. None of these studies provides an appropriate sample for investigating the outstanding questions discussed above, making plain the need for a new study of OC dwarf abundances.

This paper presents the first results of our ongoing study of oxygen abundances in OC dwarfs. This and future papers will establish a large sample of self-consistent oxygen abundances in OCs having a range in age and metallicity, with the objective of examining the [O/H] versus age and [O/Fe] versus [Fe/H] relationships, thereby testing oxygen's potential as a chemical tracer. Here we present our analysis of (1) the Hyades, a fundamental and extensively studied cluster that will provide a reference point for our metallicity and T_eff scales, and (2) NGC 752, a moderately old (∼1.5 Gyr), slightly metal-poor cluster ([Fe/H] ∼ −0.05), which is also reasonably nearby (m − M = 8.4; Anthony-Twarog et al. 2009). Photometric and membership information for NGC 752 has been compiled by Daniel et al. (1994). Despite its relative proximity, it is arguably an understudied cluster; few spectroscopic metallicity determinations exist, and only one for oxygen (see the Discussion). This paper is organized as follows: in Section 2 we describe our data and reductions, and present our radial velocity and membership determinations; Section 3 describes our abundance analysis; in Section 4 we discuss the interpretation of our results, including comparison to previous works; Section 5 summarizes our conclusions.

All data were acquired using the Hydra multi-object spectrograph on the WIYN⁶ 3.5 m telescope. The spectra cover the range 7645–7985 Å, at a dispersion of 0.2 Å per pixel, a resolution of 0.5 Å (R ∼ 16, 000 at 7774 Å), and include 11 unblended Fe i lines in addition to the O i triplet. Spectra of Hyades stars were acquired on three separate runs: 8 stars in two configurations, 1996 November 1; 7 stars in six configurations, 1997 February 13–17; and 11 stars, using a single fiber at field center, 2008 January 3. The processed Hyades spectra have S/N in the range 100–340 pixel⁻¹, with median ∼140. Spectra of 45 single stars in the field of NGC 752 were acquired 1996 September 23; 2.0 hr of total exposure were taken using one fiber configuration. The processed spectra have typical S/N in the range 60–140 pixel⁻¹, for V = 15.5–11.8. Examples of our co-added, continuum-fitted master spectra are shown in Figure 1, which includes both the 1997 and 2008 daytime-sky solar spectra, and one star of near-solar T_eff from each cluster.

Zoom In Zoom Out Reset image size

Data reduction employed standard IRAF⁷ techniques. Flat-fielding and aperture tracing used well exposed dome flats. Object spectra (only) were cleaned using the Laplacian Cosmic Ray removal routine of van Dokkum (2001). Spectra were wavelength calibrated with Th–Ar comparison spectra. Solar sky spectra were used for fiber-to-fiber throughput correction to facilitate night sky subtraction; however, we deemed sky subtraction unnecessary for our Hyads, given their brightness and the short exposure times used. All apertures of a single solar sky exposure from the 2008 January run were co-added to create a master solar spectrum, with S/N > 1400 near 7774 Å, to be used in deriving solar log gf values for the concurrent Hyades spectra. Appropriate solar spectra were not available for the other runs, and in those cases we employed an analogous master spectrum with S/N > 1300 from a 1997 November WIYN 3.5 m run. Equivalent widths were measured using the IRAF routine splot.

In this paper, we use the ID system of Van Bueren (1952) for the Hyades, and the ID system of Daniel et al. (1994) for NGC 752.

Our targets were selected based on proper motions, as summarized in Perryman et al. (1998) for the Hyades, and as measured by Platais (1991) for NGC 752. All Hyads are highly probable proper motion members according to Perryman et al. (1998), whereas NGC 752 targets are either probable or possible proper motion members, all but two of which have Platais (1991) membership probabilities ⩾77%. For the interpretation of our abundances, we wished to consider only member stars that had no convincing evidence of binarity; we thus refined our samples using previous photometry (for the Hyades), our own radial velocities, and Fourier cross-correlation tests for binarity.

