Chemical Compositions of Red Giant Stars from Habitable Zone Planet Finder Spectroscopy

HPF is a high-resolution (λ/Δλ ≡ R ∼ 55,000) near-infrared (zyJ photometric bands, 8100–12800 Å) spectrograph on the 10 m Hobby–Eberly Telescope (HET) at McDonald Observatory (Mahadevan et al. 2012, 2014). ¹¹ The main goal of HPF is to search for exoplanets of cool M-dwarf stars by detecting the orbital reflex motions of the parent stars. To achieve high velocity precision (∼1.5 m s⁻¹), the entire HPF instrument is encased in a vacuum chamber cooled to 180 K, yielding temperature stability of ∼1 mK. The instrument has input fiber scrambling (Roy et al. 2014) and precise wavelength characterization with a laser frequency comb (Halverson et al. 2014; Metcalf et al. 2019). At present HPF has generated studies of exoplanet systems and precision radial velocity (RV) techniques, but it has not been routinely applied to investigation of the atmospheres of stars warmer than spectral type M.

High-resolution, high signal-to-noise (S/N > 100) HPF spectra of 15 target stars were gathered over several months using the fully queue-scheduled HET (Shetrone et al. 2007). Table 1 gives basic data for these stars, and for warm, rapidly rotating, mostly featureless stars used as telluric divisors. In the zyJ spectral region the telluric molecular blockage ranges from near zero to almost total, varying significantly with wavelength. The dominant telluric contaminators are H₂O bands, and their contamination can change substantially on timescales of minutes, usually requiring a hot-star observation to accompany each target star. Some mitigation is provided by the fixed altitude of the HET, which on stable nights can lead to successful use of divisor star spectra acquired some time before or after the target spectra.

HPF observations of target and telluric warm stars were reduced using the HPF pipeline code Goldilocks. ¹² This package processed the raw HPF data by removing bias noise, correcting for nonlinearity, masking cosmic rays, and calculating the slope/flux and variance image using the algorithms from the pyhxrg module in the tool HxRGproc (Ninan et al. 2018). The pipeline also processed bright lamp exposures to combine into a master frame to model the trace of the calibration, sky, and science fibers as well as their respective fiber profiles. The profiles were used as weights in an optimal extraction algorithm (Horne 1986) to produce 1D spectra and their associated propagated errors for all 28 orders. The master frame was also used to correct flat-fielding features.

Goldilocks uses a single wavelength solution for all observations from the well-calibrated near-IR Laser Frequency Comb. HPF's wavelength solution is extremely stable but does drift on the order of ∼10–20 m s⁻¹ throughout the night with long-term drifts over months on the same scale. This amounts to ∼1/100 of a pixel. Although Goldilocks provides the drift correction in the observation file headers, it is only important for precision RV studies; we ignored this very small effect in our analysis.

For each stellar source, the science fiber extractions from Goldilocks include light from the target as well as OH emission from the night sky. The sky fiber, located 501 away, contains only night sky emission. In most cases, the relative flux ratio between the sky emission and stellar continuum (at the wavelengths of interest) is less than 5%, and the sky emission has almost no effect on the target spectrum. In those situations we used the spectra of targets as extracted from the science fibers without attempting sky subtraction. For a few small spectral regions in which sky subtraction was critical to accurate representation of the stellar spectra, we performed the sky subtraction manually using the pipeline-extracted star and sky spectra.

After these reduction procedures, we performed continuum normalization of the target spectra via the continuum task of IRAF, ¹³ and we used the telluric task to divide out the telluric contamination from the science target spectra. Later, individual orders were merged into a single continuous spectrum with the scombine task.

The Goldilocks pipeline produces rest-wavelength solutions that are accurate to levels far beyond what are needed for our study. To measure the RVs of our stellar targets we made use of the rv package in IRAF. First we created synthetic spectra (as outlined in Section 3.1), adopting similar T_eff, $\mathrm{log}\,{\text{}}g$, and [Fe/H] values to our program stars. Then we used rvcorrect for the heliocentric corrections, and measured the RV values with the cross-correlation task fxcor (Fitzpatrick 1993). We only worked on the wavelength regions that are not affected (or only affected minimally) by the atmospheric telluric lines: 8420–8520 Å, 8540–8640 Å, 8660–8760 Å, 10275–10400 Å, and 10450–10575 Å. We list the RVs from HPF and their standard deviations (σ) for each star in Table 2. The σ values were calculated by taking the average of RVs measured from individual regions listed above. Previous RV measurements of the same targets (Afşar18a and Gaia DR2) are also listed in Table 2 for comparison purposes.

The internal RV accuracies (line-to-line scatters) for individual stars are comparable: 〈σ_HPF〉 = 0.18 km s⁻¹ and 〈σ_opt〉 = 0.19 km s⁻¹, and the velocity scales agree: 〈RV_opt – RV_HPF〉 = +0.45 ± 0.51 km s⁻¹ (σ = 1.96 km s⁻¹). Our program stars also have RVs in the Gaia Data Release 2 Catalog (Arenou et al. 2018), and we list them also in Table 2. The agreement between HPF and Gaia is very good: 〈RV_opt – RV_Gaia〉 = −0.27 ± 0.08 km s⁻¹ (σ = 0.31 km s⁻¹).

Our program stars have all had previous extensive analyses from optical-region spectra. As described in detail by Afşar18a, the RHB high-resolution spectra were obtained with the Robert G. Tull Cross-Dispersed Echelle spectrograph (Tull et al. 1995) on the McDonald Observatory 2.7 m Harlan J. Smith telescope. The instrumental setup yielded a spectral resolving power of R = 60,000 with effective wavelength coverage for the RHB observations of 4100–9000 Å. Redward of 5000 Å the S/N values always exceeded 100. The Tull Spectrograph was also used to obtain a very high S/N spectrum of the very metal-poor HD 122563. The optical spectrum for open cluster giant NGC 6940 MMU 105 was acquired with the High Resolution Spectrograph (Tull 1998) of the Hobby–Eberly Telescope at McDonald Observatory. It was configured to cover the spectral range 5100–6900 Å, with the same resolving power as our 2.7 m Tull Spectrograph data.

