1. A New Milestone in Starburst SED Modeling: Using 30 Doradus as a Benchmark Rafael Martínez-Galarza Leiden Observatory Brent Groves (Leiden/MPIA) Bernhard Brandl (Leiden) Genevieve de Messieres (Virginia) Remy Indebetouw (Virginia) FLASH Talk. NOAO. November 19, 2010

2. Overview Starbursts in the Universe SED Modeling of Starburst The Mid-IR properties of Starbursts 30 Doradus IRS Spectral map The models Fitting routine Results FLASH Talk. NOAO. November 19, 2010

3. What is a starburst galaxy?  Burst in the Star Formation Rate (SFR) of a galaxy, lasting 10 7 -10 8 years and involving up to 5% of the total stellar mass (Larson & Tinsley, 1978).  Many HII regions concentrated in time and usually also in space.  This translates into SFRs of tens, hundreds, or even thousands of M  /yr, in the case of ULIRGs.  For comparison, the SFR of the Milky Way is 1 M  /yr. M82 has a SFR of ~10 M  /yr Credit: NASA/ESA FLASH Talk. NOAO. November 19, 2010

4. Why studying starbursts?  Starburst form many massive stars and hence:  Help us understand the conditions for massive star formation.  Are bright enough to be seen at early stages of the Universe, when they were also more common, and hence help us understanding the formation of galaxies.  Can give insight into the stellar Initial Mass Function (IMF). FLASH Talk. NOAO. November 19, 2010

5. Mid-IR properties of Starburst Galaxies  SF occurs in very dusty environments.  Lots of stellar UV radiation reprocessed by dust and emitted in the IR.  Mid-IR spectrum contains useful information on the physical conditions.  Some mid-IR spectral features:  Spectral continuum  Silicate absorption  Nebular lines  Emission from Polycyclic Aromatic Hydrocarbons (PAHs) Brandl et al., 2006 FLASH Talk. NOAO. November 19, 2010

6. Spectral Energy Distributions (SEDs) Modeling  If a galaxy is unresolved, its integrated SED is our primary source of information.  Each physical process leaves its imprint on the shape of the galactic spectrum.  Processes related to starlight dominate de UV to IR portion of the spectrum.  SED modeling is the art of predicting the SED of a starburst from a set of physical assumptions. Brandl et al., 2006 Groves et al., 2008 FLASH Talk. NOAO. November 19, 2010

7. SED Fitting  Models predict the SED based on physical assumptions.  Fitting is the opposite: given the SED, extract physical information from the model parameters that produce best fit.  However:  Quality and amount of data insufficient.  Models have intrinsic uncertainties  Lack of independent checks of the derived physical parameters.  As long as those caveats are not taken care of, SED fitting of starbursts is useless, as far as the parameter uncertainties are concerned. FLASH Talk. NOAO. November 19, 2010

8. Our goal  Build a robust fitting routine that quantifies the uncertainties in the parameters and calibrates a specific model, by solving the three mentioned issues.  First two issues: We use a Bayesian inference approach.  Third issue: We choose a well known benchmark with independent checks for its parameters. FLASH Talk. NOAO. November 19, 2010

9. Our benchmark is 30 Doradus Giant star forming region located at 53kpc, in the LMC Building block of a starburst: Star cluster + HII region + PDR Several stellar populations Different regions spatially resolved. Extensively studied at optical (Hunter et al, 1995; Walborn et al., 1995) at infrared (Indebetouw et al., 2009) wavelengths 500 pc ESO FLASH Talk. NOAO. November 19, 2010

10. The Spitzer-IRS spectral map  Low resolution (R ~ 60-120) modules.  Wavelength coverage: 5.2- 38  m  3440 slit pointings covering an area of about 40.5 square arcminutes.  Spatial resolution in the SL module is about 0.5 pc  Wavelengths shown:  33.4 mm : [SIII] nebular line  10.5 mm : [SIV] nebular line  6.2 mm : PAH emission Indebetouw et al., 2009 FLASH Talk. NOAO. November 19, 2010

