TWO LENSED LYMAN-α EMITTING GALAXIES AT z∼ 5^*

The two sources presented here were first identified as r-band dropouts in gri imaging of two different strong lensing galaxy clusters and subsequently confirmed by spectroscopy to have strong Lyman-α emission features. The two galaxy clusters were identified as part of the Sloan Giant Arcs Survey (SGAS; Hennawi et al. 2008), a blind survey for strong lensing systems in optically selected massive clusters at 0.1 ⩽ z ⩽ 0.6 detected via the Red-Sequence Cluster algorithm (Gladders & Yee 2000) adapted to run on the Sloan Digital Sky Survey (SDSS; York et al. 2000) public data release catalogs. Strong lensing clusters are identified by visual inspection of imaging in g band on 2 m to 4 m-class telescopes for 600 s in <1'' seeing, and the most spectacular systems have been followed up with multi-band imaging and spectroscopy on 8 m-class telescopes. One of the clusters discussed here, SDSS J1343+4155, appears in a recent small sample of strong lensing clusters discovered in the SDSS by Diehl et al. (2009). The other cluster, SDSS J0915+3826, does not appear in any prior published work. Imaging and spectroscopic observations were conducted with the Frederick C. Gillett Telescope (Gemini North) between the months of 2008 February and 2008 July. The GMOS imaging observations were pre-imaging conducted for the purpose of spectroscopic mask design. Both imaging and spectroscopy were part of Gemini program GN-2008A-Q-25. The primary goal of the spectroscopic observations was to obtain redshifts of arcs to facilitate strong lensing modeling. We briefly summarize the spectroscopy here and refer the reader to M. Bayliss et al. (2010, in preparation), for additional details.

Gemini North/GMOS gri photometry for both cluster fields are derived from 2 × 150 s dithered exposures, which were executed in queue mode in February and March of 2008. These pre-imaging data were used for spectroscopic mask design and target prioritization. GMOS z-band observations consisted of 6 × 180 s dithered exposures which were scheduled as follow-up, primarily in order to measure the continuum flux of the LAEs, and were executed in queue mode in February of 2010. The GMOS images were reduced using the Gemini IRAF⁵ package.

In addition to the GMOS-N photometry and spectroscopy, we obtained near-infrared (NIR) imaging of the two lensing clusters in the z'JH filters with the Near-Infrared Camera and Fabry-Perot Spectrometer (NIC-FPS) of the 3.5 m telescope at the Apache Point Observatory (APO) in New Mexico. The detector is a Rockwell Hawaii 1-RG 1024 × 1024 HgCdTe device with a pixel scale of 0273 pixel⁻¹ and a 4.58 × 4.58 arcmin unvignetted field of view. The APO/NIC-FPS z' data differ from the Gemini/GMOS z significantly due to the different wavelength responses of the NIC-FPS and GMOS detectors. The effective wavelengths of the two filters are offset by ∼1000 Åand we use the prime (') throughout this paper to distinguish the APO/NIC-FPS z' from the Gemini/GMOS z. The NIR data were taken on three different nights in the Winter and Spring of 2009. The conditions during the observing nights varied, with subarcsecond seeing for the SDSS J1343+4155 z'-band and both J-band images. The SDSS J0915+3826 z'-band and both H-band images were taken in ∼2'' seeing. The observations consisted of five-point dithers around a 40'' box and were reduced, registered, and stacked using a custom IRAF pipeline. Total exposure times are 5400 s, 6960 s, and 3500 s for z'−, J−, and H-band observations of SDSS J0915+3826 and 7800 s, 5160 s, and 2625 s for z'−, J−, and H-band observations of SDSS J1343+4155.