A few of our HPF stars have also been observed at high resolution in the infrared H and K bands with IGRINS (Yuk et al. 2010; Park et al. 2014) when this portable instrument was located at the McDonald Observatory 2.7 m telescope. The IGRINS spectra for these bright RHB stars covered a wavelength range 14800–24800 Å, had resolving power R ≃ 45,000, and achieved S/N ≥ 150. Our analysis of metallicity and abundance ratios was reported by Afşar18b.

In this exploration of HPF spectra of red giants we first sought to identify all useful atomic and molecular transitions for abundance analysis. HPF spectra cover the zyJ wavelength range 8170–12780 Å (0.82–1.28 μm). In RHB stars this spectral region contains many atomic lines, nearly all of which arise from light and Fe-group elements. In Figure 1 we show one of the HPF spectral orders that has several strong neutral-species lines of Na, Si, Ca, Fe, and Cr, one of the H i Paschen series lines, Fe ii 10862.65 Å, and Sr ii 10914.87 Å. The Sr ii line and its multiplet partners at 10036.66 and 10327.31 Å are zyJ region rarities: ionized-species neutron-capture element (Z > 30) lines. Neutron-capture element transitions are rarely strong enough in our stars for detection in this wavelength domain. A few Fe-group ionized-species lines can be seen in red giant stars, but only one Fe ii multiplet is strong enough for abundance analysis in our program stars.

Zoom In Zoom Out Reset image size

There are also many molecular features in the zyJ region: OH ground state ¹Σ⁺, CO ground state X²Π rovibrational, and CN A²Π–X²Σ⁺ red-system transitions. But in general only CN lines have large absorption strengths. They can significantly contaminate many otherwise useful atomic transitions. In practice this problem forces examination of each potential atomic line for the presence of blends.

With potential line contamination issues, we chose candidate atomic features for abundance analysis by comparing HPF data in each spectral order with synthetic spectra for RHB star HIP 114809. Detailed matching of synthetic and observed spectra was not sought at this point. For all computations of synthetic spectra and predicted single-line equivalent widths (EWs) in this paper we used the current version of the local thermodynamic equilibrium (LTE) plane-parallel line analysis code MOOG (Sneden 1973). ¹⁴ The input atomic and molecular line lists were generated by the auxiliary code linemake. ¹⁵ We adopted the model atmosphere for HIP 114809 derived for the RHB survey by Afşar18a, whose parameters are given in Table 3. By this process we identified useful transitions of 16 elements; 14 of these have only neutral-species lines. A single Ti ii line was detected, along with the very strong Sr ii lines mentioned above.

Table 3. Program Star Model Atmospheres

Star Name	Teff	$\mathrm{log}\,{\text{}}g$	Vmic	[Fe i]	σ	[Fe ii]	σ	[Fe i]	σ	#Lines	[Fe ii]	σ	#Lines
	(K)		(km s−1)
	opt a	opt	opt	opt	opt	opt	opt	HPF	HPF	HPF	HPF	HPF	HPF
Field RHB Stars
HIP 476	5109	2.68	1.36	−0.10	0.07	−0.09	0.04	−0.02	0.14	21	−0.03	0.04	4
HIP 19740	5135	2.51	1.35	−0.10	0.07	−0.15	0.03	−0.09	0.11	22	−0.12	0.03	4
HIP 27091	5001	2.37	1.56	−0.21	0.06	−0.24	0.07	−0.14	0.11	19	−0.16	0.09	4
HIP 28417	4679	2.15	1.42	−0.38	0.06	−0.43	0.05	−0.33	0.12	20	−0.27	0.12	3
HIP 29962	5118	2.92	1.19	−0.28	0.06	−0.36	0.04	−0.23	0.11	20	−0.21	0.08	4
HIP 32431	5125	2.26	1.57	−0.56	0.05	−0.63	0.05	−0.51	0.11	25	−0.53	0.03	4
HIP 33578	5118	2.92	1.19	−0.24	0.06	−0.28	0.05	−0.30	0.15	22	−0.21	0.01	3
HIP 57748	5307	2.34	1.82	−0.17	0.06	−0.19	0.05	−0.17	0.12	22	−0.13	0.05	4
HIP 99789	5054	2.41	1.12	−0.22	0.06	−0.28	0.09	−0.06	0.14	20	−0.29	0.02	4
HIP 106775	5111	2.41	1.61	−0.21	0.05	−0.26	0.06	−0.15	0.14	20	−0.15	0.07	4
HIP 113610	5106	2.39	1.55	−0.07	0.07	−0.04	0.05	−0.10	0.13	19	⋯	⋯	⋯
HIP 114809	5139	2.59	1.31	−0.33	0.06	−0.38	0.05	−0.32	0.10	23	−0.31	0.04	3
HIP 116053	5100	3.09	0.98	−0.07	0.06	−0.08	0.05	0.03	0.13	24	−0.01	0.08	3
Other Stars
NGC 6940 MMU 105	4765	2.34	1.35	−0.10	0.08	−0.15	0.07	−0.04	0.11	16	0.02	0.03	4
HD 122563	4500	0.80	2.20	−2.98	0.07	−2.98	0.07	−2.89	0.03	7	⋯	⋯	⋯

Note.

^a The optical values are taken from high-resolution optical spectroscopy obtained with the McDonald Observatory 2.7 m TS23 echelle spectrograph (Afşar16, BT16, Afşar18a).

Download table as:  ASCIITypeset image

In Table 4 we list the atomic lines chosen for this study. A fundamental limitation of abundance analyses in the zyJ spectral region is the lack of substantial sets of reliable transition probabilities based on laboratory spectroscopy. For this initial HPF abundance study of RHB stars we decided to avoid deriving empirical log(gf) values from the observed solar spectrum, because their values are difficult to disentangle from choices of solar model atmosphere, line damping parameters, and solar microturbulent velocity. The Wisconsin lab atomic physics group has published transition probabilities in the HPF domain for Ti i (Lawler et al. 2013) and Co i (Lawler et al. 2015). For Fe i a collaboration between investigators at Imperial College London Blackett Laboratory and the Wisconsin group (O'Brian et al. 1991; Den Hartog et al. 2014; Ruffoni et al. 2014; Belmonte et al. 2017) has yielded an extensive set of reliable gf values ranging from the UV to the near-IR. We used only lines from those studies in our Ti, Fe, and Co analyses. This ensured gf source consistency for these elements in HPF and optical spectra, at the expense of ignoring potentially useful transitions in the HPF region.