11. Individual Regions  R136: Central ionizing cluster  M cl ~ 5 × 10 4 M   Recently reported to have stars more massive than 150 M  (Crowther et al., 2010).  Compact, bright [SIV] source.  IR bright.  High extinction source  Prominent Mid-IR point source.  High excitation.  Spectra extracted within a square aperture of 6 SL pixels in on the side. [SIV]10.5  m map FLASH Talk. NOAO. November 19, 2010

12. Modeling expanding HII regions  Reference: Groves et al., 2008.  How it works:  Stellar synthesis: Starburst99. Kroupa IMF, M cl = 10 6 M   Radiative transfer calculated in two cases: HII region only. PDR covering HII region.  Time evolution: Mass loss expanding bubble driven by stellar wind and/or SN (Castor et al., 1975).  Add a component of UCHIIRs (embedded objects, hot dust).  Ages up to 10 Myrs. FLASH Talk. NOAO. November 19, 2010

13. Model Parameters  Metallicity, Z: fixed to the LMC value.  ISM pressure, P 0 /k: fixed to 10 5 K cm - 3  Cluster age, t: <10 Myrs  Stellar mass, M ★  ‘Embedded mass’, M emb  PDR covering, f PDR  Compactness, C  The PDR covering fraction derives from the relative contribution to the total flux from the PDR-covered models.  The compactness is related to the cluster mass (M cl ) and the ISM pressure (P 0 ) young old FLASH Talk. NOAO. November 19, 2010

14. Bayesian inference  The parameters are taken as random variables with associated probability distribution functions (PDFs).  The problem transforms: Find the PDFs given the data.  PDFs represent the complete solution to the problem.  The Bayes theorem states that: PDF(  ) ~ Likelihood * Prior  If errors are Gaussian: PDF ~ exp(-  2 /2) FLASH Talk. NOAO. November 19, 2010

15. Our priors  We introduce bounded uniform priors for M ★, M emb, f PDR and C  Boundaries are set to cover broad range of physical environments.  For example, log C 6.5 has never been measured. ParameterRangeResolution t (Myr)0-100.5 Log C3-6.50.5 f PDR 0.001-1.00.2 dex M★M★ 2 orders of magnitude0.2 dex M emb 2 orders of magnitude0.2 dex FLASH Talk. NOAO. November 19, 2010

16. Refining age priors: Nebular Line Ratios  We use line fluxes measured at high resolution with Spitzer-IRS (Lebouteiller et al., 2008).  [SIV]10.5mm/S[III]18.7mm  [NeIII]15.5mm/[NeII]12.8mm  We use Gaussian distributions with standard deviations corresponding to the age uncertainties.  Extinction might have an effect on sulfur ratios, making the source appear older. 0 Myr 2 Myr 2.5 Myr Young ages, < 2.5 Myrs FLASH Talk. NOAO. November 19, 2010

17. Continuum fitting: Integrated spectrum  With the defined priors we run the routine for continuum (Thermal + PAH) fitting.  Routine output: best fit values and PDFs calculated over the multi-dimensional parameter space.  Fit is poor at ~15  m. Dust in hot component might be hotter. IRS data Model Embedded objects HII region PDR region Best fit values t = 1.5 MyrM ★ = 2.8 × 10 5 M log C = 4.0M emb = 7.1 × 10 4 M f PDR =0.4 Martinez-Galarza et al., in prep. FLASH Talk. NOAO. November 19, 2010

18. “Embedded” component is necessary.  None of the spectra can be fitted without including this component.  Part of it could be related to the presence of embedded protostars.  Protostars have been detected at centimeter wavelengths. FLASH Talk. NOAO. November 19, 2010

19. Probability Density Functions: Integrated Spectrum Martinez-Galarza et al., in prep. FLASH Talk. NOAO. November 19, 2010

20. Summary of results  We list the results with the 1-  level uncertainties.  For C and f PDR we only provide upper or lower limits. Data at longer wavelengths needed to further constrain them. 30 DorR136[SIV] BrightHigh AV t (Myr)1.5 ± 1.52.5 ± 2.01.0 ± 1.53.0 ± 2.5 log C <4.5 <4.0 f PDR >0.1<0.3>0.2 log M ★ (M  )5.4 ± 0.43.4 ± 0.64.0 ± 0.44.1 ± 0.4 log M emb (M  )4.9 ± 0.12.7 ± 0.23.4 ± 0.12.7 ± 0.2 log M tot (M  )5.5 ± 0.43.4 ± 0.44.1 ± 0.4 f emb 0.310.200.250.04 FLASH Talk. NOAO. November 19, 2010

