3 Outline 1- What & Why? 2 – Who & How? 3 – So what? Scope & disclaimer Motivations2 – Who & How?Schools, Ingredients & Methods:- Based on indices- Based on full spectral fits3 – So what?Miscelaneous resultsCaveats
4 1 – What & Why Mission Impossible! optical spectroscopy the small printTwo of the characteristics we expect of these reviews are the need to present a very balanced overview of the theme, in a manner which informs the audience without burying it in excessive technicalities. We are confident that you will be able to fulfil these criteriaMission Impossible!Focus on:optical spectroscopystellar photons (ie, no em. line SFH diagnostics)late-type (= non-ellipticals)applications to large samples (= SDSS)
5 1 – Why study stellar pops in galaxies? To learn about galaxy evolution: SFR(t), Z*(t), cosmology...SFH(t) of the UniverseMass assembly history of a late type galaxyHeavens 04Mathis 06
6 1 – Why study stellar pops in galaxies? To clean starlight pollution from my spectra!Tremonti 04Li 05
7 1 – Why study stellar pops in galaxies? To clean starlight pollution from my spectra!z ~ 0.7 compositeScattered Broad Hb in a Sey 2Savaglio 05CF 04
8 2 – How?Spectral synthesis of integrated stellar populations: “...a subject with bad reputation. Too much has been claimed, and too few have been persuaded.” (Searle, 1986)Basic Recipe(a) Discrete x continuous representations(b) Observables: Indices x Full spectrum(c) From observables to SFH:Methods, methods & methods...
9 = S ’s (+ gas + dust + ...) Fgal(l) = S F*(l) 2 – How? The Fundamental Theorem of Population synthesis:= S ’s (+ gas + dust + ...)Individual StarsObserved clustersModel SSPsContinuous modelsFgal(l) = S F*(l)+ extinction x A(l)& kinematics x LOSVD (v*,s*,vrot)
10 2a – How? The discrete approach ≈ S SSPj j = 1...NEmpirical Pop. Synthesis: SSP = Observed Clustersx x x: Stellar evol, spectra & IMF given by Mother Nature: Incomplete coverage of (t,Z) space & l-rangeBica 88, Schmidt 91, Ahumada Pelat 97, 98, CF 01 (math)
11 2a – How? The discrete approach ≈ S SSPj j = 1...NSSP = Model “clusters” from evolutionary synthesisS xj FSSP(l ; tj,Zj ; IMF, tracks, libs...)population vector: Wider coverage of (t,Z) space & l-range: Models are always models...Models: GALAXEV, SED, STABURST99, PEGASE, Maraston, ...
12 2a – How? The discrete approach WARNING:Models look great, but there are LOTS of assumptions & tricks in this business!- tracks,- spectral libraries,- interpolation schemes,- ...Models: GALAXEV, SED, STABURST99, PEGASE, Maraston, ...
13 2a – How? The continuous approach SFR(t): More general than S bursts: Need to parametrize SF history & chemical evol.: Models are always models...Fritze-v. Alvensleben 06, Bicker 04, ...
14 2b – Observables: Indices x Full Spectrum Compare Index(t,Z,SFH) models to data to constrain SFH parameters.Instantaneous BurstsContinuous SFKauffmann 03, ...
15 2b – Observables: Indices x Full Spectrum Nolan 06Rectified spectrum (“high pass”)Mathis 06CF 05
16 2b – Observables: Indices x Full Spectrum Reichardt 01Walcher 06Mayya 06
17 2c – How? From Observables to SFH... Hypothese space(“priors”)Only 1 Z? Z = Z(t)?Al = ? Dust geometry? Al(t,Z)?Kinematics?Which basis? (clusters, models,...)Which parameters?WARNING:Impossible to review all combinations!Will browse through a few examplesObservables spaceParameter spaceMethodBrute force discrete grid search?Convex-algebra?Markov-Chains?PCA? AI-techniques?Compression on input or output?Comparions to library of models?How to deal with degeneracies?