Of the remaining 13 species, there are eight with detected lines in our HPF data that have transition probabilities given in the NIST atomic line database (Kramida et al. 2019). ¹⁶ The NIST atomic line catalog is a critical compilation, and it provides assessments of the accuracy of transition probabilities. For our lines these quality estimates range from "A+" (uncertainties less than 2%) to "D" (uncertainties less than 50%). We list the NIST line qualities in Table 4. We have included all promising lines without discrimination among NIST quality values, but caution should be observed for those lines with ratings of "C" and lower.

Finally, detected HPF transitions of Ca i, Cr i, Mn i, and Ni i lack recent lab transition work and they do not appear in the NIST database. For these lines we adopted the gf values given in the semiempirical line compendium of Kurucz (2011, 2018). ¹⁷ Decades of work by Kurucz have yielded a database of millions of transitions. Pragmatic mixing of results from laboratory and theoretical studies along with fresh computations has produced good matches between stellar observed and predicted spectra, but individual transition probabilities have widely varying accuracies; this should be kept in mind here.

S i deserves a more expanded description. There are several lines in the zyJ spectral domain, and they generally have gf values available in the NIST database. Table 1 of Spite et al. (2011) lists the most promising transitions in the optical and near-IR regions. In our HPF spectra of RHB stars we have all of these lines except the optical triplet at 6757 Å (see Costa Silva et al. 2020 for a large survey based on those lines). Detections in our spectra include the ground-state [S i] line at 10821.11 Å. Caffau & Ludwig (2007) have explored the use of this transition in solar-type stars. It is an attractive alternative to the high-excitation (χ ≳ 6.5 eV) S i lines that are normally studied, because it has little T_eff sensitivity and is essentially formed in LTE. However, there are some practical problems with this line for our abundance analyses: (a) the line is very weak; (b) it has an unidentified blend in its red wing (see Figure 1 in Caffau & Ludwig); and (c) it is located only 2 Å from the edge of its HPF spectral order. For the present HPF exploration we drop the 10821 Å line, but it should considered more carefully in future papers.

The S i triplet at 9212.9, 9228.1, and 9237.5 Å is strong and has been used in several S abundance studies (e.g., Koch & Caffau 2011; Spite et al. 2011). In our HPF spectra these lines are useful, but with two cautions. First, they are somewhat saturated in our RHB stars, and the derived S abundances have some dependence on assumed microturbulent velocity V_mic. Second, the spectral region surrounding this S i triplet has a large amount of telluric contamination, which is a potential problem with HPF located at McDonald Observatory. The papers cited here were based on data obtained at Chilean observatories, with much less water vapor contaminating the stellar spectra.

The S i triplet near 10457 Å is about half as strong as the 9200 Å one, and for metal-rich giants such as our RHBs these lines are easy to analyze. Finally, Ryde et al. (2019) has identified a line at 10635 Å as an S i transition, particularly for evolved stars with T_eff ≳ 5500 K (see their Figure 1). Preliminary syntheses confirmed its suitability for our study.

In the end, the preliminary computations of synthetic spectra outlined here yielded about 100 atomic lines worthy of further abundance analysis. Most of these were judged relatively uncontaminated by known atomic and molecular blends, and thus could be safely analyzed from their EW measurements. About 20 of the lines were either somewhat blended, or very weak, or have hyperfine structure. For these we performed synthetic spectrum calculations.

For the majority of transitions we derived abundances from EW computations. We measured EWs with the specialized software package SPECTRE (Fitzpatrick & Sneden 1987). ¹⁸ Interactive fits were made for all lines, with most of them modeled with Gaussian approximations. For the strongest lines, we used empirically determined Voigt profiles. We also set local continua interactively in this procedure, since the IRAF continuum task produces good general fits to the blaze function of each HPF order, but they are inadequate for the small spectral intervals surrounding each measured line.

The short-wavelength limit for HPF is ∼8080 Å. This overlaps with the long-wavelength limit of the optical spectrograph TS23, ∼8900 Å. ¹⁹ In the parts of this wavelength domain that are relatively free from telluric absorption (∼8300–8900 Å), we measured EWs for RHB stars HIP 33578, HIP 114809, and metal-poor HD 122563 using spectra from both instruments. In Figure 2 we show the EW comparisons. The agreement between HPF and TS23 values is at the level of our ability to reliably make EW measurements. Decisions on continuum placement and line profile fits cause EW measurement scatter at the level indicated by the line-to-line scatter. For the strongest lines, assumptions of appropriate line profile shape (Gaussian or Voigt) also can affect the EWs, but the fractional EW differences in weak and strong lines are similar. The two spectrographs yield essentially identical spectra of red giant stars.

Zoom In Zoom Out Reset image size

The model atmospheric parameters of our science targets have been adopted from our previous efforts (Afşar16, BT16, Afşar18a), in which we describe the derivation of model atmospheres in detail. In summary here, we used a semi-automated version of the MOOG code introduced in Section 3.1. This driver software begins with information from photometry and spectral line-depth ratios to generate initial model parameter estimates, and then operates iteratively using both neutral and ionized species of Fe and Ti lines to calculate the best model atmospheric parameters. They are listed in Table 3. Typical parameter uncertainties are 150 K for T_eff, 0.25 for $\mathrm{log}\,{\text{}}g$ and V_mic, and 0.1 for [Fe/H]; see Section 6.3 of Afşar18a for a discussion of these values. Optical atomic and molecular line lists that we used to derive atmospheric parameters and individual element abundances were adopted from Böcek Topcu et al. (2015) and BT16.

We determined the abundances of most elements by matching their equivalent widths with computed ones, using the stellar atmosphere models (Table 3) and their line parameters (Table 4) described above. Iterations on the output abundances and re-examinations of the spectra resulted in identification and elimination of aberrant lines in individual stars. In Table 5 we give line-by-line abundances in log units. ²⁰ Program star metallicities [Fe/H] computed from the Fe i and Fe ii lines are listed in Table 3. To compute the [Fe/H] values and subsequent abundance ratios [X/Fe] we have assumed the solar abundance set recommended by Asplund et al. (2009).

