Star Formation in Galaxies at z>6
Looking for the Sources of Reionisation
Star Formation in Galaxies through Cosmic Times NOVA PhD Fall School
1-5 October 2012
The Evolution of the Star Formation Rate Density
Hopkins & Beacom (2006)
The cosmological view
Picture Credit: NASA/ WMAP
The Universe at z>6 -- a 21st century history
Hu et al. (2002)
2 Hu et al.
Fig. 1.— Narrowband 9152/118 ˚A image (left panel) and R-band image (right panel) of the emission-line object HCM 6A, which is marked with the vertical and horizontal lines. An image in the narrowband filter with a normalized local continuum (Z band) subtracted (inset in left panel) shows that the object, which appears as two fragments, is a strong emission-line object. (We have slightly oversubtracted the Z band to completely remove the neighboring galaxy, whose core appears white.) The galaxy is not seen in the much deeper R-band image (> 8 hours on the Keck 10 m Telescope) or in a 5600 s F675W image taken with the WFPC2 camera on HST, shown in the insert in the right hand panel. The neighboring bright galaxy to the SW is a cluster member with a redshift of 0.375 (“BO #39” in Mellier et al. 1988), so that the emission is not associated with this object – in particular Hα lies shortward of the filter bandpass. The images are centered on coordinates (J2000) RA: 2h39m54.s73 Dec: −1◦33"32.""3 and are 37"" on a side. The bar in the upper left corner of the narrowband image shows a 5"" scale.
A370 emitter argues for a higher reionization redshift.
In Section 2 we summarize the narrowband observa- tions and the continuum observations at longer and shorter wavelengths together with the spectroscopic followup. The properties of HCM 6A leave little doubt about its identifi- cation as a high-z Lyα emitter. Finally, in the discussion we consider the implications for early star formation and for the evolution of the intergalactic gas.
2. observations 2.1. Narrow Band Survey
The present images were obtained using a 118 ˚A filter centered on 9152 ˚A in the LRIS camera (Oke et al. 1995) on the Keck 10 m telescopes. This wavelength lies in a very dark region of the night sky between the OH(8-4) and OH(7-3) bands, and corresponds to Lyα at a redshift of ∼ 6.5. In total, six fields have been imaged with this filter, and are summarized in Table 1. In each case the exposures were made as a sequence of spatially dithered background-limited exposures and median skys were used to flat field the images. The images were calibrated with spectrophotometric standards. The deepest exposures, to- taling about 6 hours per field, were taken on the HDF and on fields centered on the clusters A370 and A851. Over most of the area in these fields the 5 sigma limiting flux sensitivity is approximately 1.6 × 10−17 erg cm−2 s−1 but, for the massive clusters, there is a small area in the central regions where the lensing amplification provides a signifi- cant gain in sensitivity.
For each field a very deep Z band (λeff ∼ 9170 ˚A; Hu et al. 1999) was used to measure the continuum and an extremely deep R-band image was used as a shorter wave- length reference to measure the continuum break. We
searched each field for objects with observed equivalent widths in excess of 100 ˚A which were not visible in the R image. Only one such object has been found in the six fields. Finding images for HCM 6A, which lies behind the massive lensing cluster A370 in a region of significant amplification, are shown in Figure 1, which compares nar- rowband and deep R continuum images and gives coordi- nates. Insets show the continuum-subtracted image and the corresponding WFPC2 F675W image from HST. (See B´ezecourt et al. (1999) for a description of the wider field of the HST image). The following subsections describe the observations made on this object to confirm that it is a high redshift Lyα emitter rather than a lower redshift emission-line object.
2.2. Broad Band Optical Imaging
Deep multicolor images of A370 were obtained using LRIS on the Keck 10 m telescopes on UT 1999 August 11, 1999 September 9–10, 2000 August 25, and 2000 De- cember 29 and 2002 January 11–12. The data were taken as a sequence of dithered exposures, with net integration times of 2400 s in V , 27900 s in R, 4050 s in I and 9820 s in Z. A deep B (3780 s) image was obtained with ESI on Keck II on UT 2000 September 29–30. The images were processed using median sky flats generated from the ex- posures. Conditions were photometric during these obser- vations. The data were calibrated using the photometric and spectrophotometric standard, HZ4 (Turnshek et al.
