Direct Constraints on UV Attenuation, Dust- Obscured Star Formation, and Dust Mass of Z=6.5-7.5 Galaxies
Total Page:16
File Type:pdf, Size:1020Kb
Direct constraints on UV attenuation, dust- obscured star formation, and dust mass of z=6.5-7.5 galaxies! Daniel Schaerer (Geneva Observatory, CNRS)! Fréderic Boone (IRAP, Toulouse)! Other collaborators: M. Zamojski, M. Dessauges-Zavadsky, J. Staguhn, S. Finkelstein, F. Combes! ! • Introduction, motivation! • New constraints on z>6 star-forming galaxies! • Conclusions! Schaerer et al. 2014, A&A in press; arXiv:1407.5793! de Barros, Schaerer, Stark, 2014, A&A 563, A81! Schaerer & de Barros 2014, in prep! from the ratio of FIR to observed (uncorrected) FUV luminosity densities (Figure 8) as a function of redshift, using FUVLFs from Cucciati et al. (2012) and Herschel FIRLFs from Gruppioni et al. (2013). At z<2, these estimates agree reasonably well with the measure- ments inferred from the UV slope or from SED fitting. At z>2, the FIR/FUV estimates have large uncertainties owing to the similarly large uncertainties required to extrapolate the observed FIRLF to a total luminosity density. The values are larger than those for the UV-selected surveys, particularly when compared with the UV values extrapolated to very faint luminosities. Although galaxies with lower SFRs may have reduced extinction, purely UV-selected samples at high redshift may also be biased against dusty star-forming galaxies. As we noted above, a robust census for star-forminggalaxiesatz ! 2selected on the basis of dust emission alone does not exist, owing to thesensitivitylimitsofpast and present FIR and submillimeter observatories. Accordingly, the total amount of star formation that is missed from UV surveys at such high redshifts remains uncertain. Introduction! Cosmic star formation rate history! ! review of Madau & Dickinson (2014)! Figure 9: The history of cosmic star formation from (top right panel) FUV, (bottom right panel) IR, Majorand (left unknowns panel) FUV+IR rest-frame: ! measurements. The data points with symbols are given in Table 1. All UV luminosities have been converted to instantaneousFigureSFR 8: densities(Left panel) using SFR densities the factor in the FUVKFUV (uncorrected= for dust attenuation)andintheFIR. −28 • 1.15contribution× 10 (see Equation of 10), different valid for a Salpetergalaxy IMF. Thepopulations FIR data luminositie points with symbolss(8–1,000! are givenµm) in Table have 1. Allbeen UV and IR luminosities have been converted to instantaneous−44 SFR densities using the factors K =1.15 × 10−28 and K =4.5 × 10−44 (cgs units) converted to instantaneous SFRs using the factor KIR =4.5 × 10 (see Equation 11), also validFUV for a IR • dust-obscured star formation! valid for a Salpeter IMF. (Right panel)Meandustattenuationinmagnitudesasafunctionofredshift. Salpeter IMF. The solid curve in the three panels plots the beMostst-fit of the SFRD data points in Equation shown are based 15. on ultraviolet spectral slopes or stellar population model fitting. The symbol shapes and colors correspond to the data sets citedinTable1,withtheadditionofSalim • attenuation correction! et al. (2007) (cyan pentagon). Two versions of the attenuation factors are shown for UV-selected galaxies Figure 9 shows the cosmic SFH from UV and IR dataat 2 <z< following7(Reddy&Steidel2009,Bouwensetal.2012a)(o the above prescriptions, ffset slightlyintheredshiftaxisfor clarity): one integrated over the observed population (open symbols), the other extrapolated down down as well as the best-fitting function to LFUV =0(filled symbols). Data points from Burgarella et al. (2013) (olive green dots)arecalculated by comparing the integrated FIR and FUV luminosity densitiesinredshiftbins,ratherthanfromtheUV 2.7 slopes or UV-optical SEDs. (1 + z) −1 −3 ψ(z)=0.015 M! year Mpc . (15) 1+[(1+z)/2.9]5.6 Driver et al. (2008). We note that the estimates of UV attenuation in the local Universe These state-of-the-art surveys provide a remarkablyspan consi astent broad range, picture suggesting of the that cosmic more work SFH: needs to be done to firmly pin down this −2.9 <quantity< (and perhaps implying that we should be cautious about the estimates at higher arisingphase,scalingasψ(z) ∝ (1 + z) at 3 ∼redshift).z ∼ 8, Several slowing studies and of UV-selected peaking atgalaxies some at z ≥ 2(Reddy&Steidel2009,Bouwens point probably between z =2and1.5,whentheUniversewaset al. 2012a, Finkelstein∼ 3.5Gyrold,followedby et al. 2012b) have noted strong trends for less luminous galaxies as having bluer UV spectral slopes and, hence, lower inferred dust attenuation. Because the faint-end slope of the far-UV luminosity function (FUVLF) issosteepathighredshift,a 48 P. Madau & M. Dickinson large fraction of the reddened FUV luminosity density is emitted by galaxies much fainter than L∗;thisextinction–luminositytrendalsoimpliesthatthenetextinctionfortheentire population will be a