Table 5. Line Abundances ^a

Species	λ	(1) b	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)
	(Å)
C i	8335.15	8.18	8.02	8.06	8.21	8.04	7.75	7.67	8.43	7.94	8.13	8.15	8.08	8.17	⋯	⋯
C i	9061.44	8.25	7.99	8.30	8.21	8.11	7.96	7.64	8.57	8.18	8.26	8.22	8.19	8.29	⋯	⋯
C i	9094.83	8.29	8.09	8.34	8.16	8.16	8.02	7.68	8.64	8.24	8.27	8.30	8.24	8.19	⋯	⋯
C i	9405.73	8.17	⋯	⋯	8.31	⋯	7.87	7.69	8.54	⋯	⋯	⋯	8.16	8.12	⋯	⋯
C i	9658.43	8.29	8.08	8.25	⋯	7.99	7.89	7.66	8.65	8.11	8.20	8.22	8.19	8.29	⋯	⋯
C i	10123.87	8.26	8.04	8.26	8.30	8.11	7.86	7.55	⋯	⋯	8.20	8.15	8.06	8.20	⋯	⋯
C i	10683.08	8.21	8.10	8.30	8.11	8.07	7.98	7.67	8.56	8.06	8.21	8.21	8.16	8.13	⋯	⋯
C i	10685.34	8.19	8.09	8.27	8.12	8.05	7.89	7.62	8.54	8.04	8.22	8.22	8.13	8.11	⋯	⋯
C i	10691.24	8.21	8.10	8.27	8.13	8.02	7.99	7.66	8.56	8.07	8.22	8.24	8.15	8.10	⋯	⋯
C i	10707.32	8.18	8.04	8.25	8.16	8.02	7.89	7.62	8.55	8.02	8.19	8.15	8.11	8.16	⋯	⋯

Notes.

^a All abundances in this table are in log units. ^b Star identifications: (1) HIP 476; (2) HIP 19740, (3) HIP 27091; (4) HIP 28417; (5) HIP 29962: (6) HIP 32431; (7) HIP 33578; (8) HIP 57748; (9) HIP 99789; (10) HIP 106775; (11) HIP 113610; (12) HIP 114809; (13) HIP 116043; (14) NGC 6940 MMU 105; (15) HD 122563.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

HPF metallicities of RHB stars are in reasonable accord with those derived from optical data (Afşar18a). In Figure 3 we illustrate the comparisons. The mean difference for Fe i between the two metallicity sets, defined as Δ[Fe i/H] ≡ 〈[Fe I/H]_HPF − [Fe I/H]_opt〉 = 0.04 (σ = 0.06). Abundances from both spectral regions are based on the same Wisconsin/London group sources of transition probabilities. The comparison for Fe ii shows a larger offset, Δ[Fe ii/H] = 0.08 (σ = 0.05). All of the optical and HPF gf values for Fe ii are taken from the NIST database, and nearly all of them originate with a critical compilation by Raassen & Uylings (1998), who considered calculated and experimental transition probabilities in their study. The internal consistency of the Fe ii HPF lines is good, since the typical line-to-line abundance scatter in this species is small, 〈σ〉 ≃ 0.05 (Table 3). However, the relationship between transition data for these lines and those used in optical analyses should be re-examined if larger surveys are undertaken in the future.

Zoom In Zoom Out Reset image size

For a few stars in our sample we have optical, HPF zyJ, and IGRINS HK spectroscopic analyses, extending from ∼5000 Å to almost 24000 Å. In Figure 4 we plot the three abundance sets for HIP 114809. Inspection of this figure reveals agreement among the [Fe/H] values (Table 3) within the line-to-line scatter. The mean Fe abundances have similar scatter in other program stars.

Zoom In Zoom Out Reset image size

The optical and HPF Fe i abundances have been derived with laboratory transition probabilities (O'Brian et al. 1991; Den Hartog et al. 2014; Ruffoni et al. 2014; Belmonte et al. 2017). Unfortunately, there are very few recent laboratory log(gf) values for this species (and for most other species as well) for wavelengths beyond 10000 Å. Therefore the [Fe/H]_IGRINS abundances in Afşar18b were determined with transition probabilities derived from reverse solar analyses. Additionally, the formal σ values and the appearance of Figure 4 show that the line-to-line scatter is larger in the HPF abundances than in the optical ones. At least two factors may account for this effect. First, on average the near-IR Fe i lines are much weaker than optical and UV ones. They come from relatively weak branches of multiplets, and therefore they often have larger uncertainties in transition probabilities than those stronger lines at short wavelengths. ²¹ Second, all of the HPF Fe ii lines occur in the range 8200–9010 Å. In this region the CN contamination can be severe. We have derived Fe ii abundances with full synthetic spectra, and discarded lines that appear to be severely blended. It is probable that future abundance surveys can successfully extract reliable abundances from EW analyses of our chosen Fe ii lines, but caution is warranted.

Mean abundance ratios [X/Fe] are presented in Table 6, and in Figure 5 we show the trends of [X/Fe] with [Fe/H] metallicity. Carbon abundances are not plotted in this figure; they will be discussed in Section 3.7. A few general remarks are in order here. First, we re-emphasize that we have enforced choices of single-source transition probabilities for individual species, but there is gf heterogeneity from species to species; see Section 3.1 for details. Second, we have relied on standard LTE analyses. Our estimates of non-LTE corrections will be presented in Section 3.6. Third, the total metallicity domain is small, −0.5 ≲ [Fe/H] ≲ 0.0, including only thin-disk and some thick-disk stars (see Afşar18a for assessments of stellar populations of RHB stars). Meaningful discussion of Galactic trends awaits analysis of a sample over a much wider metallicity range.