Radial velocities were calculated via direct wavelength measurements of up to 36 spectral lines per star. This method allowed us greater discrimination in choosing appropriate spectral features compared to determining radial velocities via Fourier cross-correlation, and produced smaller errors, typically <0.5 km s⁻¹ (radial velocity errors are σ_μ, i.e., the standard error of the mean, unless otherwise stated), in almost all cases. Results for the Hyades are reported in Table 1, and a histogram is given in Figure 2. We see that three Hyads, vB 1, 2, and 110, are outliers from the primary envelope of the distribution, and so are inconsistent with radial velocity single-star membership; vB 1 and 2 were previously noted as possible multiple systems by Van Bueren (1952), and also have somewhat discrepant radial velocities in Perryman et al. (1998). Excluding these three, the mean radial velocity of the Hyads consistent with single-star membership is 39.4 ± 0.6 km s⁻¹. All of our Hyads are designated as members by Perryman et al. (1998), who used radial velocities, proper motions, and Hipparcos parallaxes to determine membership from six-dimensional position and radial velocity information. The radial velocities reported by Perryman et al. (1998) for the single star member Hyads in our sample yield an average of 39.9 ± 0.4 km s⁻¹; our values are thus on an identical scale, although it is important to note that Perryman et al. (1998) used proper motions and parallaxes to define radial velocities with respect to the cluster center. Thus, while our values are on a nearly identical scale on average, the individual values differ slightly from those reported by Perryman et al. (1998); the median difference between our values and those of Perryman et al. (1998) is −0.5 km s⁻¹, with the majority of our values within ±1 km s⁻¹ of theirs. We used the IRAF routine fxcor to calculate Fourier cross-correlation functions for all stars, in order to test for the possible contribution of a second set of spectral lines, and thus reveal otherwise undetected binarity; all Hyads in our sample were found to be single stars by this method.

Zoom In Zoom Out Reset image size

To provide an additional criterion for evaluating the membership of our Hyades stars, we employ the UBV photometry of Johnson & Knuckles (1955) and UBVRI photometry of Medoza (1967). Colors from these studies will also be used to determine our stellar T_eff, as described below in Section 2.3. The difference between the two systems is dominated by the number of digits reported by each study (i.e., by discrepancies in the third decimal place); thus, we adopt the average of the two for UBV and Medoza (1967) only for RI. Empirical relations were derived between (B − V) and each of the nine other unique color combinations from these five filters, using a fiducial main sequence selected from the entire photometry set (i.e., not restricted to stars in the current sample). Each color was thereby transformed to an effective (B − V), and the values averaged (up to 10 per star). The mean transformed colors and the associated standard deviation of the mean (σ_μ) for each star are included in Table 1. The standard deviation on such a mean can also measure the relative contribution of any binary companion. On this basis we exclude from further analysis stars with standard deviations greater than a few times the median value for all stars, namely vB 49 and 149; previous evidence of multiplicity for vB 149 was noted by Perryman et al. (1998). However, we include several stars classified as spectroscopic binaries by Perryman et al. (1998) which nonetheless have standard deviations comparable to the known singles, namely vB 39, 63, 102, 106, and 114. Our final evaluation of single-star cluster membership is presented in Table 1: a "1" indicates consistency with single-star membership, and a "0" indicates that there is at least some evidence to the contrary. To ensure the robustness of our cluster average abundances, stars that are to be included in our cluster averages are required to fulfill all three single-star member criteria (radial velocity, fxcor, and photometry). However, we also retain vB 1, 2, and 110 in the abundance analysis, and report abundance values for these stars, even though they will not be included in the cluster averages.