function of the faint integration limit for the sample. In Figure 8, the points from Reddy & Steidel (2009) (at z =2.3and3.05)andfromBouwensetal. (2012a) (at 2.5 ≤ z ≤ 7) are shown for two faint-end integration limits: These are roughly down to the observed faint limit of the data, MFUV < −17.5to−17.7forthedifferent redshift subsamples and extrapolated to LFUV =0.Thenetattenuationforthebrighter limit, which more closely represents the sample of galaxies actually observed in the study, is significantly larger than for the extrapolation – nearly two times larger for the Reddy & Steidel (2009) samples, and by a lesser factor for the more distant objects from Bouwens et al. (2012a). In our analysis of the SFRDs, we have adopted the mean extinction factors inferred by each survey to correct the corresponding FUV luminosity densities. Adopting a different approach, Burgarella et al. (2013) measured total UV attenuation Cosmic Star-Formation History 47 Dominguez et al. (2013)! Introduction!The Astrophysical Journal,763:145(10pp),2013February1 Dom´ınguez et al. 10 9 8 Measurement of dust attenuation! 7 β 6 H ! / α H 5 Main methods:! 4 1. Balmer decrement: Hα/Hβ ratios measured 3 2 out to z~2 (ground-based + HST WFC3 401 .040.541.041.542.042.543.0 89101112 1.01.52.02.53.03.5 1 log10(LHα/erg s− ) log10(M /M ) log (EW /A˚ ) ∗ Whitaker10 Hα et grism)" Figure 7. Balmer decrements vs. observed Hα luminosity (left panel), galaxy stellar mass (middle panel), and rest-frame Hα equivalent width (right panel). In all The Astrophysical Journal,793:115(35pp),2014October1 the panels, the open red triangles show the stacked spectra from the whole redshift rangeBouwens without et correcting al. for Hα and Hβ lineal. absorption (2013) or [N ii]contamination.! The open red squares are the Balmer decrements that include the absorption-line corrections (the uncertainties of the open symbols are omitted for clarity). The filled Heinis et al. (2013)! red stars show the stacked spectra from the whole redshift range 0.75 z 1.5, which are corrected for both Hα and Hβ absorption lines and the [N ii]doublet ! special 0!!.03-reductionsDust generated attenuation for and the SFR XDF, from HUDF09-1,z ∼ 4 to z ∼ 1. and57 ! ! HUDF09-2 data sets to ensure more optimal results. contamination (see Section 2.6 for details). All the values represented in this figure are listed in Table 2. The dashed-magenta line in the left panel shows the results z=1.5 z=3 z=4 by Hopkins et al. (2001). A dashed-blue line in the middle panel shows the relationship provided by Garn & Best (2010)forlocalgalaxies.Theresultsatz 1.5 11.5 Source selection and photometry for our z 4, presentedz 2.55, by Sobral et al. (2012)areshownwithgreencirclesintheleftandmiddlepanels.Thecontoursshowstar-formingandAGNSDSSgalaxiesforcomparison.∼ 24 29 32 16 3 ∼ 2 ∼ and z 6samplesisbasedonthesquarerootof48 35 19 17 χTheimage dotted-black line shows the intrinsic Balmer decrement assumed in our analysis of 2.86. ∼ (A color2.1 version of this figure is available in the online journal.) 11.0 79 161 (Szalay148 84 et11 al. 1999)constructedfromtheV606Y105JH140H160- 2. IR/UV ratio! 186 214 82 28 161 52 band, J125JH140-band, and J125JH140-band images, respectively. 169 366 265 140 22 1.7 3.0 The Y or H band images were not used in constructing) the ]) 10.5 105 160 FUV • O 880 691 369 113 837 271 0.75 z 1.5 Observed 320 553 570 310 61 /L 2 IR ≤ ≤ [M χ image for our z 5andz 6samples(ortheJ -band * 125 1.3 Extinction corrected ∼ ∼ 1 2.5 374 1025 1044 550 94 image for our z 27524samples),givenouruseoftheseimages2279 1007 2214 10 log(M 10.0 ∼ IRX = log(L 2 to derive β.Sourcesarerequiredtobedetectedat5σ in the χ0.9 691 1914 1980 806 107 ) 2.0 α image to ensure that they are real. β H H 9.5 1657 3788 2450 / Using the same simulations as provided in Appendix/ 0.5F, α 5007 1.5 3044 8 (H we verified that the above selection criteria resulted in very λ 9.0 10 0.1 9.5 10.0 10.5 11.0 10.0 10.5 11.0 11.510.2 10.4 10.6 10.8 11.0 11.2 11.4 similar mean redshifts for our z 4, z 5, and z log 6 log(L [L •]) log(L [L •]) log(L [L •]) [OIII] selectionsFUV O from the deepFUV XDF+HUDF09-PsO ∼ ∼ FUV dataO sets as∼ for 1.0 Figure 4. Dust attenuationour as shallower a function of stellar ERS+CANDELS mass and UV luminosity at z ∼ selections1.5(left),z ∼ 3(middle)and since thez ∼ respective4(right).Thecolor codes IRX = log(LIR/LUV)ineachcellwherethestackingmeasurementismeaningful.Filled cells indicate stacking measurements with S/N > 3inthe3SPIREbands,whileblackhatchedcellsshowmeasureselection criteria differed onlyments with slightly. at most two The SPIRE mean bands with redshiftsS/N > 3. The number we 0.5 of galaxies contributing to the stacking is indicated in eachcell.Thecontoursshowthedistributionofgalaxiesinthe(LFUV,M∗)plane. The empty circles showfound the mean stellar for massz for5 a given and UVz luminosity6 for bin, withboth the selectionsdispersion as error were bar.