New Constraints on the Lyman Continuum Escape Fraction at z~1.3
Brian Siana, Harry I. Teplitz, James Colbert, Henry C. Ferguson, Mark Dickinson, Thomas M. Brown, Christopher J. Conselice, Duilia F. de Mello, Jonathan P. Gardner, Mauro Giavalisco, Felipe Menanteau
aa r X i v : . [ a s t r o - ph ] J un New Constraints on the Lyman Continuum Escape Fraction at z ∼ . † Brian Siana, Harry I. Teplitz, James Colbert , Henry C. Ferguson , Mark Dickinson ,Thomas M. Brown , Christopher J. Conselice , Duilia F. de Mello , , Jonathan P.Gardner , Mauro Giavalisco , Felipe Menanteau [email protected] ABSTRACT
We examine deep far-ultraviolet (1600˚A) imaging of the Hubble Deep Field-North (HDFN) and the Hubble Ultra Deep Field (HUDF) to search for leakingLyman continuum radiation from starburst galaxies at z ∼ .
3. There are 21(primarily sub- L ∗ ) galaxies with spectroscopic redshifts between 1 . < z < . f esc ) for these galaxies. First, to compare with previousworks, we assume a fixed 1500 ˚A to Lyman continuum ratio, f /f , intrinsicto the stellar SEDs to convert our f limits to relative escape fractions. Second,we fit stellar population templates to the galaxies’ optical/near-infrared SEDs todetermine the starburst age and level of dust attenuation for each individual Spitzer
Science Center, California Institute of Technology, 220-6, Pasadena, CA 91125 Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218 National Optical Astronomy Observatory, 950 N. Cherry Ave., Tucson, AZ 85719 University of Nottingham, Nottingham, NG7 2RD, UK Department of Physics, Catholic University of America, 620 Michigan Avenue, Washington DC 20064 Astrophysics Science Division, Observational Cosmology Laboratory, Code 665, Goddard Space FlightCenter, Greenbelt, MD 20771 University of Massachusetts, Department of Astronomy, Amherst, MA, 01003 Department of Physics and Astronomy, Rutgers University, 136 Frelinghuysen Road, Piscataway, NJ08854 † Based on observations made with the NASA/ESA Hubble Space Telescope, obtained at the SpaceTelescope Science Institute, which is operated by the Association of Universities for Research in Astronomy,Inc., under NASA contract NAS 5-26555. These observations are associated with programs 7410, 9478, and10403 (as well as 6337 HDF-WFPC2, 7817 HDF-NICMOS, 9978,10086 HUDF-ACS,9803 HUDF-NICMOS). f /f . We findthat the two techniques do not differ significantly if the galaxies are currentlyundergoing star-formation. We show that previous high-redshift studies may haveunderestimated the amplitude of the Lyman Break, and thus the relative escapefraction, by a factor ∼
2. Once the starburst age and intergalactic HI absorptionare accounted for, 18 galaxies in our sample have limits to the relative escapefraction, f esc,rel < . f esc,rel < .
10 and a stackedlimit of f esc,rel < .
08. This demonstrates, for the first time, that most sub- L ∗ galaxies at high redshift do not have large escape fractions. When combinedwith a similar study of more luminous galaxies at the same redshift we showthat, if all star-forming galaxies at z ∼ σ ). We also show that less than 20% (3 σ ) ofstar-forming galaxies at z ∼ z ∼ z ∼ z ∼ Subject headings: cosmology: observations — galaxies: evolution — ultraviolet:galaxies
1. Introduction
Both active galactic nuclei (AGN) and young massive stars produce Lyman continuumradiation capable of ionizing hydrogen in the surrounding Intergalactic Medium (IGM). Thefraction of this photoionizing continuum which escapes into the IGM, hereafter referred toas the escape fraction ( f esc ), is a key parameter in determining how both star-formationand black hole growth drive the evolution of the ionization state of the IGM. In particular,an accurate assesment of the average f esc will help determine if star-formation, rather thanactive galactic nuclei (AGN), is responsible for hydrogen reionization at z > α forest have measured the total HI-ionizing background in the IGM as a function of redshift (McDonald & Miralda-Escud´e 3 –2001; Scott et al. 2002; Bolton et al. 2005; Fan et al. 2006). Recent measurements of theQSO luminosity functions at z < z ≥ .
5, the QSO space density falls off rapidly and recent measurements of the QSOluminosity functions at z ∼ z ∼ f esc examined low-redshift galaxies with space basedultraviolet (UV) telescopes. Leitherer et al. (1995) took UV spectra of four local starburstswhich, after accounting for absorption by HI in the Milky Way (Hurwitz et al. 1997), giveupper limits of f esc < . − .
57. Deharveng et al. (2001) took a deep FUSE spectrum ofMrk 45 and derived an upper limit of f esc < . λ ∼ z ∼ . f esc,rel , defined relative to the dust-attenuatedcontinuum flux at 1500 ˚A; see Section 2) f esc,rel < ∼ . − .
40. Steidel et al. (2001, hereafterS01) produced a composite spectrum of 29 Lyman Break Galaxies (LBGs) at z ∼ . L (1500˚ A ) /L (900˚ A ) =4 . ± . lambda rest ∼ z ∼ f esc,rel between 0.04 and 0.72.The first detection of escaping Lyman continuum from an individual galaxy was reportedby Bergvall et al. (2006) with a 16ks FUSE spectrum of Haro11, giving an escape fraction0 . < f esc < .
1, or a relative escape fraction of 0 . < f esc,rel < .
42 (though see Grimeset al. (2007, submitted) who see no detection down to f esc < .
02 with a re-reduction of thesame data). Haro11 is a luminous, metal poor, blue compact galaxy - attributes which aresimilar to LBGs at z ∼
3. However, it is somewhat underluminous ( L UV = 0 . L ∗ LBG ), mayhave a significantly larger halo mass than typical high-redshift LBGs, and was preselected ashaving a very low HI gas mass M ( HI ) ∼ M ⊙ (unpublished result noted in Bergvall et al. 4 –2006).Most recently, Shapley et al. (2006, hereafter S06) published deep optical spectra ofan additional 14 LBGs with an average f esc,rel ∼ .
14, about 4.5 times lower than theSteidel et al. (2001) result. Two of the galaxies appear to have f esc,rel ∼
1, while the remain-ing twelve are undetected. This suggests that 10-20% of LBGs have large escape fractions,or that the covering fraction of neutral HI and/or dust is 80-90% in these systems. This maybe the case if mergers or supernovae chimneys (Fujita et al. 2003) sufficiently disturb the HIreservoirs, producing low-column density lines-of-sight (LOSs) through the galaxies.With only three individual detections of escaping Lyman continuum, it is difficult todetermine whether any physical parameters (starburst age, luminosity, dust, HI mass, mor-phology) correlate with f esc . For example, the two detections from Shapley et al. (2006) donot stand out from the rest of the sample and are, in fact, very different from one another.Furthermore, most programs to date have necessarily targeted the more luminous starbursts,so the faint end of the luminosity function has not been sufficiently examined.We use deep far-UV imaging of the Hubble Deep Field North (HDFN, Williams et al.1996) and the Hubble Ultra Deep Field (HUDF, Beckwith et al. 2006) to search for theescaping ionizing flux from moderate redshift galaxies. The observations are taken with thef25QTZ filter on the Space Telescope Imaging Spectrograph (STIS) and the F150LP filteron the Solar-Blind Channel (SBC) of the Advanced Camera for Surveys (ACS) aboard theHubble Space Telescope (HST). The F25QTZ and F150LP broadband filters have effectivewavelengths of ∼ z ∼ .
