A remarkably high fraction of strong Ly_alpha emitters amongst luminous redshift 6.0<z<6.5 Lyman break galaxies in the UKIDSS Ultra-Deep Survey

Abstract

We present spectroscopic confirmation of ten highly luminous (L >= 2L*) Lyman
alpha emitters in the redshift range 6.01<z<6.49 (nine galaxies and one AGN),
initially drawn from a sample of fourteen z_phot >= 6 Lyman break galaxies
(LBGs) selected from an area of 0.25 square degrees within the UKIDSS
Ultra-deep Survey (UDS). Overall, our high rate of spectroscopic confirmation
(>= 71%) and low rate of contamination provides a strong vindication of the
photometric redshift analysis used to define the original sample. By
considering star-formation rate estimates based on the Ly_alpha and UV
continuum luminosity we conclude that our sample is consistent with a Ly_alpha
escape fraction of ~25%. Moreover, after careful consideration of the potential
uncertainties and biases, we find that 40%-50% of our sample of L >= 2L*
galaxies at 6.0<z<6.5 display strong Ly_alpha emission (rest-frame equivalent
width >= 25 Angs), a fraction which is a factor of ~2 higher than previously
reported for L <= L* galaxies at z~6. Our results suggest that, as the epoch of
reionization is approached, it is plausible that the Ly_alpha emitter fraction
amongst luminous (L >=2 L*) LBGs shows a similarly sharp increase to that
observed in their lower-luminosity (L <= L*) counterparts.

Full text

Page 1

Mon. Not. R. Astron. Soc. 000, 1–12 (2002)Printed 11 October 2011(MN LATEX style file v2.2)
A remarkably high fraction of strong Lyα emitters
amongst luminous redshift 6.0 < z < 6.5 Lyman break
galaxies in the UKIDSS Ultra-Deep Survey
E. Curtis-Lake1?, R. J. McLure1, H. J. Pearce1, J. S. Dunlop1, M. Cirasuolo1,
D. P. Stark2†, O. Almaini3, E. J. Bradshaw3, R. Chuter3, S. Foucaud4,
W. G. Hartley3
1SUPA‡, Institute for Astronomy, University of Edinburgh, Royal Observatory, Edinburgh EH9 3HJ
2Department of Astronomy, Steward Observatory, University of Arizona, 933 North Cherry Avenue, Rm N204, Tucson, AZ, 8572
3School of Physics & Astronomy, University of Nottingham, University Park, Nottingham NG7 2RD
4Department of Earth Sciences, National Taiwan Normal University, No. 88, Section 4, Tingzhou Road,
Wenshan District, Taipei 11677, Taiwan
11 October 2011
ABSTRACT
We present spectroscopic confirmation of ten highly luminous (L ? 2L?) Lyman alpha
emitters in the redshift range 6.01 < z < 6.49 (nine galaxies and one AGN), initially
drawn from a sample of fourteen zphot? 6 Lyman break galaxies (LBGs) selected from
an area of 0.25 square degrees within the UKIDSS Ultra-deep Survey (UDS). Overall,
our high rate of spectroscopic confirmation (? 71%) and low rate of contamination
provides a strong vindication of the photometric redshift analysis used to define the
original sample. By considering star-formation rate estimates based on the Lyα and
UV continuum luminosity we conclude that our sample is consistent with a Lyα escape
fraction of ? 25%. Moreover, after careful consideration of the potential uncertainties
and biases, we find that 40%−50% of our sample of L ? 2L?galaxies at 6.0 < z < 6.5
display strong Lyα emission (rest-frame equivalent width ? 25˚ A), a fraction which is
a factor of ? 2 higher than previously reported for L ? L?galaxies at z ? 6. Our
results suggest that, as the epoch of reionization is approached, it is plausible that
the Lyα emitter fraction amongst luminous (L ? 2L?) LBGs shows a similarly sharp
increase to that observed in their lower-luminosity (L ? L?) counterparts.
Key words: galaxies: high-redshift - galaxies: evolution - galaxies: formation
1INTRODUCTION
Improving our understanding of the earliest epochs of galaxy
formation and evolution relies fundamentally on the ability
to select clean samples of high-redshift galaxies, free from
significant low-redshift contamination. Traditionally, sam-
ples of high-redshift galaxies have been selected using one of
two complementary photometric techniques. Firstly, Lyman-
alpha emitters (LAEs) are selected from deep imaging us-
ing narrow-band filters centred on the redshifted Lyman-
alpha emission line (e.g. Hu, Cowie & McMahon 1999). Al-
ternatively, Lyman-break galaxies (LBGs) can be selected
from deep broad-band photometry using the Lyman-break,
?Email: efcl@roe.ac.uk
† Hubble Fellow
‡ Scottish Universities Physics Alliance
or “dropout”, technique pioneered by Guhathakurta, Tyson
& Majewski (1990).
Studies of high-redshift galaxies selected using both
techniques have made rapid progress recently, thanks to
large-area, red-sensitive, detectors on ground-based tele-
scopes and the ultra-deep near-infrared imaging now pos-
sible with WFC3/IR on-board the Hubble Space Telescope
(HST). As a result, it is now possible to obtain large, statisti-
cal, samples of LBGs/LAEs in the redshift interval 6 < z < 7
from the ground (e.g. Yoshida et al. 2006; Ouchi et al.
2008, 2010; McLure et al. 2009) and upwards of one hun-
dred LBGs have now been identified in the redshift interval
6.5 < z < 8.5 with WFC3/IR (e.g. Bouwens et al. 2010;
McLure et al. 2010, 2011; Finkelstein et al. 2010). Indeed,
recent results have demonstrated the power of these tech-
niques with the spectroscopic confirmation of LBG-selected
galaxies within the reionization epoch at 7.00 < z < 7.20
arXiv:1110.1722v1 [astro-ph.CO] 8 Oct 2011

