Identification of Absorption Lines of Heavy Metals in the Wavelength Range 0.97–1.32 μm

We investigate YJ-band spectra of 13 stars, consisting of giants and supergiants. Their spectra were obtained in 2015 and 2016 with WINERED attached to the 1.3 m Araki Telescope at Koyama Astronomical Observatory in Kyoto, Japan. The spectral resolution of WINERED with the setting of the WIDE mode and the 100 μm slit is around 28,000 (Ikeda et al. 2016). We observed dozens of supergiants, giants, and dwarfs whose temperatures were derived making use of the line-depth ratios by Kovtyukh et al. (2003, 2006) and Kovtyukh (2007). The entire set of the spectra for these objects will be considered for discussing the line-depth ratios (M. Jian et al. 2020, in preparation), but some of the spectra of supergiants and giants are used in this study (Table 1). We use the stellar parameters, including ${T}_{\mathrm{eff}}$, taken from Luck (2014) and Hekker & Meléndez (2007) because they obtained the four necessary parameters, i.e., ${T}_{\mathrm{eff}}$, log g, $[\mathrm{Fe}/{\rm{H}}]$, and the microturbulence (ξ), altogether. Their measurements are based on high-resolution spectra (R ≥ 40,000), and the obtained parameters are precise enough for our purpose; in fact, the spectra synthesized with the given parameters reproduce the observed spectra at around the 1% level, as we see below. Figure 1 presents the distribution of the 13 objects on the $({T}_{\mathrm{eff}},\mathrm{log}g)$ plane.

Zoom In Zoom Out Reset image size

Table 1.  Observed Objects and Their Parameters

HD	Name	Sp. Type	${T}_{\mathrm{eff}}$	log g	[Fe/H]	ξ	Ref.	Obs. Date	${v}_{{\text{}}b}$	γ
			(K)	(dex)	(dex)	(km s−1)		(UT)	(km s−1)
25291	HR 1242	F0 II	7171	2.19	+0.03	1.67	1	2015 Oct 28	12.65	1.103
20902	α Per	F5 Ib	6579	1.76	+0.15	3.65	1	2015 Oct 26	25.61	1.694
194093	γ Cyg	F8 Ib	6212	1.35	+0.05	4.02	1	2015 Oct 26	16.64	1.386
204867	β Aqr	G0 Ib	5511	1.54	+0.03	3.39	1	2015 Oct 26	17.60	1.535
26630	μ Per	G0 Ib	5418	1.74	+0.09	3.02	1	2015 Oct 28	17.93	1.362
159181	β Dra	G2 Ib–IIa	5291	1.87	+0.15	2.72	1	2016 Feb 3	18.10	1.198
77912	HR 3612	G7 IIa	5001	2.03	+0.12	2.16	1	2016 Mar 21	14.27	1.012
27697	δ Tau	G9.5 III	5000	3.00	+0.08	1.50	2	2015 Oct 25	12.31	0.957
19787	δ Ari	G9.5 IIIb	4875	3.05	+0.09	1.68	2	2016 Mar 11	12.07	0.917
208606	HR 8374	G8 Ib	4731	1.18	+0.25	3.49	1	2015 Oct 31	17.05	0.896
9900	HR 461	K0 II	4552	1.35	+0.19	2.46	1	2015 Oct 31	13.51	0.815
206778	Peg	K2 Ib–II	4165	0.50	−0.01	2.96	1	2015 Oct 31	14.73	1.409
52005	HR 2615	K3 Ib	4077	0.19	−0.07	2.76	1	2015 Oct 28	14.18	1.557

Note. Parameters of each target are taken from one of the two references: Ref. 1 = Luck (2014) and Ref. 2 = Hekker & Meléndez (2007). In the last two columns, ${v}_{{\text{}}b}$ indicates the broadening width including the instrumental profile, which corresponds to ∼10.7 km s⁻¹, and γ indicates the factor we used to convert the depths (see Section 3.2). Spectral types are taken from the SIMBAD database.

Download table as:  ASCIITypeset image

All the observed spectra were reduced by following the standard procedure adopted in the WINERED pipeline (S. Hamano et al. 2020, in preparation) that was established using PyRAF,¹² which calls IRAF tasks¹³ from Python scripts. The reduction steps include sky subtraction, scattered light subtraction, flat-fielding (using a halogen lamp with an integrating sphere), geometric transformation, aperture extraction, and wavelength calibration based on Th–Ar lamp spectra. In addition, small wavelength shifts¹⁴ between individual exposures were corrected, if necessary, before they were combined to give an averaged one-dimensional spectrum for each target. Throughout this paper, we use air wavelengths rather than vacuum wavelengths. Then, the one-dimensional spectrum was normalized at the continuum level, and the telluric correction was performed as described in Sameshima et al. (2018b) using the WINERED spectrum of an A0 V star taken at the same night.