3. These galaxies are typically less luminous than those of previousworks (Steidel et al. 2001; Malkan et al. 2003; Shapley et al. 2006), allowing us to probetrends with luminosity. In Section 2 we examine the various assumptions made about galaxyproperties which affect our derivation of the escape fraction. Section 3 explains the data setand sample selection. In Section 4.1 we use a simple method of applying these assumedcorrections to the f /f flux ratio to estimate the escape fraction. In Section 4.2, wefit stellar population models to the galaxies’ SEDs to determine the starburst age and dustattenuation. In Section 5, we compare the differences of these two methods and compareour limits to other surveys.Throughout the text, we use a flat ΛCDM Cosmology with H = 70 km s − Mpc − ,Ω m = 0 .
3, and Ω Λ = 0 .
7. All flux ratios use units of flux per unit frequency. 5 –
2. Determining the Escape Fraction
There are two commonly used definitions of the escape fraction in the literature. The absolute escape fraction, f esc , is simply the fraction of emitted Lyman continuum photons(typically measured at 900˚A) that escapes into the IGM. This definition is useful in the-oretical or semi-analytic models where it is used to directly translate star-formation ratesand initial mass functions (IMFs) into ionizing backgrounds. Observationally, the absoluteescape fraction is difficult to determine because the intrinsic Lyman continuum flux mustfirst be known. At low redshift, the H α line flux (corrected for reddening by measuring theBalmer decrement) is often used to approximate the number of Lyman continuum photonsemitted from massive stars. However, at high redshift this becomes increasingly difficult asthe Balmer lines shift to the near-IR. Therefore, the Lyman continuum flux is often inferredfrom the galaxy’s rest-frame UV and optical photometry, in particular the rest-frame 1500˚Aflux, f ν (1500˚A), since that is easily measured at z >
1. Unfortunately, the 1500˚A flux issubject to significant dust attenuation so this must be accurately accounted for in order touse it for determining the intrinsic Lyman continuum flux.Steidel et al. (2001) defined the relative escape fraction, f esc,rel as the fraction of escap-ing Lyman-continuum photons divided by the fraction of escaping photons at 1500˚A. Thisdefinition doesn’t require knowledge of the level of dust attenuation, and is useful at highredshift because f is easily measured and is commonly used when determining luminosityfunctions. Therefore, the relative escape fraction can be used to directly convert luminosityfunctions to ionizing backgrounds.In practice, f esc,rel is determined by comparing the Lyman continuum flux with therest-frame 1500˚A flux, f . Spectroscopic and narrow-band studies are sensitive to Lymancontinuum flux at wavelengths just short of the Lyman Break ( ∼ ∼
700 ˚A). This observed flux ratio between 1500˚A and the Lyman continuum isaffected by several factors and can be expressed as the following product( f /f LC ) obs = ( f /f LC ) stel × − . A − A LC ) × exp( τ HI − IGM ( LC )) × exp( τ HI − ISM ( LC ))(1)where LC is the wavelength at which the Lyman continuum is being observed ( ∼ f /f LC ) stel is the intrinsic flux ratio from the SED of the stellar population,( A − A LC ) is the differential dust attenuation (in magnitudes), τ HI − IGM ( LC ) is the opticaldepth of the Lyman line and continuum absorption through the IGM along the line-of-sightto that galaxy, and τ HI − ISM ( LC ) is the optical depth of the Lyman continuum absorption 6 –from HI within the observed galaxy’s interstellar medium (ISM).We can rearrange this equation to give us the relative escape fraction f esc,rel = ( f /f LC ) stel ( f /f LC ) obs × exp( τ HI − IGM ( LC )) = exp( − τ HI − ISM ( LC )) × . A − A LC ) . (2)If we can estimate the amplitude of the intrinsic stellar Lyman break, ( f /f LC ) stel , andthe average optical depth of the Ly α forest, τ HI − IGM ( LC ), then then relative escape fractioncan be computed directly from the observed flux ratio, ( f /f LC ) obs . In this section wedetail each of these factors and quantify their effects on the ( f /f LC ) obs ratio. Notethat the relative escape fraction is the product of the HI absorption and the differential reddening between 1500˚A and the Lyman continuum, whereas the absolute escape fractionis the product of the HI absorption and the total dust attenuation of the Lyman continuum.Therefore, the absolute escape fraction is equal to the relative escape fraction times the dustattenuation at 1500 ˚A, f esc = 10 − . A (1500) f esc,rel (3)where A (1500) = 10 . E ( B − V ) for a Calzetti reddening law (Calzetti 1997). The intrinsic flux decrement across the Lyman Break is dependent upon the age, star-formation history (single burst, exponential decay, constant), initial mass-function (IMF),and metallicity of the stellar population. We have used both the Starburst99 (Leitherer et al.1999) and BC03 (Bruzual & Charlot 2003) population synthesis models to quantify how theseproperties affect the size of the Lyman Break in star-forming galaxies. We note that the non-LTE models which better account for stellar winds and line blanketing in the atmospheresof massive stars (Schaerer & Vacca 1998; Smith et al. 2002) do not significantly affect theHI ionizing continuum between 700 < λ <
Since the lifetimes of the O-stars (which emit the Lyman continuum flux) are so muchshorter than the B- and A-stars that dominate the 1500˚A flux emission, the f /f flux 7 –ratio is highly dependent upon the starburst age and the star-formation history. In Figure 1we plot the expected flux ratio as a function of age for both the BC03 and Starburst99 modelswith single burst and constant star-formation histories. In the single burst systems, the dyingO-stars are not replenished with new star-formation, causing the f /f ratio to increaserapidly after a few Myrs. In the constant star-formation scenario, early in the burst, thedying O-stars are being replenished, while the B- and A-stars accumulate, slowly increasingthe f /f ratio. Eventually, the B- and A-stars begin to die as well and no longerincrease in number. Thus, the flux ratio remains constant after ∼
300 Myr. The flux ratio ofa starburst with declining star-formation rates evolves between these two extremes. If theobserved galaxy is undergoing a secondary burst, there may be significant remaining flux at1500 ˚A but only if the bursts are separated by less than a few 100 Myr. Shapley et al. (2001)find that constant star-formation scenarios fit well to the Lyman Break galaxy populationwith a median age, t sf = 320Myr. We will show in Section 5.1 that we obtain similar starformation histories for our sample. Therefore, we expect the typical intrinsic break amplitudeto be ( f /f ) stel ∼ f /f ) stel ∼ The slope of the high-mass end of the stellar initial mass funtion (IMF) determines therelative number of O-stars to B- and A-stars, and therefore impacts the f /f ratio. TheSalpeter IMF (Salpeter 1955), commonly used in studies of high-redshift galaxy formation,has a mass distribution characterized by α = − .
35, where ξ ( M ) ∝ M α . Many studies ofthe high-mass end of the IMF agree with this IMF slope, though with significant scatterbetween star clusters (Massey 1998; Scalo 1998; Kroupa 2001). No direct measurementsof the IMF can be done at high redshift, but initial indications from model fits to theUV spectra of a lensed LBG (MS1512-cB58) indicate that the high-mass IMF is nearlySalpeter and must extend above 50 M ⊙ (Pettini et al. 2000). Given the agreement of detailedlocal studies and initial indications at high-redshift, we use a Salpeter IMF in our analysis.It should be noted, however, that several studies suggest that top-heavy IMFs in somehigh-redshift star-formation may offer a possible explanation of α -element abundances seenin present-day cluster ellipticals (Faber et al. 1992; Worthey et al. 1992; Matteucci 1994;Gibson & Matteucci 1997; Thomas 1999; Nagashima et al. 2005). A top-heavy IMF wouldsignificantly decrease the expected, UV-to-LC ratio expected with a Salpeter IMF. 8 – The metallicities of galaxies such as those in our sample as well as LBGs, range from Z ∼ . − . Z ⊙ . Over this range, the amplitude of the Lyman break does not changesignificantly (see eg. Fig. 75, Leitherer et al. 1999). In our subsequent analysis we use solarmetallicity models. Along the line of sight (LOS) to any high-redshift galaxy, there are hundreds of inter-vening neutral hydrogen clouds whose Lyman line and continuum absorption significantlyaffects the measured far-UV fluxes of the galaxies. The column density and redshift dis-tributions of these HI absorbers have been well studied and have been used to model thetransmission through the IGM as a function of the observed galaxy’s redshift (Madau 1995;Bershady et al. 1999). We make use of the best available data on the column density andredshift distributions of the Ly α Forest (Kim et al. 1997), Lyman Limit Systems (LLSs,Storrie-Lombardi et al. 1994) and Damped Ly α systems (DLAs, Storrie-Lombardi & Wolfe2000) to simulate the effects of IGM Lyman line and continuum absorption on the observedfluxes of galaxies at z ∼
1. We simulate 1000 sight lines to galaxies at redshift intervals∆ z = 0 .
05 between 1 . < z < .