In the case of NGC 752, an instrumental wavelength shift of uncertain origin occurred during the night, which could not be resolved via our available calibration data, making it impossible for us to define an independent absolute radial velocity scale. We thus shifted the average of our radial velocity values to match the average from Daniel et al. (1994) of the 25 stars in common for which they report radial velocities; this adjustment results in an error in our zero point of ∼0.2 km s⁻¹. We report our full set of shifted NGC 752 radial velocities in Table 2, and present a histogram in Figure 3. The primary peak of the radial velocity distribution has a mean of 5.4 ± 0.1 km s⁻¹; Daniel et al. (1994) reports 5.5 ± 0.6 km s⁻¹ (standard deviation), for a slightly different set of stars. Pilachowski et al. (1988) report a mean of 4.9 ± 0.7 km s⁻¹ (standard deviation), in reasonably good agreement with both our and the Daniel et al. (1994) values. Thirty-seven stars are included within ±3σ (σ = 0.81 km s⁻¹) of our cluster average, and thus are consistent with single-star membership. We again used fxcor to test for binarity, and found no evidence for binarity in any of these 37 stars. Table 2 also includes the proper motion membership probabilities of Platais (1991) of these 37 stars: 25 have proper motion membership probabilities >95%, and the remaining 12 have membership probabilities from 77% to 89%. All of these 37 stars have also been designated as probable members or possible members by Daniel et al. (1994), based on proper motions and radial velocity criteria. Column 8 of Table 2 indicates our final evaluations of single-star cluster membership. One single-star member, 300, had a low S/N spectrum, which was unsuitable for the abundance analysis.

Zoom In Zoom Out Reset image size

To derive abundances from our equivalent widths, we employ model atmospheres, which require knowledge of the stellar parameters T_eff, log g, and v_turb. T_eff is derived using a color–T_eff relation and knowledge of the cluster reddenings, while log g comes from matching a star's T_eff to a suitable isochrone (see below for v_turb). The reddening of the Hyades is very small: Taylor (1980) was unable to detect any, and reported a formal limit of 0.003 ± 0.002(σ) mag. Any unsuspected reddening at the few mmag level would be negligible for our purposes. For NGC 752, we adopt E(B − V) = 0.035 ± 0.005 mag from Daniel et al. (1994). From turnoff isochrone fitting, the Hyades age is generally found to be between 600 and 700 Myr (see, for example, Perryman et al. 1998; Pinsonneault et al. 2003); we adopt an age of 650 Myr. For NGC 752, we adopt an age of 1.45 Gyr as derived by Anthony-Twarog et al. (2009). The dwarfs of interest in this study are of substantially lower mass than the turnoff, and thus are presently evolving very slowly, and thus uncertainties in the cluster ages effect negligible changes in the isochrone-derived values for log g (and v_turb). Finally, the metallicities assumed in the choices of isochrones are [Fe/H] = +0.13 and −0.06 for the Hyades and NGC 752, respectively (see Sections 3 and 4.1).

We use Kurucz (1979) model atmospheres (1993 distribution). T_eff is calculated from (B − V), using the color–temperature relation of Deliyannis et al. (2002), who employ the Carney (1983) T_eff–(B − V) relation, with the Cayrel et al. (1985) Hyades zero point and the Saxner & Hammarbach (1985) metallicity dependence (see also Thorburn et al. 1993; Deliyannis et al. 1994). For the Hyades, we use the mean (B − V) values derived as described above (and found in Table 1). We note for the Hyades that using the average (B − V) yields an identical T_eff compared to the average derived from the individual color translations, except in the case of large scatter (i.e., strong evidence for binarity). As a measure of the error in T_eff, Table 1 also shows change in T_eff resulting from changing the mean (B − V) by one σ_μ. For NGC 752, we use the photometry of Daniel et al. (1994), who combine data from numerous studies and transform the data onto a common system; their photometry and errors (standard deviation; if reported by them) are given in Table 2, along with the implied errors in T_eff. An initial [Fe/H] value is input, then iterated through the abundance analysis (see below) to consistency with the resulting cluster average. Log g is taken from the Yale-Yonsei (Y²) isochrones (Demarque et al. 2004) of appropriate age and metallicity, and is subsequently used with T_eff to get v_turb via Edvardsson et al. (1993). Atmospheric parameters are given in Tables 1 and 2.