The SDSS was used to calibrate the four optical bands and the Two Micron All Sky Survey was used to calibrate the NIR observations. Prior to making any photometric measurements, we first transform images in all bands to the reference frame of the i-band image, 01454 pixel⁻¹ (GMOS-North detector, binned 2 × 2). We then construct an empirical, normalized point-spread function (PSF) for each image based on a well-defined, non-saturated reference star. We create photometric apertures by drawing a ridge line that covers the high-redshift LAE and convolving it with the appropriate PSF for each image. Apertures are defined by an equivalent radius, which corresponds to the radius of a circular aperture that goes out to the same isophot. We make the final magnitude measurement using a detailed sequence of sky subtraction and outlier masking steps; first we subtract a general sky measurement and then compute the median pixel value and standard deviation inside annuli of fixed width at increasing radial distances from the source. Far enough out, these median values converge to zero for an accurate sky subtraction. We average the median values at large radii and subtract this average from the image to correct the general sky subtraction. Outliers are defined as pixel values that deviate by more than 5σ from the median in the respective annuli and are replaced by the median value plus an appropriate noise term. The final magnitude is measured at an equivalent radius of twice the FWHM of the PSF and corrected to an equivalent radius of 6'' based on the curve of growth of the PSF. By defining the radius as twice the FWHM, we make sure that our apertures always cover the same physical region on the sky. In the case of SGAS J091541+382655, a foreground galaxy lies very close to and partially on top of the LAE galaxy. We use the GALFIT package (Peng et al. 2002) to fit a Sersic profile to this galaxy in the imaging data where the LAE has no measurable flux and then scale the galaxy model by the peak flux in each band and subtract it out before measuring the LAE magnitudes.

When the LAEs are not detected (in wgJH for SGAS J091541+382655 and grJH for SGAS J134331+415455), we measure a limiting magnitude, derived from the total sky noise in the aperture. We define a sky noise per pixel σ_sky as the standard deviation of our overall sky measurement plus a contribution from the poison error made in this measurement. A pixel value of 2σ_sky is added in quadrature for each pixel in our final aperture with an equivalent radius equal to twice the FWHM. The limiting magnitudes are also aperture corrected to an equivalent radius of 6''. This method constrains the source to be fainter than the limiting magnitudes at 95% confidence in the relevant bandpasses.

All spectroscopic observations were carried out using the Gemini Multi-Object Spectrograph (Hook et al. 2004) using custom slitmasks that were designed to target lensed sources based on their color, morphology, and position in the GMOS imaging. After targeting all of the arc candidates, any remaining slits were placed on cluster members, easily identified by their red sequence colors. Spectra were taken using the macroscopic nod-and-shuffle (N&S) mode available on GMOS. The reasons for using N&S are threefold. First, many of the emission or absorption features that are used to determine galaxy redshifts in the range z = 1.0–3.0 characteristic of the giant arcs are in the redder part of the optical where sky lines are problematic, and N&S facilitates more accurate sky-subtraction (Glazebrook & Bland-Hawthorn 2001), particularly at low spectral resolution. Second, N&S sky-subtraction allows us to use very small 1'' × 1'' microslits that can be densely packed into the cluster core (∼30''), allowing us to target as many arcs, arclets, and cluster galaxies as possible (Gilbank et al. 2008). Third, as we are primarily interested in a number of objects around the cluster center, the density of objects and the limited field of interest are perfect for block-shuffling N&S observations. Spectra were taken with the R150_G5306 grating in first order which gives a dispersion of 3.5 Å per pixel (binned spectrally by two), with six pixels per resolution element resulting in a spectral FWHM ≃940 km s⁻¹. Although the R150 grating offers broad spectral range from the atmospheric cutoff to λ ≳ 1μm, the drop in sensitivity at the blue and red extremes, due both to the GMOS CCD and the R150 grating efficiency, results in effective spectral coverage of ∼4000–9500 Å.