5. For each sight line we randomly place absorbers with col-umn densities and redshifts which are consistent with their empirical distributions. We notethat this does not account for clustering of the absorbers, which may be important, especiallynear the observed galaxy if it is within an overdensity. We then compute the absorption lineprofiles (velocity width β = 30 km s − ) and apply continuum absorption for each absorberto compute the total transmission through the IGM as a function of wavelength (See Sianaet al. 2007 for more details). The mean transmission from 1000 LOSs is plotted in Figure 2.At the redshifts of interest, the average transmission of UV light changes from 0.57 to 0.37between 1 . < z < .
5. Therefore, our measurements are more sensitive at lower redshiftsas less of the Lyman continuum is being absorbed. However the distribution function of theIGM transmission is highly non-Gaussian. Figure 3 shows the distribution of IGM transmis-sion (as measured through our filter), indicating an almost bimodal distribution. About 20%of the simulated lines-of-sight contain large column density absorbers, resulting in very littletransmission ( < z = 1 . Although it is not necessary to know the level of dust attenuation to determine f esc,rel ,it is important for converting between the relative and absolute escape fractions and indetermining whether HI or dust is primarily responsible for the Lyman continuum absorption.If the dust extinction continues to increase at λ < λ ∼
700 ˚A (see Figure 14 in Weingartner & Draine 2001). InFigure 4 we plot the differential reddening across the Lyman Break as a function of opticalcolor excess, E(B-V), for a Calzetti reddening law (Calzetti 1997). The reddening law at λ < dk ( λ ) /dλ , where k ( λ ) = E ( B − V ) /A λ )derived at 1100 < λ < . < E ( B − V ) < . E ( B − V ) ∼ .
15 (Papovich et al. 2001; Shapley et al. 2001). Therefore, thedust alone can cause the UV-to-LC ratio to vary widely over this range (additional factorsof 1-8 at 900 ˚A and 1-20 at 700 ˚A) with typical values of the differential dust attenuation of ∼ ∼ E ( B − V ) > .
5, we cannot detect the Lyman continuum at 700 ˚A in this study.
3. Observations
We have obtained far-UV imaging of the Hubble Deep Field North and the HubbleUltra Deep Field at 1600 ˚A in three HST General Observer programs (7410, 9478, 10403).The HDF-N was observed with STIS through the FUVQTZ filter covering 1.02 arcmin (Gardner et al. 2000). Teplitz et al. (2006) imaged most of the rest of the HDF-N (3.77arcmin . ) with the F150LP filter on the Solar Blind Channel of the ACS. These two instru-ment configurations have very similar throughputs (see Figure 5), with a lower wavelength 10 –cutoff at λ ∼ λ = 1600˚A and λ = 1610 ˚A, respectively.The HUDF observations (Siana et al. 2007, in prep) use the same ACS/SBC configuration asin the HDF-N and cover 7.77 arcmin , nearly the same area as the NICMOS HUDF treasuryprogram (Thompson et al. 2005).Both the SBC and STIS use Multi-Anode Microchannel Arrays (MAMAs), which haveno read noise and are insensitive to cosmic rays. The primary source of noise is dark current,which has two components. The first component is a fairly uniform count rate that doesn’tchange with the detectors’ temperature. The second component is a temperature-dependent“glow” which arises at T >
25 C and is near the center of the SBC MAMA and one corner ofthe STIS MAMA. As the instrument warms up, the dark current from the “glow” increases,decreasing the sensitivity of the observations. Because of this, the images taken at the end ofa series of observations are significantly less sensitive than those at the beginning. Therefore,the dark current changes substantially from pointing to pointing, resulting in significantlydifferent sensitivities across the field. In addition, there is significant area where adjacentframes overlap. Weight maps have been defined which account for the total exposure timeand dark count per pixel so that accurate sensitivities can be computed. The dark subtractionand weight map production used the same procedures as in Teplitz et al. (2006). Figure 6shows the cumulative area for which we expect to obtain a 3 σ detection of a galaxy of agiven magnitude within a 1 ′′ diameter aperture. In order to avoid contamination from photons with λ rest > λ obs > z > .
1) for the STISobservations and λ obs > z > .
2) for the SBC observations. The MAMA detectorsare sensitive to wavelengths longer than λ > − times lower at 2000 ˚A than at 1500˚A and decreases rapidly toward redder wavelengths.Our sensitivity to Lyman continuum radiation falls off rapidly for galaxies with z > . λ < z ≤ .
5. The spectroscopic redshifts areprimarily from the Team Keck Redshift Survey (TKRS Wirth et al. 2004) in the HDF-N and the VLT/FORS2 spectroscopy (Vanzella et al. 2005, 2006) in the HUDF, with a few 11 –additional redshifts from Cowie et al. (2004) and Le F`evre et al. (2004). In addition to theseredshift requirements, we also ensure that there are no sources with optical spectra with high-ionization emission lines (NeV,NeIII) or with strong nuclear point sources indicative of strongAGN activity. Finally, we removed the three reddest ( B − V > .
6) galaxies from our sampleas their SEDs are indicative of older stellar populations with little ongoing star-formation.The 8 galaxies in the HDF-N and 13 galaxies in the HUDF that meet these requirementsare listed in Table 1.Recent measurements have shown that the total optical throughput of the SBC at redwavelengths is much higher than was measured before launch (3 . × − vs. 9 . × − at λ = 3500˚A, STScI Analysis Newsletter, Nov. 30, 2006 ). The total system throughput at3500 ˚A is still four orders of magnitude less than at 1500 ˚A, but we must be careful thatthe ‘leaking’ optical light does not significantly affect our sensitive far-UV measurements.As a test, we examined the effects of the red leak with two types of galaxies: a ‘typical’galaxy ( t sf = 300 Myr, E ( B − V ) = 0 .
2) and a ‘worst-case’ (ie. extremely red) galaxy( t sf = 1 Gyr, E ( B − V ) = 0 . z = 1 .
2, applied averageIGM absorption, and determined the ratio of leaking flux density (with λ rest >
912 ˚A) to themeasured flux density at λ rest > f /f > ∼ . f /f > ∼ .
01, and should thereforebe unaffected by this leaking optical light.
For the optical and near-IR photometry, we used the NICMOS-selected photometriccatalogs of Dickinson et al. (2000) in the HDF-N and Thompson et al. (2005) in the HUDF.The HUDF near-IR photometry was corrected by ∆ J = 0 .
30 and ∆ H = 0 .
18 as discussed inCoe et al. (2006). For both of these catalogs, source detection and isophotal apertures weredetermined from the summed J (F110W) and H (F160W) band NICMOS images. Thesesame apertures were used to extract fluxes from the optical images as well. Each isophotewas inspected by eye to ensure there was no contamination from nearby galaxies and thatno galaxy was improperly fragmented into several pieces.The UV photometry was extracted using isophotes defined by the optical B-band ( λ rest ∼
12 –the reasonable assumption that significant far-UV signal will not be detected where there isno measurable flux in the deeper B-band images. Because the B-band images are so sen-sitive, their faintest isophotes have large areas which contain very little flux. We find thatdecreasing the aperture to exclude the faintest 20% of the galaxy reduces the isophotal areasby factors of ∼ ∼ E ( B − V ) ∼ .
