A search for late-type M supergiants in the inner Galactic disk


Jordi Torra (1), Fernando Comerón (2), Cristina Chiappini (3),
Francesca Figueras (1), Valentin D. Ivanov (4), Salvador J. Ribas (1),
Ana E. Gómez (5)

(1) Departament d'Astronomia i Meteorologia, Universitat de Barcelona, Spain
(2) European Southern Observatory, Germany
(3) Osservatorio Astronomico de Trieste, Italy
(4) European Southern Observatory, Chile
(5) GEPI, Observatoire de Meudon, France



Abstract

Between 1998 and 2003 we carried out a series of observing campaigns using the 1.23m and 2.2m telescopes on Calar Alto with the goal of searching for distant late-type supergiants in the inner regions of the disk of our Galaxy. The observations obtained in Calar Alto have allowed the identification and spectroscopic confirmation of nearly 200 cool, luminous stars for which medium-resolution spectroscopy has been obtained at the ESO New Technology Telescope on La Silla, Chile. No less than 18 of these stars turn out to be confirmed as very likely supergiants, and another 9 are strong candidates. The rest of the cool stars are red giants and large-amplitude variables. The size of the sample and the medium-resolution spectra allow us to compare the characteristics of inner Galaxy giants and supergiants with those of similar stars in the solar neighbourhood. Interestingly, we find that the spectral features of giants and supergiants in the inner Galaxy have noticeable differences with respect to local ones, and that systematic differences suspected between local and bulge giants also apply between local and inner-Galaxy supergiants. In this regard, supergiants in the inner Galaxy more closely resemble the supergiants in the galactic center than those in the local neighbourhood. This result cautions against the modeling of central starbursts in other galaxies using spectral libraries built from solar neighbourhood templates.



Introduction

The inner regions of galactic disks contain a wealth of information on the structure, dynamics, and history of spiral galaxies. While the central parsecs of a galaxy are often dominated by intense starburst activity or by the action of a supermassive black hole on its surroundings, like in our own Galaxy (e.g. Melia & Falcke 2001), the inner few kiloparsecs provide numerous observational tracers reflecting the consequences of important mechanisms such as the presence of bars that feed gas to the central engine or rings, or enhanced star formation associated to the large-scale gasdynamical perturbations driven by resonances of the potential. They also contain a record of the metal enrichment of the galactic disk, which in turn given essential clues on its build-up process (Chiappini et al. 2001).

An additional reason to be interested in the inner regions of our own Galaxy is the accessibility of individual stars that can be observed and analyzed in detail, and the subsequent possibility of using their spectra as templates for the interpretation of the integrated spectra of much more distant, unresolved stellar populations. In this regard, one of the main tools used nowadays for deriving the recent star formation history of active galactic nuclei (AGNs) and central starbursts is the definition of appropriate spectrophotometric indices in the near-infrared, where most of the luminosity is contributed by very young, massive supergiants and older, evolved Red Giant Branch (RGB) and Assymptotic Giant Branch (AGB) stars. In principle suitably defined indices allow one to disentangle the relative contributions of both populations and to assess the evolutionary status of the more evolved population. In practice, such exercise critically depends on the choice of templates of the stellar spectra contributing to the luminosity, be it by using a library of actual stellar spectra, or by relying on model spectra produced by stellar evolution models.

The identification of a sample of late-type stars, and especially supergiants, in the direction of the inner galactic disk can address many of the points noted above. Due to the short lifetimes of their massive progenitor stars late-type supergiants highlight regions of recent star formation, and their large-scale distribution can be used to map current star formation across the galactic disk. Their high luminosities in the near-infrared, where extinction is much lower than in the visible, place late-type supergiants anywhere in the Galaxy within easy reach of medium-sized telescopes and allow high S/N spectroscopy to be obtained. Their deep CO bandheads make them easily identifiable by means of narrow-band imaging. Finally, their distribution at different distances with respect to the galactic center and the well-established metallicity gradient in the galactic disk make it possible to sample the effects of a range of supersolar metallicities on the spectral features.