Using the triplet in Hyades dwarfs, King (1993) reports a rather high [O/H] = 0.28 ± 0.10 for the Hyades, though consistent with ours to within his errors. Schuler et al. (2006b) report [O/H] = 0.25 ± 0.02, also using the triplet in Hyades dwarfs; however, using the 6300 Å [O i] line, Schuler et al. (2006a) find [O/H] = 0.14 ± 0.02 for Hyades dwarfs and [O/H] = 0.08 ± 0.02 for Hyades giants. None of these values are in good agreement with each other or with our own value. Boesgaard (2005) also uses the triplet in dwarfs, reporting [O/H] = 0.17 ± 0.01, which differs from ours by just under 2σ (quadratically combined error), and is thus in statistically marginal agreement with our value.

Iron abundances have been studied far more extensively in the Hyades than in any other cluster. Table 11 shows a compilation of modern high-resolution spectroscopic studies of [Fe/H] in Hyades dwarfs. Although these span nearly 30 years, use a variety of telescope/spectrograph/detector combinations, and employ varying analysis techniques (for example, different T_eff scales, which can affect the derived [Fe/H]), the full range of published [Fe/H] just barely exceeds 0.05 dex, with the exception of one study (Schuler et al 2006a; but note that their [Fe/H] for Hyades giants is 0.16 ± 0.02). Such concordance is remarkable, and is nearly certainly unique in the open cluster literature. Our own value for [Fe/H] agrees well with those others listed in Table 11.

Few previous abundance studies of NGC 752 exist. King (1993) reports the only previous O abundance for NGC 752, [O/H] = −0.13 ± 0.12, in good agreement with our own. Anthony-Twarog & Twarog (2006) find [Fe/H] = −0.06 ± 0.09 using uvbyCa photometry, and Anthony-Twarog et al. (2009) report [Fe/H] = −0.08 ± 0.11 from a combination of uvbyHβ photometry and moderate-dispersion spectroscopy of the giants (where the spectroscopy alone yields [Fe/H] = −0.05). Anthony-Twarog et al. (2009) suggest these supersede the Daniel et al. (1994) value of [Fe/H] = −0.15. Hobbs & Thorburn (1992) and King (1993) both find [Fe/H] = −0.09 (± 0.05, ± 0.07, respectively) spectroscopically, from high-resolution data in eight turnoff stars, and moderate-resolution data in nine dwarfs, respectively. Sestito et al. (2004) report [Fe/H] = 0.01 ± 0.04 from high-dispersion spectra of seven dwarfs, and dismiss the sub-solar values. Given the excellent agreement of most of the previous values and the sub-solar value we have presented here, the Sestito et al. (2004) value would instead seem to be the discrepant one; however, all the values summarized here are acceptably consistent within the errors, and the overall scatter is only 0.1 dex.

Figure 6 shows [O/H] versus age for our present results along with the King (1993) sample and additional OC dwarf triplet abundances from the literature (Schuler et al. 2004; Ford et al. 2005; King & Schuler 2005; Schuler et al. 2006b; Shen et al. 2007; see Section 1 for details). In Figure 6, we have recalculated the average [O/H] for the "Triplet Literature" values (green triangles) to include only prime T_eff range stars, and the error bars represent the corresponding σ_μ; the error bars on our present values represent σ_μ plus the systematic error due to a reddening error of +0.005 (reported in Table 8); and the error bars on the King (1993) values are as reported by him. Our [O/H] scale would appear to be similar to that of King (1993), as evidenced by the fact that the average difference in our cluster [O/H] for the Hyades and for NGC 752 is only 0.02 ± 0.07 dex. Perhaps this is not altogether surprising since King (1993) used a near-identical reddening for NGC 752 as we have, namely E(B − V) = 0.04 (he also assumes zero reddening for the Hyades). Our results thus also may support the apparent [O/H] versus age relationship found by King (1993), although only weakly so, given that we report on just two clusters. Some of the literature values for the younger clusters also support the high [O/H] found by King (1993), but there may be some scatter as well, which is difficult to evaluate given the inhomogeneous nature of the available literature data. In future papers we will provide self-consistent data sets for additional clusters which will shed light on these issues.

