Application of Laboratory Atomic Physics to Some Signiﬁcant Stellar Chemical Composition Questions

: This brief review highlights some current issues in Galactic stellar nucleosynthesis, and some recent laboratory studies by the Wisconsin atomic physics group that have direct application to stellar spectroscopy to advance our understanding of the chemical evolution of our Galaxy. The relevant publication history of the lab studies are summarized, and investigations into the abundances of neutron-capture and iron-peak elements in low metallicity stars are described. Finally, new initiatives in near-infrared spectroscopy are brieﬂy explored.


Introduction
Fifteen years ago, the United States National Academy of Sciences suggested a set of "Eleven Science Questions for the New Century" [1]. One consideration deemed worthy of inclusion in that list was, "How Were the Elements from Iron to Uranium Made?" This question sums up the overall goals of many efforts in fields labeled variously as Galactic chemical evolution, stellar nucleosynthesis, and cosmochronology. Most information for these fields must come from stellar surface chemical compositions, with additional contributions from solar system meteoritics, planetary nebulae, H 2 regions, and supernova remnants.
The basic tool for stellar elemental abundance studies is spectroscopy, generally obtained with wavelength resolving power comparable to the widths of stellar absorption lines. Most of the stars investigated for clues to nucleosynthesis are relatively cool, with effective temperatures T eff < 10,000 K. Their spectra are comprised almost entirely of neutral and singly-ionized species of elements with atomic numbers Z ≥ 6 and diatomic molecules mostly involving the HCNO elements. Their surface chemical compositions reflect the combination of elements synthesized and ejected into the interstellar medium by past stellar generations, and elements produced in the interiors of the stars themselves.
Determination of accurate elemental abundances from atomic lines in stellar atmospheres involves many areas of physics and astronomy, including laboratory and theoretical atomic physics, statistical and thermal physics, construction of realistic model stellar photospheres, solutions to the radiation transport problem due to a variety of absorption and scattering processes, details of stellar line formation, etc. Limits on the abundance accuracies can arise from multiple sources, and any weak link in the analysis chain can potentially destroy the usefulness of an element in helping to understand nucleosynthesis. The quest for more accurate stellar abundances must start with the parameters of individual atomic transitions. One must know their quantum properties very well, including their transition probabilities and isotopic and hyperfine substructures, to have any hope of deriving robust abundances from them.
The Vienna Atomic Line Database (VALD; [2] and references therein; http://vald.astro.uu.se/) is a major compilation of atomic transitions that are very useful to stellar spectroscopists. Many VALD data entries are from laboratory sources, but perhaps the majority of transitions come from the vast atomic and molecular line compendium of Kurucz ([3,4]; http://kurucz.harvard.edu/), which contains a mix of laboratory, theoretical, and semi-empirical transition probabilities.
These databases also have estimates of line damping parameters; these are important in deriving accurate abundances for absorption lines with large optical depths. Some progress has been made on practical application of collisional broadening theory (e.g., [5][6][7]), and further effort in this area should pay large dividends, especially in the less-studied infrared spectral domain.
Several atomic physics groups are making major contributions to improving transition data through many theoretical and laboratory approaches. These groups and some representative publications include, but are not limited to, those at: The University of Mons [8,9]; Imperial College London [10,11]; Lund University [12,13]; and Uppsala University [7,14]. Our collaboration from The University of Wisconsin-Madison and The University of Texas at Austin has concentrated on improving these basic line parameters through laboratory experiments, and applying these lab data to classes of stars most relevant to Galactic chemical evolution descriptions. This paper aims to highlight the University of Wisconsin laboratory studies of atomic species most relevant to cool-star spectroscopy.

Laboratory Transition Data
Two important astrophysical element groups have been featured: the Fe-peak elements (21 ≤ Z ≤ 30), and the neutron-capture (n-capture) elements (Z > 30). In Figure 1, we show in periodic table format the elements with published Wisconsin lab g f values, color-coded by the species that have been studied. Except for Xe I [15] and Ni II [16], all of the results include transitions that are readily accessible in the optical (3000-9500 Å) and ultraviolet (UV: 2200-3000 Å) spectral domains of cool stars 1 . There are many individual studies, and here we comment on a few aspects of these lab investigations.
