the cosmic horseshoe
DESCRIPTION
This is an overview of my research into the gravitationally lensed galaxy named the Cosmic Horseshoe. This talk includes some background information as I have aimed this presentation at physics graduate students. Without additional explanation it is probably only suitable for an expert audience.TRANSCRIPT
The Cosmic Horseshoe:The Rest-frame UV Spectrum of a z~2 LBG
Anna QuiderInstitute of Astronomy
University of Cambridge
Max Pettini (IoA), Alice Shapley (UCLA), Charles Steidel (Caltech)
Today’s Talk
‣Overview of Lyman Break Galaxies (LBGs) and rest-frame UV spectroscopy
‣Results from the Cosmic Horseshoe- Stellar spectrum- Interstellar spectrum- Lyman alpha emission feature
‣Broader conclusions from the Cosmic Horseshoe
‣Summary
Anna Quider Institute of AstronomyThe Cosmic Horseshoe: UV Spectrum
The Universe
The Universe
Galaxies from about 2 to 3 billion years after the Big Bang
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
What is a LBG?
‣ High z starforming galaxy identified by the Lyman break photometric selection technique
‣ Criteria for 2 ≤ z ≤ 2.5:R ≤ 25.5G - R ≥ -0.2G - R ≤ 0.2 (Un - G) +0.4(Un - G) ≥ (G - R) + 0.2(Un - G) ≤ (G - R) + 1.0(Steidel et al. 2004)
(Adelberger et al. 2004)
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Cosmic Context
zk 6 to z ! 3 is motivated by the apparent evolution in the UV-continuum slope over this redshift range (e.g., Stanway et al. 2005;Bouwens et al. 2006) and the correlation of UV-continuum slopewith dust extinction (e.g., Meurer et al. 1999; Reddy et al. 2006).
5. DISCUSSION
5.1. How Does the Volume Densityof UV-Bright Galaxies Evolve?
Over the past few years, there has been some controversy re-garding how theUVLFevolves at high redshift.While some stud-ies have argued that the evolution primarily occurs at the brightend (i.e., Dickinson et al. 2004; Shimasaku et al. 2005;Ouchi et al.2004a; Bouwens et al. 2006, 2007; Yoshida et al. 2006), therehave been other efforts which have argued that the evolution oc-
curs primarily at the faint end (i.e., Iwata et al. 2003; Sawicki &Thompson 2006; Iwata et al. 2007). In this work, we found ad-ditional evidence to support the fact that the most rapid evolutionoccurs at the bright end of the UV LF and that this evolution canbe approximately described by a change in the characteristic lu-minosity M "
UV of the UV LF.A significant part of the debate regarding the form the evolu-
tion of the UV LF takes at high redshift is centered on what hap-pens at the bright end of the LF (i.e.,MUV;AB P#20:5). Does itevolve or not? Iwata et al. (2007) and Sawicki & Thompson(2006) argue that there is little evidence for substantial evolutionfrom z ! 5 to z ! 3 from their work. However, a quick analysisof results in the literature indicate that such evolution is certainlyvery strong, particularly if the baseline is extended out to evenhigher redshifts (zk 6). At z ! 6, for example, the bright end ofthe UV LF (i.e., MUV;AB < #20:5) has been reported to be de-ficient by factors of !6Y11 relative to z ! 3Y4 (Stanway et al.2003, 2004; Shimasaku et al. 2005; Bouwens et al. 2006). Atz k 7, this deficit is even larger, as we can see by comparing thenumber of UV bright galaxies (i.e., MUV;AB < #20:5) in ourdropout selections with that expected from z ! 4 (Bouwens et al.2007) assuming no evolution. The comparisons are presented inTable 7. At z ! 7, the volume density of UV bright (MUV;AB <#20:5) galaxies appears to be 18$32
#11 times smaller than at z ! 4(see also Bouwens & Illingworth 2006; Mannucci et al. 2007;Stanway et al. 2008), and at z ! 9, the volume density of UVbright galaxies is at least 16 times smaller (68% confidence).
