Detailed Abundances in the Galactic Center: Evidence of a Metal-rich Alpha-enhanced Stellar Population

In order to synthesize the spectra, an accurate line list is required. In the list provided by Thorsbro et al. (2018b) wavelengths and line strengths (astrophysical log gf-values) are updated using the solar center intensity atlas (Livingston & Wallace 1991). We have also made use of recent laboratory measurements of atomic line strengths of Mg I and Sc I (Pehlivan et al. 2015; Pehlivan Rhodin et al. 2017), as well as of Si I (Pehlivan Rhodin et al., 2020) in the K-band. Of approximately 700 identified, interesting spectral lines for cool stars, about 570 lines have been assigned new values.

As molecular lines have a crucial impact in terms of blending, we include molecular lines in the line list. For CN we include the list from Sneden et al. (2016), which is the most dominant molecule apart from CO, whose lines dominate in the 2.3 μm band head region. The CO line data are from Goorvitch (1994). At the shorter wavelengths of our spectra region, SiO, H₂O, and OH are important. Line lists for these molecules are included in our line list from, respectively, Langhoff & Bauschlicher (1993), Barber et al. (2006), and Brooke et al. (2016).

To identify clean spectral lines suitable for measurement, we examine synthetic model spectra of cool M giants to identify iron lines that are not blended with molecular features. The CN line list used (Sneden et al. 2016) has been shown to reproduce the observed CN spectrum very well, and can be used to identify blends. Unknown blends might still always be present and special care has been taken when such possible mismatches are identified. Examples of lines that are not blended are illustrated in Figure 3, where an Fe I line and a Si I line are shown.

Zoom In Zoom Out Reset image size

For the determination of Fe and Si abundances the lines shown in Table 2 have been identified as good lines with little or no blending.

The effective temperatures of the stars in the MWD are obtained following Thorsbro et al. (2018a), where the temperatures from the APOGEE Data Release 14 (Blanton et al. 2017; Majewski et al. 2017; Abolfathi et al. 2018) were examined and then adopted.

To determine the temperatures of the stars in the Galactic center we rely on the method of Schultheis et al. (2016) as employed in Rich et al. (2017), which is based on using low-resolution SINFONI (Eisenhauer et al. 2003) spectra and measures an index of the CO band head at 2.3 μm. Rich et al. (2017) determined the effective temperatures for 17 of the 20 Galactic center stars in our sample. The same method is used for determining the temperature for the added stars, namely GC 8623 and GC 13727. We do not have SINFONI spectra for the GC 25 star and thus it remains undetermined using the CO band head method. The CO band head temperatures are shown in Table 3 in the "CO band" column.

To decrease the uncertainty in the determined temperatures for the stars in the Galactic center and to increase the correlation between the temperatures of the stars in the MWD stars and the stars in the Galactic center, we use the fact that there are strong Sc I lines in the K-band spectrum that are very temperature sensitive in the 3000–4000 K range (e.g., Thorsbro et al. 2018a). The effects of different scandium abundances, surface gravities, or metallicities are a second-order effect. Most of our observed stars are within this temperature range and the strengths of these lines should therefore be able to be used as a temperature indicator. We measure the equivalent widths for a set of scandium lines in the MWD stars for which the effective temperatures are known. We create four linear relations between the effective temperature and the equivalent widths of the four scandium lines at the following wavelengths in air: 21730 Å, 21812 Å, 21842 Å, and 22394 Å. Using the equivalent widths measured of the same scandium lines in the spectra of the Galactic center stars together with these empirical relations, and taking the average of the four results (having standard deviations on the order of 50 K), we can thus determine the effective temperature of the Galactic center stars. The effective temperatures determined this way are given in Table 3 under the column "Scandium."

Comparing the effective temperature values found in Table 3, and with a general uncertainty of 150–200 K, we find that the results from the different methods are in general agreement. We decide to take an average of the values, and keep a conservative estimate of 150 K uncertainty for the effective temperature values. These temperatures are given in the "Mean" column in Table 3. For GC 25, we use the temperature determined from the Sc I lines.

