LINE LISTS FOR THE A²Π–X²Σ⁺ (RED) AND B²Σ⁺–X²Σ⁺ (VIOLET) SYSTEMS OF CN, ¹³C¹⁴N, AND ¹²C¹⁵N, AND APPLICATION TO ASTRONOMICAL SPECTRA

Renewed attention to the CNO group in the solar photosphere has led to significant revisions in the abundances of all three elements. The comprehensive summary by Anders & Grevesse (1989) recommended values of log ε(C)¹⁰ = 8.56, log ε(N) = 8.05, and log ε(O) = 8.93, which were adopted by stellar spectroscopists for more than a decade. Fundamental re-examination of these values began with Allende Prieto et al. (2001), who derived a much smaller photospheric oxygen abundance from the reliable 6300.30 Å [O i] line: log ε(O) = 8.69 ± 0.05. Their study featured in-depth "three-dimensional time-dependent hydrodynamical" solar atmosphere modeling, but much of the downward abundance revision came from renewed attention to the contamination of the [O i] feature by a Ni i line at 6300.34 Å. Further studies of the photospheric O abundance (e.g., Caffau et al. 2008; Scott et al. 2009) have reached qualitatively similar conclusions. Additionally, studies of N and C features in the solar spectrum (see below) have resulted in more modest decreases in those elemental abundances. The recent review by Asplund et al. (2009) recommends adoption of photospheric abundances of log ε(C) = 8.43, log ε(N) = 7.83, and log ε(O) = 8.69, all with suggested uncertainties of ±0.05.

We employed the Holweger & Müller (1974) solar photospheric model to maintain consistency with our previous solar abundance studies that employ new atomic line data (Wood et al. 2014 and references therein) and molecular line data (Hinkle et al. 2013; Ram et al. 2014). We adopted a microturbulent velocity of v_micro = 1.1 km s⁻¹, but most molecular features in the solar spectrum are relatively weak, thus not much affected by the microturbulent velocity choice.

In principle the CNO abundances must be determined iteratively, since C and O are coupled through formation of the CO molecule, and CN features are contaminants for the [O i] lines at 6300, 6363 Å. However, molecular equilibrium computations suggest that, in line-forming regions (τ ∼ 0.5) of the solar photosphere, the CO partial pressure is small compared to those of the atomic constituents: p(CO) ≈ 0.02p(C) ≈ 0.01p(O). Thus, CO exerts negligible influence on the number densities of either atom. Additionally, CN features are detectable in the 6300 Å region, but they are very weak. To a good approximation, the CNO abundances can be determined independently for the solar photosphere.

Oxygen. Only the 6300 Å line was employed. This transition provides the primary O indicator for all stars discussed in this paper. Details of synthetic/observed spectrum matches of the [O i] feature have been discussed in other papers and will not be repeated here. We adopted transition probabilities for the [O i] and Ni i lines recommended by Caffau et al. (2008), and with them estimated log ε(O) = 8.69, identical to the value for this line derived by Scott et al. (2009) and very close to the value from multiple O abundance features recommended by Caffau et al.

Carbon. Our primary C abundance indicator is the C₂ d³Π_g­–a³Π_u Swan system. It has three detectable vibrational band sequences: Δv = +1, culminating in the blueward-degraded (1–0) P-branch head at 4737 Å; Δv = 0, with the (0–0) head at 5165 Å, and Δv = −1, with the (0–1) head at 5635 Å. We synthesized the solar spectrum in 20 Å intervals near the (0–0) and (1–0) bandheads, using the line lists developed by Brooke et al. (2013) and Ram et al. (2014). For all of our solar computations, we assumed the solar-system isotopic ratio ¹²C/¹³C = 90 (Wielen & Wilson 1997). In panels (a) and (b) of Figure 1 we show synthetic and observed spectra for small portions of these two bandhead regions. The observed spectra are part of the solar flux atlas of Wallace et al. (2011). Clearly, the individual C₂ lines are very weak, and only the pileup of lines at the bandhead has substantial absorption depth. The average synthesis/observation matches to these spectra yielded log ε(C) = 8.53 and 8.48, respectively. We also synthesized the prominent CH A²Π–X²Δ G-band in the 4270–4330 Å interval, using the Masseron et al. (2014) molecular data. We began with their original transition table quantities and translated them into a table of values suitable for input to the MOOG spectrum synthesis code in the manner described in Section 2. In panel (c) of Figure 1 we show a small section of the CH spectrum. The lines here are very strong, but they yield consistent abundances, the mean of which is log ε(C) = 8.48, in accord with the results from C₂.