It is worthwhile to add remarks on some objects. Because they are bright, there have been studies with optical spectra dating back decades, e.g., Steel (1945) for HD 20902 (α Per). In particular, HD 194093 (or γ Cyg) is an interesting supergiant. Since the early discovery by Adams & Joy (1926, 1927), it has been found that the spectrum of γ Cyg is rich in rare-earth lines, e.g., Eu and Dy (see, also, Roach 1942). The star γ Cyg is located close to the so-called γ-Cygni supernova remnant, but Johnson (1975) found that they are not connected with each other.

We used MOOG (the version released in 2017 February Sneden et al. 2012) to create synthetic spectra for selecting and confirming absorption lines of the heavy elements. We adopted the 1D plane-parallel atmosphere models compiled by R. Kurucz.¹⁵ The local thermal equilibrium (LTE) is assumed in the models we used as well as each part of the spectral synthesis. We considered two sets of stellar parameters: those for the first set are listed in Table 1 and used for making direct comparisons between the synthetic and observed spectra. We estimated the broadening of individual objects (${v}_{{\text{}}b}$, in the full width at half maximum) including the instrumental resolution based on several isolated lines in the observed spectra. For the second set, we used the ${T}_{\mathrm{eff}}$ and log g values on the grid indicated by the "+" points in Figure 1 together with the following parameters fixed: the microturbulence ξ = 2 km s⁻¹, the broadening ${v}_{{\text{}}b}=10.7$ km s⁻¹, and the chemical abundances to be solar (Asplund et al. 2009). The fixed microturbulence and broadening are not optimal for stars with broad ranges of ${T}_{\mathrm{eff}}$ and log g. The ξ of 2 km s⁻¹ is typical for red giants, but ξ tends to be larger for stars with higher luminosity (Gray et al. 2001). The macroturbulence (${v}_{\mathrm{mac}}$), or broadening in general, also shows the dependency on the luminosity class (Gray & Toner 1986). Likewise, both ξ and ${v}_{\mathrm{mac}}$ depend on ${T}_{\mathrm{eff}}$ (e.g., Ryabchikova et al. 2016). This implies that the predicted line strengths in the second set of the synthetic spectra may be biased, especially for dwarfs and supergiants. Nevertheless, those synthetic spectra are useful to predict where absorption lines in question become significant or deep in the parameter space of $({T}_{\mathrm{eff}},\mathrm{log}g)$. Incidentally, the absorption lines we discuss here are mostly shallow and not strongly saturated; therefore, ξ does not affect our predictions so much.

For both of the sets mentioned above, we assumed the solar abundance ratios of Asplund et al. (2009), $\mathrm{log}{\epsilon }_{\mathrm{Fe},\odot }=7.50$ dex for [Fe/H] = 0 and other "Photosphere" abundances listed in their Table 1, except the carbon abundance. We changed [C/H] by some amounts, between −0.31 and 0.28 dex, in order to get better agreements of the CN lines between the synthetic and observed spectra. The observed spectra in the lower temperature range show many CN lines, and those lines are not well-reproduced by the synthetic spectra of many stars without the changes. This may be caused by the surface abundance ratios modified by the first dredge-up (Luck & Lambert 1985; Takeda et al. 2013; Lyubimkov et al. 2015). The purpose of this adjustment is merely to get the synthetic spectra to match better with the observed ones. The [C/H] values we used are not necessarily accurate estimates of the carbon abundances; therefore, we do not give the [C/H] used for individual objects. In contrast, [C/H] is fixed to be solar in the second set of synthetic spectra for the grid of $({T}_{\mathrm{eff}},\mathrm{log}g)$.

In order to assess the absorption of target lines even if they are contaminated by other lines, we consider four kinds of synthetic spectra: (i) normal ones with all lines included, (ii) those with lines of only one species (i.e., each atom or atomic ion) included, (iii) those with only one target line included, and (iv) those with the target lines excluded. We consider the $\mathrm{log}{gf}$ values and other parameters of a given target line in each of the three lists. For other atomic lines, we adopted the lines and their parameters in KURZ when we considered target lines in the KURZ list, but we used the VALD line list for considering target lines in VALD or those in MB99. The KURZ list we used is the version compiled on 2017 October 8, while we downloaded the VALD3 list to use on 2019 August 27. For molecular lines, we used those listed in VALD, in which lines of C₂, CH, CN, CO, and OH molecules are included within the wavelength range of our interest. Only CN lines appear significant in the temperature ranges we investigate. In addition, we adopted FeH lines compiled by B. Plez (2019, private communication; see also Önehag et al. 2012). We adopted the FeH's dissociation energy of 1.59 eV, at 273.15 K, from Schultz & Armentrout (1991). The FeH lines appear only in a few objects with the lowest temperatures.