01, Schlegel et al. 1998).As discussed by (Gardner et al. 2000; Brown et al. 2000; Teplitz et al. 2006), the darkcurrent is the principle source of noice in these observations. The individual frames wereweighted by the square of the exposure time and divided by the total dark (primary +“glow”). As the dark count scales with exposure time, these weight maps scale linearly withthe ratio of exposure time to dark rate. Thus, the final weight maps are the square of thesignal-to-noise ratio for objects fainter than the background. For non-detections, we derive3 σ upper limits from these weight maps.
4. Results
We do not detect any of the 21 galaxies in our sample (with
S/N >
3) and the 3 σ upper limits to the far-UV flux are given in Table 1. The distribution of measured SN R (See Figure 7) is centered around zero as expected. There is one galaxy which has a largenegative flux, but that is due to poor background subtraction due to edge effects. The othersources are not affected by this. For the bluer, more luminous galaxies in our sample, weexpect to derive significant limits to the escape fraction of ionizing photons. In Section4.1 we derive f esc,rel limits by reproducing the techniques of previous high-redshift escapefraction studies. Specifically, we calculate flux ratios on either side of the Lyman limit andcompare to the average intrinsic values to deduce a relative escape fraction. In Section 4.2we fit SED models to ascertain dust and age parameters for each galaxy to better determinethe intrinsic Lyman continuum flux and thus the relative escape fraction. To replicate the same methods used in previous high-redshift escape fraction studies,the 3 σ limits to the ( f /f ) obs flux ratio were computed and are listed in Table 1. The 13 – f value is computed by interpolating the optical photometry and the f limit is derivedfrom the weight maps which account for total exposure time and dark current per pixel. Theflux ratios are converted to f esc,rel with Eqn 2. We choose to use ( f /f ) stel = 8 for ourintrinsic stellar flux ratio since the ratio varies from 6-10 between 0 . < t age < . exp ( − τ IGM ( z )) from oursimulations defined in Section 2.2.The distribution of f esc,rel limits are plotted in Figure 8 (top panel). Nineteen of thegalaxies have f esc,rel ≤ .
0, indicating either increased dust attenuation or HI absorption ofthe Lyman continuum.
In addition to deriving flux ratios, we have fit SED models to the high S/N optical/near-IR photometry. In this way, we derive better estimates of the starburst age and dust red-dening for each galaxy, rather than assuming average values for the entire sample. We usedall available optical/near-IR photometry from the deep HST images for the SED fits. Thesedata, in addition to measuring the slope of the UV continuum, span the 4000 ˚A break, al-lowing us to break the degeneracy of age and dust effects on the UV slope. The observedfar-UV (1600 ˚A) limits are not used in the fits of the SEDs.We fit a suite of nine Bruzual-Charlot models with different starburst ages varyinglogarithmically between 0 . < t age <
10 Gyr and allow dust attenuation (Calzetti Law,Calzetti 1997) to vary as a free parameter. The metallicity was fixed at the solar value and theIMF is assumed to be Salpeter (Salpeter 1955). We attempt to fit models with instantaneousstarbursts, exponentially declining star formation with an e-folding time τ = 100 Myr andconstant star-formation models.The results of the fits are given in Table 1 and a few examples are plotted in Figure9. We find that the best fit model is a constant star-formation model for 90% (19 of 21)of the galaxies, confirming that these galaxies, which are bright and blue in the rest-frameultraviolet, are actively undergoing star-formation. The fits to the two other galaxies have χ values that are not significantly better than those of the constant star-formation model.Therefore, we have chosen to list only the contant star-formation fits in Table 1 and use onlythese models in the subsequent analysis.The distribution of relative escape fractions using the ( f /f ) stel from the best-fitmodel is shown in Figure 8 (bottom panel). Eighteen of the galaxies have a relative escapefraction less than unity. We stacked the fluxes at 1500 ˚A and added the 700 ˚A errors in 14 –quadrature to achieve a summed 3 σ limit f esc,rel < .
08 at 700 ˚A, though most of the signalcomes from our brightest source.As a consistency check, we also fit to the Starburst99 (Leitherer et al. 1999) models withconstant star-formation, solar metallicity, and Salpeter IMF. There were small differences inage and dust determinations, but no systematic difference from the results derived with theBruzual & Charlot models.
This is the brightest source ( B = 22 .
41) in our sample and is undetected in the far-UV.However, there is a clear (7 σ ) detection of a compact source ∼ . ′′ to the North (Figure10). This smaller object is part of the larger isophote used in computing the optical/near-IRfluxes but is not in our smaller aperture used for far-UV photometry. It has a flat spectrum(in f ν ) across the UV/optical and a far-UV flux f = 0 . µ Jy, with no break in the SED(
F U V − U (AB)= 0.07). Therefore, it is unlikely to be at the same redshift since there is noindication of IGM absorption. However, it is interesting to note that, if this system was at z ∼ f esc,rel ∼ .
20. Thoughthis is the only object in our sample which appears to have foreground contamination, thecontamination at higher redshifts will be much higher as the foreground path is larger.Therefore, it may be important to obtain high-resolution follow-up of high-redshift escapefraction detections to ensure that the flux is not originating from low-luminosity foregroundobjects.
There is a bright ( B = 21 .
4) QSO at z = 1 .
22 (KX4, Croom et al. 2001) which wasnot included in this analysis but is detected at 4 σ in the far-UV with f = 0 . ± . µ Jy. The optical/near-IR photometry is best fit with power law slope α ν = − .
85 (where f ν ∝ ν α ), consistent with the average near-UV spectral slope of QSOs (Telfer et al. 2002).If we extrapolate this power-law and apply a correction for IGM absorption, we expect tomeasure a far-UV flux f ∼ . µ Jy. Therefore, we estimate an absolute escape fraction f esc ∼ .
02, a very small value for an unobscurred QSO. It is also possible, however, that theescape fraction is high, but that a high column density absorber lies along the line of sight 15 –to the QSO, as is expected for ∼
20% of the LOSs at this redshift (see Figure 3).
5. Discussion5.1. Comparison of SED Fits to Flux Ratios
The best fit ages and color excesses span the allowed ranges, with median values of t = 300 Myr and E ( B − V ) = 0 .