Page 2

2 E. Curtis-Lake et al.
(Vanzella et al. 2011; Pentericci et al. 2011; Schenker et al.
2011; Ono et al. 2011). Although LAEs are simply a sub-
set of the LBG population, due to the different selection
techniques employed, the study of these two high-redshift
galaxy populations has proceeded largely independently. As
a consequence, one of the key questions in the study of high-
redshift galaxies is determining how LAEs and LBGs are re-
lated and how they are connected to the galaxy populations
identified at lower redshift.
Over the last fifteen years substantial observational ef-
fort has been invested in studying the population statistics
of both LAEs and LBGs. As a result significant progress
has been made towards an understanding of the luminos-
ity function and clustering properties of the LAE and LBG
populations from z = 3 to z = 7 (e.g. Bouwens et al. 2007,
2010, 2011; McLure et al. 2009, 2010; Ouchi et al. 2008, 2010;
Reddy et al. 2008). While correlation length estimates for
LAEs/LBGs are largely consistent (r0 ? 5 Mpc) and show
little evidence for evolution, the LAE/LBG luminosity func-
tions appear to show substantial differential evolution. Most
studies now agree that the LBG luminosity function evolves
strongly with redshift, with L?dimming by a factor of ? 2
between z = 4 and z = 6 (e.g. Bouwens et al. 2007; McLure
et al. 2009, 2010; Yoshida et al. 2006). In contrast, the ob-
served LAE luminosity function does not appear to evolve
between z = 3 and z = 6 (e.g. Shimasaku et al. 2006; Ouchi
et al. 2008), although the latest results indicate a moderate
? 30% drop in L?
et al. 2010).
Studying the Lyα emission properties of LBG-selected
samples is a powerful method for improving our under-
standing of this confusing picture. Although spectroscopic
follow-up of the LBG population reveals Lyα emission in ob-
jects covering the full span of UV luminosity, the strongest
Lyα emission (equivalent width>
in fainter objects. This was first indicated by Shapley et al.
(2003), who observed the mean rest-frame Lyα equivalent
width (EW) increasing at fainter magnitudes for z ∼ 3
LBGs, and has since been noted by many authors over a
wide redshift range (eg. Ando et al. 2006; Vanzella et al.
2009; Stark et al. 2010). Furthermore, studies of how the
fraction of strong Lyα emitters evolves as a function of red-
shift offer the prospect of providing crucial information on
the dust content of LBGs and, potentially, on the neutral
fraction of the IGM. Recent work by Stark et al. (2010,
2011) has shown that, at fixed UV luminosity, the fraction
of LBGs showing Lyα emission with EW ? 25˚ A increases
by 55% between z = 4 and z = 6. More recently this analysis
has been extended to spectroscopically targeted z ∼ 7 sam-
ples (Fontana et al. 2010; Schenker et al. 2011; Pentericci
et al. 2011; Ono et al. 2011) with the results from different
groups consistently showing a drop in the Lyα fraction, ten-
tatively attributed to a change in the neutral fraction in the
IGM.
Recent theoretical studies find differing results when
trying to explain the observed fractions of Lyα emitting
galaxies amongst LBG samples. Dayal & Ferrara (2011) are
able to broadly reproduce the observed trend of increas-
ing Lyα emission with decreasing UV luminosity for an
EW>55˚ A cut at z ∼ 6 from their cosmological SPH simula-
tion of reionization. In agreement with recent observations,
Dayal & Ferrara (2011) also find a decrease in the fraction of
Lyαbetween z = 5.7 and z = 6.6 (Ouchi
∼100˚ A) is only observed
strong Lyα emitters between 6 < z < 7, though this is due
mainly to changes in dust distributions, not a change in the
neutral fraction of the IGM. However, they also conclude
that all LBGs at z ∼ 6 should display Lyα in emission with
EW>20˚ A, in disagreement with the observations of Stark
et al. (2011). Forero-Romero et al. (2011) are able to re-
produce very well the observed fraction of EW>55˚ A Lyα
emitters in the 4.5 < z < 6.0 sample from Stark et al. (2010),
by requiring the escape fraction of Lyα photons to decrease
with increasing UV-luminosity. In addition, they are are also
able to reproduce the observed drop in EW>25˚ A Lyα emit-
ters above z > 6.3 observed by Schenker et al. (2011), by
fitting to the observed evolution in the LAE luminosity func-
tion at these redshifts. Neither of these studies allow for
peculiar velocities (outflows/inflows) within the ISM which
Dijkstra et al. (2011) find to produce a large effect on the de-
tectability of Lyα, with outflows from the galaxies boosting
the observed EWs.
Despite the large amounts of effort invested in spec-
troscopically observing the properties of LBGs over recent
years, one area of parameter space which has been relatively
unexplored is that occupied by luminous LBGs at z ? 6. The
relatively poor statistics in this redshift-luminosity r´ egime
are simply a reflection of the relative rarity of L ? L?galax-
ies at z ? 6, combined with the small areas typically covered
by survey fields with the deep, multi-wavelength, imaging
necessary to reliably select such objects. In this paper we
address this issue by presenting deep, red sensitive, optical
spectroscopy of luminous (L ? 2L?) LBGs in the redshift
interval 6.0 < z < 6.5, photometrically selected from an
area of 0.25 sq. degrees within the UKIDSS (Lawrence et al.
2007) Ultra-Deep Survey (UDS). Armed with deep spectro-
scopic observations, the combination of large area and deep
optical/near-infrared imaging available in the UDS allows us
to investigate three important issues. Firstly, we are able to
investigate whether or not the photometrically selected LBG
samples used to constrain the evolution of the bright end
of the UV-selected galaxy luminosity function (e.g. McLure
et al. 2009) are significantly contaminated by low-redshift
interlopers and/or active galactic nuclei (AGN). Secondly,
we are able to place the best available constraints on the
fraction of strong Lyα emitters amongst luminous LBGs at
z ? 6. Finally, we are able to investigate whether previous
reports of a very low fraction of strong Lyα emitters amongst
samples of luminous z ? 6 LBGs have been biased due to
contamination by low-redshift interlopers or small number
statistics.
The structure of the paper is as follows. In Section 2 we
describe the initial sample selection and spectroscopic obser-
vations before proceeding to present the spectra themselves.
In Section 3 we describe our technique for accurately mea-
suring the Lyα equivalent widths (EWs) and investigate the
star-formation rates (SFRs) and Lyα photon escape frac-
tion of our LBG sample, before estimating the fraction of
luminous z ? 6 LBGs which are strong Lyα emitters. In
Section 4 we compare our results to those from the recent
literature and discuss the potential uncertainties and biases
associated with estimating the Lyα emitter fraction at high
redshift. In Section 5 we present our conclusions. Through-
out the paper we assume a cosmology with H0 = 70 km
s−1Mpc−1, Ωm = 0.3, ΩΛ = 0.7. All magnitudes are quoted
in the AB system (Oke & Gunn 1983).

Page 3

Spectrocopic confirmation of luminous LBGs at z ? 63
Figure 1. Spectra of the confirmed z > 6 objects showing the one-dimensional (1D) spectra to the left and the corresponding
two-dimensional (2D) spectra on the right. All 1D spectra are plotted on the same flux scale, except for the spectrum of the AGN
(UUDS J021627.8) which has been scaled by a factor of 0.5.
2 IMAGING DATA, SAMPLE SELECTION
AND SPECTROSCOPY
The sample of z ? 6 LBG candidates targeted for spectro-
scopic follow-up is a sub-set of that originally selected by
McLure et al. (2009) to investigate the bright-end of the
z = 6 galaxy luminosity function. Although a full descrip-
tion of the original selection process is provided in McLure
et al. (2009), in this section we briefly review the most rele-
vant details before proceeding to describe the spectroscopic
observations.
2.1Imaging data and LBG selection
The parent sample of z ? 6 LBG candidates was originally
selected via a photometric redshift analysis, which exploited
the deep optical and near-infrared imaging available within
the area covered by the UKIDSS Ultra-Deep Survey (UDS).
The UDS is the deepest of the five near-infrared surveys
being undertaken at the UK infra-red telescope (UKIRT)
which together comprise the UK infra-red Deep Sky Surveys
(UKIDSS; Lawrence et al. 2007).
The latest ESO-public data-release for the UDS (DR8)
features near-infrared imaging over an area of 0.8 square
degrees to 5σ-depths of J = 24.9,H = 24.2 & K = 24.7
(2??diameter apertures). Within the UDS field, complimen-
tary deep Subaru optical imaging is available from the
Subaru/XMM Deep Survey (SXDS), which provides opti-
cal imaging to 5σ depths (2??diameter apertures) of B =
27.9,V = 27.3,R = 27.2,i?= 27.2 & z?= 26.1 (Furusawa et
al. 2008).
The useful overlap region between the UDS near-
infrared and SXDS optical imaging covers an area of ? 0.65
square degrees, and it was from this overlap region that
McLure et al. (2009) used a photometric redshift analysis
to select luminous (z?< 26) LBG candidates in the redshift
interval 4.5 < z < 6.5. The parent sample for spectroscopic
follow-up was comprised of those candidates with a high
probability of lying at z ? 6 based on the redshift proba-
bility density function returned by the photometric redshift
analysis.
2.2Spectroscopy
The spectroscopic data analysed in this study were obtained
between October 2007 and January 2011 with the FORS2
spectrograph on the VLT as part of the systematic spec-
troscopic follow-up of the UDS obtained through the ESO
large programme ESO 180.A-0776 (UDSz; P.I. O. Almaini).
Given that full details of UDSz will be presented in Almaini
et al. (2012, in preparation), only the most relevant details
are provided here.
The UDSz programme was allocated a total of 235 hours
of observations, with 93 hours allocated for observations
with the VIMOS spectrograph and 142 hours allocated for
FORS2 observations. The primary science driver for UDSz
was to obtain spectroscopic observations of a representative
sample of K−band selected galaxies (K < 23) photometri-
cally pre-selected to lie at redshift zphot ? 1.0. These spec-
troscopic observations were designed to support the primary
science driver of the UDS survey, which is to study the as-
sembly and evolution of massive galaxies at z ? 1. Within