In comparing the observed and synthetic spectra of the 13 objects, we noticed dozens of absorption lines that are quite visible in the observed spectra but not predicted in the synthetic ones. We list such lines and discuss their characteristics in Appendix B.

We searched for absorption lines of elements heavier than the iron group elements in the three line lists, KURZ, VALD, and MB99. Combining the three lists, our target elements are from Cu to Th, i.e., 29 ≤ Z ≤ 90, with some gaps. We limited ourselves to the ionization states between I (neutral) and III (doubly ionized). We also limited the wavelength ranges to 9760–11100 and 11600–13200 Å, corresponding to the orders of 51–57th (Y band) and 43–48th (J band) of WINERED.

In the given wavelength ranges, MB99 lists 19 lines of six species in total, i.e., Zn i, Ge i, Sr ii, Y ii, La ii, and Eu ii. MB99 covers the wavelengths longer than 10000 Å only. MB99 actually lists 20 lines of our target species between 10000 and 13400 Å, but one of the lines, Ge i 11125.12, falls within the gap of the target in the wavelength range. There is dense and strong telluric absorption around the Ge i line (see, e.g., Figure 4 in Sameshima et al. 2018b), and we neglected this line. We consider the 19 lines in MB99 in the next selection step.

In contrast to MB99, VALD and KURZ list large numbers of absorption lines for various elements between Cu and Th. Hundreds of lines are listed for some elements, but we cannot expect to see so many lines of these heavy elements in real stellar spectra, as we confirm below. Before we compared the observed spectra with the synthetic ones, we limited the number of candidate lines for each species using the synthetic spectra. We considered the spectra for the grid of (${T}_{\mathrm{eff}}$, log g) with lines of only one species included at each time (see Section 2.2). We picked up lines that become deeper than 0.02 in depth in at least one of the synthetic spectra at all the grid points. We found 67 lines of 9 species in KURZ and 63 lines of 14 species in VALD to be candidates.

There are many overlaps between the lines selected with the three line lists, and there are 108 different lines of 14 species in total; 40, 35, and 5 lines are listed only in KURZ, VALD, and MB99, respectively. All of the five lines selected from MB99 only are of Ge i. In the following, we consider whether or not these lines appear significant in the observed spectra, and if they are detected, we examine their temperature dependence in comparison with the synthetic spectra.

A significant fraction of the lines we investigate are expected to be blended with other absorption lines even if they exist, and it is difficult to measure the equivalent widths. We therefore consider depths, i.e., $d\equiv 1-f({\lambda }_{c})$, where f(${\lambda }_{c}$) is the flux at the line center, ${\lambda }_{c}$, in normalized spectra.

In order to make direct comparisons between the depths in the observed and synthetic spectra, we perform the normalization to adjust the continuum levels of the two kinds of spectra. The reduction performed by the standard pipeline software for WINERED includes the continuum normalization (see, e.g., Section 3.1 in Taniguchi et al. 2018). Although the synthetic spectra we created for individual objects show reasonable agreements between the observed and synthetic spectra, there are noticeable deviations, especially within absorption lines. Such deviations can be attributed to inaccurate $\mathrm{log}{gf}$ values (Andreasen et al. 2016; Kondo et al. 2019) and maybe to inaccurate assumed abundances. Moreover, narrow flat parts outside apparent absorption lines are found to be lower than unity in some spectral regions of the synthetic spectra, but the normal continuum procedure applied to the observed spectra adjusts such regions to be around 1. We therefore match the continuum levels of the observed and synthetic spectra around each target line as follows.

First, we searched for velocity offsets to fix small offsets in the wavelength scale, if any, by minimizing the residuals between the observed and synthetic spectra. For each target line at the wavelength of ${\lambda }_{c}$, the residuals were calculated for spectral parts around the line center. We used a width of ±300 km s⁻¹ for the objects with ${T}_{\mathrm{eff}}\lt 5200$ K, and a width of ±1000 km s⁻¹ for the others for this wavelength adjustment. The large width for the warmer stars was necessary to accommodate many absorption lines, which enable us to estimate the offset precisely. When we compare the observed and synthetic spectra, the synthetic spectra were pixelized to match the pixels of the observed spectra by integrating the normalized flux within the wavelength range of each pixel of the latter, and ${f}_{\mathrm{syn}}$ indicates the counts of thus digitized spectra. We corrected the observed spectra produced by the pipeline for the velocity offsets; ${f}_{\mathrm{obs},0}$ indicates the counts of the observed spectra after this correction.