21. Independent Checks ParameterValueLiterature t 2.5 +/- 2.0 Myr ~1-2 Myr: Massey & Hunter et al. (1998), log M ★ 3.4 +/- 0.6 solar masses 4.3 for R136 alone: Walborn & Apellaniz (2002)  Time resolution is not enough to judge if the hot component of dust represented by f emb is related to embedded star formation.  We interpret it as dust that has not been pushed away by the stellar wind of the cluster and is associated to individual stars.  This component might imply that the modeling of the attenuation would be more complex than a simple dusty screen. 30 DorR136[SIV] BrightHigh AV f emb 0.310.200.250.04 Other individual sources: R136: FLASH Talk. NOAO. November 19, 2010

22. Next: NGC604 a higher metallicity environment FLASH Talk. NOAO. November 19, 2010 NGC 604 Spectral map at 8um

23. Summary  SED modeling provides a powerful tool to understand the physics of unresolved starbursts.  Interpretation of the results needs a robust fitting method that accounts for:  Lack of sufficient data  Model degeneracies  Lack of independent checks  We have presented state-of-the-art models and a fitting routine that provides a complete solution for the model parameters  We applied the fitting routine to the mid-IR spectrum of 30 Doradus and found that:  A component of ‘hot dust’ is necessary to fit the continuum slope.  We associate this component to remaining dust in the vicinity of individual stars.  Continuum fit only is insufficient for constraining the hardness of the radiation field.  Nebular line analysis necessary  Total cluster mass well constrained. FLASH Talk. NOAO. November 19, 2010

24. Summary  We applied the fitting routine to the mid-IR spectrum of 30 Doradus and found that:  A component of ‘hot dust’ is necessary to fit the continuum slope.  We associate this component to remaining dust in the vicinity of individual stars.  Continuum fit only is insufficient for constraining the hardness of the radiation field.  Nebular line analysis necessary  Total cluster mass well constrained.  Multi-wavelength analysis is necessary to fully constrain all model parameters. FLASH Talk. NOAO. November 19, 2010

25. Mid-IR SFR indicators  With the advent of IR observatories, mid-IR diagnostics of SF have been proposed.  They trace the amount of OBSCURED star formation in starbursts.  Combined with optical diagnostics, they can trace the total SF. Calzetti et al., 2007 FLASH Talk. NOAO. November 19, 2010

26. Properties of 30 Doradus PropertyValueReference Distance50 ± 2.5 kpcSchaefer, 2008 Metallicity0.4 Z  Westerlund, 1997 Mass of ionized gas8 × 10 5 M  Kennicutt, 1984 H  Luminosity 1.5 × 10 40 erg s -1 Kennicutt, 1984 FIR Luminosity4 × 10 7 L  Werner et al., 1978 Mass of molecular gasfew × 10 5 M  Johansson et al., 1998 Stellar mass of central cluster 5 × 10 4 M  Andersen et al., 2009 PhaseLocationStellar typesAge “Orion”near center (N&W)IR sources  1 Myr “Carina”center (R136)O, WN stars2-3 Myr “Scorpius OB1”everywhereOB SGs4-6 Myr “Hodge 301”3' NW of R136B/A/M SGs8-10 Myr Walborn & Blades, 1997 FLASH Talk. NOAO. November 19, 2010

27. Individual Regions IRAC 8  m Stellar Continuum HH R136High A V Bright [SIV] FLASH Talk. NOAO. November 19, 2010

28. Spectra from individual regions FLASH Talk. NOAO. November 19, 2010

29. Ingredients of the SED modeling  Stars: source of ionizing radiation UV stellar continuum.  ISM  Ionized gas: HII region  PDR material  Dust: silicates, carbonaceous material, PAHs  Time evolution: mechanical luminosity.  Multiplicity Based on sketch by Mike Bolte, Rick Waters & Brenda Wilden FLASH Talk. NOAO. November 19, 2010

30. Individual regions FLASH Talk. NOAO. November 19, 2010