Zoom In Zoom Out Reset image size

Our final C abundance gives most weight to the C₂ 5165 Å bandhead and CH: log ε(C) = 8.48. This value is only 0.05 dex larger than that recommended by Asplund et al. (2009) and is consistent with that derived by Caffau et al. (2010): log ε(C) = 8.50 ± 0.06.

Nitrogen. With our new line data we synthesized selected CN-rich regions in the solar spectrum. The CN B²Σ⁺–X²Σ⁺ violet system has well-known Δv = 0 bands led by the (0–0) and (1–1) bandheads at 3883 and 3871 Å, and Δv = −1 led by the (0–1) head at 4215 Å. In panel (a) of Figure 2 we show a small portion of our synthetic spectrum match to the (0–0) and (1–1) bandheads. There is much contamination by atomic species in this crowded spectral region, but the CN features consistently suggest log ε(N) = 8.08. We also synthesized the (0–1) bandhead, but it is very weak and blended with a strong Sr ii resonance line. We estimate from this band that log ε(N) = 7.93, but do not give this result much weight.

Zoom In Zoom Out Reset image size

The CN A²Π–X²Σ⁺ red system has detectable absorption in the solar spectrum over a broad spectral interval, from at least 5800 Å to beyond 2 μm. In panels (b) and (c) of Figure 2 we show just two examples. The 8000 Å spectral region has been observed in many cool giant stars because it is the easiest wavelength region from which to estimate ¹²C/¹³C ratios. The 10970 Å spectral region was chosen at random from among many longer wavelength intervals that show strong CN absorption. The derived abundances are log ε(N) = 8.13 and 8.03 from the CN lines shown in panels (b) and (c) of Figure 2, and other regions show little N abundance variation. From all of our syntheses, we suggest a mean abundance of log ε(N) = 8.05.

Our solar N abundance is somewhat larger than the value of log ε(N) = 7.83 recommended by Asplund et al. (2009), and log ε(N) = 7.86 ± 0.06 derived by Caffau et al. (2009). Both of those studies relied primarily on analyses of N i lines. For the absolute solar N abundance it is preferable to analyze transitions from the neutral atom, which is the dominant species in the N equilibrium. CN is a trace species in line-forming regions of the solar photosphere: our molecular equilibrium computations show that p(CN)/p(N) ∼ 10⁻⁴. Additionally, of course, any abundance derived from CN can be determined no better than the assumed C abundance. Combined with our standard modeling assumptions, it is obvious that caution should be observed in interpreting our derived mean N abundance. However, we have shown here that a variety of CN features over a very large wavelength range yield consistent abundances. CN is the molecule that provides most of the N abundances for other stars reported in the literature.

Arcturus is a very bright mildly metal-poor ([Fe/H] ∼ −0.5) red giant star with extensive spectral atlases (Griffin 1968; Hinkle et al. 2000; Hinkle & Wallace 2005).¹¹ This star has been subjected to several comprehensive abundance analyses (e.g., Mäckle et al. 1975; Peterson et al. 1993; Ramírez & Allende Prieto 2011). Special attention has been paid to the CNO elements, beginning with the discovery of an anomalously low C isotopic ratio by Lambert & Dearborn (1972): ¹²C/¹³C = 6.7 ± 1.5. This result has been revised slightly upward in subsequent studies, e.g., Lambert & Ries (1977; ¹²C/¹³C = 7.2), Pavlenko (2008; ¹²C/¹³C = 8 ± 0.5), and Abia et al. (2012; ¹²C/¹³C = 9 ± 2). However, the essential result of a very low isotopic ratio for Arcturus has remained unchanged for four decades. Recently, Abia et al. (2012) summarized investigations of the Arcturus spectrum with special emphasis on the CNO group, and determined new values for this star's C and O isotopic ratios. They found ¹²C/¹³C = 9 ± 2 and log ε(C) = 8.06 ± 0.09, assuming log ε(N) = 7.67 ± 0.13 from Ryde et al. (2009), and log ε(O) = 8.76 ± 0.17 from Ramírez & Allende Prieto (2011).

