eaa.iop.orgDOI: 10.1888/0333750888/2870 Synthetic Stellar PopulationsGuy Worthey FromEncyclopedia of Astronomy & AstrophysicsP. Murdin © IOP Publishing Ltd 2006 ISBN: 0333750888Downloaded on Thu Mar 02 23:45:56 GMT 2006 [131.215.103.76]Institute of Physics PublishingBristol and PhiladelphiaTerms and ConditionsSynthetic Stellar PopulationsENCYCLOPEDIA OF ASTRONOMY AND ASTROPHYSICSSynthetic Stellar PopulationsModern telescopes are built with ever-better angularresolution, but for galaxies and clusters outside the localgroup we generally cannot hope to study the stars in themone by one. Instead, we see the sum of the light from allof the stars. To take this integrated starlight and extractinformation on the ages or heavy-element abundancesrequires a model; a synthetic version of how many andwhat kind of stars are present. To guide the models, werely on nearby stars and star clusters where stars can bestudied one by one and also on theoretical models of stellarevolution and stellar atmospheres.ASTELLAR POPULATION is a group of stars born at thesame time and sharing the same initial composition. ASTAR CLUSTER is an example of a simple stellar population,characterized by a single age and a single abundancemixture. TheMILKY WAY is an example of a complicatedmixture of many stellar populations; old and young, metalpoor and metal rich.In the Milky Way, there is a striking concordance ofage–metallicity with spatial extent and kinematics. TheGALACTIC METAL-POOR HALO is thought to have formed in thefirst third of the Galaxy’s history and is composed of starsand clusters on randomly oriented orbits so that the neteffect is a large halo of roughly spherical shape. On theother hand, if one finds all the stars in the Galaxy youngerthan a billion years, virtually all of them are in the verythin Galactic disk. They have orderly, near-circular orbitswith all stars rotating clockwise as seen from above theGalactic north pole (seeGALACTIC THIN DISK). This dichotomygave rise to the notion of two distinct stellar populationsin the Milky Way around 1944 (W Baade is credited withthis idea). By the 1960s it became evident that therewere several, rather than two, distinct populations. Forinstance, old but metal-rich stars exist both in the diskand in the bulge. Today, we have given up counting thenumber ofstellar populations, and instead speak of theageand abundance distributions within kinematic subgroups.From distant galaxies and clusters, what informationcan one hope to derive from integrated star light?Ultimately, one could hope to find the complete starformation history of the galaxy, and at each past epochof star formation the population’s dynamic state, adistribution of heavy elements, and an initial massfunction (abbreviated IMF; the number of stars born ateach mass). Such a detailed picture is out of reach forthe present. Today, we have clues about each of thesequantities, but we are far from being able to wrestle sucha detailed picture from integrated light observations.Several types of synthetic stellar populationmodelsFigure 1 shows galaxy spectra covering the range ofspectral shapes that are seen along with a sampling ofsome of the stellar types out of which the galaxies arecomposed. From the figure one can glimpse similaritiesbetween the galaxies and the underlying stellar types.Figure 1. Example galaxy (top) and star (bottom) spectra areshown, arbitrarily scaled Fνversus wavelength in Å. Thestrongest emission lines in the starburst spectrum weretruncated for clarity, and a few important spectral features aremarked on the stellar spectra. This is the synthesis problem:adding the star spectra together in different combinations andwith different multiplicative weights, how closely can the galaxyspectrum be matched? By eye, one can see that the starburstgalaxy must have a strong O star component, and the spiral andelliptical galaxies must have many G stars. However, TiOfeatures appear in the spiral and elliptical spectra, so KM starsmust also be present.Extreme STARBURST GALAXIES emit most of their light in theultraviolet. Their spectra are matched by OB-type starswith strongly muted evidence of other spectral types.SPIRAL and ELLIPTICAL GALAXIES have minority populationsof hot stars, as seen by their small ultraviolet fluxes, andobviously have a strong G-star component. However, thestrong, broad TiO features characteristic of M-type starsare seen in the red portion of the spectrum, so very coolstars are present as well.The first type of model is this kind of star-by-starCopyright © Nature Publishing Group 2001Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998and Institute of Physics Publishing 2001Dirac House, Temple Back, Bristol, BS1 6BE, UK1Synthetic Stellar PopulationsENCYCLOPEDIA OF ASTRONOMY AND ASTROPHYSICSaddition problem. Early workers such as Spinrad andTaylor would observe, spectrophotometrically, a collectionof stars and a few galaxies. Then, by method of trialand error, they would construct a ‘model’ consisting ofthe relative proportions of the stars that best matched thegalaxy spectrum. That is, given a galaxy spectrum Lλanda collection of stellar spectra li,λ, the model consists of theweights wiby which the stars are multiplied:Lλ=iwili,λ. (1)The trial-and-error part comes in because the same setof weights must work for all λ. A useful improvementcame with the work of Faber in the early 1970s.Through a careful statement of the problem and judicioususe of technique (quadratic programming or linearprogramming) the trial-and-error portion of assigningweights was replaced by numerical optimization.This form of population modeling is often termedempirical population synthesis. Empirical populationsynthesis has a couple of drawbacks. First, solutions arenot unique. For instance, the spectrum ofa G dwarf closelyresembles that of a G giant, and replacement of one by theother probably will not noticeably affect the goodness offit. Second, information on galaxy age or metal abundancerequires interpretation. Suppose the synthesis says thatG0 dwarfs are numerous, but F8 dwarfs are rare. Theinterpretation would be that the main sequence turnoffis at spectral type G0. One would then refer to observedstar clusters or stellar evolutionary isochrones to find outthe age at which the main sequence turnoff is spectral typeG0.The connection to age and metal abundance canbe made more direct by considering, not stars, but starclusters as the