19, similar to the values found in LBGs (Shapley et al. 2001;Papovich et al. 2001). We used the best fit SEDs to compute the ( f /f ) stel flux ratioand compare with the assumed value of 8 used in section 4.1. The intrinsic flux ratio spansthe range of 6.1-11.2 with a median of 9.6, 20% higher than our assumed average value butwithin the range of uncertainty of other parameters. We conclude that simply correcting themeasured flux ratio with an average ( f /f ) stel ∼ −
10 is not unreasonable for galaxieswith ongoing star-formation (eg. LBGs). The distribution of relative escape fractions usingboth methods (constant UV-to-LC ratio and UV-to-LC ratio from the SED fits) are plottedin Figure 8, and show very similar distributions. In the subsequent analysis of individualobjects, we use the values derived from the SED fits.We have computed the UV-to-LC ratio at 900˚A, ( f /f ) stel , from our fits to comparewith the spectroscopic studies, finding a median value of 7.3. This is more than a factor oftwo larger than the value of ∼ ∼ f esc,rel . However, our fits show a factor ∼ A fundamental parameter in escape fraction studies is the relative importance of dustand HI in attenuating the Lyman continuum. If we assume a Calzetti extinction curve and 16 –extrapolate to from 1200˚A to 700 ˚A where the extinciton curve is not empirically determined,we can use our best fit values for dust extinction to determine the level of attenuation at700 ˚A. If our measured flux limit is still lower than the expected flux after dust attenuationthen there must be additional absorption by neutral Hyrdrogen. We can then derive lowerlimits to the HI column density within the ISM of these galaxies. These HI ISM transmissionlimits are given as exp( − τ HI,ISM ) in Table 1. Our large lower limits to the observed f /f ratio (after correcting for the intrinsic break and IGM opacity) can be explained entirely bydifferential dust extinction in all but four of the galaxies. The limits to the HI transmissionfor these four galaxies are 0.45, 0.61, 0.78, 0.88 and are not strong limits to the HI columndensities. Significantly deeper observations are needed to determine whether dust or HI isthe principal cause of Lyman continuum absorption within the galaxies’ ISM. One of our galaxies (J123652.69+621355.3) has a far-UV detected object within 0.3 ′′ of the aperture. These two objects would appear as one if observed with lower spatialresolution. Therefore, it is possible to have a foreground object contaminate the photometry(or spectroscopy) of a high redshift galaxy so that it appears to be emitting in the Lymancontinuum. This is especially true with ground-based studies where the spatial resolutionis low. The only detections of escaping Lyman continuum have been found at z ∼ z ∼ z ∼
3, which increasesthe likelihood of contamination. However, the comoving line-of-sight distance to z ∼ z ∼ < z < z ∼ U -band to determine the likelihood of fore-ground contamination of z ∼ < z <
3, andare emitting at wavelengths that mimic Lyman continuum of LBGs. Because the expected( f /f ) obs is so high, even very faint foreground sources ( U ∼
28) can cause L ∗ LBGs toappear to have large escape fractions. The surface density of objects with U ( AB ) <
28 is ∼ . × deg − (Dolch et al 2007, in prep). If we assume that ground-based imaging orspectroscopy can not resolve objects which lie within a 0.5 ′′ radius of each other, than wewould expect that each z ∼ ∼
2% chance of foreground contamination. Giventhis probability, there is a ∼
22% chance that one galaxy (out of 14) in the Shapley et al. 17 –(2006) study is subjected to foreground contamination, but only a ∼
3% chance that theboth detections are contaminated. Therefore, it is unlikely that foreground contaminationcan entirely explain the large escape fractions at z ∼
3. However, it is important to keep inmind that only a small percentage (maybe ∼ − ∼ − z ∼ Studies
Malkan et al. (2003, herafter M03) conducted a similar study to ours, observing 11luminous starbursts at 1 . < z < . λ eff = 1453˚A, ∼
150 ˚A shorterthan our observations. Accounting for the mean redshifts of the surveys, this correspondsto λ rest ∼
660 ˚A for M03 and λ rest ∼
715 ˚A for our study. Therefore, the flux ratios of M03are subject to larger effects from dust attenuation and star formation history than thosemeasured here. Although the STIS camera is less sensitive than the ACS/SBC, the broaderfilter increases the sensitivity so that the M03 f limits are close to our derived limits. Inaddition, they targeted more luminous starbursts, resulting in limits to the f esc,rel that are ∼ σ limits and are an additionalfactor of two too low, as the radius of the aperture used for the limits was smaller, by afactor of two, than the intended 0.5 ′′ aperture radius (M. Malkan, private communication).After applying corrections for IGM absorption (which M03 did not do), we derive limits of0 . < f esc,rel (3 σ ) < . f /f ) stel is appropriate for their sample. The STIS field-of-view is small so there are no objects (target galaxy or otherwise) detected in 9 of 11pointings. The far-UV images can not be astrometrically aligned with optical data, leavingtheir measurements subject to the intrinsic pointing uncertainties of HST. Because thesepointing uncertainties are larger than their aperture radii (1 ′′ vs. 0 . ′′ ), the stacked far-UVimages of the 11 galaxies spread the light over larger areas and do not give significantlybetter limits than the individual images.Combining with the M03 sample, we now have 28 galaxies with f esc,rel ≤ . f esc,rel < . z ∼
1. The M03 sample have UV luminosisites similar to L ∗ LBGs at z ∼
3, whereas our sample is somewhat fainter (0.1-1.0 L ∗ LBG ). These galaxies span 18 –nearly two orders of magnitude in luminosity and have a broad range in morphologies andstarburst ages, yet we see no evidence for large escape fractions at this redshift.As shown in Section 2.2, the transmission through the IGM can vary substantiallyalong different lines-of-sight. This will affect our results such that ∼
20% of our limits areeffectively meaningless as they are looking at opaque lines of sight. However, ∼
60% of ourlimits will be substantially better than the limits given in Table 1 as these lines-of-sight aremore transparent than average. To better evaluate the effects of this distribution on oursurvey, we’ve performed a Monte-Carlo simulation where we observe 32 galaxies (21 fromthis sample and 11 from M03) with the same distribution of ( f /f ) obs limits as in oursurveys. The IGM transmission is chosen randomly from the distribution plotted in Figure3. We assume an intrinsic break of ( f /f ) stel = 8 for all galaxies. We then need toassume an escape fraction for each galaxy. For this we assume that a fraction Y of star-forming galaxies at our redshifts have an f esc,rel = X , and all other galaxies have effectivelyno Lyman continuum transmission. This analysis assumes that this parameter space (X,Y)is the same for galaxies with different luminosities. We allow X and Y to vary between 0.0and 1.0 and run 10,000 iterations of our observations for each parameter set (X,Y). We thendetermine, for each parameter set, what percent of the time we achieve a null result for all32 galaxies. The results are plotted in Figure 11. The shaded regions denote the parameterspace which is excluded by our combined samples at 68, 95, and 99% confidence (from lighterto darker). If all star-forming galaxies at z ∼ . f esc,rel < .
14 at 99% confidence. Conversely, ifsome galaxies have large relative escape fractions, f esc,rel ≥ .
75, while the others have none,they must be less than 20% of the total population (at 99% confidence).
In Table 2, we summarize the escape fraction studies at all redshifts by converting the re-sults to common definitions: the UV-to-LC ratio corrected for IGM absorption, ( f /f ) corr ,and the relative escape fraction.Because observations at λ obs ∼ E ( B − V ) > .
5. If the Lyman continuum is subject tothe same dust extinction, then it would not be detectable in these measurements. The most 19 –sensitive limits at low redshift come from FUSE spectra of Mrk 54 (Deharveng et al. 2001)and Haro 11 (Grimes et al. 2007, submitted) which give UV-to-LC limits of 112 and 21respectively. Note that Bergvall et al. (2006) originally reported a strong Lyman continuumdetection, but a recent reanalysis by Grimes et al. (2007, submitted) detect no flux bluewardof the Lyman limit.The z ∼ z ∼ σ ) detection of the Lyman continuumat λ rest = 880 −
910 ˚A in a composite spectrum of 29 z ∼ . f /f ) corr = 4 . U -dropout sample, these 29 LBGs were among the bluest quartileof the entire LBG sample. Therefore, it might be expected that differential reddening is nota major issue, and the intrinsic Lyman break amplitude is somewhat smaller (due to youngerages). Regardless, as Steidel et al. (2001) point out, this is a surprising result which mustbe verified with deeper spectra.Shapley et al. (2006) published much deeper individual spectra of 14 LBGs, of whichtwo showed significant flux below the Lyman limit. These UV-to-LC ratios (2.9, 4.5) alsoimply relative escape fractions near unity, while the spectra of the remaining 12 give a 3 σ limits from f esc,rel < . − .