Second, we considered the pixel-by-pixel ratio between each observed spectrum and the corresponding synthetic one, ${r}_{\mathrm{org}}\equiv {f}_{\mathrm{obs},0}/{f}_{\mathrm{syn}}$, for adjusting the continuum level of the former to that of the latter. For the normalization around each target line, we used the adjacent parts on both sides of the line, i.e., ${\lambda }_{c}-{{\rm{\Lambda }}}_{2}\lt \lambda \lt {\lambda }_{c}-{{\rm{\Lambda }}}_{1}$ and ${\lambda }_{c}$ + Λ₁ < λ < ${\lambda }_{c}$ + Λ₂, where Λ₁ and Λ₂ are the wavelength steps corresponding to the velocity of 15 and 100 km s⁻¹, at around ${\lambda }_{c}$, respectively. The ratio shows deviations from unity caused by various errors, including the offset and slope in the prenormalized continuum level of ${f}_{\mathrm{obs},0}$ and imperfect reproduction of the stellar spectrum in ${f}_{\mathrm{syn}}$. The ratio (as a function of pixel) of the adjacent parts around each target line was fitted as a linear trend, $a(\lambda -{\lambda }_{c})+b$, by minimizing

$$\begin{eqnarray}&&{\chi }^{2}=\displaystyle \sum _{{\lambda }_{i}}{w}_{f}{w}_{\lambda }{\left\{{r}_{\mathrm{org}}({\lambda }_{i})-a({\lambda }_{i}-{\lambda }_{c})-b\right\}}^{2},\end{eqnarray} \tag{ 1 }$$

where the summation is taken over the pixels, ${\lambda }_{i}$, within the adjacent ranges mentioned above. The pixels within ±15 km s⁻¹ around the center were not included. We introduced two weight terms, w_f and w_λ,

$$\begin{eqnarray}&&{w}_{f}={[{f}_{\mathrm{syn}}({\lambda }_{i})]}^{2},\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&{w}_{\lambda }=\exp \left(-\displaystyle \frac{| {\lambda }_{i}-{\lambda }_{c}| }{{{\rm{\Lambda }}}_{2}}\right).\end{eqnarray} \tag{ 3 }$$

The former depends on the count of ${f}_{\mathrm{syn}};$ the pixels with flux closer to the continuum are weighted more than the pixels within strong absorption. The latter, w_λ, gives higher weights to the pixels closer to the target line in wavelength than to those more separated. Admittedly, we have no solid mathematical background to use the weights as given in Equations (2) and (3). Nevertheless, after some experiments, we found that they tend to give reasonable results—even if the observed spectra show deviations from the synthetic one, such as lines with clearly different depths. Next, we divided the original observed spectra by the fitted linear trend to obtain the renormalized spectra, ${f}_{\mathrm{obs}}(\lambda )={f}_{\mathrm{obs},0}(\lambda )/[a(\lambda -{\lambda }_{c})+b]$. We also calculated the weighted standard deviation of the residual between the observed and synthetic spectra,

$$\begin{eqnarray}&&e=\sqrt{\displaystyle \frac{\sum \left\{{w}_{f}{w}_{\lambda }{\left[{f}_{\mathrm{obs}}({\lambda }_{i})-{f}_{\mathrm{syn}}({\lambda }_{i})\right]}^{2}\right\}}{\sum {w}_{f}{w}_{\lambda }}}.\end{eqnarray} \tag{ 4 }$$

This e serves as the error in depth when we compare the depths of the line in the observed and synthetic spectra. The residual is attributed to various factors, including pixel-to-pixel errors in the observed spectra, the inconsistency between the observed and synthetic spectra, and the error in the normalization. The median of the e values calculated for individual ranges around the target lines is ∼0.01 or smaller for most of the 13 objects. For the two objects with ${T}_{\mathrm{eff}}$ < 4200 K, however, the median e values are nearly 0.02. We note that the error estimates described here do not include the uncertainty concerning the target lines themselves, because the normalization and the calculations of e were done without considering the pixels within Λ₁ = 15 km s⁻¹ around the lines. The residuals at the target wavelengths, if significant, would tell us the inconsistency between the depths observed and those predicted with the line lists.

Next, we measured the depths, d, at the line wavelength, λ_c, by linearly interpolating the two adjacent pixels, λ_i and ${\lambda }_{i+1}$, with λ_i ≤ λ_c < λ_i+1; ${d}_{\mathrm{obs}}$ and ${d}_{\mathrm{syn}}$ indicate the depths measured with ${f}_{\mathrm{obs}}$ and ${f}_{\mathrm{syn}}$, respectively. While the two depths of each target line can be directly compared with each other for each object, the depths for different objects are affected by differences in various stellar parameters. Here, we consider the indicator, ${d}^{* }=\gamma d$, i.e., the depths multiplied by the factor γ, which compensates for the effects of stellar metallicity and line width,

$$\begin{eqnarray}&&\gamma \equiv {10}^{-[\mathrm{Fe}/{\rm{H}}]}\left(\displaystyle \frac{{v}_{{\text{}}b}}{{v}_{\mathrm{inst}}}\right),\end{eqnarray} \tag{ 5 }$$

