Understanding Lyα Emission Using LBGs (and vice versa)

Transcription

1 Understanding Lyα Emission Using LBGs (and vice versa) Alice Shapley (UCLA) Collaborators: Kathy Kornei, Kristin Kulas, Juna Kollmeier, Dawn Erb, Anna Quider, Max Pettini, Chuck Steidel

2 z>1.5 Galaxy Identification and Spectroscopy (Steidel et al. 2004) (Smail et al. ) (Franx et al. 2003) (Gronwall et al. 2007) Explosion of z>1.5 surveys. Unlike traditional magnitudelimited surveys, new results utilize several complementary selection techniques for finding high-z galaxies: UV (LBG), DRG, BzK, submm, Lyα emitters (alphabet soup!!) All color-based techniques are incomplete wrt star formation and stellar mass. NB and submm selection also only yield incomplete slices of the galaxy population at any given redshift.

3 Rest-frame UV selection LBG technique, and extensions to both lower and higher redshift (figure shows z~ ), tuned to find galaxies with ongoing star-formation, and little to moderate extinction in the rest-uv (< 100). (Steidel et al. 2004) We have learned much about the large-scale spatial distribution, luminosity function, dust content, stellar populations, and starformation properties of UV-selected galaxies from z~1.5 to z>6. Spectroscopic redshifts have been obtained for >2000 UV-selected galaxies at z~ with R AB 25.5.

4 Rest-frame UV selection LBG technique, and extensions to both lower and higher redshift (figure shows z~ ), tuned to find galaxies with ongoing star-formation, and little to moderate extinction in the rest-uv (< 100). We have learned much about the large-scale spatial distribution, luminosity function, dust content, stellar populations, and starformation properties of UV-selected galaxies from z~1.5 to z>6. Spectroscopic redshifts have been obtained for >2000 UV-selected galaxies at z~ with R AB 25.5.

5 Overview The connection between Lyα and galaxy stellar populations. Lyα as a probe of the dusty ISM in star-forming galaxies. Lyα as a probe of large-scale outflows.

6 LAE/LBG Stellar Populations: A Controlled Comparison

7 LBG Lyα EWs The LBG technique yields objects spanning a broad range of Lyα properties, from strong absorption to strong emission. ~25% of LBGs at z~3 have rest-frame EW(Lyα)>20Å, the typical limit for NB-selected LAEs at the same redshift. Therefore, correlation of Lyα emission strength with other properties can be examined down to fixed magnitude/sfr/stellar mass limit. (Shapley et al. 2003) Properties of strong emitters can be differentiated as well.

8 z~3 LBG Stellar Populations & Lyα (Shapley et al. 2001) Based on a sample of 74 LBGs with Keck/NIRC UGRJKs photometry, young ( 35 Myr) and old ( 1 Gyr) rest-uv composite spectra constructed (N=16 for each composite). Younger galaxies characterized by weaker Lyα emission, redder UV continua, than older galaxies. Note: average composite spectrum of young/old objects not quite equivalent to average age as a function of EW(Lyα).

9 z~2 LBG Stellar Populations & Lyα On the other hand, based on sample of ~100 UV-selected galaxies with stellar population models, Erb et al. (2006) constructed 6 composite spectra based on stellar mass. Found that 1/3 (N=30) of sample w/ lowest stellar mass has stronger Lya emission, lower metallicity than 1/3 (N=28) with highest stellar mass. (Erb et al. 2006)s Age and mass are strongly correlated. Higher stellar mass bin significantly older stellar population ( 5).

10 z~2 LBG Stellar Populations & Lyα (Erb et al. 2006) Reddy et al. (2008) consider the stellar populations of strong (EW(Lyα) 20 Å) emitters vs. the rest of the population. 139 z~2 UV-selected galaxies, 14 of which have strong Lyα emission. K-S tests indicate no significant differences in the stellar pops of strong emitters, relative to rest of sample.