The N&S technique employed in our program is nonstandard in that it involves a nod distance on the sky that is half the size of the macroscopic shuffle. The mask is designed to incorporate two submasks, each of which is a set of slits covering an area one third the size of the detector. Slits for the two submasks overlap on the sky in an area that is one sixth the size of the detector (because the nod distance is set to half the shuffle distance). This design allows us to place science slits for the primary target—the core of a strong lensing cluster in this case—on both submasks and obtain useful spectra for the entire duration of the N&S exposure, whereas standard macroscopic N&S results in science spectra for only half of the total exposure time. Our integration times were 2400 s resulting in a 1200 s effective integration for each of the two submasks. Two exposures were taken for each target. Thus, if an arc was targeted on both submasks (typical for the most prominent arcs) the total integration time was 4800 s. N&S facilitates straightforward sky subtraction by differencing two sections of the detector—each 1/3 the size of the full detector. The spectra presented here were wavelength calibrated, stacked, extracted, and analyzed using a custom pipeline based on the XIDL software package.⁶

SDSS J0915+3826 is a new strong lensing cluster that was discovered in the SGAS, discussed above. We measure redshifts for 16 cluster members in our Gemini/GMOS spectroscopy and combine this new data with a redshift for the brightest cluster galaxy (BCG) measured in the SDSS to identify a mean cluster redshift of z = 0.397 and a velocity dispersion of 870 ± 169 km s⁻¹. The most prominent strong lensing feature around this cluster is a bright blue arc that we identify as a background galaxy at z = 1.501 from O [ii]λ3727, C ii]λ2328, C iii]λλ1907, 1909 in emission, as well as A iiiλλ1855, 1863  in absorption. Pre-imaging of this cluster reveals a source near the cluster core that exhibits a dramatic drop in flux from the i− to r-band, as shown in Figure 1. With Gemini/GMOS-N, in griz, we measure this source at g_AB ⩾ 26.07, r_AB = 24.86 ± 0.25, i_AB = 23.34 ± 0.09, and z_AB = 23.29 ± 0.13. Slits were placed on this source in both submasks for our N&S observation of this cluster and the spectra exhibit bright emission at 7539 Å which we interpret as Lyman-α λ1216 Å at z = 5.200.

Zoom In Zoom Out Reset image size

The LAE galaxy in this case is located on the sky very near to a foreground galaxy, and so we took care to account for possible contamination in the LAE spectrum by light from the galaxy that falls into the slit, as this could mimic a continuum break in the source spectrum. The separation between the LAE box slit and the potentially contaminating foreground galaxy is 25. Examination of the radial profile of the potential contaminant in the pre-imaging observation shows that the i-band flux within the LAE slit due to the galaxy is on average 1.5 × 10⁻³ of the peak flux of that galaxy. The mean i-band flux of the LAE within the spectroscopic aperture is 5.1 × 10⁻² compared to the peak of the contaminating galaxy, and thus assuming comparable seeing between the imaging and spectroscopic observations suggests that this object's contribution to the i-band flux within the LAE aperture is approximately 3% of the total light. Note that given the separation between the contaminant and the slit the seeing of the spectroscopic and imaging observations would have to be grossly different to produce significant contamination; such a mismatch is neither expected due to execution of the observations in specified conditions in queue mode nor suggested by the spatial width of stellar spectra within the alignment boxes of the spectroscopic observations.

The relevant part of the extracted spectrum is displayed in Figure 2, and despite the low dispersion grating used for our observations the line is measurably asymmetric. We also detect continuum signal redward of the emission line and no significant signal blueward of it. This source has a strong photometric r − i break—even after subtracting the emission line flux from the i-band photometry. It is also very blue in i − z and is not detected in NIC-FIPS imaging to a 2σ limiting magnitude of J_AB ⩾  22.49, indicating that the continuum spectral slope redward of the emission feature is blue. The blue continuum redward of the emission feature and the strong r − i break along with the non-detection in g and the asymmetry of the line confirm the interpretation of the source as a Lyman-α emitting galaxy at high redshift. The most likely candidate emission lines which confuse searches for high-redshift LAEs are O [ii]λ3727 Å, [O iii] λ5007 Å and H-α, but for each of these other candidate lines we would expect accompanying emission features to fall within the wavelength coverage of our spectroscopy. We measure the equivalent width of the Lyman-α emission by fitting a simple power law to the continuum spectrum redward of the emission line over a wavelength range of 7575–8275 Å and subtracting the underlying continuum flux from the emission line. Over the relatively small wavelength range used the continuum level is consistent with a constant value. We note that our spectral flux calibration uses a standard from the Gemini archive to correct roughly for the detector sensitivity as a function of wavelength but is not suitable for measuring absolute fluxes from the spectral data. Taking into account the underlying continuum and the source redshift, we measure a rest-frame EW for the Lyman-α emission of 25 ± 4 Å and combine the EW measurement with our i−band photometry to calculate an observed Lyman-α line flux of 2.1 ± 0.7 × 10⁻¹⁶ erg s⁻¹ cm⁻².