Zoom In Zoom Out Reset image size

With these cautions in mind, it is clear in Figure 5 that Fe-group elements Ti, Cr, Mn, Co, and Ni all have solar abundance ratios, [X/Fe] ≃ 0, from HPF spectra as they do in abundance surveys in other spectral regions. Forming averages of these elements for each star, and defining [Fe-group/Fe] ≡ [〈Ti i, Ti ii, Cr i, Mn i, Co i, Ni i〉/Fe], for the whole Fe group, the mean for the RHB set of 13 stars is 〈[Fe‐group/Fe]〉 = −0.01 (σ = 0.04). The mean Fe-group abundance for each star is shown in the bottom right panel of Figure 5, and shows no trend with metallicity, as expected from previous surveys of Galactic disk abundance (e.g., Reddy et al. 2003, 2006; Bensby et al. 2014, Afşar18a)

The α elements Mg, Si, S, and Ca show a small trend of increasing [X/Fe] values with decreasing [Fe/H]. Defining [α/Fe] ≡ [〈Mg, Si, S, Ca〉/Fe], we plot the mean α values for the RHB stars in the bottom left panel of Figure 5. At [Fe/H] ≃ 0.0, [α/Fe] ≃ 0.0, and for stars with [Fe/H] ≲ −0.2, [α/Fe] ≃ +0.2, again in agreement with the trends found in the survey cited above.

LTE can be an adequate approximation for abundance computations in red giants for ionized majority species, e.g., the lanthanide ions. However, almost all of our HPF transitions arise from neutral species. Most of their parent elements are heavily ionized in line-forming atmospheric levels, and their neutral transitions in the HPF domain arise from high-excitation levels (averaging ∼4.5 eV in our line list), and many of them have deep line cores formed in shallow layers of the atmosphere where photon losses are strong. We expect that some non-LTE corrections are necessary to produce accurate abundances for most of our HPF transitions.

New detailed non-LTE abundance corrections for 13 elements have been published by Amarsi et al. (2020), including Na, Mg, Si, K, and Ca. The abundance corrections discussed in that study were limited to the optical transitions of most interest to the GALAH survey (De Silva et al. 2015). For this paper we have estimated the non-LTE abundance shifts for as many of our HPF transitions as possible using the departure coefficients given by Amarsi et al.

The LTE and non-LTE synthetic spectra were computed with PySME, the Python version of the spectroscopic analysis code SME (Piskunov & Valenti 2017; Wehrhahn & Valenti 2019). The computations here are initial estimates of the magnitudes of non-LTE effects for each HPF line in a typical RGB star. A more detailed investigation will be carried out in the future.

For now, we feel more confident with the general magnitude of the non-LTE corrections, and so in Table 7 we present mean values, standard deviations, and numbers of lines involved in the calculations for six neutral species for a typical RHB star and for HD 122563 (to be discussed in Section 4.4). The entries in this table suggest that for HPF lines in RHB stars the typical non-LTE abundance correction is 〈Δ_corr〉 ≃ −0.15 with a line-to-line scatter σ ≃ 0.10. In Figure 5 we illustrate the 〈Δ_corr〉 values with green arrows in the panels for Na, Mg, K, and Ca. It is apparent that in each of these cases the estimated non-LTE shift significantly reduces the apparent LTE-based overabundances.

Our RHB stars have few CNO strong abundance indicators in the zyJ spectral range. There are some easily identified C i lines, but this species has played only a very minor role in the extensive literature on C abundances in evolved cool stars. Additionally, there are no strong O i or OH lines in this wavelength domain, thwarting any attempt to derive O abundances. Two prominent CN red-system (0–0) bands are seen in the metal-rich stars of Figure 1: the R-branch head at ≃10871 Å and the Q-branch head at ≃10925 Å. But N abundances derived from these bands are dependent on derived or assumed C and O abundances.

Even with these limitations some headway on CNO can be made with HPF spectra. In Table 8 we summarize optical and HPF abundances for the RHB stars and NGC 6940 MMU 105. The data from optical spectra will be reported by Z. Bozkurt et al. (2021, in preparation). That work will give details of the abundance derivations and will include discussion of uncertainties. The final [C_mean/Fe] values are based on multiple C indicators: the CH G-band in the 4280–4330 Å range, and C₂ Swan bands with bandheads near 5160 and 5630 Å. Abundances from C i are determined and tabulated by Bozkurt et al. but they do not participate in the C abundance means. The optical N abundances have been determined using C_mean abundances derived earlier in concert with the O abundances (from the [O i] line). Carbon isotopic ratios come mostly from the 8004 Å CN feature that is the basis for most ¹²C/¹³C estimates in red giants.

Table 8. LiCNO Abundances

Star	log (Li)	[CC I/Fe]	[Cmean/Fe] a	[NCN/Fe] b	[OO I/Fe]	12C/13C
Optical Spectra
HIP 476	0.72	−0.35	−0.37	0.62	−0.05	⋯
HIP 19740	<0	−0.47	−0.54	0.66	−0.11	17
HIP 27091	1.28	−0.25	−0.42	0.73	0.02	20
HIP 28417	<0	−0.02	−0.25	0.33	0.07	12
HIP 29962	<0	−0.23	−0.36	0.61	0.12	20
HIP 32431	<0	−0.12	−0.05	0.58	0.31	20
HIP 33578	<0	−0.72	−0.88	0.67	0.17	3
HIP 57748	<0	−0.04	−0.24	0.40	−0.07	6
HIP 99789	2.61	−0.37	−0.51	0.66	−0.17	13
HIP 106775	<0	−0.22	−0.43	0.76	0.07	13
HIP 113610	<0	−0.46	−0.53	0.53	−0.12	20
HIP 114809	<0	−0.24	−0.35	0.57	0.11	⋯
HIP 116053	0.42	−0.21	−0.30	0.34	−0.01	17
NGC 6940 MMU 105	<0	−0.27	−0.28	0.41	−0.14	15
HPF spectra
HIP 476	⋯	−0.22 ± 0.07	⋯	0.54 ± 0.05	⋯	>20
HIP 19740	⋯	−0.30 ± 0.06	⋯	0.70 ± 0.08	⋯	⋯
HIP 27091	⋯	−0.09 ± 0.12	⋯	0.58 ± 0.08	⋯	${15}_{-3}^{+5}$
HIP 28417	⋯	0.09 ± 0.08	⋯	0.37 ± 0.08	⋯	${10}_{-2}^{+2}$
HIP 29962	⋯	−0.17 ± 0.07	⋯	0.61 ± 0.05	⋯	⋯
HIP 32431	⋯	−0.02 ± 0.07	⋯	0.58 ± 0.05	⋯	${10}_{-3}^{+5}$
HIP 33578	⋯	−0.50 ± 0.05	⋯	0.68 ± 0.05	⋯	${3}_{-2}^{+2}$
HIP 57748	⋯	0.25 ± 0.09	⋯	0.37 ± 0.05	⋯	⋯
HIP 99789	⋯	−0.30 ± 0.08	⋯	0.52 ± 0.05	⋯	⋯
HIP 106775	⋯	−0.10 ± 0.07	⋯	0.70 ± 0.05	⋯	${9}_{-2}^{+4}$
HIP 113610	⋯	−0.17 ± 0.07	⋯	0.61 ± 0.10	⋯	⋯
HIP 114809	⋯	0.00 ± 0.07	⋯	0.56 ± 0.05	⋯	${10}_{-3}^{+10}$
HIP 116053	⋯	−0.31 ± 0.07	⋯	0.24 ± 0.08	⋯	${25}_{-5}^{+10}$
NGC 6940 MMU 105	⋯	−0.15 ± 0.09	⋯	0.39 ± 0.05	⋯	${15}_{-2}^{+5}$

Notes.

