Astronomical Science

Red Supergiants as Cosmic Abundance Probes

Ben Davies1 to the shallower potential well, prevent- infrared (NIR). At these wavelengths, the Rolf-Peter Kudritzki2,3 ing them from reaching high metallicities, gain from adaptive optics will be the most Maria Bergemann4 and gives rise to the functional form of profound, as the increase in sensitivity Chris Evans5 the mass–metallicity [Z] relation we see scales with the fourth power of the mirror Zach Gazak2 today. State-of-the-art cosmo­logical sim- diameter. Carmela Lardo1 ulations (e.g., Shaye et al., 2015) make Lee Patrick6 specific predictions about the quantitative Bertrand Plez7 nature of the mass–metallicity relation, Red supergiants: A novel, NIR-based Nate Bastian1 which makes it possible to test the current technique for extragalactic metallicity ­theories of galaxy evolution under the studies dark energy, cold dark matter framework. 1 Astrophysics Research Institute, To this end, in recent years we have been ­Liverpool John Moores University, Such tests, however, require reliable and developing a technique to obtain extra­ United Kingdom robust measurements of a galaxy’s galactic abundances using NIR spectros- 2 Institute for Astronomy, University of chemical abundances. Historically this copy of red supergiants (RSGs). These 5 Hawaii, USA has been done using emission lines from stars are extremely bright (> 10 LA), 3 Munich University Observatory, Germany extragalactic H II regions. The “direct” young (< 30 Myr), and have the advan- 4 Max Planck Institute for Astronomy, method requires measurements of auro- tage that their fluxes peak at NIR wave- Garching, Germany ral lines, such as [O III] 436.3 nm, which lengths. A challenge for abundance work 5 UK Astronomy Technology Centre allows the observer to uniquely determine using these stars is that they are cool, (UKATC), Edinburgh, Scotland the gas temperature and, crucially, the with effective temperatures around 6 Institute for Astronomy, University of abundance of the element. Unfortunately, 4000 K, and so their spectra are domi- Edinburgh, Scotland these auroral lines are often very weak, nated by absorption from thousands of 7 University of Montpellier, France especially at high metallicity. This means molecular lines. Historically this has that one must attempt to derive abun- meant that high spectral resolving powers dances based on only the strong emission were needed to study individual lines, By studying a galaxy’s present-day lines (e.g., [O III] 500.7 nm), which contain making these stars unsuitable for extra- chemical abundances, we are effec- no temperature information on their own, galactic work due to the large integration tively looking at its star-forming history. by calibrating against either results from times required. Cosmological simulations of galaxy the direct method or from photoionisation evolution make predictions about the models. The limitation of these strong- In order to adapt these techniques to be relative metal contents of galaxies as a line methods is that they have well-known useful at extragalactic distances, in function of their stellar mass, a trend systematic offsets from one another, Davies, Kudritzki & Figer (2010) we identi- known as the mass–metallicity relation. ­particularly at high metallicity, where two fied a spectral region around 1.2 μm that These predictions can be tested with different calibrations may give results is relatively free of molecular lines, the observations of nearby galaxies.­ How- ~ 0.8 dex apart (Kewley & Ellison, 2008). dominant features being those of neutral ever, providing reliable, accurate abun- metallic atoms. We then demonstrated dance measurements at extragalactic In order to solve this problem, our team that metallicities relative to Solar (defined distances is extremely challenging. has been pioneering new ways to obtain as [Z]) may be determined from moderate In this project, we have developed a the metallicities of external galaxies by spectral resolutions of only R ~ 3000, technique to extract abundance infor- studying their constituent stars individu- making this technique viable at large dis- mation from individual red supergiant ally. The stars chosen as targets must tances, and well-suited to the latest stars at megaparsec distances. We be: (a) young, so that their abundances multi-object spectrographs (MOSs), such are currently exploiting this technique reflect those of H II regions; and (b) lumi- as the K-band Multi-Object Spectrograph using the unique capabilities of KMOS nous, so that they can be observed at (KMOS) on the Very Large Telescope on the VLT. megaparsec distances. This means that (VLT). Using instrument simulators, we the obvious targets of choice are super- then demonstrated the potential with the giants — massive, post main-sequence E-ELT, showing that we could measure [Z] A well-known correlation exists between stars, which can outshine entire globular at distances of 75 Mpc (Evans et al., 2011). the mass of a galaxy and its metal con- clusters and dwarf galaxies. Substantial tent, whereby more massive galaxies recent progress has been made using tend to be more metal-rich (Lequeux et optical spectroscopy of blue supergiants Building the tools for the job al., 1979). This is understood qualitatively (BSGs; see Kudritzki et al. [2014] and as being due to progressive enrichment ­references therein). However, to really Spurred on by the potential of this tech- by stellar and supernova nucleosynthesis, take advantage of the next generation of nique, we set about developing the nec- moderated by the infall of pristine gas telescopes such as the European essary analysis tools. The initial task in and outflow of enriched material via Extremely Large Telescope (E-ELT), it is this project was to build a suite of model winds. This enriched gas escapes more crucial to reduce our dependence on spectra from which we can derive high- easily from lower-mass galaxies due optical techniques and move to the near- precision abundances. We generated a