The only analysis to find mild evolution in the volume densityof UV bright galaxies from z ! 7Y10 to z ! 3Y4 is that of Richardet al. (2006) and involves a search for z % 6 galaxies behind lensingclusters. However, as we argue in Appendix C, it seems likelythat a substantial fraction of the UV-bright galaxies in the Richardet al. (2006) z ! 6Y10 selection are spurious and therefore thebright end of their UVLFs ismuch too high. The evidence for thisis rather striking: not only are none of the z % 6 candidates in theRichard et al. (2006) selection detected (<2 !) in significantlydeeper (!1Y2 mag) NICMOS+IRAC data where available (11 oftheir z % 6 candidates), but also their reported prevalence is >10times higher than what we measure in searches for similar can-didates behind lensing clusters (Bouwens et al. 2008; see alsoAppendix C).
TABLE 6
UV Luminosity Densities and Star Formation Rate Densities
log10 SFR Density(M& Mpc#3 yr#1)
Sample hzilog10L
(ergs s#1 Hz#1 Mpc#3) Uncorrected Correcteda
z .............. 7.3 25.32 ' 0.21 #2.58 ' 0.21 #2.40 ' 0.21
J.............. 9.0 <25.14b <#2.76b <#2.58b
B ............. 3.8 26.18 ' 0.05 #1.72 ' 0.05 #1.29 ' 0.05
V ............. 5.0 25.85 ' 0.06 #2.05 ' 0.06 #1.75 ' 0.06
i .............. 5.9 25.72 ' 0.08 #2.18 ' 0.08 #2.00 ' 0.08
Notes.—Integrated down to 0.2 L"z(3. Based on LF parameters in Table 4 (seex 4.3).
a The adopted dust-extinction corrections are 0.4 mag, 0.6 mag, and 1.1 magat zk 6, z ! 5, and z ! 4, respectively. Our use of an evolving dust correctionfrom zk 6 to z ! 3 is motivated by the apparent evolution in UV-continuumslope over this redshift range (e.g., Stanway et al. 2005; Bouwens et al. 2006) andthe correlation of UV-continuum slope with dust extinction (e.g., Meurer et al.1999; Reddy et al. 2006).
b Upper limits here are 1 ! (68% confidence).
Fig. 9.—Estimated star formation rate density as a function of redshift (inte-grated down to 0.2 L"z(3 as in Fig. 8). The lower set of points give the SFR densitywithout a correction for dust extinction, and the upper set of points give the SFRdensity with such a correction. This is also indicated with the shaded blue and redregions, respectively, where the width of these regions show the approximate un-certainties estimated by Schiminovich et al. (2005). At lower redshift (zP 3), weadopt the dust correction suggested by Schiminovich et al. (2005). At zk 6, weadopt the dust correction obtained by Bouwens et al. (2006; see also Stark et al.2007a and Stanway et al. 2005) at z ! 6 from the UV-continuum of i-dropoutsand the Meurer et al. (1999) IRX-" prescription. At z ! 4 and z ! 5, we inter-polate between the estimated dust extinctions at z ! 3 and z ! 6. The symbols arethe same as in Fig. 8.
TABLE 7
Comparison of the Observed Numbers of Dropoutswith No-Evolution Extrapolations from z ! 4
Number
Sample Observed z ! 4 Prediction Evolutionary Factora
Bright (MUV;AB < #20:5)
z-dropout .............. 2 35.9 18$32#11
J-dropout .............. 0 17.9 >16b
Faint (MUV;AB > #20:5)
z-dropout .............. 6 24.7 4$3#2
J-dropout .............. 0 8.0 >7b
Notes.—Extrapolations assume no evolution in the UV LF from z ! 4(Bouwens et al. 2007) and were calculated using eq. (1). The sizes and UVcolors of UV-bright galaxies at zk 7 assumed for these estimates are the sameas those given in Appendix B.
a Ratio of the volume density of galaxies at z ! 4 (Bouwens et al. 2007) insome luminosity range to that in our higher redshift dropout samples.
b Lower limits here are 1 ! (68% confidence).
z ! 7Y10 GALAXIES IN HUDF AND GOODS FIELDS 241No. 1, 2008
(Bouwens et al. 2008)
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Cosmic Context
zk 6 to z ! 3 is motivated by the apparent evolution in the UV-continuum slope over this redshift range (e.g., Stanway et al. 2005;Bouwens et al. 2006) and the correlation of UV-continuum slopewith dust extinction (e.g., Meurer et al. 1999; Reddy et al. 2006).