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The second lensed LAE presented here is located near the strong lensing cluster SDSS J1343+4155, which was previously identified as a strong lensing cluster in the SDSS by Wen et al. (2009) and Diehl et al. (2009). We combine our Gemini/GMOS spectroscopy with two cluster member redshifts from the SDSS to measure a mean cluster redshift of z = 0.418 and a velocity dispersion of 1077 ± 266 km s⁻¹ from seven cluster member galaxies. The bright arc around this lens is spectroscopically identified by Diehl et al. (2009) as a background galaxy at z = 2.091. Pre-imaging of this cluster reveals a source near the cluster core that exhibits a dramatic drop in flux from the i− to r-band, shown in Figure 1. We measure this source at g_AB ⩾ 25.97, r_AB ⩾ 25.64, i_AB = 23.78 ± 0.12, and z_AB = 24.24^+0.18_−0.16. We placed slits on this source in each of the two submasks for our spectroscopic N&S observations of this cluster. The spectra corresponding to these slits exhibit a bright emission line at 7289 Å, which we interpret as Lyman-α λ1216 Å at z = 4.994.

Using the measured equivalent widths and i magnitudes, we also calculate lensed isotropic Lyman-α line luminosities of L_Ly−α = 5.9 ± 1.1 × 10⁴³ erg s⁻¹h²_0.7 and L_Ly−α = 5.4 ± 0.9 × 10⁴³ erg s⁻¹h²_0.7 for SGAS J091541+382655 and SGAS J134331+415455, respectively. These are not the true isotropic luminosities because both sources are lensed by foreground galaxy clusters and are therefore significantly magnified. To measure the intrinsic luminosities of the LAEs, we estimate the magnification due to strong lensing by the intervening clusters. The mass models are constructed using the publically available software, LENSTOOL (Jullo et al. 2007), with Monte Carlo Markov Chain minimization in the source plane, and are shown in Figure 4. For SDSS J1343+4155, we compute a simple mass model using as constraints the giant blue arc at z = 2.09, the LAE at z = 4.994, paired with a counter image candidate that we have identified. The cluster and the BCG are represented by pseudo isothermal elliptical mass distributions (PIEMD; Jullo et al. 2007). We allow all the parameters of the cluster halo to vary. For the BCG, we follow the light distribution for the positional parameters and vary the core and cut radii and the velocity dispersion. Our best-fit model reproduces the locations and orientations of the observed lensed images and is consistent with the measured velocity dispersion to within the uncertainty. Based on this model, the magnification at the location of the LAE is m ∼ 12, which we use to calculate the intrinsic isotropic Lyman-α line luminosity for SGAS J134331+415455 to be L_Ly−α = 4.5 ± 0.7 × 10⁴² erg s⁻¹h²_0.7. Assuming Case B recombination and taking the prescription from Kennicutt (1998), we calculate the star formation rate (SFR) for SGAS J134331+415455 to be SFR_Ly−α = 4.1 ± 0.6 M_☉ yr⁻¹h⁻¹_0.7. Applying a magnification correction to the apparent magnitude in z-band, we find the intrinsic z-band AB magnitude to be z = 27. We can also estimate the SFR from the UV continuum flux density at 1500 Å according to Equation (1) from Kennicutt (1998), which results in SFR_UV = 18 ± 3 M_☉ yr⁻¹h⁻¹_0.7.