^a The optical mean C abundances are based on CH and C₂ features. ^b Both optical and HPF N abundances assume the C and O abundances determined from optical [O i] transitions.

Download table as:  ASCIITypeset image

Abundances from HPF C i transitions in Table 8 are repeated from Table 6. Mean values for the stellar sample have not been computed because star-to-star abundance differences for C, N, and ¹²C/¹³C are natural in evolved giants. The three sets of C abundances in Table 8 are offset from each other but they do correlate well: 〈[C_{C I,opt}/Fe] − [C_mean,opt/Fe]〉 = +0.10, σ = 0.08; 〈[C_{C I,HPF}/Fe] − [C_{C I,opt}/Fe]〉 = +0.19, σ = 0.08; and 〈[C_{C I,HPF}/Fe] − [C_mean,opt/Fe]〉 = +0.30, σ = 0.12. The C i HPF lines clearly yield aberrantly high abundances compared to the other values. The uncertainties in the mean differences are consistent with uncertainties in the observations and abundance computations involving C i lines. This issue should be investigated in the future, with attention to accuracy and consistency of sources of transition probabilities, dependences on stellar atmosphere parameters, and non-LTE sensitivities of these high-excitation (χ ≳ 7.5 eV) transitions.

Abundances of N were determined via synthetic/observed spectrum matches of the CN (0–0) Q-branch wavelength regions, with the redward-degrading bandhead culminating at 10925 Å. The synthetic spectra were generated assuming the optical O and C_mean abundances of Table 6. Formally the optical and HPF N abundances agree very well: $\langle [{{\rm{N}}}_{\mathrm{CN},\mathrm{HPF}}/\mathrm{Fe}]-[{{\rm{N}}}_{\mathrm{CN},\mathrm{opt}}/\mathrm{Fe}]\rangle$ = +0.01, σ = 0.05. However, the CN transition data used here all come from one recent extensive laboratory study (Brooke et al. 2014), and Sneden et al. (2014) have shown that various CN bands from blue through the near-IR yield internally consistent abundance results. The red-system (0–0) bands in the HPF region are much stronger than the (2–0) bands near 8000 Å that have dominated CN studies in the optical spectral region. Probably this will enable CN detection in warmer RHB stars than those included in this study.

Finally, we attempted to estimate ¹²C/¹³C ratios from the ¹³CN (0–0) Q-branch bandhead that occurs at 10923 Å, almost 3 Å blueward of the ¹²CN bandhead. However, in our RHB stars the ¹³CN lines are very weak, and this small spectral region is contaminated by telluric lines that are difficult to remove accurately. Therefore Table 8 lists only nine isotopic ratios, and a couple of these have large estimated uncertainties. For most stars the best conclusion is that the HPF ¹²C/¹³C values are consistent with, but probably inferior to, those derived from the optical (2–0) transitions.

HIP 33578 is an RHB field star that was an unremarkable member of the Afşar18a survey. But that study concentrated on kinematics, Fe metallicities, and limited Fe-group and α-element abundance ratios of its RHB sample. Z. Bozkurt et al. (2021, in preparation) will provide much more extensive abundance information on the LiCNO elements that are sensitive to stellar interior fusion cycles and envelope mixing, along with results for many other element groups. In this work we have noticed that HIP 33578 is one of the rare metal-rich field giants with a very low carbon isotopic ratio. Fortunately we were also able to obtain an IGRINS H- and K-band spectrum of this star. The optical, IGRINS, and now HPF spectra present a solid observational case for the extreme ¹²C/¹³C value, as we illustrate in Figure 6. For all three wavelength regions, the C abundance is the same while the N abundance has been allowed to vary by <0.1 dex to best match the observed ¹²CN and ¹²CO in each small spectral interval. The best estimate from synthetic/observed spectrum matches from each of these molecular bands is ¹²C/¹³C = 3 ± 1. This ratio at the surface of HIP 33578 is lower than, but consistent within the observational and theoretical uncertainties of, the interior CN-cycle ¹²C/¹³C value presented and first discussed in detail by Caughlan & Fowler (1962).

Zoom In Zoom Out Reset image size

Further support comes from the C and N abundances. From Table 8 the mean values for RHB stars excluding HIP 33578 are 〈[C/Fe]〉 = −0.36 (σ = 0.14) and 〈[N/Fe]〉 = 0.57 (σ = 0.14). The optical results for HIP 33578 are [C/Fe] = −0.88 and [N/Fe] = 0.67 (in accord with the HPF N abundance). The 0.5 dex extra depression in C and more modest enhancement of N are consistent with the very low ¹²C/¹³C of this star.