Zoom In Zoom Out Reset image size

Our NGC 752 sample exhibits a prominent triplet [O/H] overabundance trend for T_eff < 5500 K (see Figure 5), whose origin we may attempt to investigate. This trend may be analogous to those found by Schuler et al. (2004), King & Schuler (2005), and Schuler et al. (2006b) for the Pleiades, UMa moving group, and the Hyades, respectively. Furthermore, our sample has the largest T_eff range for OC triplet abundances to date (although the Schuler et al. 2006b Hyades sample had better sampling density), and so provides a unique means with which to investigate the overabundance trend. Of note is the extraordinarily high [O/H] of the coolest star in our sample, star 748 at 4209 K. This abundance, [O/H] = 2.85 ± 0.02, is suspicious because (1) the implied oxygen abundance is 700 times solar and thus unphysical to an extent which calls into question the suitability of the analysis, and (2) we detect all three triplet lines in star 748, whereas we detect only the 7771 Å line, i.e., the strongest line, in all other stars cooler than 5550 K. Note that this star is a kinematic member and has an [Fe/H] abundance which, while low, is not discrepant with respect to the cluster. Given the paucity of data between this star and the next hottest one, which is fully 400 K hotter, and thus given the absence of information about the shape of the trend in this temperature range, it is difficult to know whether this star is actually peculiar or whether the trend simply goes right through it. Having noted this uncertainty, we will confine attention to the NGC 752 overabundance trend as defined by stars between 5500 and 4500 K, to be conservative.

Schuler et al. (2006b) compared the overabundance trend exhibited by their Hyades dwarfs to that seen in the Pleiades (Schuler et al. 2004) and in the UMa moving group (King & Schuler 2005), and found that the UMa trend appeared very similar to that of the Hyades, while the Pleiades trend was clearly much steeper (see their Figure 6). Schuler et al. (2006b) concluded that this indicated an age rather than metallicity effect, given that the UMa moving group may be of similar age to the Hyades (600 Myr; King & Schuler 2005) but more than 0.2 dex lower in [Fe/H] (−0.09; Boesgaard & Friel 1990), while the Pleiades is a factor of ∼5 younger than the Hyades and of solar metallicity (King 2000). NGC 752, which is a factor of ∼2 older than the Hyades, is an ideal cluster with which to test the suggested age correlation. We did not measure [O/H] for Hyades dwarfs with T_eff  ≲  5400 K, so we compare our NGC 752 results to the Schuler et al. (2006b) Hyades results, as plotted in Figure 7; we note that in Figure 7 we have shifted both data sets so that the average [O/H] of the "prime" members is zero. The NGC 752 dwarfs that have T_eff between 5500 and 4500 K lie above the Hyades dwarfs of similar T_eff, and the NGC 752 trend is either simply offset relative to that of the Hyades, or alternatively may have an even steeper slope. Regardless, the NGC 752 overabundance trend lies above that of the Hyades, and thus, does not appear to support the age correlation proposed by Schuler et al. (2006b).