The project described here combines two campaigns of observations at Calar Alto in which we imaged through narrow-band filters a narrow strip, 15 degrees long, along the galactic equator through narrow-band filters that sample the drop in flux due to the first CO(2,0) bandhead at 2.293 microns in cool stars. This was followed by low resolution spectroscopy in both Calar Alto and La Silla to confirm the cool temperatures of the candidates detected via imaging, including high S/N spectroscopy that enables us to examine the behavior of the molecular and atomic bands seen in the infrared H and K bands. Here we present and discuss the highlights of our results, and refer the reader to our paper (Comerón et al. 2004) for a more complete and detailed presentation.



Imaging observations

The first set of imaging observations was obtained in 1998 using MAGIC at the 1.23m telescope. A second set of observations, obtained in 2001 with the same telescope and instrument, completed the imaging survey covering three parallel strips of 15 degrees length between galactic longitudes 6 and 21 degrees, and galactic latitudes centered on b = -4', 0, +4', each strip being approximately 5' wide. Each strip was imaged through the J, H, CO continuum and CO on-band filters, down to completeness limits J=15.5, H=15, CO continuum (supposed to be well approximated by K) = 13.0. Of more than 435,000 stars detected, over 23,000 had their CO index (defined as the magnitude difference between the on- and off-band magnitudes in the proximity of the CO(2,0) bandhead) measured with an accuracy better than 0.1 mag. Due to the narrow CO on-band filter this accuracy was limited to stars brighter than K=10.0, approximately. The rather complicated process of data reduction and calibration of the imaging observations has been described step-by-step by Comerón et al. (2004). We show in Figure 1 a sample fragment of a color composite of the mosaic formed by combining the individual dark-subtracted, flat-fielded, and bias-subtracted images that compose two parallel strips.


Figure 1: Fragment of the image obtained by combining the reduced frames obtained along two parallel strips in the J, H, and CO on-band filters, showing the crowdedness of the fields and the wide spread of reddenings along the line of sight. The reddish colors on the right edge of the strip are due to a slight offset in the pointing of the H-band image, which does not cover exactly the same area as the images in the other two filters. The frame shown here is approximately 20 arcminutes across.




Spectroscopic observations and spectrophotometric indices

Stars were selected for subsequent spectroscopy on the basis of the extreme value of their CO indices, indicative of a deep drop of flux at 2.293 microns caused by strong CO absorption bands. A first set of spectroscopic observations were obtained in a number of runs between 2000 and 2003 using MAGIC the 1.23m and 2.2m telescopes, as shown in Figure 2.


Figure 2: Authors Gómez (left) and Torra (right) discuss details of the data reduction process of low-resolution spectra obtained with MAGIC in the control room of the 1.23m telescope.


The spectra obtained with MAGIC confirm indeed that the photometry-based criterion for the selection of cool star candidates is able to detect with almost certainty stars characterized by their deep CO bands. They also clearly show in some cases the existence of strong wings of the water band centered around 1.9 microns, between the H and K atmospheric windows. The water feature is an excellent criterion to separate cool supergiants, which have either weak or absent water absorption wings, from large-amplitude, Mira-like AGB stars with extended atmospheres, which also possess strong CO features in their spectra (Lançon & Wood 2000). Our sample also selected a large number of RGB stars, whose CO bands are generally weaker but can reach strengths similar to those of supergiants. Objects suspected to be in each of these three classes on the basis of their low-resolution (R = 240) MAGIC spectra are shown in Figure 3.


Figure 3: Representative sample of spectra obtained with MAGIC for stars suspected to be (a) late-type supergiants (top), characterized by deep CO bands and no noticeable water band wings; (b) large-amplitude variables (middle), with similarly deep CO bandheads but clear wings of the water feature centered at 1.9 microns; and (c) RGB stars, with shallower CO bandheads and again no signs of water absorption wings. The steep rise of the continuum towards the red is due to the strong extinction in the direction of the stars whose spectra are plotted here.