1990; Oke 1990), and faint Landolt standard stars in the SA 95-42 field (Landolt 1992).
2.3. Near-infrared Imaging
Narrowband filter @ 9152 AA
• galaxy at z=6.56 lensed by cluster Abell 370
• SDSS quasar at z=6.28 (J103027.10+052455.0) -- Fan et al. (2001)
The power of the HST -- the GOODS fields
GOODS-South -- ACS/WFC composite
Picture Credit: www.hubblesite.org
Two fields of 180 sq. arcmin each mapped in B,V,i,z (z~26 AB) + J and H bands
Reminding the Lyman-break selection technique
Lyman-break selection technique
Steidel et al. (1996) to select star-forming
galaxies at z~3
Picture Credit: http://www.astro.ku.dk/~jfynbo/LBG.html
Lyman-break technique at z~6
Stanway et al. (2003)
Lyman Break Galaxies and the Star Formation Rate at z ≈ 6 3
5000 6000 7000 8000 9000 10000 λ(obs)
0.0 0.1 0.2 0.3 0.4
Ly α z=5.5 Ly α z=6.0 Ly α z=6.5
Figure 1. The ACS-i! and -z! bandpasses overplotted on the spectrum of a generic z = 6 galaxy (solid line), illustrating the Lyman break technique.
such data is available, one (object 2) is detected in F606W v-band (vAB = 27.0±0, .2) and has colours consistent with being an ERO contaminant. All other candidates have vAB > 28.0 (our 3 σ detec- tion limit), although object 9 also has red near-IR colours. Object 5 has a colours which suggests that it is likely to be a low mass star: (zAB! − JVega)=3.12 & (J − Ks)Vega = 0.2 (an L- or T- dwarf). Comparison with the sample in Hawley et al. (2002) indi- cates a spectral type around T3. However, of the remaining objects for which no reduced infrared data is yet available, one is now con- firmed as a high redshift object (see ‘Note Added in Proof’) and the others are all fully resolved as shown in Figure 5. The contamina- tion rate would thus appear to be about 25 per cent.
3.3 Limiting Star Formation Rates
We are sensitive to objects above z ≈ 5.6 and in principle could identify objects out to z ≈ 7.0 where the Lyman-α line leaves the z! filter. The rapid falloff in sensitivity of the ACS detectors above 9000 ˚A, however, combined with the transition of the Lyman-α break through the z! filter and the resulting incomplete coverage of that filter, means our sensitivity to star formation drops rapidly past z ≈ 6.0, and renders us unlikely to detect anything past z ≈ 6.5 (Figure 6). The effect of luminosity-weighting of the redshift range also affects our effective survey volume, and we quantify this in Section 4.1.
The observed luminosity function of Lyman break galaxies (LBGs) around λrest = 1500 ˚A is m∗R = 24.48 at #z$ = 3.04 and m∗I = 24.48 at #z$ = 4.13, from Steidel et al. (1999), with a faint end slope α = −1.6 and normalization φ∗ ≈ 0.005 h370 Mpc−3. If there is no evolution in the luminosity function from z = 3 (as is found to be the case at z = 4) this would predict an apparent mag- nitude of z!AB = 25.6 for an L∗LBG at z ∼ 6. Thus our complete catalogue of zAB! < 25.6 would include galaxies down to L∗LBG at z ∼ 6 if there is no evolution in the LBG luminosity function.
The relation between the flux density in the rest-UV around
0 1 2 3 4 5 6 7
−0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
(i’ − z’)AB
Figure 2. Model colour-redshift tracks for galaxies with non-evolving stel- lar populations (from Coleman, Wu & Weedman 1980 template spectra).
The ‘hump’ in (i!−z!) colour seen at z ≈ 1 − 2 is due to the 4000 ˚A break redshifting beyond the i!-filter.