32 The Messenger 161 – September 2015 grid of MARCS (Model Atmospheres in  Figure 1. A spectrum of Radiative and Convective Scheme) model 1 an RSG in Per OB1 observed with Subaru atmospheres (Gustafsson et al., 2008), and IRCS in black, and computed under the assumptions of the best-fitting model spherical symmetry and local thermody- (red), at bottom. The namic equilibrium (LTE), which span the 1 other spectra are the  same data but degraded parameter ranges in effective tempera- to lower resolution and ture, surface gravity, microturbulence and S re-fit, showing that the RD metal abundances appropriate for RSGs. EE 1 diagnostic lines are still We use these models to compute syn- identifiable down to ­resolutions of R = 3000. W N thetic spectra with full non-LTE radiation T Taken from Gazak et al.  transport for the diagnostic chemical ele- Ck (2014a). ments Fe, Si, Mg and Ti (Bergemann et 1 al., 2012; 2013; 2015). Fitting the models to the data is done by a χ2-minimisation process of matching the observed -NQL@KHRD 1 strengths of the diagnostic metal lines,  described in detail in Davies et al. (2015).

Road-testing in the Milky Way %D 3H 3H %D 2H 2H 2H  The next phase of our project was to         demonstrate that our technique works at 6@UDKDMFSGƅL high precision in the Solar Neighbour- hood. For our target we chose the nearby Perseus OB1 association, which contains The added value of the X-shooter data precision absorption-line spectroscopy several RSGs, and which we can safely was that, in addition to the coverage of of the highest possible quality, it is there- assume all have the same metallicity. our spectral window at 1.2 μm, we fore essential that we take the greatest Using high-resolution spectroscopy from obtained spectra covering the whole of care in removing sky emission and telluric the Infrared Camera and Spectrograph the optical and near-infrared, from 0.3– features from our data. After extracting (IRCS) on the Subaru Telescope, we 2.5 μm. This allowed us to verify that the the calibrated datacubes using the obtained a metallicity averaged over all temperatures we obtain from our narrow KMOS data reduction pipeline, we found RSGs in the region consistent with Solar, window in the J-band were consistent that variations in spectral resolution and [Z] = –0.02 ± 0.08, as would be expected with those from fits to the spectral energy wavelength calibration were introducing (Gazak et al., 2014a). Further, by degrad- distributions. The agreement for the noise into the final spectra. In Figure 2, ing the spectral resolution of the data and SMC stars was excellent, while the LMC we demonstrate these variations across repeating the analysis, we showed con- stars were also similar to within the the field of view of one integral field unit clusively that this technique is stable down errors, aside from the two stars in our (IFU), as measured from the sky emission to resolutions of R < 3000, typical of MOS sample with the highest temperatures. lines. We corrected for this effect in a instruments (see Figure 1). process we call KMOGENIZATION, whereby we smooth the spectra at each Beyond the Local Group with KMOS: A spaxel down to a level which is common Extragalactic baby steps: The Magellanic study of NGC 300 to all spaxels to be extracted. This is usu- Clouds ally defined to beR = 3000–3200. The Our first target beyond the Local Group spatially varying wavelength calibration With the technique now proven at Solar was chosen to be the Sculptor Group effects are minimised by only extracting metallicities, the first sub-Solar metallicity member NGC 300. This galaxy is a rela- spectra within a 1 spaxel radius of the systems to be studied were the Large tively nearby (2 Mpc) face-on spiral flux peak. Although this discards a small and Small Magellanic Clouds (LMC and with a strong abundance gradient, as amount of flux, this is more than com­ SMC). The data here were provided by measured from H II regions with the pensated for by the reduction in artefacts X-shooter on the VLT (Davies et al., 2015). direct method, and with BSGs (Bresolin that can be introduced when removing From analysis of ten and nine stars in the et al., 2009; Kudritzki et al., 2008). It is sky and telluric features. Repeated ob­­ LMC and SMC respectively, we measured therefore an excellent testbed to provide servations of the same target, for which average metallicities of [Z]LMC = –0.37 ± one more validation of our technique. the wavelength calibration may vary 0.14 and [Z]SMC = –0.53 ± 0.16, consistent throughout the night, were co-added with contemporary results from obser­ For these observations, we employed the onto a master wavelength scale without vations of hot supergiants (see Davies et new VLT NIR multi-object spectrograph resampling, so as to minimise systematic al. [2015] and references therein). KMOS. Since our project requires high- noise.