5. DISCUSSION
5.1. How Does the Volume Densityof UV-Bright Galaxies Evolve?
Over the past few years, there has been some controversy re-garding how theUVLFevolves at high redshift.While some stud-ies have argued that the evolution primarily occurs at the brightend (i.e., Dickinson et al. 2004; Shimasaku et al. 2005;Ouchi et al.2004a; Bouwens et al. 2006, 2007; Yoshida et al. 2006), therehave been other efforts which have argued that the evolution oc-
curs primarily at the faint end (i.e., Iwata et al. 2003; Sawicki &Thompson 2006; Iwata et al. 2007). In this work, we found ad-ditional evidence to support the fact that the most rapid evolutionoccurs at the bright end of the UV LF and that this evolution canbe approximately described by a change in the characteristic lu-minosity M "
UV of the UV LF.A significant part of the debate regarding the form the evolu-
tion of the UV LF takes at high redshift is centered on what hap-pens at the bright end of the LF (i.e.,MUV;AB P#20:5). Does itevolve or not? Iwata et al. (2007) and Sawicki & Thompson(2006) argue that there is little evidence for substantial evolutionfrom z ! 5 to z ! 3 from their work. However, a quick analysisof results in the literature indicate that such evolution is certainlyvery strong, particularly if the baseline is extended out to evenhigher redshifts (zk 6). At z ! 6, for example, the bright end ofthe UV LF (i.e., MUV;AB < #20:5) has been reported to be de-ficient by factors of !6Y11 relative to z ! 3Y4 (Stanway et al.2003, 2004; Shimasaku et al. 2005; Bouwens et al. 2006). Atz k 7, this deficit is even larger, as we can see by comparing thenumber of UV bright galaxies (i.e., MUV;AB < #20:5) in ourdropout selections with that expected from z ! 4 (Bouwens et al.2007) assuming no evolution. The comparisons are presented inTable 7. At z ! 7, the volume density of UV bright (MUV;AB <#20:5) galaxies appears to be 18$32
#11 times smaller than at z ! 4(see also Bouwens & Illingworth 2006; Mannucci et al. 2007;Stanway et al. 2008), and at z ! 9, the volume density of UVbright galaxies is at least 16 times smaller (68% confidence).
The only analysis to find mild evolution in the volume densityof UV bright galaxies from z ! 7Y10 to z ! 3Y4 is that of Richardet al. (2006) and involves a search for z % 6 galaxies behind lensingclusters. However, as we argue in Appendix C, it seems likelythat a substantial fraction of the UV-bright galaxies in the Richardet al. (2006) z ! 6Y10 selection are spurious and therefore thebright end of their UVLFs ismuch too high. The evidence for thisis rather striking: not only are none of the z % 6 candidates in theRichard et al. (2006) selection detected (<2 !) in significantlydeeper (!1Y2 mag) NICMOS+IRAC data where available (11 oftheir z % 6 candidates), but also their reported prevalence is >10times higher than what we measure in searches for similar can-didates behind lensing clusters (Bouwens et al. 2008; see alsoAppendix C).