Rich et al. (2017) demonstrated that in spite of the distance to the stars in the Galactic center being known to a good degree, the photometrically determined surface gravities are plagued with large uncertainties due to the patchy extinction. To determine surface gravity, $\mathrm{log}g$, we instead benefit from the fact that isochrones of old (>1 Gyr) M giants are quite insensitive to the age of the star. This reduces isochrones to depend on surface gravity, effective temperature, and metallicity. We thus use a grid of Yonsei–Yale isochrones (Demarque et al. 2004) to determine $\mathrm{log}g$ given an effective temperature and metallicity. The construction and validation of this method is described in more detail in Rich et al. (2017).

The microturbulence, ξ_micro, that takes into account the small scale, nonthermal motions in the stellar atmospheres, is important for saturated lines, influencing their line strengths. We estimate this parameter from an empirical relation with the surface gravity based on a detailed analysis of spectra of five red giant stars (0.5 < log g < 2.5) by Smith et al. (2013), as described in Rich et al. (2017).

The estimated uncertainty of 150 K for the effective temperatures propagates into an uncertainty of 0.3 for $\mathrm{log}g$ and 0.15 km s⁻¹ for ξ_micro.

The metallicities of our stars are important not only for identifying the model atmosphere for a certain star but also for the metallicity distribution function (MDF), which is an important diagnostic for the discussion of the origin of the nuclear star cluster, see Section 4.

In order to determine metallicities we synthesize spectra and compare them to the observed spectra. We have chosen to use the spectral synthesis code Spectroscopy Made Easy (SME; Valenti & Piskunov 1996, 2012), which interpolates in a grid of one-dimensional (1D) MARCS atmosphere models (Gustafsson et al. 2008). These are hydrostatic model atmospheres in spherical geometry, computed assuming LTE, chemical equilibrium, homogeneity, and conservation of the total flux (radiative plus convective, the convective flux being computed using a mixing-length recipe). This code has the advantage that it includes a flexible χ² minimization tool for finding the solution that fits an observed spectrum the best in a prespecified spectral window. It also has a powerful continuum normalization routine. In cool star spectra, extra care is needed to normalize the spectrum in the region of a spectral line under consideration. The ubiquitous molecules and unavoidable residuals from the telluric line division necessitates a careful analysis line by line.

We recognize that using 3D atmospheric models would be an improvement over 1D atmospheric models (Scott et al. 2015b, 2015a). However, for this differential study we expect to see the same systematics to be present in both the Galactic center and the MWD stellar samples.

As a result of our updated temperatures and careful work on the line list, especially the new analysis of lines blended with CN, we reanalyze the [Fe/H] values given in Rich et al. (2017). The new [Fe/H] values are shown in Table 1, with a general estimated uncertainty of 0.15 dex. A comparison of our new [Fe/H] distribution function (MDF) with that of Rich et al. (2017) is shown in Figure 4.

Zoom In Zoom Out Reset image size

Principally due to our new treatment of CN blending, our iron abundances have changed compared to earlier work; for example, the most metal-rich star now is [Fe/H] = +0.5, compared to +0.64 in Rich et al. (2017). We consider the present abundance distribution reported in this work to be the best current measurement of [Fe/H] and its range in the NSC.

The α-element abundances can potentially be measured from several lines from different species. However, considering the line strengths and number of unblended lines, there are only a few elements possible in the available wavelength region at the spectral resolution considered. The available Mg I lines are heavily blended with CN molecular lines and the strengths of the Ti I and Ca I lines increase conspicuously with cooler effective temperature, broadly similar to the Sc I lines discussed in detail by Thorsbro et al. (2018a). This behavior argues strongly for non-LTE effects affecting the line strengths more than the abundances. Since no non-LTE corrections are available for these elements in the temperature range that our stars span, these lines cannot yet be used for determining the abundances. Si I, however, has several suitable lines and most importantly, non-LTE departure coefficient calculations show that departure from LTE is not important for Si I lines in the stellar parameter space of the stars in our sample. We verified this by examining the departure coefficients in MARCS model giant stars, using the atomic model presented in Amarsi & Asplund (2017), which utilizes the rather efficient inelastic hydrogen collisions of Belyaev et al. (2014).