The Messenger 161 – September 2015 33 Astronomical Science Davies B. et al., Red Supergiants as Cosmic Abundance Probes

(%4 2GHESOHWDK 1DRNKTSHNMh6h                

8 8 8    

  l   

   

  l         7 7 7

Figure 2. Output from our KMOGENIZATION soft- One step beyond: RSG-dominated star RSG lifetime (< 1 Myr) is short compared ware. The left panel shows the wavelength-col- clusters to the age of the cluster, all the RSGs lapsed image of one KMOS IFU, in this case IFU#1, which has been positioned on a star cluster in should have roughly the same mass and NGC 4038. The centre and right panels show the When observing stars at very large dis- hence gravity. Secondly, since the stars spatial variations in wavelength calibration and spec- tances, the issues of crowding and formed from the same molecular cloud, tral resolution respectively. We correct for these vari- source blending must always be consid- they should have the same metallicity. ations to minimise systematic noise in our extracted spectra. ered. For this project, RSGs are so much Finally, we know from our previous stud- brighter than any other class of star in ies that RSGs span only a narrow range the near-infrared that the only blending in temperature and microturbulence. Our After the spectra were extracted and ana- sources to worry about are other RSGs. prediction was therefore that the spectrum lysed, the RSG-based metallicity gradient­ It is well known that massive star clusters of an unresolved cluster of 100 RSGs was constructed (Gazak et al., 2015), and with ages between 7–50 Myr may contain would be indistinguishable from that of is shown in Figure 3. The agreement dozens of RSGs. If such a cluster was at one of its constituent stars, but be 100 between the RSG gradient and that from a distance so great that it was totally unre- times brighter. This then means we could observations of BSGs is remarkable — solved and appeared as a point source, use our technique on massive star clus- the gradients are identical to within the and we mistook it for a single abnormally ters at distances ten times further than errors, and with a systematic offset of bright RSG, what would we see? for individual stars. only 0.05 dex. The agreement between the RSG/BSG gradients and that of the Our hunch was that an unresolved star direct H II region measurements is cluster dominated by RSGs would have Figure 3. Radial metallicity gradient in NGC 300, as equally striking, see Bresolin et al. (2009). the spectral appearance in the J-band of measured from RSGs (red points), BSGs (blue points), and direct method (auroral line) H II region a single RSG. Firstly, since the RSGs in observations, illustrating the stunning agreement the cluster are effectively coeval, and the between the different methods. A survey of the mass–metallicity relation in the local Universe with KMOS   12&R With all the validation and road-testing !2&R phases complete, we are now beginning   '((R to construct a mass–metallicity relation for galaxies in the nearby Universe based only on observations of RSGs. The   weapon of choice for this project is KMOS,

with its multiplexing, spectral range and < l  9 resolving power being ideally suited to : our technique. l  We began with a recently published study of NGC 6822, which found a mean abun- l  dance of [Z] = –0.52 ± 0.21 (Patrick et al., 2015). This will continue with a study l  of Sextans A (data taken April 2015), and a recently approved KMOS programme l  to observe NGC 55, WLM and IC1613 in          

Period 96. 11

34 The Messenger 161 – September 2015   ,  

  

@W     NEL

W   < 9 %KT

:  

 