0, depending on depth (when using a Lyman break amplitudeof 6). These limits still allow for non-zero escape fractions, but do not display the smallUV-to-LC ratios exhibited in the two detection or in the stack of Steidel et al. (2001).Several other groups have published escape fraction limits of LBGs at z ∼
3, findingno detections. Once converting the limits to 3 σ and using a more conservative Lymanbreak amplitude, ( f /f ) stel = 6, the data from Giallongo et al. (2002) and Inoue et al.(2005) do not significantly constrain the relative escape fraction (ie. limits of f esc,rel > ∼
1, SeeTable 2). Fern´andez-Soto et al. (2003) have analyzed broadband photometry of 27 galaxiesbetween 1 . < z < . σ limit of f esc < . f /f ) stel = 8 (the value listed in Table 2 accounts for this). Secondly, thefilter used to measure the escaping Lyman continuum contains significant flux from redwardof the Lyman break, necessitating very accurate determinations of the UV continuum fluxin order to estimate the level of Lyman continuum in the same filter. Once these effectsare accounted for, their stacked limit is above the average detection of S06. In summary,Shapley et al. (2006) is the only large study at z ∼ z ∼ z ∼ detections at z ∼ σ , with our results, implyingthat there may be an evolution in the relative escape fraction with redshift. However, manyof the galaxies in our combined sample are somewhat less luminous than the z ∼ L UV = 0 . − . L ∗ UV for z ∼ z ∼ z ∼ f /f ) stel ratio. If there has beenany significant decrease in star-formation within t <
10 Myr, f will be significantly lowerthan f , while having little significant effect on the broadband SED redward of the LymanBreak. Therefore, it is possible that some of our galaxies will have a larger ( f /f ) stel ,thus weakening our limits on f esc,rel . 21 –
6. Summary
We have examined deep far-ultraviolet (1600˚A) imaging of the HDF-North and HUDFto search for escaping Lyman continuum flux from 21 star-forming galaxies at 1 . < z < . ∼
50% scatter in the intrinsic stellar Lyman Break due to the starburst age, it is reason-able to assume a constant ( f /f ) stel ∼ f /f ) stel ∼ f /f ) stel = 3assumed in many previous studies. Assuming a reasonable extrapolation of the extinctioncurve below the Lyman limit, we show that the observed flux decrement at 700 ˚A can beattributed to dust attenuation and does not require large column densities of HI within theISM of the galaxies. Deeper observations are required to determine the relative importanceof dust and HI to the escape fraction.We obtain 3 σ limits better than f esc,rel < . f esc,rel < . f esc,rel < .
10. Our stacked fluxes give a combined limitof f esc,rel < .
08, similar to the sensitivity achieved by Malkan et al. (2003, corrected withcommon assumptions with this study) for more luminous starbursts at the same redshift.This is the first study to achieve these sensitivities on high redshift starbursts which are lessluminous than typical L ∗ LBG . These stacks give the deepest escape fraction limits achieved atany redshift and demonstrate a paucity of ionizing emissivity in most starbursts at z ∼ . z ∼ . f esc,rel < f esc,rel < . z ∼ f esc,rel < . z ∼ ∼ − z ∼ z ∼ z ∼ σ ),suggesting a possible decrease in the escape fraction with redshift. We cannot yet rule outthe possibility that the discrepancy is the result of differeint observing methods which probe 22 –different regions of the Lyman continuum (700 ˚A vs. 900 ˚A).Further investigation of the evolution of the escape fraction requires larger samples of z ∼ z ∼ z ∼ Facilities:
Hubble(ACS)
REFERENCES
Arnouts, S., Schiminovich, D., Ilbert, O., Tresse, L., Milliard, B., Treyer, M., Bardelli,S., Budavari, T., Wyder, T. K., Zucca, E., Le F`evre, O., Martin, D. C., Vettolani,G., Adami, C., Arnaboldi, M., Barlow, T., Bianchi, L., Bolzonella, M., Bottini, D.,Byun, Y.-I., Cappi, A., Charlot, S., Contini, T., Donas, J., Forster, K., Foucaud,S., Franzetti, P., Friedman, P. G., Garilli, B., Gavignaud, I., Guzzo, L., Heckman,T. M., Hoopes, C., Iovino, A., Jelinsky, P., Le Brun, V., Lee, Y.-W., Maccagni, D.,Madore, B. F., Malina, R., Marano, B., Marinoni, C., McCracken, H. J., Mazure, A.,Meneux, B., Merighi, R., Morrissey, P., Neff, S., Paltani, S., Pell`o, R., Picat, J. P.,Pollo, A., Pozzetti, L., Radovich, M., Rich, R. M., Scaramella, R., Scodeggio, M.,Seibert, M., Siegmund, O., Small, T., Szalay, A. S., Welsh, B., Xu, C. K., Zamorani,G., & Zanichelli, A. 2005, ApJ, 619, L43Beckwith, S. V. W., Stiavelli, M., Koekemoer, A. M., Caldwell, J. A. R., Ferguson, H. C.,Hook, R., Lucas, R. A., Bergeron, L. E., Corbin, M., Jogee, S., Panagia, N., Robberto,M., Royle, P., Somerville, R. S., & Sosey, M. 2006, AJ, 132, 1729 23 –Bergvall, N., Zackrisson, E., Andersson, B.-G., Arnberg, D., Masegosa, J., & ¨Ostlin, G. 2006,A&A, 448, 513Bershady, M. A., Charlton, J. C., & Geoffroy, J. M. 1999, ApJ, 518, 103Bolton, J. S., Haehnelt, M. G., Viel, M., & Springel, V. 2005, MNRAS, 357, 1178Brown, T. M., Kimble, R. A., Ferguson, H. C., Gardner, J. P., Collins, N. R., & Hill, R. S.2000, AJ, 120, 1153Bruzual, G. & Charlot, S. 2003, MNRAS, 344, 1000Calzetti, D. 1997, AJ, 113, 162Coe, D., Ben´ıtez, N., S´anchez, S. F., Jee, M., Bouwens, R., & Ford, H. 2006, AJ, 132, 926Cowie, L. L., Barger, A. J., Hu, E. M., Capak, P., & Songaila, A. 2004, AJ, 127, 3137Croom, S. M., Warren, S. J., & Glazebrook, K. 2001, MNRAS, 328, 150Deharveng, J.-M., Buat, V., Le Brun, V., Milliard, B., Kunth, D., Shull, J. M., & Gry, C.2001, A&A, 375, 805Dickinson, M., Hanley, C., Elston, R., Eisenhardt, P. R., Stanford, S. A., Adelberger, K. L.,Shapley, A., Steidel, C. C., Papovich, C., Szalay, A. S., Bershady, M. A., Conselice,C. J., Ferguson, H. C., & Fruchter, A. S. 2000, ApJ, 531, 624Faber, S. M., Worthey, G., & Gonzales, J. J. 1992, in IAU Symposium, Vol. 149, The StellarPopulations of Galaxies, ed. B. Barbuy & A. Renzini, 255–+Fan, X., Strauss, M. A., Becker, R. H., White, R. L., Gunn, J. E., Knapp, G. R., Richards,G. T., Schneider, D. P., Brinkmann, J., & Fukugita, M. 2006, AJ, 132, 117Fern´andez-Soto, A., Lanzetta, K. M., & Chen, H.