It is possible to quantify the CO and H2O indices in the low-resolution spectra and use them to ascertain the contents of our sample, by comparing the position of our stars with those of different classes of cool, bright stars in various published spectral atlases. While it is possible to define a virtually reddening-free CO index by just choosing two narrow adjacent passbands, the broader wavelength coverage needed for the measurement of the water index makes necessary to use an ad-hoc definition that becomes reddening-free when adopting an appropriate form for the extinction law. Such an index, I(H2O), is discussed in Comerón et al. (2004). Plotting the equivalent width of the CO feature versus I(H2O) for the sample observed at higher resolution (described below), and comparing them to the same values measured on the spectra of the Lançon & Wood (2000) atlas, we obtain the diagram plotted in Figure 4.


Figure 4: I(H2O) vs. CO equivalent width diagram, comparing the positions of the stars in our sample (eight-pointed asterisks) with non-pulsating RGB stars (triangles), supergiants (squares), bulge giants (crosses), and oxygen-rich variables, separated between those having strong (circles) and weak (three-pointed asterisks) water absorption.


This diagram shows that our sample is dominated by non-pulsating stars, although some stars with spectroscopic characteristics of AGB long-period variables are also present in it. While the bulk of stars in our sample have relatively strong CO bands, very few reach the extreme values (equivalent width greater than 26 Angstrøm, and reaching up to 30 Angstrøm in the passbands used here) characteristic of local supergiants. In principle this may be taken as an indication that our sample is strongly dominated by RGB stars with mostly low-to-moderate CO bands, with few if any supergiants. Nevertheless, it is important to investigate whether the apparent lack of supergiants in our sample hinted at by Figure 4 is real or rather points towards intrinsic differences between the spectroscopic properties of supergiants in the inner galactic disk and in the solar neighbourhood, which may invalidate the strict application of selection criteria derived from local samples to more distant stars having formed in a different environment, particularly regarding metallicity. It should be possible to further investigate possible differences by using the weaker metallic lines in the spectra of these stars, as different metallic and molecular species show different dependencies on the temperature, surface gravity, and metallicity (Ivanov et al. 2004). Unfortunately, the resolution of the spectra obtained with MAGIC is in most cases insufficient to allow an accurate determination of strengths of the much fainter metallic features sufficient for the investigation of systematic effects in them. For this reason we obtained time at the ESO New Technology Telescope (NTT) in July 2003 using the SOFI near-infrared camera and spectrograph, yielding a wavelength coverage similar to that of the setup used with MAGIC but with a fourfold improvement in spectral resolution and the greater light-gathering power of the 3.5m mirror of the NTT.



Supergiants in our sample

A first hint that supergiants are actually represented in our sample, albeit not dominating it, comes from the 2MASS color-magnitude diagram of all our stars, plotted in Figure 5. The AGB stars indentified on the basis of their strong water band wings occupy a band in this diagram whose tilt is due to the effects of both distance and extinction, since more distant stars are expected to be generally more reddened. The absolute K-band magnitudes of AGB stars are expected to lie between -6.4 and -8.2 (Knapp et al. 2003), thus being brighter than most RGB stars and overlapping with the luminosities of supergiants (from which they are otherwise well separated spectroscopically). The band that they define in the (H-K, K) diagram thus marks the region where we expect to find also the supergiants that there may exist in our sample. And indeed, it turns out that the region is not devoided of objects from our sample. It is remarkable in particular the existence of a clump centered near (H-K) = 2.4, K = 9.5, of extremely red stars with weak or absent water absorption in their spectra, which may be distant, highly reddened, bright supergiants.


Figure 5: 2MASS color-magnitude diagram for the sample observed with SOFI at the NTT on La Silla. The filled circles indicate the stars with strong water band wings in the spectrum, taken to be bona-fide AGB stars. A sizeable number of stars from our sample without water absorption overlap with the band defined by the AGB stars, suggesting similar absolute magnitudes and therefore luminosities characteristic of supergiants, rather than of RGB stars. The straight line is the extinction vector showing the displacement of a star obscured by 10 magnitudes in the visible. The band defined by the AGB stars has a steeper slope than the extinction vector, since more obscured stars are also more distant on the average.