TABLE 6
UV Luminosity Densities and Star Formation Rate Densities
log10 SFR Density(M& Mpc#3 yr#1)
Sample hzilog10L
(ergs s#1 Hz#1 Mpc#3) Uncorrected Correcteda
z .............. 7.3 25.32 ' 0.21 #2.58 ' 0.21 #2.40 ' 0.21
J.............. 9.0 <25.14b <#2.76b <#2.58b
B ............. 3.8 26.18 ' 0.05 #1.72 ' 0.05 #1.29 ' 0.05
V ............. 5.0 25.85 ' 0.06 #2.05 ' 0.06 #1.75 ' 0.06
i .............. 5.9 25.72 ' 0.08 #2.18 ' 0.08 #2.00 ' 0.08
Notes.—Integrated down to 0.2 L"z(3. Based on LF parameters in Table 4 (seex 4.3).
a The adopted dust-extinction corrections are 0.4 mag, 0.6 mag, and 1.1 magat zk 6, z ! 5, and z ! 4, respectively. Our use of an evolving dust correctionfrom zk 6 to z ! 3 is motivated by the apparent evolution in UV-continuumslope over this redshift range (e.g., Stanway et al. 2005; Bouwens et al. 2006) andthe correlation of UV-continuum slope with dust extinction (e.g., Meurer et al.1999; Reddy et al. 2006).
b Upper limits here are 1 ! (68% confidence).
Fig. 9.—Estimated star formation rate density as a function of redshift (inte-grated down to 0.2 L"z(3 as in Fig. 8). The lower set of points give the SFR densitywithout a correction for dust extinction, and the upper set of points give the SFRdensity with such a correction. This is also indicated with the shaded blue and redregions, respectively, where the width of these regions show the approximate un-certainties estimated by Schiminovich et al. (2005). At lower redshift (zP 3), weadopt the dust correction suggested by Schiminovich et al. (2005). At zk 6, weadopt the dust correction obtained by Bouwens et al. (2006; see also Stark et al.2007a and Stanway et al. 2005) at z ! 6 from the UV-continuum of i-dropoutsand the Meurer et al. (1999) IRX-" prescription. At z ! 4 and z ! 5, we inter-polate between the estimated dust extinctions at z ! 3 and z ! 6. The symbols arethe same as in Fig. 8.
TABLE 7
Comparison of the Observed Numbers of Dropoutswith No-Evolution Extrapolations from z ! 4
Number
Sample Observed z ! 4 Prediction Evolutionary Factora
Bright (MUV;AB < #20:5)
z-dropout .............. 2 35.9 18$32#11
J-dropout .............. 0 17.9 >16b
Faint (MUV;AB > #20:5)
z-dropout .............. 6 24.7 4$3#2
J-dropout .............. 0 8.0 >7b
Notes.—Extrapolations assume no evolution in the UV LF from z ! 4(Bouwens et al. 2007) and were calculated using eq. (1). The sizes and UVcolors of UV-bright galaxies at zk 7 assumed for these estimates are the sameas those given in Appendix B.
a Ratio of the volume density of galaxies at z ! 4 (Bouwens et al. 2007) insome luminosity range to that in our higher redshift dropout samples.
b Lower limits here are 1 ! (68% confidence).
z ! 7Y10 GALAXIES IN HUDF AND GOODS FIELDS 241No. 1, 2008
(Bouwens et al. 2008)
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
(Bouwens et al. 2008)
Local Galaxies
2
Figure 1: HST/ACS images of 25 galaxies at z ~ 2 in the GOODS-N field, each box is 3 arcsec on
a side (from Law et al. 2007b). Note the complex morphologies of these systems.
2. Kinematics:
Given the complications of such morphological studies, invaluable additional information can be
gleaned from the nebular emission lines (e.g., H!, H", [O III], [O II], and [N II]) which provide
good kinematic tracers of the ionized gas surrounding active star forming regions. Such nebular
emission-line spectroscopy has been used (e.g.) to trace the evolution in the Tully-Fisher relation
out to z ~ 1.2 (Weiner et al. 2006; Kassin et al. 2007) and suggests the growing importance of
non-circular motions to this relation with increasing redshift. In combination with new adaptive
optics (AO; e.g. Wizinowich et al. 2006) and integral-field unit (IFU) technologies it has
additionally become possible for ground-based telescopes to overcome the limitations imposed
by atmospheric turbulence and “dissect” galaxies with spectroscopy on hitherto-unprobed sub-
kiloparsec scales.