The α-element abundance is therefore determined from the unblended Si I lines available (Table 2) with a general estimated uncertainty of 0.15 dex. Typical spectra are shown in Figure 6. The silicon abundances are shown in Figure 5 with the Galactic center stars marked with red triangles and squares, the latter for those not identified to be in the NSC (see Section 3.6), and the MWD stars are marked with blue disks.

Zoom In Zoom Out Reset image size

We now examine more closely those metal-rich stars with Si enhancement. Figure 6 shows our observed spectra for the metal-rich subset giving an extra sanity check on the abundance results. The plot indicates stronger silicon features when comparing the Galactic center stars to the MWD stars, but the Fe I line does not show the strong difference we see in the Si I line. This figure shows that, independent of any abundance analysis, we can see that Si appears to be enhanced in the Galactic center for this subset of stars.

Zoom In Zoom Out Reset image size

We examine the stellar dynamics of the three stars added to our data set to verify if they belong to the NSC. Second, we also check with "metal-rich" stars, [Fe/H] > 0.1, that differ in their position or dynamics.

As in Ryde et al. (2016a) and Rich et al. (2017) we use the H–K color, which is nearly only caused by extinction for testing whether the stars are currently in the NSC in the line-of-sight direction. We find that the ${v}_{\mathrm{rad}}^{\mathrm{LSR}}$ for the added stars GC 25, GC 8623, and GC 13727 are, respectively, 83, −20, and −125 km s⁻¹ from the Keck spectra. These velocities are consistent with the lower precision values in Fritz et al. (2016). For the proper motions we use Fritz et al. (2016) as we did previously. The new stars are all likely members of the NSC; see Figure 7.

Zoom In Zoom Out Reset image size

As is visible in Figure 7 we do not find differences between the three added stars or the different metallicity groups. All stars seem to belong to the same hot system with a weak rotation aligned with the Galaxy. The sample, especially of stars with proper motions, is still too small to detect minor differences. We note that three stars, GC 10812, GC 7104, and GC 11473, have a preferred location in the NSD, of which the latter two are in the metal-rich set. Two of the stars, GC 6227 and GC 10195, could possibly be background stars, depending on if they are intrinsically reddened. While effects like motion can lead to exchange between the NSC and the NSD, it is still prudent to account for the possibility that some of these stars may reside in the NSD when explaining the enrichment pattern. However, there is still a need for a Galaxy model that includes all four central components (Sgr A*, the NSC, the NSD, and the bar/bulge) to better constrain the location of the stars. Mixing of stars is also a concern to take note of, at the projected distance of 2 pc from Sgr A* there may be about one NSD star for every three to four NSC stars (Gallego-Cano et al. 2020).

In the line of sight it seems that there could be a larger metallicity range in the NSD than in the NSC, since, of the three stars there, none are normal metallicity stars. Given the low number of stars, this could be possible by pure chance since Do et al. (2015) and Feldmeier-Krause et al. (2017) find metal-poor stars that are at least projected NSC members.

Assuming that the enhancement in Si is also present in other α-elements, the origin of the NSC populations is even more intriguing and we can speculate on possible formation and chemical enrichment scenarios. Some of the key questions for this investigation is ages of stars and the source of gas for star formation.

As per the discussion on Terzan 5 above, the subsolar and supersolar metallicity stars of the NSC can be seen as two groups of stars, and it would be interesting to ponder if there is a clear age dichotomy between them. Unfortunately, isochrone dating of the NSC is not possible because currently we can only observe bright giants near the tip.