For our computations we adopted the Arcturus model atmosphere parameters derived by Ramírez & Allende Prieto (2011): T_eff = 4286 K, log g = 1.66, [Fe/H] = −0.52, v_micro = 1.74 km s⁻¹ as well as their individual abundances for elements other than the CNO group. These parameters were employed to generate a model atmosphere from the Kurucz (2011)¹² grid, using software kindly provided by A. McWilliam and I. Ivans. Models for the other stars are also interpolated from the Kurucz grid. The observed spectrum is the NOAO Arcturus Spectral Atlas (Hinkle et al. 2000). To derive the abundances, we used the line lists described in Section 3 or developed lists for new spectral regions in the same manner. We initially assumed ¹²C/¹³C = 7 in the synthetic spectrum calculations; further comments on the isotopic ratio are given below.

Oxygen. In panel (a) of Figure 3 we show synthetic and observed spectra of the [O i] 6300 Å line. Transitions of CN can be easily identified in this panel by their strength responses to changes in O abundance. Specifically, a smaller O abundance decreases the 6300.30 Å [O i] absorption, and also decreases the number of C atoms that are tied up in CO. For typical line forming atmospheric depths (τ ∼ 0.5) in Arcturus, using our final CNO abundance set, we estimate that p(C)/p(CO) ∼ 0.5 and p(O)/p(CO) ∼ 5. A decrease in O liberates C from CO, resulting in more CN formation. This leads to increases in the strengths of prominent nearby CN features seen at 6297.5, 6298.6, 6304.1 Å, and in the many weaker transitions of this molecule. All elements of the CNO group take part in producing this and other spectral regions of Arcturus. Therefore, it is necessary to derive the CNO abundances iteratively until satisfactory observed/synthetic matches are achieved. Iteration is needed especially to match the 6300.30 Å [O i] transition, which is contaminated both with the Ni i 6300.34 Å line discussed in Section 3.1 and a CN line at 6300.27 Å that can be ignored in the solar photosphere but not in Arcturus. Our final O abundance is log ε(O) = 8.63, well within the uncertainty suggested by Abia et al. (2012). We also synthesized the weaker 6363.78 Å [O i] line and derive an O abundance in excellent agreement with that derived from the 6300 Å line. However, there is significant CN absorption contaminating this [O i] feature, and this result should be treated with much caution.

Zoom In Zoom Out Reset image size

Carbon. We again used the C₂ (0–0) and (0–1) bandheads as primary C abundance indicators, and used the CH G band to confirm this result. Observed and computed C₂ spectra are illustrated in Figure 3 of Ram et al. (2014). From both of these bandheads we derived log ε(C) = 8.02, in good agreement with Abia et al. (2012). The CH lines are very strong and contaminated with many other strong atomic lines. Nevertheless, we estimate log ε(C) ∼ 8.0 from CH, in accord with the more trustworthy result from C₂.

Nitrogen. The CN red system produces detectable features in all spectral regions redward of 6000 Å in Arcturus. We surveyed the CN absorption in several large spectral intervals: 6000–6100, 6280–6380, 8000–8100, and 10900–11000 Å. A small portion of the last of these regions is shown in panel (b) of Figure 3. All CN red system transitions considered here yield very similar abundances, the mean of which is log ε(N) = 7.66, with band-to-band scatter approximately ±0.05 dex. This result is supported by syntheses of the 4215 Å violet CN (0–1) bandhead, which we illustrate in panel (c) of Figure 3. In this very complex region it is difficult to find any clean CN transitions, but from consideration of the total absorption we derive log ε(N) ∼ 7.6, matching the more solid result from the CN red system, and in good agreement with the value recommended by Abia et al. (2012).