Of primary importance is the ability of stellar spectroscopists to have access to single-source internally consistent transition data for as many lines as possible for a given species. A few decades ago, stellar abundance studies had to rely on a patchwork set of relatively inaccurate lab and theoretical line parameters, or to resort to differential analyses of stellar and solar spectra. Neither of these approaches could produce abundances that are accurate enough to meaningfully compare to stellar nucleosynthesis predictions. Happily, the huge increase in quality and quantity of transition data means that, for most spectroscopically-accessible species, the basic line data are now very small factors in abundance error budgets.
Among the Fe-group elements, accurate transition probabilities have been published recently for Fe I [10,17,18] through the combined efforts of the groups at Imperial College London, Uppsala, and Wisconsin. Almost all of the transitions reported in these papers arise from states with lower excitation energies χ > 2.2 eV. For Fe I lines with smaller excitation energies, the older lab study of O'Brian et al. [19] still provides trustworthy transition probabilities. Combining lab data from these studies and applying them to the solar photosphere and metal-poor stars appears to yield reliable Fe abundances with relatively small line-to-line scatter [20]. Unfortunately, no single comprehensive lab study of Fe II has been published recently; see the summary of sources up to 2016 in [20].
The neutral species for Ti through Ni have been extensively studied [21][22][23][24][25][26], and the ionized species for these elements except Ni (see above footnote) [24,[27][28][29][30]. Application to the solar photospheric spectrum and to the spectrum of the well-studied metal-poor field star HD 84937 results in accurate, 1 Only the Xe I 1470 Å resonance line was investigated by Anderson et al. [15]; seven Ni II lines in the 1450-1750 Å were included by Fedchak and Lawler [16].
internally consistent abundances for each species. More importantly, the abundances for both species (when available) are in agreement (see Figure 7 of [20]), strongly suggesting that the derived Fe-group abundances are true chemical composition markers, relatively unaffected by line formation and stellar atmospheric modeling uncertainties.
An earlier series of lab studies concentrated on near-complete coverage of the ionized species of "rare earth" lanthanides. Only the ionized species of lanthanides are usually detectable in stellar spectra. Despite their relatively small abundances in most stars in comparison to the Fe-peak group, there are many ionized-species lanthanide lines detectable in the blue and near-UV spectrum of cool stars. In Figure 1, it is clear that most Wisconsin lab studies of n-capture elements outside the lanthanide domain have been on neutral-species transitions. In many cases, this mirrors their stellar spectroscopic accessibility. Neutral species lines of these elements mostly are few and far between, and rarely do ionized species lines become detectable.

Some Applications of Laboratory Data to Low Metallicity Stars
The onset of Galactic nucleosynthesis produced stars of very low overall metallicity 2 . Many of these objects exhibit large departures from the solar-system abundance distribution. Among the n-capture elements, there are outstanding examples of stars that have been born with or acquired from a companion extreme enhancements of elements (isotopes) that were generated in fusion environments that were conducive to slow neutron captures (the s-process) or rapid-blast ones (the r-process).
There has been renewed interest in the r-process production site following the recent discovery of gravitational wave event GW170817, so far unique in providing electromagnetic signals as well in multiple wavelength domains. The totality of "multi-messenger" information from GW170817 strongly suggests that this event arose from the merger of two neutron stars [32]. Additionally, photometric and spectroscopic observations [33,34] suggest the presence of r-process-rich ejecta from this event, in particular of unstable lanthanide isotopes. It is important to determine accurate abundances for as many n-capture elements in r-process-rich stars as possible to quantitatively constrain what the output of a binary neutron star merger must match.