Zoom In Zoom Out Reset image size

One might wonder whether the NGC 752 overabundances correlate with other parameters, such as rotation or rotation-related (and perhaps rotation-driven) effects. Rotation is an interesting parameter because, as is well-known, G and K dwarfs spin down during main-sequence evolution (Skumanich 1972; Kawaler 1988; Cardini & Casatella 2007; Reiner & Mohanty 2012), and in particular, Pleiads rotate faster than Hyads do (e.g., Meibom et al. 2009), so perhaps the Schuler et al. (2006b) decrease of the oxygen overabundances from the Pleiades to the Hyades has something to do with rotation. While considerable effort has been put forth to study rotation in the Hyades (e.g., Delorme et al. 2011, and references therein), unfortunately there is almost no such information about NGC 752. From our spectra, we can place conservative upper limits in vsin i of 15 km s⁻¹, based upon our instrumental resolution limit, for all stars in our NGC 752 sample that have T_eff < 6200 K. In comparison, this upper limit is well above the equatorial rotation rates of about 3.5 km s⁻¹ for Hyads near 5000 K (converted from the Delorme et al. 2011 periods). Thus, while the expectation is that NGC 752 dwarfs should rotate more slowly than Hyads of comparable temperature, which is entirely consistent with our limit of 15 km s⁻¹, the data also allow for them to rotate up to about 3–5 times faster than the Hyades. Thus, we cannot directly evaluate whether the higher cool dwarf O overabundances in NGC 752 are correlated with rotation. If the NGC 752 cool dwarfs do rotate faster than the Hyads, there would also be implications for stellar spin-down.

We can nevertheless proceed to test whether the triplet overabundance trend is related to certain rotation-related phenomena, such as X-ray activity (stellar coronal activity is driven by rotation; e.g., Giardino et al. 2008, and references therein). As noted in the Introduction, the work of Shen et al. (2007) appears to support a correlation between [O/H] abundances and X-ray activity for the case of IC 4665, albeit based on only three stars. To test for such a correlation with our current [O/H] results, we first turn to the most comprehensive and most recently published study of X-ray activity in the Hyades, that of Stern et al. (1995), who used ROSAT data. This study provides X-ray data for 20 of the 22 Hyads for which we report [O/H]. Figure 8 shows [O/H] versus X-ray luminosity for these Hyads, where the X-ray luminosity has been normalized by the bolometric luminosity, as taken from the Y² isochrones, for each star; also included in this plot are the two stars from Schuler et al. (2006b) for which Giardino et al. (2008) report values. Unfortunately, this combined sample contains only one star with T_eff < 5400 K, which is insufficient to investigate the low-T_eff overabundance trend. The available data can be interpreted as exhibiting somewhat increased scatter in [O/H] for X-ray luminosity ratios ≳ 3 × 10⁻⁵, and indeed, the standard deviation of [O/H] for stars above this limit is approximately twice that of stars below it. However, the difference in the means of the two groups is 0.045 ± 0.033 (the error is the combined σ_μ), which is of questionable statistical significance; furthermore, a linear least-squares fit to the entire sample yields a slope which is consistent with zero. Without even taking into account the substantial errors on the individual data points, we can conclude that there is no significant trend of [O/H] versus X-ray luminosity.

Zoom In Zoom Out Reset image size

In the case of NGC 752, there have been two studies of X-ray emission, Belloni & Verbunt (1996), who used ROSAT data, and Giardino et al. (2008), who used both Chandra and XMM-Newton data. Giardino et al. (2008) provide data for four of our NGC 752 stars: 701, 859, and 953 from Chandra data, and 783 from XMM-Newton data; unfortunately, these stars cover a rather limited T_eff range of 5650–6000 K. Figure 9 shows [O/H] versus X-ray counts (in the 0.5–2.0 keV band) normalized by bolometric luminosity (as done for the Hyades) for these four stars. Three of these stars show no correlation between [O/H] and X-ray luminosity. The fourth is star 783, which has both a higher [O/H] than other prime range members and an unusually high X-ray luminosity. Giardino et al. (2008) note that this star is not identified as binary or rapid rotator, either of which could increase X-ray luminosity (and recall that our own conservative upper limit on vsin i is 15 km s⁻¹ for the NGC 752 stars in our sample). They also note that the ROSAT X-ray luminosity reported for 783 in Belloni & Verbunt (1996) is three times higher than reported by Giardino et al. (2008), indicating it may have been flaring during the ROSAT observation, which implies this star is an X-ray variable. Due to the paucity of stars in our sample for which X-ray data are available, we cannot evaluate how this star might reflect any underlying scatter in [O/H] versus X-ray luminosity. In short, the available X-ray data for our two clusters do not provide any evidence for a correlation between [O/H] and X-ray activity.