Page 4

4E. Curtis-Lake et al.
Table 1. The spectroscopic sample of z ? 6 UDS LBGs. Column 1 lists the UDS object IDs (incorporating the J2000
coordinates). Column 2 gives the spectroscopic redshift and errors derived from the FORS2 spectra, taking the spectroscopic
redshift from the peak of the Lyα line. Columns 3 & 4 give the quality of the spectroscopic redshift (A-C, see text for
details) and the photometric redshift derived in Section 2.3 respectively. The remaining columns list the z?-band and z921-
NB observed magnitudes (corrected to total) plus errors, absolute UV magnitude derived from the z921-NB photometry
and the measured Lyα fluxes and EWs (derived from spectra, see text for more details).
IDzspec
Qualityzphot
mz
m912
MUV
Lyα fluxLyα EW
(/˚ A) (/10−18ergs s−1cm−2)
UUDS J021800.90-051137.8
UUDS J021616.53-050217.7
UUDS J021807.14-045841.5
UUDS J021816.33-051116.6
UUDS J021735.34-051032.6
UUDS J021838.90-050944.0
UUDS J021653.00-044623.3
UUDS J021922.01-045536.3
UUDS J021841.02-051247.4
UUDS J021701.44-050309.4
6.027 ± 0.002
6.046 ± 0.004
6.050 ± 0.003
6.114 ± 0.009
6.120 ± 0.003
6.190 ± 0.014
6.213 ± 0.004
6.216 ± 0.009
6.475 ± 0.003
6.487 ± 0.004
A
A
A
C
A
A
A
A
A
A
6.0+0.1
−0.1
5.9+0.2
−0.1
6.0+0.2
−0.1
6.1+0.3
−0.3
6.0+0.1
−0.1
6.0+0.2
−0.1
6.5+0.4
−0.4
6.4+0.8
−0.8
5.9+0.4
−0.2
6.0+0.3
−0.3
25.36 ± 0.08
25.35 ± 0.15
25.02 ± 0.12
25.55 ± 0.15
25.21 ± 0.12
25.21 ± 0.11
25.75 ± 0.21
25.79 ± 0.22
25.80 ± 0.23
25.91 ± 0.19
25.30 ± 0.09
25.84 ± 0.13
24.99 ± 0.10
25.42 ± 0.14
25.47 ± 0.15
25.13 ± 0.11
25.34 ± 0.09
26.25 ± 0.20
25.46 ± 0.16
25.03 ± 0.10
−21.40
−20.86
−21.72
−21.29
−21.25
−21.62
−21.41
−20.50
−21.36
−21.79
44.9 ± 7.4
17.8 ± 3.1a
29.4 ± 2.6
7.3 ± 1.7
31.4 ± 2.6
36.4 ± 3.4
9.5 ± 1.1
16.8 ± 2.2
18.2 ± 3.1a
10.2 ± 1.8a
54.7 ± 11.3
36.8 ± 8.5
27.6 ± 4.1
11.6 ± 3.1
46.8 ± 8.4
39.4 ± 6.0
12.5 ± 3.1
50.7 ± 12.8
27.8 ± 6.8b
10.6 ± 2.3b
aLyα flux calibration tied to the z−band photometry using an average correction factor, see text for details.
bEW values potentially affected by Lyα contributing to the NB photometry used to derive the UV continuum value.
this context, the large allocation of FORS2 time (20 MXU
masks, 5 hours of integration each) was designed to pro-
vide sensitive, red optical, spectra of massive galaxies in the
redshift interval 1.0 < z < 1.5.
Each of the 20 FORS2 masks covers an area of 6.8?×
6.8?and typically allowed for the allocation of 30-35 science
slits (based on a slit length of 8??). During the mask design
process targets from the core sample of K < 23 galaxies
were allocated with the highest priority. However, in order
to fill each mask, several other samples of additional sci-
ence targets were also allocated slits on a best-effort basis.
The list of high-redshift LBG candidates from McLure et al.
(2009) with a high probability of lying at z ? 6 was included
as one of the samples of additional science targets. A total
of fourteen z ? 6 LBG candidates were observed as part of
UDSz, each receiving a total of 5-hours of on-source integra-
tion with the GRS 300I grism (6000˚ A< λ < 10000˚ A with
R = 660). Data reduction was performed using a modified
version of the FORS2 pipeline, details of which are provided
in Pearce et al. (2011, in preparation).
2.2.1 Spectra of confirmed z ? 6 sources
Of the fourteen galaxies for which spectra were obtained
we are able to assign secure redshifts to eleven sources ac-
cording to Lyα line detections, placing them at z > 6. The
1D spectra for the spectroscopically confirmed z > 6 Lyα
emitters are displayed in Figure 1 and the measured line
fluxes and EWs are reported in Table 1. Many of the spec-
tra show continuum red-wards of the emission line as well as
the distinctive asymmetrical line profile indicative of Lyα.
Each redshift is assigned a quality flag, defined as in Vanzella
et al. (2009), with values of A (unambiguous detection), B
(likely detection) and C (uncertain detection). It should be
noted that the redshifts for the 11 confirmed z > 6 sources
are measured from the position of peak flux within the Lyα
line, and are hence susceptible to shifts in the peak position
due any velocity structure of the neutral hydrogen within
the galaxy or the surrounding IGM.
One of our targets (UUDS J021627.80-045534.2) is an
active galactic nuclei (AGN) at a redshift of z = 6.01. This
object was initially identified as a potentially ultra-luminous
z ? 6 LBG candidate by McLure et al. (2006) and was
subsequently targeted as a high-redshift quasar candidate
by Willott et al. (2009). In Willott et al. (2009) this source is
referred to as CFHQS J021627−045534, and was confirmed
as a low-luminosity quasar at z = 6.01 based on a four-hour
spectrum obtained with the GMOS spectrograph on Gemini
(Lyα FWHM ? 1600 km s−1). However, it is important to
note that CFHQS J021627−045534 was the only z ? 6 AGN
identified by Willott et al. (2010) over a search area of 4.47
sq. degrees, down to a limit of z?= 24.5. Based on this
fact, and the steepness of the z ? 6 UV-selected galaxy
luminosity function, it seems reasonable to conclude that
the AGN contamination of 26 ? z?? 25 LBG samples at
z ? 6 is negligible.
2.2.2 Spectra of unconfirmed sources
Four of the z ? 6 LBG candidates which were spectro-
scopically observed did not produce spectra which allow
us to assign a robust redshift. One of these four objects
(UUDS J021816.33-051116.6) does display a probable Lyα
emission line but is assigned a quality flag of C since the
emission line sits within a region of poorly subtracted sky-
line residuals. One further object does not display a promi-
nent Lyα emission line, but does display a continuum break
consistent with a redshift of z = 5.83. However, the signal-
to-noise in the final FORS2 spectrum is too low to be con-
fident of the redshift.
The spectra of the remaining two candidates show only
a very faint continuum, with a signal-to-noise ratio too low
to reliably distinguish between a Lyman break and low-
redshift interloper. We note that three of the four uncon-
firmed candidates lie on the same FORS2 mask. However,
although this mask is slightly poorer than average, in terms
of redshift completeness for primary targets at 1 < z < 1.5,
there is no particular reason to suspect that the spectra of
these three high-redshift candidates have been unduly af-
fected.