20 21 22 23 24 25 26 z’ (AB)
-2 -1 0 1 2 3
Figure 3. A colour-magnitude diagram for our data with the z!-band limit of zAB! < 25.6 and the colour-cut of (i! − z!)AB = 1.5 shown (dashed lines). Candidate objects are marked with circles, or 3 σ lower-limits on the (i! − z!) colour where objects are undetected in i!. Note: A catalogue based on a simple colour cut is contaminated by stars, EROs and wrongly identified extended objects as described in the text.
– 7 –
Fig. 1.— (i775 −z850)AB vs. (z850 −J)AB colour-colour diagram illustrating the position of our RDCS1252-2927 i-dropout sample (shaded region) relative to the photometric sample as a whole. Tracks for a 108 year starburst with various amounts of extinctions have been included to illustrate both the typical redshifts (labelled for z = 5.5 and z = 6) and SED types included in the selection window. The low-redshift (0 < z < 1.2) tracks for typical E, Sbc, and Irr spectra have been included as well to illustrate the region in colour-colour space where possible contaminants might lie. There is a clear separation between the i − z > 1.3 point-like objects (crosses) and i − z > 1.3 extended objects (solid squares) along the z − J axis. The distribution of objects in color-color space led us to adopt (i − z) > 1.5 as our generalized i-dropout selection criteria. In all cases, error bars represent 2 σ limits. The clump at (i775−z850)AB ∼0.9 and (z850−J)AB ∼1 are early-type galaxies from the cluster.
Bouwens et al. (2003)
Lyman-break technique at z~7-8– 9 –
Fig. 1.— (z850 − J110)AB/(J110 − H160)AB color-color diagram showing the position of our z850-dropouts (selection region is shaded gray) relative to the UDF photometric sample (cyan squares). Objects included in the source list (Table 1) are shown as black squares (2σ lower limits are indicated by vertical arrows). These objects are not detected in the optical V606 and i606 bands. The cyan squares that lie in the selection region have clear V606 and i775 detections (> 2σ) and so are not candidate z850-dropouts; representative error bars for these objects are shown at the right of this diagram. The color-color tracks of both lower redshift interlopers (red lines) and 108 yr starburst SEDs with different reddenings (blue lines) are plotted as a function of redshift. The position of M, L, and T dwarfs are also shown (green cross hatched region) (Knapp et al. 2004). Error bars on the z850 − J110 and J110 − H160 colors are 1σ.
Bouwens et al. (2004)
Galaxies selected on HST / WFC3 images
Lyman-break galaxies at z~7-8 – 10 –
Fig. 2.— Postage stamps images (V606
bands) of our 5 z850
-dropout candi- dates. Also shown (UDF-640-1417: bottom row) is one very red (z850
= 1.1 object which nearly met our selection criteria and could be a reddened starburst at z ∼ 6.5 (or a reddened early type at z ∼ 1.6) (also found by Yan & Windhorst 2004). While our best-fit to UDF-387-1125 is a z ∼ 6.8 starburst spectrum, this object is also consistent with being a compact z ∼ 1, 0.01L∗
dust-reddened early type galaxy. Magnitudes are those measured in a
-diameter aperture. The three rightmost panels show the combined J110
each object, its position in color-color space, and an SED fit to the broadband fluxes. The derived redshift is also provided in the rightmost panel. The ACS cutouts here are shown at a much higher contrast than the NICMOS cutouts, demonstrating the significance of the optical non-detections. The postage stamps are 2.9""
in size. A linear stretch is used for scaling the pixel fluxes.
Bouwens et al. (2004)
Galaxies in the very young Universe
The rest UV luminosity function at z~7-8
10 R.J.. McLure, et al.
Figure 5. Our new determinations of the ultraviolet (UV) galaxy luminosity function at z ! 7 and z ! 8. The left-hand panel shows our V /Vmax estimate of the redshift z = 7 luminosity function. The absolute UV magnitudes have been calculated at a rest-frame wavelength of 1500 ˚A. The filled circles are the data-points from this study, where the horizontal error bars indicate the width of the magnitude bins adopted (either ∆m = 0.5 or ∆m = 1.0 depending on signal-to-noise) and the vertical error bars indicate the uncertainty due to simple Poisson statistics. The open squares are taken from the ground-based study of the Subaru Deep Field and GOODS-North field by Ouchi et al. (2009). The dotted and dashed curves are the Schechter function fits to the z = 5 and z = 6 luminosity function from McLure et al. (2009). The solid line is our best-fitting Schechter function to the z = 7 luminosity function. The right-hand panel shows our corresponding estimate of the luminosity function at redshift z = 8. In this panel the dashed and dot-dashed curves are the z = 6 and z = 7 Schechter function fits from McLure et al. (2009) and this work respectively. The solid line is the best-fitting z = 7 Schechter function with φ! reduced by a factor of two (see text for a discussion).