11 z>3.5 LBG Stellar Populations & Lyα Mass E(B-V) Age Sample of 47 LBGs at z= from the GOODS-S survey with multiwavelength photometry and stellar population fits. 19 have Lyα emission; 28 have absorption or nothing. Emitters significantly less massive and younger than non-emitters. Also less dusty. Emitters: EW(Lyα) 0 Non-Emitters: EW(Lyα)<0 (Pentericci et al. 2007)

12 z>3.5 LBG Stellar Populations & Lyα Closer examination of galaxies with Lyα emission at z= in the GOODS-S field. All 68 have EW(Lyα) 0. 38/68 have EW(Lyα) 20 Å. Most fit by young stellar populations, but small, significant fraction fit by ~1 Gyr models. Small fraction have stellar mass >10 10 M. Lack of high stellar mass objects with strong emission. (Pentericci et al. 2009) Strong Lyα emitters are a diverse population.

13 z>3.5 LBG Stellar Populations & Lyα Closer examination of galaxies with Lyα emission at z= in the GOODS-S field. All 68 have EW(Lyα) 0. 38/68 have EW(Lyα) 20 Å. Most fit by young stellar populations, but small, significant fraction fit by ~1 Gyr models. Small fraction have stellar mass >10 10 M. Lack of high stellar mass objects with strong emission. (Pentericci et al. 2009) Strong Lyα emitters are a diverse population.

14 More systematic study at z~3 248 z~3 LBGs with both UV spectra and near/mid-ir photometry (stellar population fits). Considered (1) correlations between EW(Lyα) and stellar populations; (2) binary comparisons between LAEs and non-laes. Correlations between EW(Lyα) and E(B-V), age, stellar mass, SFR. (Kornei et al. 2010) EW(Lyα) E(B-V) % EW(Lyα) Age EW(Lyα) SFR % EW(Lyα) uncorr. with M star

15 More systematic study at z~3 248 z~3 LBGs with both UV spectra and near/mid-ir photometry (stellar population fits). Considered (1) correlations between EW(Lyα) and stellar populations; (2) binary comparisons between LAEs and non-laes. Composite spectra as a function of E(B-V), age, stellar mass, SFR. (Kornei et al. 2010) EW(Lyα) E(B-V) % EW(Lyα) Age EW(Lyα) SFR % EW(Lyα) uncorr. with M star

16 More systematic study at z~3 248 z~3 LBGs with both UV spectra and near/mid-ir photometry (stellar population fits). Considered (1) correlations between EW(Lyα) and stellar populations; (2) binary comparisons between LAEs and non-laes. Average E(B-V), age, stellar mass, SFR as a function of EW(Lyα). (Kornei et al. 2010) EW(Lyα) E(B-V) % EW(Lyα) Age EW(Lyα) SFR % EW(Lyα) uncorr. with M star

17 More systematic study at z~3 248 z~3 LBGs with both UV spectra and near/mid-ir photometry (stellar population fits). Considered (1) correlations between EW(Lyα) and stellar populations; (2) binary comparisons between LAEs and non-laes. Relative age distributions of LAEs and non-laes. (Kornei et al. 2010)

18 More systematic study at z~3 248 z~3 LBGs with both UV spectra and near/mid-ir photometry (stellar population fits). Considered (1) correlations between EW(Lyα) and stellar populations; (2) binary comparisons between LAEs and non-laes. Average SEDs for LAEs, non-laes, and other, fainter LAE samples (Nilsson, Gawiser). (Kornei et al. 2010)

19 Other results at z~3 Pentericci et al. (2010) have sample of 130 U-drops in GOODS-MUSIC survey, at z~ Define 66 emitters and 66 nonemitters simply based on the presence or absence of Lya line in the spectrum (2 classified as both). Find that emitters are less massive (emitters black, non-emitters red), with bluer UV-continuum slopes (emitters green, non-emitters black). (Pentericci et al. 2010)

20 Other results at z~3 Pentericci et al. (2010) have sample of 130 U-drops in GOODS-MUSIC survey, at z~ Define 66 emitters and 66 nonemitters simply based on the presence or absence of Lya line in the spectrum (2 classified as both). Find that emitters are less massive (emitters black, non-emitters red), with bluer UV-continuum slopes (emitters green, non-emitters black). (Pentericci et al. 2010)