The first extremely r-process-rich star, CS 22892-052, was recognized on a noisy high-resolution spectrum and analyzed with weak laboratory data [35]. A total of 54 lines contributed to abundances of 12 n-capture species. Contrast that with the study of the recently-discovered r-process-rich star J09544277+5246414 (J0954+5246) [36], which reported abundances based on 134 lines of 28 species. In Figure 2, we reproduce their Figure 1, a comparison of abundances of that star and scaled solar-system r-process-only abundances. The bottom panel of this figure also includes in the n-capture abundances for CS 31082-001 [37], a very metal-poor r-process-rich star. This Galactic halo member is notable especially for its Th and U abundances, which are larger than the extrapolated scaled solar-system values. CS 31082-001 became the first of such "actinide boost" stars, which now includes J0954+5246.
For the present review, a more important point is the match between the abundance distribution in the lanthanide domain to the solar-system r-process-only abundances. It is near-perfect for the two stars in Figure 2 and in many other r-process-rich stars (e.g., [38,39]). Spectroscopically, this means that atomic data parameters for a large number of detectable n-capture elements are now good enough that they play only a very small part of the overall elemental abundance uncertainties. Astrophysically, this means that the creation of r-process elements has been essentially the same over Galactic history, and possibly limited to one type of production site. For low metallicity stars, the Fe-group element abundances have not had the same level of atomic physics attention as the n-capture elements. It has been known for decades that relative abundances among this element group do not conform exactly to the solar-system distribution [40][41][42] Figure 1); • precise abundances from 17 Fe-peak species (mean line-to-line scatters < σlog > = 0.07), with good agreement (mean log differences of 0.03 dex) between neutrals and ions of seven elements ([20] Figure 7); • little sign of over-ionization of neutral species, which would be a signature of departures from local thermodynamic equilibrium (LTE) in the Fe-group neutral species, which could be a fortuitous result that can be realized as well by a combination of more realistic three-dimensional model atmospheres and detailed non-LTE computations (e.g., [45]); • no evidence for depletion of Cr and enhancement of Co at this metallicity; and • a curious overabundance of the three lightest Fe-peak elements Sc, Ti, and V.
This work is being extended to lower metallicities [46,47] to investigate Fe-peak abundance ratios for any solid evidence of other departures from the solar-system distribution.
Lab studies of Fe-peak elements are continuing. The HD 84937 study detailed above did not include recent transition studies of Sc II, Cr II, Co II, Cu I or Zn I. New data for Cr II [28] yields consistent abundances from neutral and ionized species in the solar photosphere. The same agreement is seen in HD 84937 as long as the anomalously small abundances (0.2-0.3 dex lower than the rest of the Cr transitions) from the Cr I resonance lines at 4254, 4274, and 4289 Å are ignored. This appears to be a clear signal of departures from LTE in the neutral species. We suspect that this may explain the earlier reports of Cr deficiencies in very metal-poor stars, since the Cr I resonance lines have played a major role in most Fe-peak abundance surveys.
Recently, Lawler et al. [29] reported new lab data for Co II transitions, almost all of which arise in the UV spectral region. In Figure 3, we summarize the HD 84937 abundances from that study. Translating these log values into relative abundances with respect to the Sun yields [Co/H] = −2.24 or [Co/Fe] = +0.08, essentially the solar ratio within the observational uncertainties. There is no obvious enhancement in the relative Co abundance at this metallicity. Interestingly, the abundance from Co I lines is 0.08 larger; one looks for over-ionization of an element as evidence for non-LTE effects, but in HD 84937 we derive an under-ionization. This anomaly is being further tested with our more metal-poor stellar sample [46,47]; preliminary evidence supports the notion that Co is not overabundant in very metal-poor stars, i.e., [Co/Fe] ∼ 0 for stars with [Fe/H] ≤ −3.
The vacuum-UV spectral region, λ < 3000 Å, presents the ideal stellar laboratory to make progress in understanding Fe-peak abundances in low metallicity stars. Hundreds of potentially useful neutral and ionized species transitions can be easily identified in the vacuum UV. Unfortunately, the required high-resolution spectra may only be acquired with the Hubble Space Telescope STIS instrument. At present, there are less than about 20 metal-poor ([Fe/H] < −2) stars with good HST/STIS UV spectra. Given the ongoing demand for HST time, one cannot reasonably expect a large expansion of this meager database. A more feasible goal would be to use ground-based large telescopes with good near-UV high resolution spectrographs (e.g., VLT/UVES and Keck/H IRES) to access more Fe-peak transitions in the 3000-4000 Å spectral range.