Page 5

Spectrocopic confirmation of luminous LBGs at z ? 65
Figure 2. SED fits to the spectroscopically confirmed z ? 6 LBGs showing the BV RizJHK photometry (red data points with error
bars), the best fitting template with a fixed spectroscopic redshift and including the measured Lyα contribution (black line and blue
data points) and the best-fitting template assuming no Lyα flux and leaving redshift as a free parameter (grey lines). The best-fitting
SED templates at low redshift (zphot? 2) are plotted as dashed grey lines. The inset panels show the distribution of χ2versus redshift.
2.3 Photometric redshift selection and
spectroscopic completeness
From the sample of fourteen z ? 6 LBG candidates which
were selected for spectroscopic follow-up, a total of ten
objects have been robustly spectroscopically confirmed as
z ? 6 objects. As a consequence, regarding the other four
spectra as contaminants, the lower limit to the spectroscopic
completeness for our sample is 71%. However, as described
above, two of the remaining spectra display strong evidence
of being genuine high-redshift galaxies at z ? 5.8 and have
only failed to make our robust sample due to insufficient
signal-to-noise. If we regard these two objects as also be-
ing confirmed z ? 6 objects, then the spectroscopic com-
pleteness of our sample rises to 86%. Irrespective of the na-
ture of the two uncertain candidates, the high spectroscopic
completeness provides a strong vindication of the original
method of selecting high-redshift targets on the basis of an
SED-fitting photometric redshift analysis.
To illustrate this point, in Fig. 2 we show the latest
photometry available for each of the spectroscopically tar-
geted z ? 6 objects, along with the results of our updated
photometric redshift analysis. For each object the observed
photometry is plotted along with the best-fitting SED model
returned when redshift is kept as a free parameter, as well
as the best-fitting model using the fixed spectroscopic red-
shift and adding a Lyα emission line with the correct EW.
The photometric redshifts were calculated using LePhare
photometric redshift code (Ilbert et al. 2006 ). The galaxy
templates were made using the Bruzual & Charlot (2003)
stellar population synthesis models, with exponentially de-
creasing star formation rates with e-folding times in the
range 0.1 Gyr < τ < 30 Gyr and a Chabrier (2003) ini-
tial mass function (IMF). Reddening by dust is added using
the Calzetti et al. (2000) extinction law with E(B-V) values
ranging from 0.0 to 0.5. For the fitting with fixed redshift,
Lyα is added to the models after extinction is applied, using
the flux from the template convolved with the narrow-band
filter as the estimate of the continuum to translate the EW
into the Lyα flux.

Page 6

6E. Curtis-Lake et al.
Figure 3. SED fits for the three unconfirmed objects and UUDS J021816.33-051116.6, which is assigned a quality ’C’ emission line
redshift (see text) and is not treated as a secure Lyα detection. For the three objects without an assigned spectroscopic redshift, only
the primary and secondary photometric redshift solutions are shown as grey solid and dashed lines respectively. For these objects the
model fluxes are calculated from the primary redshift solution (blue points).
In Fig. 3 the insets showing the distribution of χ2versus
photometric redshift clearly demonstrate, as expected, that
the primary photometric redshift solutions are all at z>
Crucially, it can also be seen that any competing low-redshift
solution (shown as the dashed grey SED fits) can be ruled-
out at high statistical significance. The fundamental reason
for this is the additional information on the spectral slope
long-ward of Lyα provided by the UDS near-infrared pho-
tometry. Without this additional information any sample of
bright z ? 6 LBGs candidates is vulnerable to substantial
contamination by low-redshift interlopers, a point we will
return to in Section 4.
∼6.
3LYMAN ALPHA
3.1Lyα equivalent widths
The Lyα equivalent widths were obtained by measuring the
observed line flux, and estimating the UV continuum at Lyα
from Subaru NB921 narrow-band (NB) imaging available in
the UDS (Ouchi et al. 2009).
The line flux was measured by subtracting any contin-
uum from the line profile, where the continuum is estimated
from regions either side of the line unaffected by sky line
residuals, then integrating the flux over the pixels contribut-
ing to the line. To ensure that the calibration of the spectra
matches the photometry, they are convolved with the Sub-
aru z?-band filter profile and the flux scaled to match the
z?-band photometry (see Fig. 4). To ensure that sky resid-
uals do not dominate the derived flux from the convolved
spectrum, the continuum level is estimated just red-wards
Figure 4. An illustration of the method adopted to calculate the
Lyα EWs. The spectrum is of object UUDSJ 021735.34-051032.6
scaled to the height of the Lyα line. Over-plotted as the red solid
line is the throughput of the Subaru z?-band filter. The spectrum
is convolved with this filter and scaled to match the z?-band pho-
tometry. To ensure that sky residuals do not dominate the con-
volution the continuum is modelled as a simple step function,
as shown by the black long dashed line, where the continuum is
derived from regions either side of the emission lines (shown by
the points with associated error bars). The green dot-dashed line
shows the Subaru narrow-band NB921 filter profile. Photometry
from this filter is used to determine the rest-frame UV flux red-
ward of Lyα emission. This is used with the calibrated line flux
to determine the Lyα EW.