-W. 2003, MNRAS, 342, 1215Fontanot, F., Cristiani, S., Monaco, P., Nonino, M., Vanzella, E., Brandt, W. N., Grazian,A., & Mao, J. 2007, A&A, 461, 39Fujita, A., Martin, C. L., Mac Low, M.-M., & Abel, T. 2003, ApJ, 599, 50Gardner, J. P., Brown, T. M., & Ferguson, H. C. 2000, ApJ, 542, L79Giallongo, E., Cristiani, S., D’Odorico, S., & Fontana, A. 2002, ApJ, 568, L9Gibson, B. K. & Matteucci, F. 1997, MNRAS, 291, L8 24 –Hoopes, C. G., Heckman, T. M., Salim, S., Seibert, M., Tremonti, C. A., Schiminovich,D., Rich, R. M., Martin, D. C., Charlot, S., Kauffmann, G., Forster, K., Friedman,P. G., Morrissey, P., Neff, S. G., Small, T., Wyder, T. K., Bianchi, L., Donas, J., Lee,Y.-W., Madore, B. F., Milliard, B., Szalay, A. S., Welsh, B. Y., & Yi, S. K. 2006,ArXiv Astrophysics e-printsHunt, M. P., Steidel, C. C., Adelberger, K. L., & Shapley, A. E. 2004, ApJ, 605, 625Hurwitz, M., Jelinsky, P., & Dixon, W. V. D. 1997, ApJ, 481, L31+Inoue, A. K., Iwata, I., & Deharveng, J.-M. 2006, MNRAS, 371, L1Inoue, A. K., Iwata, I., Deharveng, J.-M., Buat, V., & Burgarella, D. 2005, A&A, 435, 471Kim, T.-S., Hu, E. M., Cowie, L. L., & Songaila, A. 1997, AJ, 114, 1Kroupa, P. 2001, MNRAS, 322, 231Le F`evre, O., Vettolani, G., Paltani, S., Tresse, L., Zamorani, G., Le Brun, V., Moreau, C.,Bottini, D., Maccagni, D., Picat, J. P., Scaramella, R., Scodeggio, M., Zanichelli, A.,Adami, C., Arnouts, S., Bardelli, S., Bolzonella, M., Cappi, A., Charlot, S., Contini,T., Foucaud, S., Franzetti, P., Garilli, B., Gavignaud, I., Guzzo, L., Ilbert, O., Iovino,A., McCracken, H. J., Mancini, D., Marano, B., Marinoni, C., Mathez, G., Mazure,A., Meneux, B., Merighi, R., Pell`o, R., Pollo, A., Pozzetti, L., Radovich, M., Zucca,E., Arnaboldi, M., Bondi, M., Bongiorno, A., Busarello, G., Ciliegi, P., Gregorini, L.,Mellier, Y., Merluzzi, P., Ripepi, V., & Rizzo, D. 2004, A&A, 428, 1043Leitherer, C., Ferguson, H. C., Heckman, T. M., & Lowenthal, J. D. 1995, ApJ, 454, L19+Leitherer, C., Schaerer, D., Goldader, J. D., Delgado, R. M. G., Robert, C., Kune, D. F., deMello, D. F., Devost, D., & Heckman, T. M. 1999, ApJS, 123, 3Madau, P. 1995, ApJ, 441, 18Malkan, M., Webb, W., & Konopacky, Q. 2003, ApJ, 598, 878Massey, P. 1998, in Astronomical Society of the Pacific Conference Series, Vol. 142, TheStellar Initial Mass Function (38th Herstmonceux Conference), ed. G. Gilmore &D. Howell, 17–+Matteucci, F. 1994, A&A, 288, 57McDonald, P. & Miralda-Escud´e, J. 2001, ApJ, 549, L11 25 –Nagashima, M., Lacey, C. G., Okamoto, T., Baugh, C. M., Frenk, C. S., & Cole, S. 2005,MNRAS, 363, L31Papovich, C., Dickinson, M., & Ferguson, H. C. 2001, ApJ, 559, 620Pettini, M., Steidel, C. C., Adelberger, K. L., Dickinson, M., & Giavalisco, M. 2000, ApJ,528, 96Richards, G. T., Croom, S. M., Anderson, S. F., Bland-Hawthorn, J., Boyle, B. J., DePropris, R., Drinkwater, M. J., Fan, X., Gunn, J. E., Ivezi´c, ˇZ., Jester, S., Loveday,J., Meiksin, A., Miller, L., Myers, A., Nichol, R. C., Outram, P. J., Pimbblet, K. A.,Roseboom, I. G., Ross, N., Schneider, D. P., Shanks, T., Sharp, R. G., Stoughton,C., Strauss, M. A., Szalay, A. S., Vanden Berk, D. E., & York, D. G. 2005, MNRAS,360, 839Salpeter, E. E. 1955, ApJ, 121, 161Scalo, J. 1998, in Astronomical Society of the Pacific Conference Series, Vol. 142, The StellarInitial Mass Function (38th Herstmonceux Conference), ed. G. Gilmore & D. Howell,201–+Schaerer, D. & Vacca, W. D. 1998, ApJ, 497, 618Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525Scott, J., Bechtold, J., Morita, M., Dobrzycki, A., & Kulkarni, V. P. 2002, ApJ, 571, 665Shapley, A. E., Steidel, C. C., Adelberger, K. L., Dickinson, M., Giavalisco, M., & Pettini,M. 2001, ApJ, 562, 95Shapley, A. E., Steidel, C. C., Pettini, M., Adelberger, K. L., & Erb, D. K. 2006, ApJ, 651,688Siana, B., Polletta, M., Smith, H. E., Lonsdale, C. J., Gonzalez-Solares, E., Farrah, D., &Babbedge, T. S. R. 2007, ApJSmith, L. J., Norris, R. P. F., & Crowther, P. A. 2002, MNRAS, 337, 1309Steidel, C. C., Pettini, M., & Adelberger, K. L. 2001, ApJ, 546, 665Storrie-Lombardi, L. J., McMahon, R. G., Irwin, M. J., & Hazard, C. 1994, ApJ, 427, L13Storrie-Lombardi, L. J. & Wolfe, A. M. 2000, ApJ, 543, 552 26 –Telfer, R. C., Zheng, W., Kriss, G. A., & Davidsen, A. F. 2002, ApJ, 565, 773Teplitz, H. I., Siana, B., Brown, T. M., Chary, R., Colbert, J. W., Conselice, C. J., de Mello,D. F., Dickinson, M., Ferguson, H. C., Gardner, J. P., & Menanteau, F. 2006, AJ,132, 853Thomas, D. 1999, MNRAS, 306, 655Thompson, R. I., Illingworth, G., Bouwens, R., Dickinson, M., Eisenstein, D., Fan, X.,Franx, M., Riess, A., Rieke, M. J., Schneider, G., Stobie, E., Toft, S., & van Dokkum,P. 2005, AJ, 130, 1Vanzella, E., Cristiani, S., Dickinson, M., Kuntschner, H., Moustakas, L. A., Nonino, M.,Rosati, P., Stern, D., Cesarsky, C., Ettori, S., Ferguson, H. C., Fosbury, R. A. E.,Giavalisco, M., Haase, J., Renzini, A., Rettura, A., Serra, P., & The Goods Team.2005, A&A, 434, 53Vanzella, E., Cristiani, S., Dickinson, M., Kuntschner, H., Nonino, M., Rettura, A., Rosati,P., Vernet, J., Cesarsky, C., Ferguson, H. C., Fosbury, R. A. E., Giavalisco, M.,Grazian, A., Haase, J., Moustakas, L. A., Popesso, P., Renzini, A., Stern, D., & TheGoods Team. 2006, A&A, 454, 423Weingartner, J. C. & Draine, B. T. 2001, ApJ, 548, 296Williams, R. E., Blacker, B., Dickinson, M., Dixon, W. V. D., Ferguson, H. C., Fruchter,A. S., Giavalisco, M., Gilliland, R. L., Heyer, I., Katsanis, R., Levay, Z., Lucas, R. A.,McElroy, D. B., Petro, L., Postman, M., Adorf, H.-M., & Hook, R. 1996, AJ, 112,1335Wirth, G. D., Willmer, C. N. A., Amico, P., Chaffee, F. H., Goodrich, R. W., Kwok, S.,Lyke, J. E., Mader, J. A., Tran, H. D., Barger, A. J., Cowie, L. L., Capak, P., Coil,A. L., Cooper, M. C., Conrad, A., Davis, M., Faber, S. M., Hu, E. M., Koo, D. C.,Le Mignant, D., Newman, J. A., & Songaila, A. 2004, AJ, 127, 3121Worthey, G., Faber, S. M., & Gonzalez, J. J. 1992, ApJ, 398, 69
This preprint was prepared with the AAS L A TEX macros v5.2. able 1. Far-UV limits and best-fit SED parameters.