The LiCNO abundances and carbon isotopic ratio of HIP 33578 are in the domain of the "weak G-band" (wkG) disk giants (Sneden et al. 1978; Adamczak & Lambert 2013; Palacios et al. 2016). For HIP 33578 Z. Bozkurt et al. (2021, in preparation), supported by our HPF results, will derive log (Li) < 0.0, [C/Fe] = −0.7, [N/Fe] = 0.7, [O/Fe] = 0.2, and ¹²C/¹³C = 3, while approximate mean values from Adamczak & Lambert (2013) and Palacios et al. (2016) are $\langle \mathrm{log}\epsilon (\mathrm{Li})\rangle$ ≃ 1.3, 〈[C/Fe]〉 ≃ − 1.4, 〈[N/Fe]〉 ≃ 1.1, 〈[O/Fe]〉 = 0.0, and 〈¹²C/¹³C〉 ≃ 3 to 10. HIP 33578 has evidence of less C → N conversion. This star's Li abundance is very much smaller than the typical wkG giant, but the samples of both Adamczak & Lambert (2013) and Palacios et al. (2016) have stars with no obvious Li enhancement. Our single star cannot provide much new insight into the evolutionary history of this small red giant subclass. However, the RHB status of HIP 33578 is shared by many wkG stars. From its T_eff = 5118 K and its photometry and parallax given in Afşar et al. (2018a) we suggest that log(L/L_⊙) ≃ 2.5, in the middle of the set of warmer wkG stars shown in the H-R diagrams of Palacios et al. (2016) (see their Figures 7 and 8). We tentatively assign HIP 33578 to the wkG subclass.

In contrast to HIP 33578, the RHB star HIP 99789 has ordinary CNO abundances and ¹²C/¹³C for our RHB sample (Table 8) but an extremely high Li content: log (Li) ≃ 2.6, close to the value in the interstellar medium. No other abundance anomaly is evident in this star. However, we have detected an unusually strong He i absorption line at 10830 Å. In Figure 7 we show a montage of our RHB spectra in the wavelength region surrounding the 10830 Å line. This transition arises from a very high-excitation (19.8 eV) metastable level of He i. It has been used extensively in past studies to trace solar/stellar chromospheric activity and wind outflows, and has gained recent attention as a potential indicator of outflows from exoplanet atmospheres (e.g., Ninan et al. 2020). Inspection of Figure 7 suggests that more than half of our RHB sample, 8 out of 13 stars, have extremely little or no detectable He i absorption at 10830 Å. Another three stars, HIP 476, HIP 33578, and HIP113610, have modest He i lines. But HIP 29962 and HIP 99789 have strong 10830 Å absorptions, suggesting substantial chromospheric activity in their atmospheres. We cannot identify any other distinguishing spectroscopic signatures in HIP 99789, but the possible connection between high Li abundance and He i activity is intriguing and should be pursued with a dedicated set of HPF observations of a larger sample of high-Li stars.

Zoom In Zoom Out Reset image size

Z. Bozkurt et al. (2021, in preparation) will show that HIP 99789 is the most Li-rich member of the Afşar18a sample. However, its appearance among the RHBs may match a recent discovery about evolved stars with high Li. Singh et al. (2019) have determined Li abundances in red giants whose evolutionary status can be assigned to either the first ascent of the RGB or the red clump on the basis of their asteroseismological properties determined from the Kepler satellite (Mathur et al. 2017). Significant Li abundances can only be found in the He core-burning red clump stars (see their Figure 4). The standard red clump is cooler than the RHB; the highest T_eff reported by Mathur et al. for their Li-rich stars is 4999 K, and almost all other stars are cooler than 4900 K. Most of our RHBs have T_eff ≳ 5100 K. But for HD 99789 Afşar18a derived T_eff = 5054 K and $\mathrm{log}\,{\text{}}g$ = 2.41 (Table 3). The temperature is not radically different than those of the Li-rich sample of Mathur et al., and the gravity is in the middle of their $\mathrm{log}\,{\text{}}g$ distribution. HIP 99789 seems to support their assertion that Li enhancement is strongly associated with the He core-burning phase of late stellar evolution.

However, many decades ago Alexander (1967) speculated that red giants could temporarily enhance their Li content by ingesting companion planets in their bloated envelopes. Some evidence connecting Li-rich giants to planetary systems has accumulated recently, e.g., Adamów et al. (2018) and references therein. HIP 99789 may have one or more stellar or planetary companions, as indicated by its abnormally large apparent spatial acceleration. Close binary companions can influence unresolved spectra by diluting line depths of the primary star, adding in new lines, and producing anomalous RV offsets. None of our targets show signs of being double-line spectroscopic binaries from their RV cross-correlation functions. But companions with a low mass ratio or longer period could still be present in these systems. Recently, Brandt et al. (2019) cross-matched astrometry from Hipparcos and Gaia DR2 to produce the Hipparcos–Gaia Catalog of Accelerations (HGCA), which provides systematics-corrected proper motions spanning a ≈24 yr baseline between the two missions. This catalog provides a convenient way to search for tangential accelerations (changes in proper motion) that can be attributed to long-period giant planets, brown dwarfs, low-mass stars, or white dwarfs depending on the amplitude of the acceleration and orbital separation (e.g., Bowler et al. 2020).

Following Brandt et al. (2019), we make use of the proper motions from Gaia and Hipparcos–Gaia scaled positional differences to calculate accelerations in our sample. Among the 17 stars we observed, all except for NGC 6940 MMU 105 have Hipparcos astrometry and therefore have entries in the HGCA. Four out of these 16 stars have significant astrometric accelerations: HIP 29962 (149 ± 22 m s⁻¹ yr⁻¹), HIP 99789 (670 ± 19 m s⁻¹ yr⁻¹), HIP 113610 (255 ± 18 m s⁻¹ yr⁻¹), and HIP 114809 (76 ± 8 m s⁻¹ yr⁻¹). Interestingly, HIP 29962, HIP 99789, and HIP 114809 also have the largest discrepancies (≈4 km s⁻¹) between the RVs measured from the optical Tull data and those from our HPF spectra, offering further evidence that these stars have companions based on radial accelerations. The limited astrometric and RV sampling makes it challenging to constrain the nature of the companions, but the amplitude of these accelerations suggests that the companions are most likely low-mass stars or perhaps white dwarfs, both of which are not expected to meaningfully impact the spectra and abundances we derive in this study. While no solid connection can be argued between the large Li of HIP 99789 and its abnormally large apparent acceleration, it should be studied in the future.