IFU studies at z ~ 1.5 (Wright et al. 2007, 2009) confirm a slight increase in non-circular
motions relative to the local universe, but also find evidence of organized rotation within their
galaxy sample. At higher redshifts, numerous studies at z ~ 2 - 3 (e.g., Förster-Schreiber et al.
2006; Genzel et al. 2008; Law et al. 2007a, 2009; Nesvadba et al. 2008) have found that galaxies
have extremely large velocity dispersions (# ~ 80 km s-1
) as compared to their rotational velocity
(V) about a preferred kinematic axis. While exact values of the ratio V/# vary from less than
unity up to about 4 – 5, there is clearly a dynamical difference from disk galaxies in the nearby
universe, which typically have V/# ~ 15 – 20 (e.g., Dib et al. 2006).
Early Galaxies
galaxy?
Stars
Dust
Gas
Also: planets, comets, asteroids...but too small to see
What’s visible in a galaxy?
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
galaxy?galaxy?
How can you study galaxies?Using Spectroscopy!
Light is split into its component wavelengths so that we can directly study the stars and gas in the galaxy
Light from stars
Hot, ionized gas close to stars
Cold gas between stars
Text
Studying Galaxies Using Spectroscopy
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Stars are BlackbodiesVery red (cool)
Very blue (hot)
• Different wavelengths probe different stellar populations
• Rest-frame UV spectroscopy probes the most massive, youngest stars
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Aside: Astronomy Naming Conventions
Any element heavier than Hydrogen or Helium is called a “metal” (e.g. C, N, O, Fe, Ni, Si, etc.)
O I = neutral oxygenO II = singly ionized oxygenO III = doubly ionized oxygenetc.
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Rest-frame Optical Spectra
H II region emission lines are very visible
and therefore are relatively easy to
study
(Erb et al. 2006a)
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Rest-frame Optical Spectra
(Erb et al. 2006a)
SFR ~23 MO/yr
Z 0.4 to 1.0 ZO
E(B-V) 0.15
σ ~100 km/s
Age 570 Myr
M❋ 2 x 1010 MO
Median Values for z~2 LBGs
(Erb et al. 2006a,b,c)
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Rest UV Spectral Features
A mix of features from hot OB stars, low and high ionization interstellar gas, and the H II regions
(Shapley et al. 2003)
Low Ion IS AbsHigh Ion IS AbsStellar AbsNebular EmStellar EmH I Em/Abs
Anna Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
What about a detailed study of the stars, interstellar gas, and H II regions
in an individual LBG?
The answer: Strongly gravitationally lensed LBGs!
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Gravitational Lensing
Gravity distorts and magnifies light from distant galaxies
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
MS1512-cB58
Nitrogen
‣ Serendipitously found in cluster MS1512+36 (z=0.37)
‣ zcB58 = 2.7276
‣ Magnified ~ 30x and L ~ L*
(www.eso.org)
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
MS1512-cB58
‣ Stellar population (metallicity, IMF)
‣ Interstellar abundances
‣ Large-scale outflows
‣ Lyman-α feature morphology
Nitrogen
α-capture
Fe-peak
Rel
ativ
e Fl
ux
Relative Velocity (km s-1)
Relative Velocity (km s-1)
Figures from Pettini et al. 2002
Rel
ativ
e Fl
ux
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Cosmic Horseshoe
‣ 10” Einstein Ring‣Discovered by Belokurov et al. (2007) in SDSS‣ 24±2x magnification (Dye et al. 2008)‣ L ≈ 2.4L*
‣ zCH = 2.38115
From rest-frame optical spectrum:‣ SFR = 100 MO/yr‣Mvir ≈ 1.4 x 1010 MO
‣Z ≈ 0.5-1.5 ZO
(Hainline et al. 2009)
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
ESI Spectrum
‣Echellette Spectrograph and Imager (ESI) spectrum
-Keck II telescope
- 4000 - 10000 Å coverage(1184 - 2959 Å restframe)
- 11.4 km s-1 pixel-1 resolution
- 36100s total exposure
- Spectra of two knots
(Quider et al. 2009; courtesy of Dr. Lindsay King)
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of AstronomyAnna Quider
10 ly
A high z star-forming galaxy has many regions like this