21 LBG/LAE stellar populations Results at z~2, 3, 4 and above apparently in contradiction? At least partially due to non-uniformity of investigations: independent/ dependent variable, EW(Lyα) range, sample selection effects, methods of modeling SEDs. Our in-depth analysis at z~3 shows that LBG/LAEs are older on average (relative lack of objects with t 100 Myr). Non-parametric results (G-R and R-K colors of LAEs and non-laes). Suggests that, among more luminous galaxies, evolution towards stronger Lyα emission. Strongest result at z 3: EW(Lyα) and E(B-V) anti-correlated (more coming up).

22 Lyα: a probe of the dusty ISM

23 Lyα Radiative Transfer & Dust In a uniform, neutral, dusty medium, Lyα will be preferentially attenuated relative to UV continuum. Resonant scattering of Lyα photons leads to effectively longer path length, greater probability of destruction. (Neufeld 1991) In multiphase medium (neutral, dusty clouds plus ionized, dust-free phase), in particular where dust is segregated from neutral gas, Lyα photons spend most time in dust-free intercloud medium. Higher probability of escape EW(Lyα) enhanced. Theory: Neufeld, Hansen & Oh

24 Lyα Radiative Transfer & Dust Finkelstein et al. have advanced model of clumpy dust to explain high Lyα EWs of several z~4.5 LAEs that also appear reddened by dust. Cite evidence from Giavalisco et al. (1996): IUE spectra of 21 local starburst galaxies for which there is no significant correlation between EW(Lyα) and β (UV continuum slope). (Giavalisco et al. 1996) How does this evidence compare with what we measure in LBGs at z 3?

25 Lyα Radiative Transfer & Dust In z~3 LBGs, there is a significant correlation between EW(Lyα) and E(B-V), such that EW(Lyα) increases as E(B-V) decreases. See also work by Pentericci et al. (2009), consistent results. Would such a correlation be expected if clumpy dust were dominant form of ISM? (Shapley et al. 2003, Kornei et al. 2009)

26 Lyα Radiative Transfer & Dust Also consider Lya escape fraction, f esc,lyα : f esc,lyα = L Lyα,obs /L Lyα,int Measure L Lyα,obs. Infer L Lyα,int based on SFR (i.e. SFR N ion L Hα L Lyα,int ). Average value of f esc,lyα is ~10%. Strong correlation between f esc,lyα and E(B-V). See also comparisons of SFR(Lyα) vs. SFR(UV) in Gronwall et al. (2007) and Pentericci et al. (2009). (Kornei et al. 2009) Models of Verhamme et al. (2008) predict correlation between f esc,lyα and E(B-V).

27 Lyα Radiative Transfer & Dust Can also investigate the relative extinction of Lyα and UV continuum photons, or relative escape fraction. Correct L Lyα,obs using E(B-V) derived from UV-continuum slope or SED fitting. f esc,rel = f esc E(B-V)κ(1216) If Lyα and UV continuum are attenuated with the same E(B-V), relative escape fraction is 1. (Kornei et al. 2010) We find average value of relative escape fraction is ~25%, i.e. Lyα is significantly more attenuated than UV continuum, consistent with simple expectations of Lyα radiative transfer.

28 Lyα Radiative Transfer & Dust If correct, extra attenuation of Lyα photons is inconsistent with a model in which dust is segregated in neutral clumps (Finkelstein et al.). We know that ISM is not simply singlephase mixture of dust and gas. However, result suggests that dust and gas are well-mixed among different phases of inhomogeneous ISM. (Kornei et al. 2010)

29 Lyα Radiative Transfer & Dust What if Lyα photons simply aren t in our slit? Calculation of f esc and f esc,rel based on assumption that Lyα and UV continuum have same spatial distribution and same fraction of light in the spectroscopic slit. In study of LAEs, Hayashino et al. (2004) find extended halo of Lya emission, going out to 5 in radius (40 kpc), significant flux in ring between 2 and 4. (Hayashino et al. 2004) Redo analysis of f esc in fields with large samples of LBGs with narrowband Lyα observations (i.e. protoclusters with objects all at the same redshift SSA22a, q1549, etc.).