Page 7

Spectrocopic confirmation of luminous LBGs at z ? 67
and blue-wards of Lyα and is modelled by a simple step
function. Over the small wavelength range within the filter,
differences in true slope red-wards of Lyα produce minimal
differences to the final convolved flux.
Absolute calibration of the spectra in this way is not ef-
fective if the signal-to-noise is so low in the continuum that
it is effectively measured as zero. In these cases (noted in
Table 1) the median correction factor measured from all the
other spectra is used and the standard deviation of correc-
tion factors is folded into the final error for these line fluxes.
The final error in the Lyα flux measurement includes
pixel error estimates from the noise array, estimated error
in the subtracted continuum measurement, as well as the
estimated error in the derived calibration. This calibration
error includes the error in the convolved flux, assuming er-
rors per pixel derived from the standard deviation of pixel
values for those contributing to continuum estimates, as well
as the errors in the z?-band photometry.
The NB921 photometry samples the continuum red-
wards of Lyα for z < 6.5 (see Fig. 4) and to estimate the
continuum at Lyα a flat UV continuum in Fν (Fλ ∝ λ−2)
is assumed. The NB photometry is used to estimate the UV
continuum rather than direct measurements made from the
spectra because sky subtraction hinders accurate continuum
measurements in the spectra, as well as the signal-to-noise
in any detected continuum being very low (< 2) in many
of the spectra. The errors in the derived EWs include the
errors in NB photometry3. No attempt was made to correct
either the UV continuum or Lyα line flux for dust extinction
during the EW calculation.
Both the z?-band and narrow-band fluxes were cor-
rected to total using measured aperture corrections from
stars within the individual SXDS pointings. The measured
equivalent widths were compared to estimates taken directly
from the spectra themselves and are found to be in good
agreement, indicating that the quoted equivalent widths are
secure.
3.2Lyα and UV-derived SFRs
Both the UV continuum and Lyα emission-line luminosity
are commonly used to to derive SFR estimates. Both of these
indicators are sensitive to the presence of dust and Lyα is
also sensitive to neutral hydrogen within the galaxy and
the surrounding IGM. In Fig. 5 we compare the SFR esti-
mates derived from the UV continuum luminosity using the
Madau et al. (1998) formula (corrected to a Chabrier IMF),
with those derived from the measured Lyα luminosity (e.g.
Nilsson et al. 2009), with no correction for dust extinction:
LUV (/ergs−1Hz−1) = 4.8 × 1027× SFRUV (/M?yr−1) (1)
LLyα(/ergs−1) = 9.7 × 1041× SFRLyα(/M?yr−1)
The constant in equation 1 is based on the continuum lu-
minosity at λrest = 1500˚ A and has been converted to a
Chabrier IMF using a ratio of 1.65, the asymptotic ratio
(2)
3For the two highest redshift LBGs, the Lyα emission line lies
within the extreme blue-end of the NB921 filter. Although the
throughput of the filter is extremely low at these wavelengths, it is
possible that the resulting EW measurements are underestimated
by ? 5%.
Table 2. Star-formation rates derived from Lyα emission-line
luminosity and UV continuum luminosity (without any correction
for dust extinction). The UV-derived SFR (SFRUV) is estimated
from the UV luminosity at λrest = 1500˚ A using equation 1 and
the Lyα-derived SFR (SFRLyα) is estimated from the luminosity
of the Lyα emission using equation 2.
IDSFRLyα
(/M? yr−1)
13.8 ± 1.3
7.4 ± 1.3
12.3 ± 1.1
13.4 ± 1.1
16.0 ± 1.5
4.2 ± 0.5
7.5 ± 1.0
8.8 ± 1.5
5.0 ± 0.9
SFRUV
(/M? yr−1)
30.2 ± 2.6
18.4 ± 1.7
40.6 ± 3.8
26.3 ± 3.6
37.0 ± 3.8
30.5 ± 2.4
13.2 ± 2.4
29.1 ± 4.4
43.3 ± 4.0
UUDS J021800.90-051137.8
UUDS J021616.53-050217.7
UUDS J021807.14-045841.5
UUDS J021735.33-051032.6
UUDS J021838.90-050944.0
UUDS J021653.00-044623.3
UUDS J021922.01-045536.3
UUDS J021841.02-051247.4
UUDS J021701.44-050309.4
between the number of ionizing photons produced by a con-
stant SFR with a Salpeter compared to a Chabrier IMF
predicted by Bruzual & Charlot (2003).
Although there is a large scatter between the two SFR
measurements, it can immediately be seen from Fig. 5 that
the SFR estimate provided by the Lyα luminosity is system-
atically lower than that provided by the UV continuum lumi-
nosity, with a mean value of SFRLyα/SFRUV = 0.36±0.05.
We note that this is very close to the value of 0.40 which
would be predicted by the recent study of the redshift evo-
lution of the Lyα escape fraction by Hayes et al. (2011). At
low redshift Hayes et al. (2011) calculate fLyα
ing the observed Lyα luminosity function with a prediction
based on the extinction corrected Hα luminosity function
(assuming case B recombination). Above z ∼ 2.2, Hα is
no longer easily observable and the intrinsic Lyα luminosity
function is predicted from the UV, with extinction estimates
being made from SED fitting. This calculation is essentially
taking the ratio of Lyα and UV star formation rate densities,
˙ ρ? (equation 3) where the star-formation rate densities are
derived from integrating over the Lyα and UV luminosity
functions respectively, i.e.
esc
by compar-
fLyα
esc =
˙ ρObs
? Lyα
100.4EB−VkUV× ˙ ρObs
Where kUV = 10.3 (Calzetti et al. 2000). Based on this
formalism, Hayes et al. (2011) find that the escape frac-
tion varies as fLyα
esc
∝ (1 + z)ξwith ξ = 2.57+0.19
fLyα
esc
? 0.25 at z ? 6. If we substitute fLyα
˙ ρObs
? UV = 0.36 into equation 3 we obtain a predic-
tion that EB−V ? 0.04. This estimate of EB−V is in ex-
cellent agreement with the Hayes et al. (2011) fit to an ob-
served anti-correlation between EB−V and fLyα
at z ∼ 2. This fit gives fLyα
(with CLyα = 0.445 and kLyα = 12), providing an estimate
of EB−V ? 0.05 for fLyα
results of Hayes et al. (2011), we conclude that our sample
of L ? 2L?LBGs at z ? 6 are consistent with a Lyα escape
fraction of fLyα
Finally, we note that in Fig. 5, while most of the sample
agree well with the prediction of Hayes et al. (2011), the two
? UV
(3)
−0.12, giving
= 0.25 and
esc
? Lyα/ ˙ ρObs
esc
for sources
esc
= CLyα × 10−0.4EB−VkLyα
esc ? 0.25. Consequently, based on the
esc ? 25%.

Page 8

8E. Curtis-Lake et al.
Figure 5. Lyα-derived SFRs vs. UV-derived SFRs. The short
dashed red line shows the average Lyα:UV SFR ratio (0.36) and
the black long dashed line shows the ratio of 0.4 predicted from
Hayes et al. (2011) (for fLyα
esc
= 0.25 at z ∼ 6 with E(B-V) =
0.05, see text). Objects with z > 6.2 are plotted as red triangles.
largest outliers are drawn from the high-redshift end of the
sample at z > 6.2 and suggest a lower value of fLyα
esc .
4THE LYMAN-ALPHA EMITTER FRACTION
In this section we use our sample of spectroscopically con-
firmed LBGs to provide entirely new information on the Lyα
emitter fraction of L ? 2L?LBGs at z ? 6. As previously
described, the original sample of spectroscopic candidates
was selected using the redshift probability density function
derived using a photometric redshift analysis (McLure et al.
2009). However, although this strategy makes optimal use
of the available data, in order to perform a comparison be-
tween our results and those in the literature it is necessary
to re-engineer our candidate selection in terms of traditional
colour-cut criteria. Consequently, all of the parent sample
considered for spectroscopic follow-up also satisfy the fol-
lowing criteria:
z?< 26(4)
i?− z?? 2.0
z?− J ? 0.8
where all magnitudes are measured within a 2??diameter
aperture and we also require each object to be a non-
detection in the BV R bands at the 2σ level. The choice
of a stringent i?−z?? 2 colour cut is motivated by the fact
that any galaxy at z ? 6 is predicted to display a drop of at
least two magnitudes between the Subaru i?and z?filters as-
suming the Madau (1995) prescription for IGM absorption.
The z?− J ? 0.8 colour-cut is motivated by the desire to
effectively exclude low-redshift interlopers at z ? 1.5, while
maintaining sensitivity to genuine z ? 6 galaxies with mod-
erate reddening.
Within the overlap region between the optical and near-
infrared imaging in the UDS the surface density of objects
(5)
(6)
Figure 6. A greyscale representation of the region with over-
lapping Subaru (optical) and UKIRT (near-infrared) data in the
UDS. Over-plotted in blue are the locations of the 20 FORS2
masks from the UDSz programme and the area imaged by
HST/WFC3 as part of the CANDELS survey (Grogin et al. 2011)
is shown in red. The small black squares show all the objects in
the field that are plausible z ? 6 candidates following the criteria
described in the text. The green squares show the spectroscop-
ically confirmed z ? 6 LAEs, while the green circles show the
objects for which the FORS2 spectra do not provide an unam-
biguous redshift.
obeying the above selection criteria is 0.023 ± 0.003 per
square arcmin Therefore, within the total area covered by
the FORS2 pointings (20 pointings, each covering an area
of 6.8?×6.8?), we would expect ∼ 15−20 candidates. From
Fig. 6, which shows the region of optical/near-infrared over-
lap in the UDS with the positions of the FORS2 masks and
the candidates satisfying the above criteria over-plotted, it
can be seen that there are 19 candidates in this area, with
14 of those having been targeted for spectroscopy. Conse-
quently, we conclude that the sample of candidates targeted
for spectroscopy appears to be consistent with being drawn
randomly from the parent population of candidates satis-
fying the colour-cut criteria described above (although see
section 4.2.4 for further discussion).
From our sample of fourteen objects we find seven which
display strong Lyα emission (EW?25˚ A), which suggests a
minimum Lyα emitter fraction of 50%. However, if we ex-
clude the AGN from our sample this fraction increases to
54%. Either way, these are very high Lyα fractions com-
pared to previous literature results for luminous LBGs and
clearly require some discussion.
4.1 Comparison with the literature
We now look to other studies of spectroscopically targeted
LBGs at z ∼ 6 to see whether this Lyα fraction is com-
parable. Stark et al. (2011) present a sample of i?-band