Figure 6: Equivalent widths of the CO(2,0) and NaI features of the stars in our sample, compared to different populations mostly in the solar neighbourhood. The stars we observed are noted with asterisks and, in the case of those with strong water absorption, with filled circles. The reference sample is composed of supergiants (squares), non-pulsating red giants (triangles), bulge giants (crosses), and long-period variables with strong (open circles) and weak (three-pointed asterisks) water absorption. There is a clear offset between the locus occupied by most of our stars and that defined by the different local samples. The reference sample is taken from Lançon & Wood (2000), including also a few bulge giants. A more abundant sample of high-metallicity bulge giants from Ramirez et al. (2000) shows a much better overlap with our sample.


The nearly 200 high quality spectra obtained with the NTT at a resolution of nearly R = 1000, of which we show some samples in Figure 7, reveal in turn interesting properties when comparison is made to the local sample. The most obvious systematic differences concern the relationship between the most prominent atomic features, due to NaI and CaI, and the strength of the CO bands: cool stars in the inner Galaxy tend to have stronger atomic features for a given CO strength, even exceeding those measured among the coolest local supergiants. This trend can be seen also among the bulge giants observed by Ramírez et al. (2000), and hints at metallicity effects as a likely cause. Most interestingly, spectroscopy of M supergiants at the Galactic center presented by Carr et al. (2000) show not only an enhancement of the metallic lines similar to those that we notice in our sample, but also a corresponding decrease in the strength of the CO feature when compared to bright supergiants in the solar neighbourhood, which brings it well within the range of our measured CO(2,0) equivalent widths. Similar conclusions regarding the lack of strong CO absorption among supergiants near the galactic center has been found by Schultheis et al. (2003). Together with the indications of high luminosities of some stars in our sample, this leads us to conclude that our sample does indeed contain a moderate number of supergiants, but that the typical criteria of identifying supergiants as those stars possessing extreme CO absorption may not be generally applicable. On the other hand, the existence of supergiants among the stars with strongest CO features in our sample is suggested too by the plot, shown in Figure 8, of the CO(6,3) feature appearing in the H band of our spectra, and a measurement of the distance of the star to the band defined by the AGB stars in the color-magnitude diagram shown in Figure 5, K0, which is an indicator of luminosity (with negative values of K0 denoting the highest luminosities). For low values of the CO(6,3) equivalent width, its correlation with K0 traces the temperature-infrared luminosity relationship for giants, with luminosity increasing with decreasing temperature. The relationship nevertheless flares up at the highest CO(6,3) strengths, where the scatter in K0 becomes much larger. This hints at a wide range of luminosities at the coolest temperatures in our sample, with the brightest stars having luminosities similar to those of the brightest AGB stars, and thus typical of supergiants.


Figure 7: Representative spectra of the different classes of objects included in our sample obtained with the NTT. The numbers accompanying each spectrum correspond to the list number in Comerón et al. (2004). The spectral resolution is approximately R=1000.





Figure 8: Plot of the equivalent width of the CO(6,3) feature at 1.62 microns measured on the stars in our sample and the quantity K0, which measures the difference in magnitudes between the K magnitude of a star and the lower boundary of the band defined by AGB stars at the same (H-K) color. The filled circles represent AGB stars. The large scatter in K0 among the stars with strongest CO(6,3) absorptions suggest that the coolest stars display a wide range of luminosities, which in some cases reach those of the brightest AGB stars.


Rather than using the CO(2,0) bandhead as a strict criterion for the identification of supergiants in our sample using thresholds based on the supergiants of the solar neighbourhood (which would in fact exclude the galactic center supergiants), we have preferred a combination of criteria combining different spectroscopic features and photometric properties. We thus select likely supergiants as those stars having:

In this way we could identify 18 very good supergiant candidates is our sample fulfilling all these criteria. Taking into account moreover that the I(H2O) criteria is quite restrictive, because some of the coolest nearby supergiants do show some weak water band wings, we could identify 9 aditional candidates with I(H2O) between 0.0 and -0.1. No previous identification exists for any of these stars, most of which are heavily reddened by extinctions reaching up to 35 mag in the visible.