¹²C/¹³C. Transitions of ¹³C¹⁴N are easy to detect in all spectral regions with relatively strong CN lines, λ ⩾ 8000 Å. Three small wavelength intervals with prominent ¹³CN features are displayed in Figure 4. Panel (a) of this figure includes the triplet of ¹³CN lines that have been employed most often in C isotopic studies of red giant stars. The C, N, and O elemental abundances are those derived above (Table 3). We have not performed a detailed analysis of these features, but our syntheses always suggest that ¹²C/¹³C = 7 ± 1, in accord with the literature sources cited above.

Zoom In Zoom Out Reset image size

HD 122563 is the only very metal-poor star in the Bright Star Catalog (Hoffleit & Jaschek 1991). As such, its spectrum has been scrutinized at high resolution many times, beginning with Wallerstein et al. (1963). Sneden (1973) reported the first C and N abundances, finding [C/Fe] ∼ −0.5 and [N/Fe] ∼ +1.2. Following the assertion by Conti et al. (1967) that [O/Fe] > 0 in very metal-poor stars, Lambert et al. (1974) detected the [O i] line in HD 122563, deriving [O/Fe] ≈ +0.6. Subsequently, Lambert & Sneden (1977) used CH G-band lines to deduce ¹²C/¹³C = 5 ± 2, near the CN-cycle equilibrium value of 3–4 (e.g., Caughlan 1965). The HD 122563 abundances have been confirmed by subsequent studies. The most recent complete CNO determination is that of Spite et al. (2005), who derive [Fe/H] = −2.82, [C/H] = −3.29, [N/H] = −2.12 from CN or −1.72 from NH, and [O/H] = −2.20. Their CNO abundances translate to log ε(C) = 5.55, log ε(N) = 5.80 or 6.20, and log ε(O) = 6.77. The very low ¹²C/¹³C ratio in HD 122563 has also been confirmed in other investigations, e.g., Sneden et al. (1986) and Spite et al. (2006). The phenomenon of anomalous CNO-group abundances and C isotopic ratios in low-metallicity evolved stars (especially globular cluster giants) is now solidly established.

For our analysis we used an interpolated model with T_eff = 4600 K, log g = 1.30, [Fe/H] = −2.70, v_micro = 2.5 km s⁻¹, close to the mean of parameters for HD 122563 in papers gathered in the SAGA database (Suda et al. 2011).¹³ The extreme metal deficiency leads to very few CNO-containing spectroscopic features that are available for abundance analysis. The spectrum used for our study was obtained with the Tull Echelle Spectrograph (Tull et al. 1995) of McDonald Observatory's 2.7 m Smith Telescope. The spectrograph resolving power is R = 60,000.

Oxygen. Our synthesis of the [O i] 6300 Å transition is displayed in panel (a) of Figure 5. The Ni i and CN contaminants are undetectably weak in HD 122563; only O contributes at 6300.3 Å. We derived log ε(O) = 6.73, recovering the results of previous studies. We also detected the very weak 6363 Å [O i] line. Its absorption at line center is only 1% of the continuum, but our syntheses yield an abundance that is in agreement with that determined from the 6300 Å line.

Zoom In Zoom Out Reset image size

Carbon. The C₂ Swan bands are undetectably weak in this very low metallicity star; we must rely only on the CH G band for the C abundance. We employed the same CH line list that was used for the Sun and Arcturus syntheses. A portion of synthesis/observation matches near the 4300 Å bandhead for HD 122563 is shown in panel (b) of Figure 5. A single C abundance works very well throughout the 4225–4375 Å spectral range: log ε(C) = 5.21.

Nitrogen. The CN red system is undetectably weak in the HD 122563 spectrum, and the very weak CN violet system (0–1) bandhead is lost among strong atomic absorption lines. The sole opportunity for deriving an N abundance lies with the 3883 Å (0–0) and 3871 Å (1–1) bandheads. Our observed/synthetic spectrum match of (0–0) lines is shown in panel (c) of Figure 5. Only very close to the bandhead can the CN absorption be established reliably. The derived abundance, log ε(N) = 5.85, also provides a good match to the very weak (1–1) bandhead in this star.

¹²C/¹³C. Test syntheses of CH spectral regions away from the bandhead (near 4230 and 4370 Å), where the ¹³CH and are significantly offset in wavelength from the ¹²CH lines, confirm previous work cited above that ¹²C/¹³C ≈ 5. The CN bands are too weak to permit detection of the ¹³C¹⁴N features.