where the first term compensates for the metallicity effect and the second term converts the depths in spectra broadened by the width of ${v}_{{\text{}}b}\,(\simeq \sqrt{{v}_{\mathrm{mac}}^{2}+{v}_{\mathrm{inst}}^{2}})$ into those in the spectra merely broadened by the instrumental line width, ${v}_{\mathrm{inst}}=10.7$ km s⁻¹, of the WIDE mode of WINERED (i.e., with the broadening by macroturbulence, ${v}_{\mathrm{mac}}$, ignored). The γ factor for each target is listed in Table 1. It should be noted that d* does not necessarily agree with the depth of a line in the spectrum of a star with $[\mathrm{Fe}/{\rm{H}}]=0$ dex broadened by the instrumental width, if the line is blended with other lines. Nevertheless, the conversion from d to d* makes it easier to compare the depths of 13 targets directly and examine the dependence on ${T}_{\mathrm{eff}}$. The d* may still be affected by log g and ξ. We will take into account the effect of log g by considering the depths in the synthetic spectra obtained for the grid of $({T}_{\mathrm{eff}},\mathrm{log}g)$ that was described in Section 2.2. On the other hand, ξ has little impact on shallow lines, for which we need precise measurements in order to confirm the line identification.

Finally, we evaluate the indicator that measures the absorption of the target lines themselves. As mentioned in Section 2.2, we synthesized the spectra with the target lines excluded. Their depths at the wavelengths in question are also measured and denoted as ${d}_{\mathrm{syn}\dagger }$. They are expected to be zero if there is no blending line, but they are often not zero. By subtracting ${d}_{\mathrm{syn}\dagger }$ from the normal depth, we can estimate the contribution of a target line to absorption. We calculated this indicator, α, for the depths in both observed and synthetic spectra; ${\alpha }_{\mathrm{obs}}\equiv {d}_{\mathrm{obs}}-{d}_{\mathrm{syn}\dagger }$ and ${\alpha }_{\mathrm{syn}}\equiv {d}_{\mathrm{syn}}-{d}_{\mathrm{syn}\dagger }$. If ${\alpha }_{\mathrm{obs}}$ is large with respect to e, the absorption by the target line is suggested to be significant. When we compare the measurements for the 13 objects, we consider the conversion by considering the effects of metallicity and broadening, ${\alpha }^{* }\equiv {d}^{* }-{d}_{\mathrm{syn}\dagger }^{* }=\gamma (d-{d}_{\mathrm{syn}\dagger })$, where ${\alpha }^{* }$ and d* are considered for both the observed and synthetic spectra. In addition, we use $\beta \equiv {d}_{\mathrm{syn}\dagger }/{d}_{\mathrm{syn}}$ as the estimate of the blend; β varies from zero (no blend) to unity (fully contaminated). We define β = 1 when ${d}_{\mathrm{syn}}=0$, but we are not interested in such cases, because our purpose is to confirm or reject predicted lines.

We measured the depths, ${d}_{\mathrm{obs}}$ and ${d}_{\mathrm{syn}}$, and also the errors, e, for the 108 candidate lines (selected in Section 3.1) in the spectra of the 13 objects. While many lines were not seen, there are dozens of lines showing large ${\alpha }_{\mathrm{obs}}/e$ values. We considered the lines with ${\alpha }_{\mathrm{obs}}/e\gt 2$ to be significant unless they are overly affected by spurious noises. We then examined d*, ${\alpha }^{* }$, and β plotted against ${T}_{\mathrm{eff}}$ (Figure 2), and checked if the dependence on ${T}_{\mathrm{eff}}$ is consistent with the expectation from the synthetic spectra. For some lines, the gravity effect on the depths measured for the targets with different log g is significant, and we took it into account by comparing the ${T}_{\mathrm{eff}}$ trends expected for the two log g values, 0.5 and 2.5, illustrated in Figure 2. In addition, we checked the appearance of the spectra, comparing the synthetic spectra with the target lines removed with the observed spectra (see Figure 3), as well as how the telluric absorption could have left noises after the correction.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (23 panels)

Tarball of figure set images

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (6 panels)

Tarball of figure set images

Among the 108 lines we examined in the observed spectra (Section 3.1), we detected 23 lines (Table 2). Figure 3 presents the spectra of some stars with the line detected. We here summarize the characteristics of the lines we identified, but more details on the lines of individual species are given in Appendix A. We detected two Zn i, three Sr ii, and five Y ii lines included in all the three lists, KURZ, VALD, and MB99, but no other lines of these species were confirmed. We also detected one MB99 line of Eu ii (10019.52 Å), together with Eu ii 9898.27 and 10165.56, none of which are included in MB99. In addition, among the candidate lines that are not in MB99, we identified four Zr i, one Ba ii, two Ce ii, two Sm ii, and one Dy ii lines.

