1. Obscured Star Formation in Small Galaxies out to z<1 MPIA: Xianzhong Zheng, Eric F. Bell, Hans-Walter Rix Steward Obs.: George H. Rieke, Casey Papovich, Emeric Le Floch, Pablo G. Perez-Gonzalez Extreme Starbursts: Near and Far, Lijiang, August 19, 2005

2. Motivation A complete understanding of galaxy formation and evolution: massive + small Is star formation hidden by dust in intermediate- or high-z small galaxies? High z: we need relatively precise dust extinction correction for UV-Optical SFR estimators. To measure SFR (L IR + L UV ), L IR is essential. However, IR observation is limited by confusion noise. Stacking technique allows us to detect mean fluxes below the individual detection limit.

3. CDFS: COMBO-17 + MIPS(24µm) M* --- LFs of VVDS Survey (Ilbert et al. 2005)

4. Test stacking Confusion noise decreases as inverse root square of number of images stacked.

5. Examples of stacked images

6. Average SEDs (M B, z) Nine redshift slices over 0.1<z<1 are given top- bottom and left-right. In each panel, average SEDs are shown for subpopulations with different B-band magnitude relative to characteristic M*. Massive galaxies show a redder optical color and a higher 24µm luminosity than small galaxies. Three templates normal spiral (Sbc1987), starburst (M82) and ULIRG (Arp220) are used to convert 24µm to total IR luminosity. -16 -18 -20 -22

7. How much of population-averaged 24µm luminosity is ascribed to unresolved sources? In each panel: Black region: individually- resolved; Blank region: unresolved (stacking) Left side: Ind.ly-resolved + mean stacking (upper limit); right side: Ind.ly-resolved + median stacking (lower limit) More than 50% are unresolved sources. Stack number: from ~50 to ~500 Detection by stacking is crucial to fully understand IR emission from even bright galaxies. massivesmall

8. Dust obscuration L IR /L UV as a function of M B over 0.1<z<1 In the local universe, it is well known that L IR /L UV is correlated with M B. The correlation reveals that relatively more UV lights escape from smaller galaxies. massive small Left figure shows that this relation holds out to 0.8 (where our sample is complete). Wang & Heckman (1996)

9. A universal correlation between Dust obscuration L IR /L UV and SFR L IR +L UV ⇒ SFR Dotted line is the local relation (Martin et al. 2005). To first order, correlation between L IR /L UV and SFR is consistent with the local relation, independent from luminosity and redshift. L IR /L UV ∝ Z metal × ∑ gas ×α geom. SFR ∝ ( ∑ gas ) 1.4 Higher SFR and IR/UV driven by evolution in gas density?

10. Estimate the total IR luminosity density from the B-band LF

11. Total IR luminosity density vs redshift Power law index for IR luminosity function:1.2  0.3 Local value: 1.23

12. Conclusions Over 0.1<z<1, dust obscuration L IR /L UV is of dependence on luminosity, i.e. more UV lights escape from smaller galaxies. A universal correlation between L IR /L UV and SFR holds up to z~0.8, suggesting that gas amount plays a more important role than metallicity. Total IR luminosity density independently derived from the B-band LF is excellently consistent with that from the IR LF. The result suggests a flat end slope of the IR LF with a power law index of 1.2  0.3.