Why are cool stars in the inner Galaxy different?

As we have already stressed the characteristics of the cool stars of our sample of the inner Galaxy, including those that we identify as the best supergiant candidates, are markedly different from those of the solar neighbourhood. Closer analogs are found among galactic bulge giants and galactic center giants and supergiants. Interestingly, detailed chemical composition studies of the latter carried out by Carr et al. (2000) show that the different spectral characteristics of galactic center giants are not due to a metallicity significantly different from the solar value, but rather to altered surface abundances of CNO-processed material indicating a deeper penetration of the convective zone. This not only explains the weaker CO bands due to the reduced surface abundance of carbon and oxigen, but also the stronger atomic features given the important and eventually dominant contribution of CN bands to them when observed at low-to-medium resolution.

Having the constraint of a solar metallicity for the galactic center supergiants, the altered CNO abundances are attributed by Carr et al. (2000) to higher rotation velocities, perhaps as a consequence of the unique conditions in which star formation near the center of the Galaxy takes place. Such an explanation is extremely unlikely both for bulge giants and for the stars in our sample, which must have formed in an environment much more similar to the solar neighbourhood than to the galactic center. However, contrarily to the situation at the galactic center, the existence of higher metallicities in the inner regions of the Galaxy and the bulge is well established (see Chiappini et al. (2001) for a review of gradient determinations for different elements). Both increased metallicity and increased rotation velocity operate in the same sense of increasing the depth of the convection zone; in the case of high metallicity this is caused by the higher opacity in the mantle, which forces the outwards transport of the energy produced at the core to take place by convection from deeper inside the star. To a lesser extent, also the higher mass loss rates associated to higher metallicity contributes in the same direction (G. Meynet, priv. comm.). The medium-resolution spectroscopic characteristics of our stars can thus be at least qualitatively explained by their metallicities, although it would be desirable to confirm that this is the correct explanation by future atmosphere analyses directly measuring metallicity indicators unaffected by nuclear processing at the core.



Conclusions

Our search of late-type supergiants in the inner Galaxy has turned up 18 excellent candidates and 9 additional very good candidates, as well a large number of RGB and AGB stars. Spectroscopy at medium resolution shows that these stars have significant differences with respect to local samples, most probably due to their higher metallicities: the CaI, NaI features are stronger, and the CO bands probably weaker. Being drawn from a higher metallicity environment that is representative of the conditions of star formation in the central regions of other galaxies, we consider that these stars provide much more suitable templates than local giants and supergiants for the interpretation of the integrated spectra of central starbursts. The systematic differences also caution against the use of thresholds in the strengths of features derived from stars in the solar neighbourhood to decide on the nature of cool stars elsewhere. In this respect we may consider that our work extends existing studies comparing supergiants in the Magellanic Clouds and our Galaxy (e.g. Elias et al. 1985) in the opposite direction, by providing samples of supersolar metallicity.

The execution of our work also illustrates the usefulness of dedicated infrared imaging surveys with small and medium-sized telescopes in the post-2MASS era, showing the power of narrow-band imaging of extended areas of the sky to identify objects of particular interest for which broad-band imaging alone does not provide adequate selection criteria. In the case of late-type supergiants, an extension of a narrow-band survey such as the one presented here complemented with existing 2MASS data has the potential of revealing most of the cool supergiants of our Galaxy, including those that are more distant and obscured, thanks to their extremely bright absolute magnitudes. Low-resolution spectroscopy may then be used to confirm candidates, to be followed up in detail with observations at large telescopes. Given the brightness of the targets of interest the exposure times per field or per object are short, and the demands on observing time are thus mostly driven by operational overheads such as pointing, offsetting, and instrumental setup, which are generally similar for larger telescopes. Such studies thus highlight a niche where modest equipment by nowadays standards can make valuable contributions to current problems in galactic and extragalactic astronomy, by giving access to extended periods of observing time that would be hardly available at larger facilities.



References