Table 2.  Absorption Lines Identified in the Observed Spectra

Species	${\lambda }_{\mathrm{air}}$	EP	$\mathrm{log}{gf}$ (dex)
	(Å)	(eV)	KURZ	VALD	MB99
Zn i	11054.26	5.796	−0.20	−0.30	−0.50
Zn i	13053.60a	6.655	+0.39	+0.34	+0.13
Sr ii	10036.65	1.805	−1.22	−1.31	−1.10
Sr ii	10327.31	1.839	−0.25	−0.35	−0.40
Sr ii	10914.89	1.805	−0.57	−0.64	−0.59
Y ii	10105.52	1.721	−1.63	−1.89	−1.89
Y ii	10186.46	1.839	−2.65b	−2.65b	−1.97
Y ii	10245.22	1.738	−1.82	−1.82	−1.91
Y ii	10329.70	1.748	−1.51	−1.76	−1.71
Y ii	10605.15	1.738	−1.71	−1.96	−1.89
Zr i	9822.56	0.623	−1.44	−1.20	⋯
Zr i	10654.18	1.582	−1.32	⋯	⋯
Zr i	11612.67	1.366	−0.88	⋯	⋯
Zr i	11658.07	1.396	−0.56	⋯	⋯
Ba ii	13057.72c	5.251	+0.34b	+0.34b	⋯
Ce ii	9805.49	0.322	⋯	−3.23b	⋯
Ce ii	9853.11	0.704	⋯	−2.49b	⋯
Sm ii	9850.67	1.971	⋯	−0.46b	⋯
Sm ii	9936.51	1.890	⋯	−0.61	⋯
Eu ii	9898.30	2.108	−0.76b	−0.07	⋯
Eu ii	10019.52	2.091	−0.66b	−0.66b	−0.30
Eu ii	10165.56	2.108	−0.78	−0.78	⋯
Dy ii	10523.39	1.946	⋯	−0.45	⋯

Notes.

^aThe wavelength in the KURZ list, 13053.559 Å, is slightly different from the counterparts in the VALD, 13053.627 Å, and the MB99, 13053.64 Å. ^b The measured depths differ from those predicted with the given $\mathrm{log}{gf}$ by ∼0.3 dex or more. ^cThe wavelength in the KURZ list, 13057.716 Å, is shorter than the counterpart in the VALD, 13058.015 Å.

Download table as:  ASCIITypeset image

A few lines have been already reported in previous papers, aside from MB99, based on high-resolution spectra. Although the temperature of the object is different from ours, Sameshima et al. (2018a) identified the two Sr ii lines at 10327.31 and 10914.89 Å in a WINERED spectrum of an A0 V-type star. The same lines were used for measuring abundances of red supergiants by Origlia et al. (2013), and the one at 10914.89 Å was also identified by Hubrig et al. (2012) in peculiar A-type stars. In addition to these Sr ii lines, we also confirmed the one at 10036.65 Å (a strong line listed in MB99). The Dy ii line at 10835.94 Å reported by Hubrig et al. (2012) with spectra of peculiar A-type stars was not confirmed in any of our targets.

We could not confirm 85 candidate lines in the observed spectra. They include some lines listed in MB99: Zn i 13196.62, La ii 11874.19, and six Ge i lines (see the notes on the individual elements in the Appendix). We found that there may be absorption by some of the unconfirmed lines, but rejected them either because the lines are not deep enough to be confirmed (${\alpha }_{\mathrm{obs}}/e\lt 2;$ e.g., Ce ii 9774.63) or the telluric absorption seems to leave spurious noises (e.g., Sm ii 9788.96). In addition, we found quite strong absorption at the wavelength of Dy ii 10305.36, but did not conclude that the line was confirmed because the observed ${T}_{\mathrm{eff}}$ trend is not consistent with the prediction. The Dy ii line, if it exists, is at least contaminated by an unidentified line.

We examined $\mathrm{log}{\alpha }_{\mathrm{obs}}^{* }-\mathrm{log}{\alpha }_{\mathrm{syn}}^{* }$ and judged whether the $\mathrm{log}{gf}$ values given in the line lists reproduce the depths of significant lines ($\mathrm{log}{\alpha }_{\mathrm{obs}}^{* }-\mathrm{log}{\alpha }_{\mathrm{syn}}^{* }\simeq 0$) or not. The superscript c in Table 2 indicates that the measured depths differ from those predicted with the given $\mathrm{log}{gf}$ by ∼0.3 dex or more. We note, however, that our conclusions on the $\mathrm{log}{gf}$ cannot be final, because we only made comparisons with the synthetic spectra created with the solar [X/Fe] values assumed.