Page 9

Spectrocopic confirmation of luminous LBGs at z ? 69
Figure 7. The evolution with redshift of the fraction of lumi-
nous LBGs (−21.75 < MUV < −20.25) which display Lyα line
emission with rest-frame EW? 25˚ A. The black squares show the
measured fractions from Stark et al. (2010, 2011) and the black
diamond is the measured fraction at z ∼ 7 estimated by Ono
et al. (2011) from the combination of a number of spectroscopic
samples. The red triangle shows the expected value at z ∼ 7 ex-
trapolated from the trends at lower redshift (Stark et al. 2011),
with the fitted linear relationship and uncertainties plotted as the
red solid and dashed lines, respectively. The blue star shows the
results of this work based on the sample selected from the UV-
continuum flux and with upper and lower limits as described in
the text.
dropouts selected with a i775 − z850 > 1.34colour cut,
which is a considerably bluer cut than that adopted here.
This sample, when compared to results at lower redshift
(Stark et al. 2010), provides strong evidence that the frac-
tion of objects showing Lyα in emission increases from
z = 4 to z = 6 for galaxies in the absolute magnitude
range −20.25 < MUV < −18.75. However, in contrast, lit-
tle evolution in the Lyα fraction is seen for brighter ob-
jects with UV luminosities more comparable to our sample
(−21.75 < MUV < −20.25). In this luminosity range, Stark
et al. (2011) find that the fraction of objects showing Lyα
in emission is 20%±8.1% and 7.4%±5.0% for objects with
EWs greater than 25˚ A and 55˚ A respectively.
Considering the small size of our sample, at the higher
EW? 55˚ A threshold our results are in reasonably good
agreement with Stark et al. (2011), given that one of our
objects has EW ? 55˚ A (i.e. 1/9). However, at the lower
EW? 25˚ A threshold our sample shows a substantially larger
fraction of LBGs with Lyα emission (? 54%).
4.2Potential biases
Understanding or reconciling the difference in the measure-
ments made by these two studies is difficult due to the dif-
ferent selection mechanisms used and the small numbers of
bright objects available for study in fields smaller than the
UDS. In fact, it is not clear that a direct comparison be-
tween our results and those of Stark et al. (2011) is actually
4i775 and z850 are the F775W and F850LP HST/ACS filters
respectively.
meaningful, given that the vast majority of our sample are
confined to the bright half of the luminous absolute magni-
tude bin (−21.75 < MUV < −20.25) defined by Stark et al.
(2011). However, given that the trend reported by Stark et
al. (2011), is for a decreasing fraction of strong Lyα emitters
at z ? 6 with increasing UV luminosity, our results indicat-
ing a high fraction of Lyα emitters amongst luminous z ? 6
LBGs clearly requires some explanation. In this section we
investigate several effects which could have potentially bi-
ased our determination of the Lyα emitter fraction.
4.2.1 Reddening
First, we consider whether our colour selection criteria are
selecting against galaxies with internal reddening by dust
that may produce lower Lyα EWs. The i?− z?? 2.0 colour
cut does not select against reddened galaxies as this colour
is dependent on the fraction of flux attenuated in clumpy HI
regions in the intervening IGM, which is independent of the
galaxy properties. Reddening tends to increase this colour
since there is more flux at redder wavelengths still contribut-
ing to the z?−band flux. It is possible that the true IGM
absorption for individual galaxies may scatter the colours of
genuine z ? 6 galaxies below the colour cut, but this scat-
ter should be independent of the intrinsic galaxy properties.
The z?− J colour cut may well exclude a few redder z ? 6
galaxies from our sample. This colour cut is designed to ex-
clude low-redshift galaxies and cool galactic stars which are
able to satisfy the i?− z?colour cut. However, because we
are insisting on a strict i?− z?? 2.0 colour, only five addi-
tional objects are excluded on the basis of the z?− J ? 0.8
criterion over the entire UDS field, so we do not expect our
sample to be greatly biased towards dust-free galaxies.
However, it is important to note that if a bluer i?− z?
colour cut is adopted, the potential for contamination in-
creases due to a population of massive red galaxies at
z ? 1.5 with a surface density which increases significantly
at z?? 26. As a result, any samples selected without near-
infrared data and a bluer i?− z?colour cut will be signifi-
cantly contaminated by low-redshift interlopers at z?? 26.
Of course, until the present study, selecting significant sam-
ples of z ? 6 objects with z?? 26 has not been possible due
to insufficient area.
4.2.2Contamination from low-redshift interlopers
To further explore how low-redshift interlopers could con-
taminate samples of bright z ? 6 LBGs, we have investi-
gated different selection criteria using the deep HST data
available in GOODS-S. All of our objects were found to sat-
isfy the HST colour selection of i775−z850 > 1.7. This lower
colour cut is due to the difference in filter profiles between
the HST z850−band compared to the Subaru z?band, as well
as slightly increased overlap between the HST i775 and z850
band filters. With the new HST/WFC3 near-infrared data
available as part of the CANDELS survey (Grogin et al.
2011; Koekemoer et al. 2011) we can investigate how our
z?− J cut may affect these samples. We find that only two
objects in the field satisfy i775− z850 > 1.7 with z850 < 26,
both of which are excluded by a z850 − J125 < 0.75 crite-
rion (consistent with our z?−J < 0.8 colour cut). However,