Extending the Wavelength Domain to the Near-Infrared
The near-infrared spectral domain, 1.0-2.4 µm, is enjoying new high-resolution spectroscopic access, with such telescope/instrument combinations as VLT/CRIRES, SDSS/APOGEE, and McDonald − DCT − Gemini/IGRI NS now able to provide quality spectra of many faint (K ≤ 14) targets in star clusters, the general Milky Way field and the Galactic bulge. Laboratory atomic physics so far has not been able to keep pace in the near-IR, mostly due to intrinsic difficulties in deriving accurate atomic transition probabilities. Upper levels that only radiate in the infrared (IR) are not currently accessible to TRLIF (Time-Resolved Laser-Induced Fluorescence) radiative lifetime measurements. The lack of an IR equivalent to an optical-UV photomultiplier (PMT) is a problem. The combination of PMT gain, electronic bandwidth, and detector area is not yet available in the IR. Progress is needed, and here we describe some aspects of the efforts of the Wisconsin group in this spectral domain.
Lawler et al. [21] reported accurate transition probabilities for nearly 950 Ti I lines, and derived new Ti abundances for the solar photosphere and very metal-poor star HD 84937. However, the absorption strengths of Ti I lines, similar to those of most species, decrease substantially with increasing wavelength, and there are relatively few measurable ones in warm stars such as the Sun and especially in metal-poor stars. Such stars cannot be used to test the parameters of many atomic lines redward of ∼8000 Å. For this task, we have recently turned to the bright, mildly metal-poor ([Fe/H] −0.5) red giant star Arcturus. This star has both excellent high-resolution spectroscopic atlases and careful model atmosphere studies. The reader is referred to [48] for citations to these studies and a discussion of V I transitions in the red spectral region of Arcturus. In Figure 4, we present abundances for 168 Ti I lines in Arcturus from 5200 Å to 2.4 µm. The mean abundance, log (Ti) 4.6, suggests that [Ti/H] −0.4 or [Ti/Fe] +0.1. The transition probabilities for these analyses all come from [21], and in general they yield a consistent Ti abundance. There is a very small upward drift in derived abundances with increasing wavelength, but it may arise from a variety of possible causes. The important conclusion is that for Ti I the optical and near-IR data essentially agree, lending confidence to the lab transition probabilities for the long wavelength lines. Stellar spectroscopists usually focus on using the best available lab g f -values for abundance work. However, as the spectroscopic datasets attain ever-increasing resolution and signal-to-noise quality, some secondary atomic data considerations become important for the most accurate abundances. Odd-Z elements and odd-A isotopes often yield transitions with significant hyperfine structure. Examples of this have been well documented in recent Fe-peak studies [22,[28][29][30]. Less attention has been paid to isotopic displacement of spectral lines of some elements. The isotopic wavelength offsets are undetectably small in the UV and most of the optical spectral region, but they grow with increasing wavelength.
Ni has five naturally-occurring isotopes, three of which have non-negligible percentages in solar-system material: 58 Ni, 68.1%; 60 Ni, 26.2%; and 62 Ni, 3.6%. For Ni I, the presence of significant isotopic splitting for several lines redward of 7000 Å has been demonstrated by Wood et al. [26] (see their Figure 10). Hyperfine and isotopic subcomponents of spectral lines will always act to spread the absorption in wavelength, desaturating the overall feature. Neglect of transition substructure can lead to over-estimation of abundances, and for strong lines this can be a significant effect.
The isotopic broadening of optical Ti I lines is negligible, but detectable in the near-IR. As an example, in Figure 5, we show one near-IR transition in the lab and in the solar photosphere. Isotopic wavelength shifts are not resolved into individual components, but the extra line broadening is evident. As with Ni, this effect can become important in deriving accurate elemental abundances when Ti I lines become strong, as they surely will in cool red giant stellar spectra. Note finally that detection of isotopes in stellar line broadening does not generally lead to much leverage in determining the isotopic fractions for most elements. For Ti, it is more profitable to do isotopic studies with molecular TiO, as in [51].