In the following, we discuss the implications that individual elements with the detected line(s) may have for the chemical evolution of our Galaxy and other nearby galaxies. We here focus on the chemical imprints one would find in Galactic-disk stars of around solar metallicity. For stars with different metallicities or those in different systems, some elements may be explained by different origins. For example, Ba is an s-process element at around the solar metallicity, but the contribution of the r-process gets stronger at the low-metallicity range (Burris et al. 2000). Moreover, there are proposed n-capture processes that show patterns different from those of the main s- and r-processes (e.g., Burris et al. 2009; Hampel et al. 2016). Our main targets, supergiants, are young and expected to be relatively metal-rich, and previous explanations on the origins of the heavy elements in the Sun and our solar system (Burris et al. 2000; Sneden et al. 2008) gives at least an approximate idea on the origins of individual elements in the supergiants.

Zn is often included in the iron peak elements and considered as the heaviest one of this group. In fact, in a broad range of the metallicity down to $[\mathrm{Fe}/{\rm{H}}]\simeq -2$ dex, Zn in stars in the solar neighborhood shows a concentration around [Zn/Fe] = 0, and thus seems to be created along with Fe and other iron peak elements (Sneden et al. 1991). However, it has become clear that systematic differences in the [Zn/Fe] trends are found in different systems (thin and thick disks, bulge, halo, and dwarf galaxies), thanks to the efforts of various authors (Duffau et al. 2017; Hirai et al. 2018; Ji & Frebel 2018, and references therein). The origins of Zn remain elusive, and its implications upon Galactic chemical evolution may be unique, compared to other, better-understood elements (see, e.g., Tsujimoto & Nishimura 2018).

The s-process elements are created mainly in low- and intermediate-mass Asymptotic Giant Branch stars. Good reviews of the process are found, e.g., in Busso et al. (1999) and Karakas & Lattanzio (2014). Sr, Y, and Zr are grouped as light s-process elements, although they show slightly different trends from each other in disk stars (Delgado Mena et al. 2017). Ba and Ce are often called heavy s-process elements, together with Nd. The abundance trends of the three heavy s-process elements are, however, not necessarily common. For example, Andrievsky et al. (2013, 2014) reported that the radial gradient of Ba has a negligible slope, while Lemasle et al. (2013) and da Silva et al. (2016) found significant slopes of the gradient for the other heavy s-process elements, Ce and Nd, as well as for Y and Zr (i.e., light s-process elements). As noted by Andrievsky et al. (2014) and Luck (2014), only strong Ba lines were used with non-LTE calculations, which may leave significant uncertainties. Readers are referred to other studies, such as D'Orazi et al. (2009), Bensby et al. (2014), and references therein, concerning the Ba abundances by using the strong lines in the optical, but their targets are dwarfs and giants. Ba ii 13058.01 found in this study is, in contrast, expected to be weak; therefore, this line may add an important constraint on the Ba abundances. On the other hand, the three Sr ii lines in the Y band are very strong, and the same problem concerning the non-LTE effect may prevent us from obtaining accurate Sr abundances.

Eu is representative of r-process elements. Sm and Dy are considered to be formed mainly through the r-process around the solar metallicity (Burris et al. 2000). The contribution of the s-process to Sm can be larger, especially for some isotopes, but the trend of Sm seen in the disk stars with about the solar metallicity is similar to that of Eu (Battistini & Bensby 2016). After the discovery of the gravitational wave and the electromagnetic counterpart of GW170817 (Abbott et al. 2017), it has become more evident that the neutron star mergers accompanied by kilonovae are major contributors to r-process elements (Smartt et al. 2017; Tanaka et al. 2017). However, explaining the chemical evolution of r-process abundances seen in disk stars requires further investigations concerning some complications due to, e.g., the delay-time distributions of the neutron star mergers (Hotokezaka et al. 2018), the initial mass functions (Tsujimoto & Baba 2019), and potential contribution from other origins (Siegel et al. 2019). New observational constraints on the r-process enrichment can be expected if the near-infrared diagnostic lines of Eu and Dy are used for measuring the abundances of stars in unexplored regions of the Galactic disk behind severe interstellar extinction.

The lines we report here will be useful diagnostic lines of detailed chemical abundances available in the YJ bands. In order to give a rough idea of the $({T}_{\mathrm{eff}},\mathrm{log}g)$ range in which the detected lines are seen, we present in Figure 4 the contours of ${\alpha }_{\mathrm{syn}}={d}_{\mathrm{syn}}-{d}_{\mathrm{syn}\dagger }$ based on the synthetic spectra obtained for the $({T}_{\mathrm{eff}},\mathrm{log}g)$ grid (see Figure 1). Some lines are blended, and it may be difficult to measure their abundances even if α becomes significant. Also, note that the depths refer to those expected for the spectra with the WINERED resolution, R = 28,000, for stars with solar metallicity, $[\mathrm{Fe}/{\rm{H}}]=0$ dex, and sufficiently small broadening. Nevertheless, the contours will be useful, e.g., for selecting targets to study the abundances discussed in this paper and making observational plans to achieve the detection of those lines.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (23 panels)