Page 10

10E. Curtis-Lake et al.
in a sample selected using i775− z850 > 1.3; z850 < 27, we
find that > 90% of objects survive the z850− J125 < 0.75
colour cut. It is therefore clear that while a colour-cut of
i775− z850 > 1.3 should be sufficient to select a clean sam-
ple of z ? 6 LBGs at z850 ? 27, at magnitudes brighter
than z?? 26 suitably deep near-infrared photometry is es-
sential to avoid significant contamination from low-redshift
interlopers. In fact, this point is very well illustrated by the
SED template fits to our spectroscopically confirmed LBGs
shown in Fig. 2.
4.2.3EW measurements and cosmic variance
It is worth noting at this point that two of our objects with
strong Lyα emission have EWs that are quite close to the
EW? 25˚ A cut. Although our method of EW calculation is
quite robust, not all fields have narrow band imaging just
red-wards of Lyα to supply a continuum estimate, and it is
difficult to get high signal to noise in the continuum from
the actual spectra at these redshifts. It is certainly feasible
that with a different continuum estimator, or a lower limit
derived from the noise in the spectrum, that these EWs
could be measured to be lower, pushing down the measured
Lyα fraction.
Moreover, it is also important to consider the problem
of cosmic variance. Even selecting over an area of 0.25 sq.
degrees, our sample is clearly limited by small number statis-
tics. As illustrated by Fig. 6, previous samples of luminous
z ∼ 6 LBGs selected over the GOODS fields (the CANDELS
imaging in the UDS covers the area of one GOODS field)
will include very few, if any, objects as luminous as those in
our sample. Again, this simply highlights the point that our
sample is exploring a new area of parameter space.
4.2.4Photometric redshift selection
One potential issue which needs to be addressed is the fact
that our original object selection was performed using the
Subaru z?-band imaging in the UDS which, for objects at
z ? 6, is affected by both IGM absorption and a varying Lyα
emission line contribution. As a result, we can probably re-
gard our estimate of 54% for the fraction of luminous z ? 6
LBGs which display Lyα emission with EW? 25˚ A as an up-
per limit. The fundamental reason for this is that, due to
our original photometric redshift selection our spectroscopic
sample is not an entirely random sampling of the parent pop-
ulation satisfying the colour-cut criteria listed previously. In
fact, because we targeted objects with the highest probabil-
ity of being at z ? 6, the location of the Lyα emission line
within the selection filter introduces a bias towards preferen-
tially targeting objects with strong Lyα emission. In short,
the presence of strong Lyα emission leads to an exaggerated
Lyman break and an apparently bluer UV spectral slope,
both of which can conspire to produce a more robust pho-
tometric redshift solution at z ? 6 than would otherwise be
the case.
4.3An unbiased estimate of the Lyman-alpha
fraction using UV continuum selection
Given the complications introduced by selecting our original
z ? 6 sample in the Subaru z?−band, we can take advantage
of the availability of the NB921 imaging data to investigate
the Lyα fraction in a UV-continuum selected sub-sample. In
general, for z < 6.4 where the narrow-band imaging is clear
of the Lyman-break, we expect the narrow-band fluxes to
be the same or brighter than the z?−band fluxes for any ob-
ject with no Ly-α emission (this is reversed if there is strong
Lyα emission contributing to the z?-band). A comparison
of the z?−band and NB921 photometry for our sample sug-
gests that our original selection limit of z?? 26 provides a
complete sample of objects with z921 ? 25.7 (all objects sat-
isfying equations 5 & 6 with z921 ? 25.7 also obey z?? 26).
This narrow-band magnitude cut corresponds to an absolute
UV magnitude of MUV = −21.0 at z = 6 and MUV = −21.1
at z = 6.5 (assuming a flat UV slope in fν). At this limit we
find 45% of objects (5/11) with Lyα EW ? 25˚ A, confirming
a high fraction of Lyα emitters at MUV ? −21.1.
4.4A lower limit to the Lyman-alpha fraction
Using our spectroscopic sample it is also possible to estimate
a hard lower limit to the Lyα fraction amongst luminous
z ? 6 LBGs, if we consider the unlikely scenario in which our
EW? 25˚ A Lyα detections are the only strong Lyα emitters
out of the 19 objects which could have been spectroscopi-
cally targeted. Even in this extreme scenario the fraction of
galaxies showing Lyα with EW?25˚ A is 37%±14% (7/19),
where the quoted error is the poisson uncertainty.
However, in reality, this situation is better described by
a binomial distribution which predicts that the probability
of finding 7 or more galaxies with EW ? 25˚ A from a sample
of 14 is only p = 0.01, if the chance of “success” in each trial
(i.e. finding EW ? 25˚ A) is p = 0.20 (as is the case for the
most luminous z ∼ 6 objects in Stark et al. 2011). However,
given the potential for bias, perhaps a more reasonable cal-
culation is to determine the probability of finding 7 or more
galaxies with EW ? 25˚ A from all 19 potential spectroscopic
targets. This returns a probability of p = 0.07, meaning that
we are only able to exclude the hypothesis that the true frac-
tion of EW? 25˚ A Lyα emitters within our sample is 20%
at the ? 93% confidence level.
4.5 Dust free luminous LBGs
Our Lyα fraction results are plotted in Fig. 7 along with the
results of Stark et al. (2010, 2011) and Ono et al. (2011).
It is clear from this plot that our best estimate of ? 45%
for the Lyα fraction amongst luminous LBGs at z ? 6 is
significantly higher than found by Stark et al. (2011) at the
same redshift and by Ono et al. (2011) at z ? 7 (although we
note that, before combining their results with other studies
taken from the literature, they find a fraction of ? 33%).
As far as we can tell, this result is not significantly biased
by our sample selection and, as can be seen from Fig. 7, is
in good agreement with a simple extrapolation of the Lyα
fraction derived by Stark et al. (2010) for luminous LBGs
at z = 4 and z = 5.
One possibility is that by sampling the bright end of

Page 11

Spectrocopic confirmation of luminous LBGs at z ? 6 11
the −21.75 < MUV < −20.25 bin, we are starting to see
the effects of targeting the extreme tail of the population
of the less-dusty, highly star forming systems. As we move
to brighter and brighter objects, the extremely steep slope
of the luminosity function would seem to require that sam-
ples eventually become dominated by objects with low dust
reddening and, presumably, a correspondingly high escape
fraction of Lyα photons.
4.6 A UV continuum selected sample
As discussed previously, selecting z ? 6 galaxy samples
based on z?−band photometry involves the added complica-
tions of varying amounts of IGM absorption and Lyα con-
tamination. In this study, we were able to correct any possi-
ble biases by using narrow-band imaging data red-wards of
Lyα to define a sample complete to a given depth in the UV
continuum. However, to achieve this at higher redshifts it
is clearly necessary to perform the primary sample selection
in the near-infrared. Within this context, the new Cosmic
Assembly Near-infrared Deep Extragalactic Legacy Survey
(CANDELS; Koekemoer et al. 2011; Grogin et al. 2011) will
supply deep, near-infrared, imaging (5σ depth of H160 ? 27)
over a total area of ∼ 700 square arcmin. The combination
of depth and area provided by CANDELS will allow the Lyα
fraction amongst large, unbiased, samples of z > 7 LBGs to
be studied in the near future.
5 CONCLUSIONS
Targeting a sample of z ? 6 galaxy candidates originally
selected using a photometric redshift analysis, we obtained
a very high redshift completeness, with 11/14 objects pro-
viding robust spectroscopic redshifts from the detection of a
Lyα emission line. Comparing the star-formation rate esti-
mates based on UV continuum and Lyα luminosity we find
that our sample is consistent with a Lyα escape fraction of
fLyα
esc
? 25%, in agreement with the recent study of Hayes
et al. (2011).
Based on our sample of L ? 2L?LBGs at z ? 6 we
derive estimates for the maximum (54%±20%) and mini-
mum (37%±14%) fraction of Lyα emitters with EW? 25˚ A.
Testing whether these fractions are biased by our z?-band
selection criteria, we use the available NB photometry to
calculate the fraction of EW? 25˚ A Lyα emitters amongst a
complete, UV continuum selected, sub-sample. This calcu-
lation returns a fraction of 45%±15%, consistent with our
previous estimates.
Our estimate of the EW? 25˚ A Lyα emitter fraction in
the magnitude range −21.75 < MUV < −21.25 is a factor
of ? 2 larger than previous estimates (i.e. 20 ± 8%; Stark
et al. 2011). Indeed, we calculate that our sample is incon-
sistent with a Lyα emitter fraction of 20% at the ? 93%
confidence level. However, we also note that our sample ex-
plores a higher redshift and luminosity range than previous
studies.
In summary, our results suggest that, as the epoch of
reionization is approached, it is plausible that the Lyα emit-
ter fraction amongst luminous (L ? 2L?) LBGs shows a
similarly sharp increase to that observed in their lower-
luminosity (L ? L?) counterparts.
ACKNOWLEDGMENTS
ECL would like to acknowledge financial support from the
UK Science and Technology Facilities Council (STFC) and
the Leverhulme Trust. RJM would like to acknowledge the
funding of the Royal Society via the award of a University
Research Fellowship and the Leverhulme Trust via the award
of a Philip Leverhulme research prize. JSD acknowledges the
support of the Royal Society via a Wolfson Research Merit
award, and also the support of the European Research Coun-
cil via the award of an Advanced Grant. MC acknowledges
the award of an STFC Advanced Fellowship. DPS acknowl-
edges support from NASA through Hubble Fellowship grant
#HST-HF-51299.01 awarded by the Space Telescope Sci-
ence Institute, which is operated by the Association of Uni-
versities for Research in Astronomy, Inc., for NASA under
contract NAS5-26555. HJP, RC, and EB acknowledge the
award of an STFC PhD studentships. WGH acknowledges
the award of an STFC PDRA. The authors would like to
thank the HiZELS team for supplying the Subaru NB921
imaging data and the staff at UKIRT for operating the tele-
scope with such dedication under difficult circumstances.
REFERENCES
Ando M., Ohta K., Iwata I., Akiyama M., Aoki K., Tamura
N., 2006, ApJ, 645, L9
Bouwens R. J., Illingworth G. D., Franx M., Ford H., 2007,
ApJ, 670, 928
Bouwens R. J., Illingworth G. D., Oesch P. A., Labb´ e I.,
Trenti M., van Dokkum P., Franx M., Stiavelli M. et al.,
2011, ApJ, 737, 90
Bouwens R. J., Illingworth G. D., Oesch P. A., Stiavelli
M., van Dokkum P., Trenti M., Magee D., Labb´ e I. et al.,
2010, ApJ, 709, L133
Bouwens R. J., Illingworth G. D., Oesch P. A., Trenti M.,
Stiavelli M., Carollo C. M., Franx M., van Dokkum P. G.,
Labb´ e I., Magee D., 2010, The Astrophysical Journal, 708,
L69
Bruzual G., Charlot S., 2003, MNRAS, 344, 1000
Calzetti D., Armus L., Bohlin R. C., Kinney A. L., Koorn-
neef J., StorchiBergmann T., 2000, The Astrophysical
Journal, 533, 682
Casali M., Adamson A., Alves de Oliveira C., Almaini O.,
Burch K., Chuter T., Elliot J., Folger M. et al., 2007,
A&A, 467, 777
Chabrier G., 2003, Publications of the Astronomical Soci-
ety of the Pacific, 115, 763
Dayal,Pratika Ferrara,Andrea 2011, eprint arXiv:1102.1726
Dijkstra M., Mesinger A., Wyithe J. S. B., 2011, Monthly
Notices of the Royal Astronomical Society, 414, 2139
Finkelstein S. L., Papovich C., Giavalisco M., Reddy N. A.,
Ferguson H. C., Koekemoer A. M., Dickinson M., 2010,
ApJ, 719, 1250
Fontana A., Vanzella E., Pentericci L., Castellano M., Gi-
avalisco M., Grazian A., Boutsia K., Cristiani S. et al.,
2010, ApJ, 725, L205
Furusawa H., Kosugi G., Akiyama M., Takata T., Sekiguchi
K., Tanaka I., Iwata I., Kajisawa M. et al., 2008, ApJS,
176, 1
Forero-Romero J. E., Yepes G., Gottloeber S., Prada F.,
2011, eprint arXiv:1109.0228