18 2c – How? From Observables to SFH... Pelat 97, 98, Moultaka 00, 04A very elegant method, yet largely overlooked because (?) of complex math (convex algebra) & few applications.Observables: 2 EWsParameters:5 light fractionsReconstructed spectrumBoisson 00
19 2c – How? From Observables to SFH... 5 indices: D4000, Hb, Hd+Hg, [MgFe]’ & [Mg2Fe]Bayesian comparison to a large library of modelsGallazzi 05PDF of light weighted mean age
20 2c – How? From Observables to SFH... F(l) fitting with MOPEDMultiple Optimized Parameter Estimation & Dta cmpsnMass & Z in N ~ 10 time-binsM*Mass assembly histories: M(t)Panter 03Mathis 06
21 2c – How? From Observables to SFH... F(l) fitting with STARLIGHT- Light (Mass) in N ~ 100 time & Z SSPs- Compress outputDownsizingM*M(t) & Z(t) of Star-Forming galaxiesPop. vector=SFHCF , 06, Mateus 06, Asari 07
22 2c – How? From Observables to SFH... UCBD galaxiesCorbin 06
23 2c – How? From Observables to SFH... Many other methods!STEllar Content via Maximum A Posteriori – Ocvirk 05 + Koleva + ...Active Instance-Based Machine Learning – Solorio 05Bayesian Latent Variable modelling – Nolan 06Principal Component Analysis – Li 05, WildDirect fitting – Tadhunter 05, Holt 06, Moustakas MacArthur...Brute Force – Bush 01, 02, 03, 04, 05, 06, 07, 08...Diversity in:Math / elegance / speed1000 “Technicalities” (masks, kinematics, extinction, ...)Physical ingredientsInput & Output...
24 3 – So What? A few miscelaneous results (a) Global relations:Synthesis parameters X other things:<t>, <Z>, <SFR>, <SSFR>, ...Zgas, M*, Mdyn, environment, ...(b) Daring one step further: SFH(t)!(c) Sanity checksCaveats(d) Closing words
25 <Z> & <t> x Mass 3a – Global relationsZ(gas) x Mass<Z> x Z(gas)<t> x Z(gas)<t> x Z(gas)<Z> x Z(gas)<Z> & <t> x MassM*Tremonti 04, Gallazzi 05CF 05, Mateus 06, Asari 07
26 3b – Going one step further: Evolution AGNSFZ(gas)Idea: Dissect the SFH = SFR(t) & Z*(t)along the left wing of the Seagull(normal SF galaxies)
27 3b – Going one step further: Evolution ResultLow Z(gas) galaxies are much slower in their mass assembly and chemical evolution
29 3c – Sanity checks: good news Different ingreedients yield ~ similar result !!Panter 06SFR(Synt) ~ SFR(Ha)Asari 07
30 3c – Sanity checks: good news Ha/HbNebular exctinction – NaD ISM
31 3c – Sanity checks: good news Ha/HbAV (Balmer) ~ 2 AV (Stellar)
32 a–enhancement is not only an E-gal problem... 3c – Warning: Ellipticalsa–enhancement is not only an E-gal problem...SF-galaxiesaAsari 07Sodre 05
33 AZD still present in full spectral fits 3c – Warning: AZD still present in full spectral fitsCF 05, Gomes 05
34 3c – Residuals: ~ Within errors, but ... EllipticalsSF-galaxiesHb–troff: Low amplitude, but systematic. ~ 100 Myr pops. STELIB?
35 3 – So What? Conclusions(a) Ingreedients & methods have matured a lot!(b) Global properties<t>, <Z>, SFR, SSFR, ...in very good shape (b) Evolution ... Looking good!(c) Caveats & Futurea/Fe issueRealistict dust models...
36 Spectral synthesis of integrated stellar populations: “ Spectral synthesis of integrated stellar populations: “...a subject with bad reputation. Too much has been claimed, and too few have been persuaded.” (Searle, 1986)
37 (At least we managed to fool Scott!) Spectral synthesis of integrated stellar populations: “...a subject with a not so bad reputation anymore. By not claiming too much, we’re now able to convince quite a few people.”(At least we managed to fool Scott!)(A bunch of us, 2006)
38 Public version of STARLIGHT + results for SDSS galaxies &
41 STARLIGHT & its many applications HE – The “homeless” QSOCaII Triplet velocity dispersionsMerritt et al 2006Vega 2004, Garcia-Rissman et al 2005
42 2 – The SF-History of Sey 2 nuclei CF, Gu, Melnick, Terlevich2, Kunth, Rodrigues Lacerda, Joguet 2004, MNRASStrong FC in this Sey 1 79 galaxies65 Sey 2s~ 200 pcBase = BC03 + FC
43 6 – Synthesis of 582k SDSS galaxies N Asari, J Gomes, W Schoenell, J P Papaqui (UFSC)A Mateus (IAG), L Sodré (IAG) & G Stasinska (Meudon)The SEAGal Collaboration: Semi-Empirical Analysis of Galaxies