our sample, when deriving the luminosity function we adopt the method outlined in McLure et al. (2009) whereby, rather than sim- ply adopting the primary photometric redshift solution, each ob- ject is represented by its normalized redshift probability density function. In addition to making better use of the available infor- mation, this method makes it possible to construct the luminosity function using the classic V /Vmax
estimator of Schmidt (1968), and deals with the problem of multiple redshift solutions in a trans- parent fashion.
In Fig. 5 we show our estimates of the z = 7 and z = 8 galaxy luminosity functions, which were calculated within the redshift in- tervals 6.5 < z < 7.5 and 7.5 < z < 8.5 respectively. In both panels the data-points from this study are shown as the filled cir- cles, and in the left-hand panel we also plot the z = 7 data-points at bright magnitudes from the ground-based study of the Subaru Deep Field and GOODS-North field by Ouchi et al. (2009). Although a full analysis of the evolving UV-selected luminosity function is de- ferred until the remainder of the deep HUDF09 WFC3/IR data are obtained over the coming year, we have performed a simple fit to the z = 7 luminosity function, including the Ouchi et al. (2009) data-points at the bright end. The best-fitting Schechter function is shown as the solid line in the left-hand panel of Fig. 5 and yields the following parameter values: M1500!
= −20.11, φ!
and α = −1.72. Due to the large uncertainties on the data-points it is not possible to place meaningful constraints on either M1500!
or α at z = 7. However, a comparison with the best-fitting Schechter luminosity function parameters at z = 6 (M1500!
= −20.04 ± 0.12, φ!
= 0.0018 ± 0.0005 Mpc−3
and α = −1.71 ± 0.11) from McLure et al. (2009), suggests that the z = 6 and z = 7 luminosity functions have very similar shapes, and can be reconciled via an evolution in number density by a factor of " 2.5.
The limited dynamical range in absolute luminosity makes it impossible, at present, to place any meaningful constraints on the shape of the UV-selected galaxy luminosity function at z = 8. All that can be said with the current dataset is that, at a characteristic luminosity of M1500
" − 19.5, the number density of high-redshift galaxies is a further factor of " 2 smaller than at z = 7. To illus- trate this point we show in the right-hand panel of Fig. 5 that the best-fitting Schechter function at z = 7 can be brought into good agreement with the z = 8 data-points simply by reducing φ!
by a factor of two. Although the uncertainties associated with the cur- rent dataset, along with degeneracies between Schechter function parameters, prevent any definitive conclusions, it appears that the z = 7 and z = 8 luminosity functions are consistent with having the same overall shape as at z = 6, but with φ!
a factor of " 2.5 and " 5 lower respectively.
5.3 Implications for re-ionisation
Given the apparent decline in the number density of star-forming galaxies over the redshift range z " 6 − 8, it is prudent to examine the ability of the observed population of galaxies to achieve reioni- sation. We adopt the method of Bolton & Haehnelt (2007), convert- ing the observed luminosity functions at z > 6 to ionising emissivi- ties assuming a Salpeter IMF, a metallicity of 0.2 Z"
, and an escape fraction for ionising photons of 20%. For simplicity, we ignore any possible mass dependence in the escape fraction. We consider rela- tively low values for the hydrogen clumping factors in our analysis (C " 2−5), following the results of recent simulations (e.g. Bolton
& Haehnelt 2007). At z < 6, we adopt the comoving ionising pho- ton density implied by the Lyα opacity data presented in Bolton &
Haehnelt (2007), where ˙ Nion
. For the luminosity function at z " 7 and 8, we adopt the
" 0000 RAS, MNRAS 000, 000–000