The final abundance set for HD 122563 given in Table 3 is typical for ordinary evolved low-metallicity stars in the Galactic halo. Large-sample studies typically suggest that 〈[O/Fe]〉 ≈ +0.4 to +0.8 (e.g., Gratton et al. 2000; Spite et al. 2005); our value of [O/Fe] ≈ +0.7 equals the mean of the Spite et al. sample. The low C and high N abundances are also in accord with other work. We derive [N/C] ≈ +1.2 for HD 122563, while Gratton et al. obtain 〈[N/C]〉 ≈ +1.0 for the upper red giant branch stars in their sample, and Spite et al. obtain 〈[N/C]〉 ≈ +1.4 for their chemically mixed stars.

HD 196944 and HD 201626 are two well-studied CEMP stars. Their low metallicities and large [C/Fe] ratios lead to dominance by C-containing molecular transitions. These CEMP stars should present good tests of the new C₂ and CN line lists. Spectra were obtained with HIRES at Keck Telescope, with a resolving power of 48,000 and 103,000 for, respectively, HD 196944 and HD 201626.

Začs et al. (1998) performed an extensive study of HD 196944, and recommended T_eff = 5250 K, log g = 1.7, [Fe/H] = −2.45, and v_micro = 1.9 km s⁻¹. This star was included also in the CEMP high-resolution abundance analyses of Van Eck et al. (2003) and Aoki et al. (2002). All three investigations agree that the neutron-capture elements are very overabundant, e.g. [La/Fe] ∼ [Ba/Fe] ∼ +1.0. In our work we adopt the model atmosphere parameters of Aoki et al.: T_eff = 5250 K, log g = 1.8, [Fe/H] = −2.25, and v_micro = 1.7 km s⁻¹. Those authors derived [C/Fe] = +1.2 and [N/Fe] = +1.3.

Vanture (1992a, 1992b, 1992c) studied the CNO element group in HD 201626, suggesting atmospheric parameters T_eff = 5190 K, log g = 2.25, [Fe/H] = −2.1, and v_micro = 2.0 km s⁻¹. Van Eck et al. (2003) adopted these model parameters. Both studies determined very large neutron-capture overabundances for this star, e.g. [La/Fe] ∼ +1.8. However, the V − K color for HD 201626 in the SIMBAD¹⁴ stellar database is 2.42, indicative of a lower temperature than HD 196944, for which V − K = 1.87. If both of these stars are unreddened, then the temperatures derived from the Ramírez & Meléndez (2005) color calibration suggests T_eff ≈ 4700 K for HD 201626 and 5200 K for HD 196944. Using the Schlegel et al. (1998) or Schlafly & Finkbeiner (2011) extinction maps¹⁵ we derive total Galactic reddening estimates E(B − V) ≈ 0.15 or E(V − K) ≈ 0.41 for HD 201626; the numbers are 0.04 and 0.11 for HD 196944. Then if both stars are suffering the total Galactic sightline reddening, their temperatures are really T_eff ≈ 5050 K and 5420 K, respectively. But these objects do not reside at the edge of the Galaxy, and in particular the parallax of HD 201626 suggests that it is roughly only 250 pc away. Therefore we adopted here a compromise T_eff = 4800 K for HD 201626, and used the Padova isochrones (Girardi et al. 2010) to estimate log g = 1.3 for a low-metallicity star of this temperature. The other atmospheric parameters were assumed to be approximately [Fe/H] = −1.9 and v_micro = 2.0 km s⁻¹. We verified that the adopted metallicity is appropriate for this star by synthesizing the 6210–6270 Å spectral region, which has many strong Fe-group transitions and relatively weak CN absorption. This star is in need of a new comprehensive atmospheric analysis in the future.