The Cosmic Horseshoe: UV Spectrum Institute of AstronomyAnna Quider
To Earth
Hot, young stars
Cavity caused by stellar wind
10 ly
A high z star-forming galaxy has many regions like this
The Cosmic Horseshoe: UV Spectrum Institute of AstronomyAnna Quider
To Earth
Hot, young stars
Cavity caused by stellar wind
Gas being ionized by the young, hot stars
(a H II region)
10 ly
A high z star-forming galaxy has many regions like this
The Cosmic Horseshoe: UV Spectrum Institute of AstronomyAnna Quider
To Earth
Hot, young stars
Cavity caused by stellar wind
Gas being ionized by the young, hot stars
(a H II region)
Cold interstellar gas(interstellar absorption)
10 ly
A high z star-forming galaxy has many regions like this
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Stellar Photosphere
(Quider et al. 2009)
“1425” Index:
‣ Blend of:- Si III 1417- C III 1427- Fe V 1430
‣ ZOBstars = 0.5ZO
“1978” Index:
‣ Only Fe III
“1425” “1978”
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Stellar Wind
C IV Wind Feature
‣ Complex superposition ‣Wind due to most massive O stars
‣ Starburst99 models- Continuous SF- 100Myr old- Salpeter IMF
- P-Cygni broad emission/absorption
- photospheric broad absorption
- narrow interstellar absorption
- narrow nebular emission
(Quider et al. 2009)
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Stellar Wind
ZOstars = 0.6ZO
MS1512-cB58 and Cosmic Horseshoe have very similar
winds!
(Quider et al. 2009)
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of AstronomyAnna Quider
Nor
mal
ized
flux
Interstellar Gas Absorption
The Cosmic Horseshoe: UV Spectrum Institute of AstronomyAnna Quider
Nor
mal
ized
flux
Interstellar Gas Absorption
Gas moving away from the stars
Gas moving towards the stars
Structure of the interstellar absorption lines is interpreted as being due to large-scale outflows of gas away from the stars
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Interstellar Gas Absorption
‣ -800 km/s to +250 km/s
‣ Same for low and high ionization gas
‣ Evidence for only ~60% coverage of stars by outflowing gas
(Quider et al. 2009)
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Lyman α Emission Feature
‣Lyman α is from H I gas
‣Double-peaked emission
‣Kinematic structure:- Peak 1 at +115 km/s- Peak 2 at +275 km/s- Red wing to +700 km/s
(Quider et al. 2009)
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Structure matches well with outflow
model from Verhamme et al.
(2006).
NHI ~ 7 x 1019 cm-2
a
b
c
λ
λ
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
What broader conclusions can we draw from studying
the Cosmic Horseshoe?
‣Comparison between different metallicity indicators
‣Possible candidate for Lyman continuum photon leakage
‣A cautionary note on over-interpreting Lyα profiles
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Metallicity Indicators
Good agreement among different metallicity indicators
(Quider et al. 2009)
14 Quider et al.
Table 5. METALLICITY COMPARISON
Method Element(s) Z/Za! Comments
R23 O 1.5 H II regionsb
N2 O 0.5 H II regionsb
O3N2 O 0.5 H II regionsb
1425 C, Si, Fe 0.5 Photospheric, OB starsc
1978 Fe . . . Photospheric, B starsc
C IV C, N, O, Fe ! 0.6 Stellar wind, O starsd
a Abundance relative to solar (on a linear scale), using thecompilation of solar abundances by Asplund et al. (2005).b As reported by Hainline et al. (2009).c This work (Section 4.2).d This work (Section 4.3).
stellar data from Section 4 with the nebular measurements by Hain-line et al. (2009). The first three entries in the Table give the oxy-gen abundance deduced from strong emission line indices, respec-tively the R23 index of Pagel et al. (1979), for which Hainline etal. (2009) used the calibration by Tremonti et al. (2004) based onobservations of tens of thousands of nearby galaxies in the SDSS,and the N2 and O3N2 indices of Pettini & Pagel (2004). The lastthree entries in the Table are from our analysis in this paper of UVspectral features from OB stars in the Horseshoe.