Name z spec fuv a Age E(B-V) f ( f /f ) int b ( f /f ) int b ( f /f ) obs a exp(- τ IGM ) exp(- τ HI ) f esc,rel [ µ Jy] [Gyr] [ µ Jy]J033233.46-274712.4 1.298 0.007 0.300 0.44 0.10 9.94 7.29 14.76 0.51 26.73 1.32J033234.66-274728.0 1.438 0.007 0.300 0.10 0.69 11.21 7.29 93.06 0.41 0.61 0.29J033234.83-274722.1 1.316 0.017 0.300 0.25 0.61 10.02 7.29 35.54 0.50 3.04 0.57J033235.80-274734.8 1.223 0.011 3.000 0.44 0.09 9.47 7.35 8.67 0.55 34.87 1.97J033236.56-274640.6 1.414 0.007 0.300 0.15 0.26 10.80 7.29 36.35 0.43 1.98 0.70J033236.90-274726.2 1.318 0.017 1.000 0.20 0.53 10.10 7.35 31.13 0.49 2.50 0.66J033237.07-274617.3 1.273 0.023 0.030 0.35 0.43 7.62 5.64 18.47 0.52 7.90 0.79J033237.73-274642.7 1.307 0.009 0.300 0.15 0.42 9.98 7.29 44.85 0.50 1.21 0.44J033238.24-274630.1 1.216 0.008 0.300 0.40 0.14 9.56 7.29 17.32 0.56 12.69 0.99J033239.92-274606.9 1.295 0.027 0.300 0.35 0.56 9.94 7.29 20.39 0.51 9.90 0.95J033240.93-274823.6 1.244 0.008 0.300 0.15 0.42 9.61 7.29 54.25 0.54 0.88 0.33J033241.32-274821.1 1.318 0.042 0.300 0.15 0.85 10.03 7.29 20.28 0.49 2.72 1.00J033244.16-274729.5 1.220 0.010 3.000 0.25 0.28 9.65 7.35 29.60 0.56 2.91 0.59J123643.41+621151.6 1.241 0.012 0.300 0.15 1.28 9.58 7.29 106.89 0.53 0.45 0.17J123647.18+621342.0 1.314 0.009 0.100 0.35 0.40 9.33 6.80 43.17 0.50 4.54 0.43J123649.44+621316.6 1.238 0.008 0.300 0.44 0.33 9.56 7.29 42.88 0.54 7.61 0.41J123649.95+621225.5 1.204 0.006 1.000 0.20 0.24 9.58 7.35 39.92 0.57 1.52 0.42J123652.69+621355.3 1.355 0.024 0.010 0.30 3.20 6.11 4.38 132.09 0.47 0.78 0.10J123656.13+621329.7 1.242 0.018 1.000 0.15 0.66 9.66 7.35 37.54 0.54 1.28 0.48J123656.60+621252.7 1.233 0.013 1.000 0.10 0.38 9.58 7.35 28.06 0.55 1.22 0.62J123656.73+621252.6 1.231 0.010 3.000 0.44 0.12 9.55 7.35 12.33 0.55 26.01 1.41 a σ upper limits. b Derived from SED fits to the optical/near-IR data. able 2. Compilation of flux ratio limits (corrected for average IGM absorption) andrelative escape fractions. The IGM absorption is redshift dependent so, in order tofacilitate comparison between surveys at different redshifts, we have accounted for this bymultiplying the flux ratios by the average IGM transmission at the corresponding redshift.Limits are in parentheses and have been converted to 3 σ . The z ∼ f = 1 . × f was used based on the constant star-formation SEDs. A( f /f ) stel = 6 or ( f /f ) stel = 8 ratio has been assumed to convert ( f /f ) obs or( f /f ) obs to f esc,rel . The Fern´andez-Soto et al. (2003) limits have been multiplied byeight to account for the intrinsic Lyman break of the stellar population.Redshift Sample ( f /f ) corr f esc,rel z ∼ z ∼ . z ∼ < . > ∼ · · · (0.32)Inoue et al. (2005) (2.6,4.0) (2.3,1.4)Shapley et al. (2006) detections 2.9,4.5 ∼ < > f /f and f /f intrinsic flux ratios as a function of time since onsetof star-formation using the BC03 (solid) and Starburst99 (dashed) models . 30 –Fig. 2.— The average transmission through the IGM of UV light from galaxies at redshift 1.2,1.3, 1.4, and 1.5 (left to right). Each curve is computed from simulations along 1000 lines ofsight through the Ly α forest. The corresponding Lyman limits are denoted by solid verticallines. As a reference, the transmission curve is plotted for the ACS/SBC F150LP filter usedin our observations (dot-dashed line). The vertical dotted line is the pivot wavelength of thefilter curve. 31 –Fig. 3.— The distribution of IGM transmission within the F150LP filter for 1000 lines ofsight towards galaxies at z = 1 .
3. The spike at zero transmission is due to Lyman LimitSystems (LLSs) and Damped Ly α systems (DLAs) within ∆ z ∼ . − . A (1500) − A ( LC )) where LC is the either 900 or 700 ˚A) as a function of dust reddeningfor a Calzetti reddening law extrapolated to λ = 700 ˚A. The dotted vertical line shows themedian value of E ( B − V ) = 0 .
155 for Shapley et al. (2001) for Lyman Break Galaxies derivedfrom broadband photometry. The tail of the reddening distrubution for LBGs extends outto E ( B − V ) = 0 .
4. The 700 ˚A flux is affected far more than at 900 ˚A showing that ourmeasurements are more sensitive to dust than those measurements just blueward of theLyman limit. 33 –Fig. 5.— Total system throughput for the ACS+SBC+F150LP (solid line) and STIS+FUV-MAMA+F25QTZ (dashed) configurations. The ACS/SBC is nearly 3x more sensitive thanthe STIS configuration. The dashed vertical lines denote the location of the Lyman Limitat z=1.1 and z=1.2, the low redshift cutoffs for our sample selection for each configuration. 34 –Fig. 6.— Detectable area as a function of limiting magnitude (AB, 3 σ ) within a 1 ′′ diametercircular aperture. The solid line is the total area from all three surveys. The dashed, dotted,and dash-dotted lines are the areas of the HUDF, HDFN-STIS, and HDFN-SBC surveysrespectively. The HUDF survey is much larger, allowing us to detect more objects, whilethe HDF-SBC survey is deeper. We are are sensitive to fainter galaxies if their extractionisophotes are smaller. 35 –Fig. 7.— Histogram of measured signal-to-noise ratios for the 29 galaxies in our sample.There are no detections above SN R >
3. The dashed vertical line is the average < SN R > = − .
40. The standard deviation of this distribution, σ = 1 .
7, is slightly larger than expected.This is due to small errors in “background” subtraction of a few sources (namely the objectat
SN R = −
5) since the dark current is non-planar. 36 –Fig. 8.— Histogram of the f esc,rel limits of our sample. The top panel limits are derivedassuming a UV-to-LC ratio ( f /f ) stel = 8, whereas the lower panel show the limitsderived when taking the UV-to-LC ratio from the best fit model. Limits of f esc,rel < . f esc,rel ≤ . f esc,rel < . σ limits are plotted with downward arrows. Note that our ability to detect leaking Lymancontinuum is largely dependent upon the level of dust attenuation. 38 –Fig. 10.— The HST/WFPC2 B-band (f450, left) and far-UV (F150LP, right) images ofJ123652.69+621355.3. The orientation and alignment is the same in both images and thepointers are in the same location on the sky. The pointers are each 1 ′′ in length and pointto the faint source to the North of the target galaxy. The source is clearly detected in thefar-UV image. 39 –Fig. 11.— The parameter space excluded by a Monte-Carlo analysis of the combined limits ofthis work with Malkan et al. (2003). The x-axis is the relative escape fraction and the y-axisis the fraction of galaxies which have this escape fraction. The other galaxies are assumedto have negligible escape fractions. The shaded regions are excluded(at 3, 2, 1 σσ