The cool red giant MMU 105 is a member of the open cluster NGC 6940, and we have previously reported on its chemical composition from optical (BT16) and IGRINS (BT19) spectra. The instrumental setups and observations were like those discussed in Section 2.2. Model atmosphere parameters, metallicities, and abundance ratios for 12 NGC 6940 giants were determined in similar fashion to that used for the RHB field stars (Afşar18a), with the added advantage of having a reliable color–magnitude diagram (CMD) for the cluster. The optical-region metallicity of MMU 105 ([Fe/H] = −0.15, ²² σ = 0.07) is somewhat smaller than the cluster mean [Fe/H] = −0.02 (σ = 0.06). Note that MMU 105 is the coolest (4765 K) of the 12 RGs. The IGRINS H- and K-band iron abundance is larger, 〈[Fe/H]〉 = −0.04; our new HPF value is more consistent with the optical result. The Fe-group and α elements in this cluster have the abundances of slightly metal-poor disk stars. For elements with 21 ≤ Z ≤ 30, 〈[Fe‐group/Fe]〉 = 0.01 ± 0.14, and for lighter elements, 〈[α/Fe]〉 = 0.17 ± 0.07. The neutron-capture elements (Z > 30) appear to be slightly overabundant by an average of about 0.2 dex. From IGRINS data, 〈[Fe‐group/Fe]〉 = 0.15 ± 0.07, and [α/Fe] = [〈Mg, Si, S, Ca〉/Fe] = 0.11 ± 0.08. The HPF abundances listed in Table 6 concur with the optical means for these element groups.

The conclusion of BT16 and BT19 from photometric (CMD) and spectroscopic (abundance) evidence is that all 12 RGs including MMU 105 are helium-burning clump stars. The abundance argument begins with carbon isotopic ratios: ¹²C/¹³C_optical = ${15}_{-2}^{+3}$, ¹²C/¹³C_IGRINS = 23 ± 3, and now we can add ¹²C/¹³C_HPF = ${15}_{-2}^{+5}$. Both optical and IGRINS data reveal underabundances of C and overabundances of N. The HPF N abundances from the (0–0) bandheads agree well with the previous values, and the HPF C i abundances are in accord with the optical and IGRINS abundances after accounting for the systematic C i offset discussed above. The isochrones discussed in BT16 and BT19 suggest that all NGC 6940 RGs are He core-burning red clump stars that already have undergone canonical first dredge-up. The HPF C, N, and ¹²C/¹³C values should not be the primary data for assessing mixing in NGC 6940 MMU 105, but clearly they are in agreement with our previous optical and IGRINS studies.

The red giant HD 122563 (HIP 068594) is the only very low-metallicity member of the Yale Catalog of Bright Stars (Hoffleit & Jaschek 1991). It was first noted as a star with high proper motion (Roman 1955), and was one of the first very metal-poor stars subject to detailed abundance analysis (Wallerstein et al. 1963). Among many noteworthy chemical composition features, HD 122563 has a very large relative oxygen abundance ([O/Fe] ≃ +0.6; Lambert et al. 1974), very low carbon isotopic ratio (¹²C/¹³C = 5 ± 2; Lambert & Sneden 1977), and weak neutron-capture abundances in an r-process pattern (Sneden & Parthasarathy 1983; Honda et al. 2006).

The portion of the HD 122563 spectrum displayed in Figure 1 has identifications of three Si i lines and one each of H i, Ca i, and Sr ii. Neglecting the H i Paschen line, these five transitions represent a significant fraction (5/34) of the total number of lines that we were able to employ in the HD 122563 abundance analysis. This star simply has a metallicity about 500 times lower than the other stars of our study; most absorption lines simply fade into undetectability.

In Table 9 we list the lines employed in our analysis. The same EW and synthetic spectrum analyses described earlier were applied to the HPF spectrum of HD 122563. We imposed the same restrictions on transition probabilities for this star that we used for the other stars of our sample (see Section 3.1). We adopted the same model atmosphere that was used in the IGRINS spectrum analysis of Afşar et al. (2016). The modest results are presented in Table 10.

Table 9. HD 122563 Line Abundances

Wavelength	Species	χ	log(gf)	log	Source a
(Å)		(eV)
10683.08	C i	7.477	0.08	5.38	NIST D
10691.24	C i	7.482	0.35	5.53	NIST B
8806.76	Mg i	4.343	−0.14	5.29	NIST A
10288.94	Si i	4.916	−1.48	5.08	NIST C+
10371.26	Si i	4.926	−0.71	5.01	NIST C+
10660.91	Si i	4.916	−0.32	5.19	NIST C+
10689.72	Si i	5.949	0.09	5.07	NIST C+
10694.25	Si i	5.959	0.26	5.05	NIST C+
10727.41	Si i	5.979	0.42	5.05	NIST B
10749.38	Si i	4.926	−0.27	5.20	NIST C+

Note.

^a As defined in Table 4.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

The HD 122563 HPF metallicity, [Fe i _HPF/H] = −2.87 (σ = 0.07), is consistent with the optical value: [Fe i _opt/H] =−2.92 (σ = 0.12), Of the other species entered in Table 10, we draw attention to Si i, S i, and Sr ii.

The abundances from Si i lines in all three spectral domains are displayed in Figure 8. The transition probabilities for all the transitions were taken from the NIST database. Si abundances derived from HPF lines perhaps are the most reliable set. Certainly, as discussed in Section 4.2 of Afşar et al. (2016), the optical abundances may be the least trustworthy: the two strong lines in the blue (3905 Å and 4102 Å) are beset with blending and analytical difficulties, and the yellow/red lines are all very weak (log RW ≡ log(EW/λ) ≲ −5.6). The zyJ Si i lines in the HPF spectrum yield a mean Si abundance about 0.1 dex higher than do the lines in the IGRINS H- and K-band region. Resolution of this small discrepancy should begin with a comprehensive new study of transition probabilities.

Zoom In Zoom Out Reset image size

We have detected S i lines in HD 122563, as we illustrate in Figure 9. The triplet of lines near 10457 Å is strong in metal-rich giants such as HIP 114809 (top panel), and the 10455 Å line is blended enough that its abundance must be derived by synthetic spectrum computations. The two strongest lines of the triplet are clearly seen in HD 122563 (bottom panel). We have derived an S abundance from two of these lines and all three of the 9220 Å lines. Our LTE abundance is [S/Fe] = +0.75 (σ = 0.26), and if we apply the non-LTE correction of −0.17 computed by Spite et al. (2011) we get [S/Fe] = +0.58, somewhat larger than their value of +0.42. Further investigation of this abundance with a completely self-consistent model and computational techniques should be pursued in a future investigation.

Zoom In Zoom Out Reset image size