It is difficult to quote an error for each of the entries in Table 5;a realistic estimate should take into account the S/N of the data,the scatter in the adopted index calibration and the systematic biaswhich may affect it. However, most estimates evidently convergetowards a metallicity of about a half solar (which, incidentally, isquite typical of star-forming galaxies at these redshifts—see Erbet al. 2006a). It is reassuring that stellar and nebular abundancesare generally in good mutual agreement. The only deviant measureappears to be from the R23 method which is known, however, tooverestimate the oxygen abundance in the near-solar regime, par-ticularly when the calibration by Tremonti et al. (2004) is used (e.g.Kennicutt, Bresolin & Garnett 2003; Kewley & Ellison 2008).
Less satisfactory aspects of the comparison in Table 5 are: (i)the poorly understood failure of the 1978 index of Rix et al. (2004)to provide a meaningful estimate of metallicity in this particularcase; and (ii) the current limitations in the use of the wind linesas accurate abundance diagnostics. Point (i) can only be addressedwith a larger set of high quality UV spectra of high redshift galax-ies. Until it is resolved, however, it may be unwise to use this indexto investigate the possibility of a differential enrichment of Fe-peakelements compared to the products of Type-II supernovae (Pettiniet al. 2002; Halliday et al. 2008). Concerning point (ii), it is frustrat-ing that while the strong wind lines are in principle one of the mosteasily measured abundance diagnostics—at least in data of suffi-cient resolution to resolve stellar and interstellar components—westill lack a comprehensive library of empirical ultraviolet spectra ofOB stars of different metallicities to realise their full potential. TheHST survey of the Magellanic Clouds by Leitherer et al. (2001)went some way towards remedying this situation, but their sam-pling of the upper H-R diagram is still too sparse to assemble sepa-rate sets of Large and Small Magellanic Cloud stars. The resultinghybrid library, obtained by combining all the available spectra intoone set, can only give an approximate measure of metallicity, giventhe factor of ! 2 difference in the oxygen abundance of the twoClouds.
8.2 Escape of ionizing photons
It is still unclear what determines fLyCesc , the fraction of hydrogen
ionizing photons which escape from star-forming galaxies into theintergalactic medium. Direct detection of these Lyman continuumphotons has proved problematic until recently (e.g. Shapley et al.2006; Iwata et al. 2009), and yet fLyC
esc must have been large atvery high redshifts for the Universe to be reionized by the star-formation activity thought to have taken place at z > 6 (e.g. Bolton& Haehnelt 2007; Ryan-Weber et al. 2009).
Observations of gravitationally lensed galaxies may offer in-sights into the factors that control fLyC
esc . By fully resolving the in-terstellar absorption lines in our ESI spectrum of the Cosmic Horse-shoe, we reached the conclusion that the interstellar gas only cov-ers ! 60% of the stellar UV light, as viewed from Earth. This is apromising prerequisite for large values of fLyC
esc , making the Horse-shoe a high priority candidate in searches for LyC emission fromhigh-z galaxies. On the other hand, if ! 40% of the photons fromthe stars and surrounding H II regions really had a clear path out ofthe galaxy, we may have expected to see a strong and narrow Ly!emission line centred at zH II , whereas no such feature is present inour spectrum. Possibly, some 40% of the stars are located behindneutral gas of too low a column density to give discernible absorp-tion in the metal lines, and yet capable of scattering most of theLy! photons out of the line of sight (Hansen & Oh 2006). Whethersuch gas would be optically thick to LyC photons remains to beestablished.