Zoom In Zoom Out Reset image size

Another possibility is that triplet [O/H] abundances might be affected by chromospheric activity. As noted in the Introduction, Schuler et al. (2004) found no strong correlation between triplet [O/H] and Hα emission or Ca ii emission; however, their abundance and activity data were not contemporaneous. Since activity levels can change over time, possible correlations between abundances and activity levels might have been missed. To eliminate such potentially confusing effects of changing activity, R. M. Maderak et al. (2014, in preparation) obtained and studied contemporaneous spectra of the O i triplet, the Ca ii infrared triplet (8498 Å, 8542 Å, 8662 Å), and Hα spectral regions in 45 Pleiades dwarfs that are known to have a variety of chromospheric activity levels and rotation rates. R. M. Maderak et al. (2014, in preparation) found a strong anti-correlation between total Ca ii triplet emission and T_eff, as expected if Ca ii triplet emission is correlated with Ca ii HK activity (e.g., Pace & Pasquini 2004, and references therein). In addition, the Pleiads with vsin i ≳ 30 km s⁻¹ have Ca ii emission that is elevated by an additional ∼50% above the underlying trend with T_eff. This suggests that Ca ii emission has both a chromospheric and rotational component, as discussed by Shen et al. (2007). R. M. Maderak et al. (2014, in preparation) find a clear correlation between elevated triplet [O/H] and increased Ca ii triplet emission. Interestingly, the cool Pleiads (T_eff < 5400 K) that have both the largest Ca ii emission and largest [O/H] overabundances appear not to be rotating rapidly; rather, they have vsin i < 30 km s⁻¹. This suggests that chromospheric activity, rather than rotation, is responsible for the triplet [O/H] overabundance trend in cool dwarfs, at least in the Pleiades. But if this were also to be true for the Hyades and NGC 752, a new mystery would be manifested, namely why would the cool dwarfs in the older NGC 752 have higher activity levels than in the cool Hyads?

Of importance is the R. M. Maderak et al. (2014, in preparation) finding that [O/H] is flat versus total Ca ii emission for the prime-T_eff Pleiads with vsin i ≲ 30 km s⁻¹. This suggests that the triplet [O/H] abundances of slowly rotating prime-temperature dwarfs are not affected by chromospheric activity, even though activity likely causes the cool dwarf overabundance trend, and therefore, slowly rotating prime-temperature dwarfs, like those used in the present work to define our cluster averages, yield reliable triplet abundances.

Finally, we note the difference in metallicity between the Hyades and NGC 752 of 0.194 ± 0.017 and speculate as to whether this might be related to the O overabundance differences in the cool dwarfs of the two clusters. A promising possibility is the effect of UV photons causing overphotoexcitation which enhances the strengths of the O i lines and higher excitation potential Fe i lines. Such a mechanism might also lead to overphotoionization enhancing strengths of Fe ii lines; anomalously high Fe abundances derived from Fe ii lines are indeed observed in cool Hyades dwarfs (Schuler et al. 2009). The key feature of this model would be that the lower-metallicity photospheres present smaller opacity to the UV photons (whatever their source). Such a mechanism has been inferred to be operating in cool pre-main-sequence stars on the basis of comparisons of Li i based abundances derived from the λ6708 Li resonance lines and the subordinate λ6104 feature (Bubar et al. 2011).