30 Other f esc (Lyα) Studies (Atek et al. 2009) (Scarlata et al. 2009) In sample of 24 GALEX-selected LAEs at z~0.3, Atek et al. (2009) find anticorrelation between f esc and E(B-V) (based on Hα/Hβ ratio). Relative escape fraction is not uniformly < 1, indicating Lyα is not always attenuated more than UV continuum. Scarlata et al. 2009, 31 z=0.3 LAEs, find that Lyα/Hα always less than Case B value no evidence for clumpy dust distribution. Which sample is representative of high-z LAEs? What are the implications for the structure of the ISM in high-redshift starforming galaxies?

31 Other f esc (Lyα) Studies (Hayes et al. 2010) Hayes et al. (2010), blind NB survey for both Lyα and Hα emitters at z=2.2. For 67 objects (55 HAE, 18 LAE, including 6 common detections), find anti-correlation between fesc(lyα) and E(B-V). Find that extinction governing Lyα is significantly higher than that of the continuum. Use extinction-corrected Hα emission, and L(Lyα)/L(Hα)=8.7 to estimate f esc. HAE and LAE are disjoint populations.

32 Escape of LyC Radiation (Nestor et al. 2010)

33 LBG Lyα Profiles and Radiative Transfer Models

34 Feedback/Outflows M82: Nearby starburst galaxy Red is Hα emission from ionized gas above the plane of the galaxy. outflow speed: >500 km/s. outflows observed when Σ SFR >0.1M /yr/kpc 2 (Heckman et al. 2002). Ubiquitous in well-studied SF gals at high redshift. (Subaru Image)

35 Tour of an LBG Spectrum Features UV cont: O & B stars Lyα forest decrement Stellar: photospheric and wind Outflow-related: Lyα (Δv=+360), low and high ions (Δv=-170) Nebular emission Fine-structure emission (Shapley et al. 2003)

36 Kinematic signature of outflow Redshifts measured for interstellar absorption, Lyα emission, and systemic (nebular emission) differ. Low-ionization absorption features typically blueshifted (Δv=-150 km/s), Lyα emission typically redshifted (Δv= km/s) Observed at z~2-3. (Steidel et al. 2010) Latest: 90 galaxies at <z>=2.27 with NIRSPEC systemic redshifts.

37 Schematic Picture of an Outflow Kinematic signatures of large-scale motions are generic feature of z>2 starforming galaxies. Schematic diagram, courtesy of Max Pettini

38 Lyα Radiative Transfer Models LBGs exhibit a wide variety of Lyα profile shapes. Range from symmetric to asymmetric emission, to emission/absorption, to pure absorption, to nothing. Velocity structure is sometimes complex, with multiple peaks. Explaining this diversity in Lyα profiles within the context of a unified model for Lyα radiative transfer is an important goal. In rare cases where spatial information is available, it offers additional constraints.

39 Lyα Radiative Transfer Models (Verhamme et al. 2008) Much recent progress in detailed radiative transfer models of Lyα emission. Model from Verhamme et al. (2006, 2008): 3D Monte Carlo radiative transfer code for predicting Lyα profiles given arbitrary gas density and kinematics. Comparison of models with 11 Lya profiles at z=3-5 from Tapken et al. (2007). Spectra modeled in the context of expanding thin, cold shells of HI, with varying column densities, expansion velocities, and dust content. Good fits obtained.