Another necessary condition for the escape of LyC photons isa weaker H! emission line than expected on the basis of the UVcontinuum luminosity and reddening (admittedly in the idealisedcase of continuous star formation at a constant rate). Such a dispar-ity would arise from a ‘matter-bounded nebula’, where not all ofthe LyC photons emitted by the stars are absorbed and reprocessedwithin the H II region. As discussed in Section 6.3, in the CosmicHorseshoe SFR(H!) ! SFR(UV), although a number of differentfactors, apart from leakage of LyC photons, can affect this com-parison. In conclusion, it would definitely be worthwhile to searchfor LyC emission in the Cosmic Horseshoe at wavelengths below3085 A (the redshifted value of the Lyman limit at z = 2.38115).Furthermore, it would be of interest to check for partial coveringof the stars by the foreground interstellar medium in more galaxiesamong the newly discovered strongly lensed sources, ideally in-cluding galaxies with a range of Ly! equivalent widths. With suchdata in hand, we should be in a better position to understand theconditions that determine the escape fraction of LyC photons andthe morphology of the Ly! line.
8.3 Comparison with MS 1512-cB58
One of the motivations for the present study was to establish howtypical are the properties of the galaxy MS 1512-cB58, the onlyprevious case where the gravitational lensing boost was sufficient toallow a detailed look at the spectrum of a high-redshift star-forminggalaxy. While we have now doubled the ‘sample’ of high-z galax-ies with good-quality ESI spectra, it would clearly be prematureto draw general conclusions on the basis of just two objects. Nev-ertheless, one cannot help but being struck by how closely thesetwo galaxies resemble each other in many of their properties. Theyare very similar in their overall metallicity and probably in theirdetailed chemical composition, indicating that they have reachedcomparable stages in the conversion of their gas reservoirs intostars. Their young stellar populations are largely indistinguishable
c" 2009 RAS, MNRAS 000, 1–17
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Lyman Continuum Photons
‣10-15% of Lyman alpha photons escape
‣60% covering of stars by interstellar gas may provide a route for the escape of Lyman continuum photons
C II 1334 O I 1302 Si II 1304
Anna Quider
The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Lyman α Morphology
Similarities
Z~0.5ZO
Salpeter IMF
ΔvISM ~1000 km/s
SFR~50-100 MO/yr
Mvir~1-1.5x1010 MO
Relative Velocity (km s-1)
Rel
ativ
e Fl
ux Relative Velocity (km s-1)
MS1512-cB58top 25% of LBGsLyα absorption
Cosmic Horseshoetop 50% of LBGs
Lyα emission
(Pettini et al. 2002)
(Quider et al. 2009)
Anna Quider
Summary
• LBGs are high z starforming galaxies whose spectra show a wide variety of Lyα profiles, ISM trends with Lyα strength, young stellar populations, and gas with outflow speeds v ~ 200 km s-1
• Highly lensed LBGs are key to understanding the detailed chemical, kinematic, and structural properties of LBGs, as evidenced by the work on MS1512-cB58 and the Cosmic Horseshoe
• More galaxies need detailed study to determine the range of properties of high redshift starforming galaxies: stay tuned for the Cosmic Eye, Cosmic Clone, and 8 o’clock Arc!
The Cosmic Horseshoe: UV Spectrum Institute of AstronomyAnna Quider
A. Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Rest UV Spectral Features‣ Stellar Component:
Best-fit Starburst99 model is 300Myr, continuous star formation, Z = 0.25 ZO
‣ Interstellar Component:
Absorption strength and Δvem-abs vary with Lyα emission strength for low-ionization transitions but are constant for high-ionization transitions
‣ Physical Picture:
Patches of neutral gas are embedded in a continuous shell of high-ionization gas, all of which is outflowing.
(Shapley et al. 2003)
(Steidel et al. 2003)
A. Quider The Cosmic Horseshoe: UV Spectrum Institute of Astronomy
Interstellar Gas Absorption
(Quider et al. 2009)
Column densities and ~0.5ZO imply N(H I) ≈ 6x1020 cm-2