We use chemical evolution models (Matteucci 2012; Grieco et al. 2015) to explore the chemical enrichment history of the NSC as traced by the subsolar metallicity stars with about solar-scaled alphas and supersolar metallicity stars with enhanced α-elements. The models assume that at the very early epoch of the galaxy and bulge formation a star formation episode occurred with punctual (and probably local) high star formation rates, i.e., mini-starburst activity. The type-II supernovae formed during such a burst would then inject α-elements within a few megayears The depth of the gravitational potential well of the Galactic center could help retain or rapidly reaccrete this enriched gas (Emsellem et al. 2015), and possibly form the population presented above, before the outbreak of SNe Ia (that would lower the [α/Fe] content of the interstellar medium). This early and massive star-forming episode could be followed by a recent starburst that occurred a few gigayears ago (Pfuhl et al. 2011; Nogueras-Lara et al. 2019a, R. Schödel et al. 2020, in preparation), but the origin of the gas and its retention in the NSC remain a challenge.

We first explore a model where a starburst is initiated by a sudden accretion of gas, either primordial or slightly enriched, which results in a dilution of chemical abundances (e.g., Fe). The model is illustrated in Figure 9, where the dilution of chemical abundances appears on the plot as a horizontal line followed by an increase of the [α/Fe] ratio. This is due to the contribution of core-collapse SNe II in the starburst, followed by a decrease due to the Fe production from SNe Ia. This behavior appears as a loop, as one can see in Figure 9. This kind of behavior has been shown also by Johnson & Weinberg (2019) studying starbursts and by Spitoni et al. (2019) and Calura & Menci (2009) in connection with the Galactic disk. In this scenario the super-solar NSC stars could be either as old or as young as the subsolar stars.

Zoom In Zoom Out Reset image size

To trigger a recent starburst episode, gas needs to be accreted either from other regions of the Galaxy or from an extragalactic origin. In the former case, torques from the bar drive gas inflows toward the center. It is, however, not clear why such inflow should be intermittent (as required to explain a sudden burst). Simulations of the Milky Way show that discrete accretion events could be due to the fragmentation of the inflowing gas from the central molecular zone (CMZ) at the inner Lindblad resonance (Renaud et al. 2013; Krumholz & Kruijssen 2015). The other possibility is the accretion of gas from extragalactic origin, for instance from a satellite galaxy tidally stripped during a passage near the Milky Way's center, i.e., on a rather radial trajectory. The gas stripped from a low-mass galaxy would have very low metallicity, whereas the gas from the CMZ could be near or above solar metallicity (Morris & Serabyn 1996).

Alternatively, we explore a different chemical evolution model where we suppose a pause in the star formation (see Figure 10 where pauses of several lengths are assumed). In these cases, no extra infall event is assumed, but the star formation simply starts again after the stop. As one can see from Figure 10, in this case there is also a loop but not a horizontal decrease of [Fe/H]. Here, the loop results from α-elements decreasing during the pause as Fe continues to be produced by SNe Ia, although less than in the star formation regime. When star formation reignites the [α/Fe] ratio increases and then decreases, as explained above.

Zoom In Zoom Out Reset image size

The main differences between the two scenarios are the extra infall (starburst) and the efficiency of star formation. In the burst case, there is extra infall and the efficiency is high both before and during the starburst, whereas in the case of the pause in the star formation the efficiency is lower especially at the beginning but high after the reignition. The reason for this is that with a high star formation efficiency, as in some models of the classical bulge/bar (Matteucci et al. 2019), the gas is consumed quickly leaving too little to trigger star formation after the pause (a similar behavior was suggested by Matteucci et al. 2019 for the bulge). In the case of the bulge it was suggested that buckling of the bar could be the agent responsible for the pause in star formation.

Further investigations into modeling the alpha enrichment of supersolar metallicity stars in the NSC and vicinity is called for, and we will investigate this in an upcoming paper.