Page 12

12E. Curtis-Lake et al.
Grogin N. A., Kocevski D. D., Faber S. M., Ferguson H. C.,
Koekemoer A. M., Riess A. G., Acquaviva V., Alexander
D. M. et al., 2011, eprint arXiv:1105.3753
Guhathakurta P., Tyson J. A., Majewski S. R., 1990, ApJ,
357, L9
Hayes M., Schaerer D.,¨Ostlin G., Mas-Hesse J. M., Atek
H., Kunth D., 2011, The Astrophysical Journal, 730, 8
Hewett P. C., Warren S. J., Leggett S. K., Hodgkin S. T.,
2006, MNRAS, 367, 454
Hu E. M., Cowie L. L., McMahon R. G., 1999, The Hy-
Redshift Universe: Galaxy Formation and Evolution at
High Redshift, 193
Ilbert O., Arnouts S., McCracken H. J., Bolzonella M.,
Bertin E., Le F` evre O., Mellier Y., Zamorani G. et al.,
2006, A&A, 457, 841
Koekemoer A. M., Faber S. M., Ferguson H. C., Grogin
N. A., Kocevski D. D., Koo D. C., Lai K., Lotz J. M. et
al., 2011, eprint arXiv:1105.3754
Lawrence A., Warren S. J., Almaini O., Edge A. C., Ham-
bly N. C., Jameson R. F., Lucas P., Casali M. et al., 2007,
MNRAS, 379, 1599
McLure R. J., Cirasuolo M., Dunlop J. S., Foucaud S.,
Almaini O., 2009, MNRAS, 395, 2196
McLure R. J., Dunlop J. S., Cirasuolo M., Koekemoer
A. M., Sabbi E., Stark D. P., Targett T. A., Ellis R. S.,
2010, MNRAS, 403, 960
McLure R. J., Dunlop J. S., de Ravel L., Cirasuolo M., Ellis
R. S., Schenker M., Robertson B. E., Koekemoer A. M.
et al., 2011, eprint arXiv:1102.4881
McLure R. J., Jarvis M. J., Targett T. A., Dunlop J. S.,
Best P. N., 2006, MNRAS, 368, 1395
Madau P., 1995, ApJ, 441, 18
Madau P., Pozzetti L., Dickinson M., 1998, ApJ, 498, 106
Nilsson K. K., Tapken C., Mø ller P., Freudling W., Fynbo
J. P. U., Meisenheimer K., Laursen P.,¨Ostlin G., 2009,
A&A, 498, 13
Oke J. B., Gunn J. E., 1983, ApJ, 266, 713
Ono Y., Ouchi M., Mobasher B., Dickinson M., Penner K.,
Shimasaku K., Weiner B. J., Kartaltepe J. S. et al., 2011,
eprint arXiv:1107.3159
Ouchi M., Shimasaku K., Akiyama M., Simpson C., Saito
T., Ueda Y., Furusawa H., Sekiguchi K. et al., 2008, ApJS,
176, 301
Ouchi M., Mobasher B., Shimasaku K., Ferguson H. C.,
Fall S. M., Ono Y., Kashikawa N., Morokuma T., Naka-
jima K., Okamura S., Dickinson M., Giavalisco M., Ohta
K., 2009, The Astrophysical Journal, 706, 1136
Ouchi M., Shimasaku K., Furusawa H., Saito T., Yoshida
M., Akiyama M., Ono Y., Yamada T. et al., 2010, ApJ,
723, 869
Pentericci L., Fontana A., Vanzella E., Castellano M.,
Grazian A., Dijkstra M., Boutsia K., Cristiani S. et al.,
2011, eprint arXiv:1107.1376
Reddy N. A., Steidel C. C., Pettini M., Adelberger K. L.,
Shapley A. E., Erb D. K., Dickinson M., 2008, ApJS, 175,
48
Schenker M. A., Stark D. P., Ellis R. S., Robertson B. E.,
Dunlop J. S., McLure R. J., Kneib J. P., Richard J., 2011,
eprint arXiv:1107.1261
Sekiguchi K., Akiyama M., Furusawa H., Simpson C.,
Takata T., Ueda Y., Watson M. W., The Sxds Team
2005, Multiwavelength Mapping of Galaxy Formation and
Evolution. ESO Astrophysics Symposia, Springer-Verlag,
Berlin/Heidelberg
Shapley A. E., Steidel C. C., Pettini M., Adelberger K. L.,
2003, ApJ, 588, 65
Shimasaku K., Kashikawa N., Doi M., Ly C., Malkan M. A.,
Matsuda Y., Ouchi M., Hayashino T. et al., 2006, PASJ,
pp 313–334
Stark D. P., Ellis R. S., Chiu K., Ouchi M., Bunker A.,
2010, MNRAS, 408, 1628
Stark D. P., Ellis R. S., Ouchi M., 2011, ApJ, 728, L2
Vanzella E., Giavalisco M., Dickinson M., Cristiani S., Non-
ino M., Kuntschner H., Popesso P., Rosati P. et al., 2009,
ApJ 695, 1163
Vanzella E., Pentericci L., Fontana A., Grazian A., Castel-
lano M., Boutsia K., Cristiani S. et al., 2011, ApJ, 730,
L35
Willott C. J., Delorme P., Reyl´ e C., Albert L., Bergeron J.,
Crampton D., Delfosse X., Forveille T. et al., 2009, The
Astronomical Journal, 137, 3541
Willott C. J., Delorme P., Reyl´ e C., Albert L., Bergeron J.,
Crampton D., Delfosse X., Forveille T. et al., 2010, The
Astronomical Journal, 139, 906
Yoshida M., Shimasaku K., Kashikawa N., Ouchi M., Oka-
mura S., Ajiki M., Akiyama M., Ando H. et al., 2006, ApJ,
653, 988