40 Lyα Radiative Transfer Models (Kulas et al. 2010) Most Lyα model comparisons are done with crucial observational information missing: zeropoint of velocity field (i.e. systemic redshift); intrinsic width of Lyα (i.e. Hα or [OIII] velocity dispersion); escape fraction of Lyα photons (ratio of Lyα to Hα or other nebular lines). Sample of 18 z~2-3 UV-selected galaxies with multiple-peaked Lyα lines. Keck/ NIRSPEC measurements of Hα (z~2) or [OIII] (z~3) redshift, line-width and flux. All crucial inputs for model comparisons; dramatically reduce input assumptions.

41 Gravitationally Lensed LBGs Keck/ESI z=2.73 Gravitationally lensed LBGs provide an excellent opportunity for testing Lya radiative transfer models, because of the superior S/N and spectral resolution. For 10 years, MS1512-cB58 (z=2.73) was the single best example of a stronglylensed LBG. Lyα profile modeled by Schaerer et al. (2008). There are limitations with a sample of one, when you want to generalize to the full population i.e. how typical is cb58? Within the last few years, the number of strongly-lensed z 2 galaxies has increased dramatically. New lensed galaxies are from SDSS and cluster surveys.

42 Gravitationally Lensed LBGs New search technique in SDSS: look for multiple faint blue companions around luminous red galaxies. At least ~10 similar candidates, a few already spectroscopically confirmed, including the Cosmic Horseshoe, the Clone, 8:00 arc, Cosmic Eye. Cosmic horseshoe, z=2.38, x25 magnification, R-mag ~19.5 (Belokurov et al. 2007; Dye et al. 2008), has Lyα in emission. (INT imaging, SAO spectroscopy)

43 High-res Spectra of the Horseshoe Cosmic Horseshoe (z=2.38) (Discovery spectrum from Belokurov et al. 2007) In a very superficial sense, the restframe UV spectrum of the Horseshoe is strikingly different from that of cb58. Lyα in emission; IS absorption lines weaker. Non-unity (~60%) covering fraction of cool gas is inferred. In particular, Lyα is double-peaked. Offers opportunity to study the source of variations in rest-frame UV spectra in LBGs.

44 Lyα Probe of Outflows red blue In Verhamme models, different components of emergent multi-peaked profile correspond to photons that took different paths through an expanding shell. 1a, 2, 3: 0 backscatterings; 1b: 1 backscattering; 1c: 2+ backscatterings. Generic feature of models that yield multiple peaks: expanding shell has small velocity dispersion relative to expansion velocity.

45 Lyα Probe of Outflows (Quider et al. 2009) (Verhamme et al. 2006) Lyα is seen prominently in emission in Horseshoe spectrum. Profile doublepeaked, and invariant between 2 apertures. Expanding thin shell model from Verhamme et al. (2006) reproduces multipeaked form of emission line, and gradual fall-off towards higher velocities. (Compare red curve). Yields N(HI), crucial for estimating mass outflow rate!!! v

46 Lyα Probe of Outflows (Quider et al. 2009) (Verhamme et al. 2006) v Lyα is seen prominently in emission in Horseshoe spectrum. Profile doublepeaked, and invariant between 2 apertures. Expanding thin shell model from Verhamme et al. (2006) reproduces multipeaked form of emission line, and gradual fall-off towards higher velocities. (Compare red curve). Yields N(HI), crucial for estimating mass outflow rate!!!

47 Lyα Probe of Outflows (Quider et al. 2009) (Verhamme et al. 2006) BUT: IS absorption lines suggest range of velocities in outflowing interstellar gas at least comparable to outflow speed, extends over ~1000 km/s. In such cases, the models predict peaks become blurred, don t resemble Horseshoe Lyα spectrum. ALSO: covering fraction < 100% (model has 100%). Fundamental problem with Lyα radiative transfer modeling. Shell model is not correct for high-redshift outflows! We need to do better!

48 Summary Stronger Lyα emission is correlated with a bluer UV continuum, with no evidence for clumpy dust. Definitive results await about stellar mass and age, and an evolutionary scenario for LBGs regarding Lyα emission. At z~3, LBGs appear to evolve towards stronger Lyα emission on average. Gravitationally lensed LBGs offer excellent test of Lyα radiative transfer models, calling the thin, expanding shell model into question.