Carbon. CEMP stars usually have [C/Fe] ≫ 0, so their C abundances can be determined independently of the assumed O abundances. We determined these abundances from the (0–0) and (0–1) C₂ bands and the CH G band. In Figure 6 we show observed and computed spectra of C₂ lines in the 5148–5168 Å region. Dominance of C₂ can be easily seen in the relative weakness of the Mg i, Fe i blend at 5167.4 Å, which overwhelms the C₂ absorption in K-giant stars with ordinary CNO abundances (see Figure 3 of Ram et al. 2014). The C₂ lines are much stronger in HD 201626 than in HD 196944, due to its 450 K lower T_eff and its larger C abundance; C₂ absorption is a sensitive function of both of these parameters. All C-containing features in both stars yield internally consistent abundances, and we obtain log ε(C) = 7.36 for HD 196944 and log ε(C) = 7.88 for HD 201626.

Zoom In Zoom Out Reset image size

Oxygen. The 6300 Å [O i] transition is detected in both stars, but it is only about 3% deep in HD 196944. The line suffers some blending by a telluric O₂ feature. We were able to cancel the contaminant to first order, but we also compared our synthetic spectra to observed spectra with and without the O₂ feature, and derived a similar O abundance in both cases. The [O i] line is deeper (∼8%) in HD 201626, and is free from neighboring O₂ contaminants. We derived log ε(O) = 7.03 for HD 196944 and 7.28 for HD 201626. These values, translated into [O/Fe] ratios (Table 3) are consistent with the mean [O/Fe] overabundances of halo red giant stars as discussed in Section 3.3. The large production of C by the progenitors of these stars was not accompanied by any enhancement of O.

Nitrogen. The CN violet 4215 Å bandhead provides the most reliable N abundance in both stars. Synthetic and observed spectra for this spectral region are shown in Figure 7. For HD 196944, the 3883 Å band provides corroborating evidence, but the CN red system lines are too weak on our spectra to be useful. For HD 201626, the 3883 Å band is extremely strong, and provides no useful abundance information. However, in compensation, the CN red system is easily detected in this star. From the 4215 Å bandhead we derived log ε(N) = 6.90, while from CN lines scattered throughout the 6200–6400 Å region the N abundances ranged from 6.95 to 7.05. The final abundance in Table 3 reflects our greater confidence in the CN red transitions.

Zoom In Zoom Out Reset image size

¹²C/¹³C. Many of the branches of CH, C₂, and CN band systems have significant isotopic wavelength shifts, and our CEMP stars have such high relative C abundances ([C/Fe] ≈ +1.35, Table 3) that ¹²C/¹³C may be estimated from many spectral features. Ram et al. (2014) suggested that the C₂ Swan (0–0) bandhead region could be used to estimate the isotopic ratios for CEMP stars; see their Figure 1. Inspection of the gaps between the main ¹²C₂ lines reveals the presence of ¹²C¹³C lines when ¹²C/¹³C is low. The same effect can be seen here in Figure 6. The C₂ absorption is much less in HD 196944 than in HD 201626, but the strength ratio of the main and minor absorptions is different between the two. The CH features near 4230 and 4370 Å were used also for the CEMP stars. All syntheses yielded the same results of ¹²C/¹³C ≈ 6 for HD 196944 and 25 for HD 201626.

Our new results are in reasonable agreement with previous studies of HD 196944; Začs et al. (1998) derived [C/Fe] = +1.4, in good agreement with our value in Table 3. Their O abundance is much larger than ours: [O/Fe] = +1.1. However, their value was derived from an LTE analysis of the high-excitation O i triplet. Many studies have shown that severe departures from LTE attend the formation of the triplet lines in low metallicity stars: Fabbian et al. (2009) suggest that calculated [O/Fe] LTE abundance ratios are too large by 0.5–0.9 dex. If so, our much lower abundance deduced from the [O i] line is consistent with their result.

For HD 201626, our absolute CNO abundance scale is about 0.5 dex lower than that of Vanture (1992b). However, his abundance ratios are in better agreement with ours. From his Table 3, ε(C)/ε(O) = 4.0 and ε(C)/ε(N) = 5.6, while the data from Table 3 yield ε(C)/ε(O) = 4.0 and ε(C)/ε(N) = 8.0. The difference in C/N ratio may simply be due to a difference in CN line lists, and is not large enough to pursue further. Additionally, we concur with Vanture's (1992a) assessment that ¹²C/¹³C ⩾ 25 for this star.