              6@UDKDMFSGƅL

Figure 4. The relative contributions to the overall flux l  of a 15 Myr old star cluster by the RSGs (red) and the           rest of the stars (blue). Our J-band spectral window 11 for deriving metallicities is shown in grey. The figure  shows that 95 % of the flux atJ comes from the RSGs, which are only 0.1% of the stars. From Gazak Figure 5. The metallicity of an RSG-dominated star et al. (2014b). cluster close to the centre of M83 (red), as measured in Gazak et al. (2014b), plotted against galactic radius. The blue points show a recent study of the same This is a prediction that was confirmed in show that the RSGs account for > 95 % ­galaxy using BSGs (­Bresolin, Kudritzki & Gieren, in Gazak et al. (2014a; 2014b), where the of the flux in theJ -band (Figure 4), and prep). The agreement between the two methods is spectra of the individual RSGs in Per OB1 that neither the age of the cluster nor the excellent, pro­viding further evidence that the RSG- cluster method can determine accurate metallicities. were combined to make a mock star contribution to the J-band flux by the cluster spectrum. This spectrum was other stars has a significant effect on the Figure 6. Hubble Space Telescope Wide Field then analysed in the same way as that of inferred metallicity. ­Camera 3 J-band image of the Antennae Galaxies, a single star, with an almost identical illustrating the locations of the observed star clusters that contain dozens of RSGs. Their KMOS spectra metallicity retrieved. We also performed This RSG cluster based technique was are shown on the right (black), along with their best- population synthesis experiments to first utilised on bright objects in M83 and fitting models (red) and abundances.

  ( ( ( ( ( ( 2H 2H 2H 2H ( 3H 3H ( K K K ,F K K ,F %D %D %D %D %D   :9< Ÿ           

   DR

  :9< Ÿ  BLHMTS

TW   @Q

k   S l    ERD   -NQL .E  

:9< Ÿ    l           

            6@UDKDMFSGƃL      l  l  .EERDS@QBLHMTSDR

The Messenger 161 – September 2015 35 Astronomical Science Davies B. et al., Red Supergiants as Cosmic Abundance Probes

NGC 6946 in Gazak et al. (2014b), using merging galaxies (Lardo et al., in prep; References the Infrared Spectrometer And Array see Figure 6). Bergemann, M. et al. 2012, ApJ, 751, 156 Camera (ISAAC) on the VLT and the SpeX Bergemann, M. et al. 2013, ApJ, 764, 115 medium resolution spectrograph on the Bergemann, M. et al. 2015, ApJ, 804, 113 NASA Infra-Red Telescope Facility (IRTF) Prospects Bresolin, F. et al. 2009, ApJ, 700, 309 respectively. Our result of a super-Solar Davies, B., Kudritzki, R.-P. & Figer, D. F. 2010, MNRAS, 407, 1203 metallicity in the centre of M83 is in excel- We have shown in this article that RSGs Davies, B. et al. 2015, ApJ, 806, 21 lent agreement with recent obser­vations are extremely powerful cosmic abun- Evans, C. et al. 2011, A&A, 527, A50 of BSGs in the same galaxy (­Bresolin, dance probes, and in combination with Gazak, J. Z. et al. 2014a, ApJ, 788, 58 Kudritzki & Gieren, in prep.; see Figure 5). instruments like KMOS can provide Gazak, J. Z. et al. 2014b, ApJ, 787, 142 Gazak, J. Z. et al. 2015, ApJ, 805, 2 ­high-precision metallicities at distances Gustafsson, B. et al. 2008, A&A, 486, 951 Our first study of a sample of massive of several megaparsecs. The next stage Kewley, L. J. & Ellison, S. L. 2008, ApJ, 681, 1183 young clusters across a galaxy is to be of our project is to provide an RSG-based Kudritzki, R.-P. et al. 2014, ApJ, 788, 58 submitted shortly. We obtained KMOS mass–metallicity relation that will have Kudritzki, R.-P. et al. 2008, ApJ, 681, 269 Lequeux, J. et al. 1979, A&A, 80, 155 observations of three clusters in the inter- the power to directly test cosmological Patrick, L. R. et al. 2015, ApJ, 803, 14 acting galaxy NGC 4038 (one half of simulations of galaxy formation. The first Schaye, J. et al. 2015, MNRAS, 446, 521 the Antennae system), showing that the set of KMOS obser­vations required to galaxy has a flat abundance gradient — complete this are scheduled for the com- consistent with theoretical predictions for ing semester.

The nearby dwarf irregular galaxy NGC 6822 (Barnard's Galaxy) is shown in an MPG/ESO 2.2-metre ­telescope image taken with the Wide Field Imager (WFI) with B, V and R broad band filters and a narrow Hα ­filter. Young shell H II regions with prominent Hα emission are visible in the outer regions of the galaxy. See Release eso0938 for more details.

36 The Messenger 161 – September 2015