Tarball of figure set images

Only two species (Zn i and Zr i) among the nine with the lines detected are neutral. In contrast, the others are singly ionized. The first ionization potentials of the elements with the detected lines are low, 5.2–6.2 eV, compared to Zn (9.4 eV) and Zr (6.6 eV). The lines of the ionized species are sensitive to log g and strong in low-gravity stars, as expected. Furthermore, Figure 4 indicates that supergiants with ${T}_{\mathrm{eff}}\sim 5000$ K and $\mathrm{log}g\lesssim 1.5$ dex show the strongest absorption of the lines of Sr ii, Y ii, Ba ii, Sm ii, Eu ii, and Dy ii. The contours of Zn i are similar, to some extent, to those of the above species, the peak being around 5000 K and stronger toward the lower surface gravity, but the sensitivity to the gravity is not so high as the ionized species. In contrast, the lines of Ce ii require both low temperature (${T}_{\mathrm{eff}}$ < 4000 K) and low surface gravity (log g < 1) to be significant. In the case of Zr i, the line strengths grow rapidly toward low ${T}_{\mathrm{eff}}$ but are insensitive to surface gravity.

Cepheids have intermediate temperatures, 4500–6500 K, and low surface gravities, log g ≲ 1.5 dex (see, e.g., Genovali et al. 2014). Their locations on the Hertzsprung–Russell diagram (HRD) thus agree with the peaks of the strengths for most of the lines reported in this paper, although Zr i may be seen only in Cepheids with the lowest ${T}_{\mathrm{eff}}$ among this group of pulsating stars in the Cepheid instability strip. Therefore, YJ-band spectra of Cepheids would allow us to measure various heavy elements, including both s- and r-process elements and also those with mixed origins such as Zn.

Cepheids are young, 10–300 Myr old, and provide valuable information on chemical abundances of the present Galaxy. It is known that their metallicities are anticorrelated with the distances from the Galactic center (${R}_{\mathrm{GC}}$), known as the abundance gradient (Genovali et al. 2014). Observing Cepheids in a wide range of ${R}_{\mathrm{GC}}$ thus allows investigating the abundance patterns from the metal-rich end ([Fe/H] ∼ +0.2 at ${R}_{\mathrm{GC}}\,\leqslant 6$ kpc) to the metal-poor end ([Fe/H] ≲ −0.3 at ${R}_{\mathrm{GC}}\geqslant 14$ kpc) of the current Galactic disk. The youth of Cepheids offers a crucial advantage in studying the disk chemical evolution. The radial migration of the disk (Sellwood & Binney 2002) affects the current abundance distribution of various elements for stars in the disk; e.g., see a recent study on the r-process abundance in the solar neighborhood by Tsujimoto & Baba (2019)). Cepheids are, however, considered to be located almost at their birth positions in terms of ${R}_{\mathrm{GC}}$, and they tell us, at least approximately, the current abundances of star-forming gas. Recent large-scale surveys have found a large number of new Cepheids spread in a large range of ${R}_{\mathrm{GC}}$ (Udalski et al. 2018; Chen et al. 2019; Dékány et al. 2019; Skowron et al. 2019), some of which are heavily reddened in the Galactic disk. Near-infrared spectroscopic observations are desired to obtain their abundances.

Bright red giants, roughly ${T}_{\mathrm{eff}}$ ≲ 5000 K and 1 ≲ log g ≲ 2 dex, are often targeted in order to investigate the heavy elements in nearby dwarf galaxies (e.g., McWilliam et al. 2013; Ji et al. 2016; Ji & Frebel 2018). Sr ii and Zr i lines are expected to be strong, while Y ii lines are probably weak, ≲0.05 in depth. Some Eu ii lines may also be seen, but the abundance measurements would require high signal-to-noise ratios (S/Ns), with the typical depths expected to be around ∼0.02 or smaller. Observing the lines of other heavy elements would be even more challenging. Bright red giants are important because they are usually the easiest targets in stellar systems like the Galactic bulge and dwarf galaxies, which are dominated by old stars. Measuring their Eu abundances, if possible, would be very useful for studying the chemical evolution of those systems, together with the s-process abundances traced with Sr and Zr.

At the lower parts of the HRD, log g ≲ 3, only a small fraction of the lines reported in this study are expected to be visible. The three Sr ii lines are expected to still be strong, ≳0.2, for a wide range of ${T}_{\mathrm{eff}}$. In fact, Caffau et al. (2016) detected the three lines and estimated the abundances by taking into account the non-LTE effect. The two Zn i lines may be seen in stars with intermediate temperatures around 5500 K, while Zr i 9822.56 is expected to be visible only in late-type stars with ${T}_{\mathrm{eff}}\lt 5000$ K. None of the other species are likely to be significant unless the abundances are highly enhanced.