Zoom In Zoom Out Reset image size

The strong lensing model for SDSS J0915+3826 is constrained by the positions of the components of the giant arc. We do not identify a clear counter image for the LAE and therefore do not use it as a constraint. We represent the lens with a single PIEMD and allow all its parameters to vary, which produces a velocity dispersion estimate that is similar to the value we measure from spectroscopy of cluster members (see Section 2.2). The critical curves for our best-fit model are plotted in Figure 5. Since the available constraints are internal to the radial projection of the LAE, the predicted LAE magnification is highly uncertain. In particular, if we compute a set of models with parameters in the range allowed by the uncertainties, the location of the critical curve for the LAE varies significantly and there is practically no upper limit for the implied magnification. Since in all cases the magnification is higher than ∼10, we adopt this number as our working order-of-magnitude estimate and determine an upper limit for the intrinsic isotropic Lyman-α line luminosity for SGAS J091541+382655 of L_Ly−α ⩽ 5.9 ± 1.1 × 10⁴² erg s⁻¹h²_0.7 and a corresponding SFR_Ly−α ⩽ 5.3 ± 1.0 M_☉ yr⁻¹h⁻¹_0.7. Correcting the z-band magnitude for the lensing magnification, we recover an intrinsic magnitude of z ⩾ 25.75 mag, from which we measure SFR_UV ⩽ 51 ± 7 M_☉ yr⁻¹h⁻¹_0.7 using Equation (1) from Kennicutt (1998).

Zoom In Zoom Out Reset image size

We note that SFR_Ly−α and SFR_UV differ significantly for both sources. For SGAS J134331+415455 SFR_Ly−α is lower than SFR_UV by a factor of 4.5, which is consistent with previous studies finding that the SFR calculated from Lyman-α emission is often systematically lower than other SFR metrics—such as UV continuum flux density at 1500 Å —by a factor of ∼5 or more (Tapken et al. 2007). The disagreement between the two SFR estimates is more extreme for SGAS J091541+382655, with SFR_UV ∼9× larger than SFR_Ly−α. In a study of LAEs at z = 3.1 identified in the Chandra Deep Field South, Gronwall et al. (2007) find on average, SFR_UV/SFR_Ly−α ∼ 3 with a large scatter about the mean, including several objects with SFR_UV/SFR_Ly−α as large as we measure for SGAS J091541+382655. Significant differences between these two SFR calibrations are understandable considering the potential systematic errors involved in measuring the SFR from both UV continuum and Lyman-α emission. For example, resonant scattering of Lyman-α photons off of neutral hydrogen can suppress observed Lyman-α emission if there is even a very small amount of dust in the source galaxy, thus producing an underestimate of SFR_Ly−α. Additionally, Kennicutt (1998) identifies several caveats associated with the SFR_UV calibration, including assumptions about the IMF and the timescale and manner of star formation (e.g., constant vs. burst). Specifically, Equation (1) in Kennicutt (1998) recovers SFR from the rest-frame UV flux by assuming that a galaxy has undergone continuous star formation for 10⁸ years or longer and originates from a Salpeter IMF with mass limits of 0.1 and 10 M_☉. If the properties of the underlying stellar populations in either of these galaxies differ significantly from the parameters assumed in Equation (1) from Kennicutt (1998), then there will be systematic differences between the SRFs estimated from the rest-frame UV flux and the true SFRs.

We have also considered the possibility that the Lyman-α emission in one or both of these objects could be due all or in part to active galactic nucleus (AGN) activity. Given the lack of detection of significant emission from highly ionized species, such as N[V] and C[IV], we conclude that it is unlikely that AGNs are playing a significant role in the observed Lyman-α emission. The disagreements between the two SFR estimates for the two sources discussed here highlight the difficulty in making robust SFR measurements from source-frame UV observables.

Having accounted for the magnification due to gravitational lensing, we can also compare the luminosities of these two sources to large samples of LAEs and LBGs at comparable redshifts (e.g., Hu et al. 2004; Shimasaku et al. 2006; Bouwens et al. 2007; Dawson et al. 2007; Murayama et al. 2007; Ouchi et al. 2008; McLure et al. 2009). Studies of luminosity functions at high redshift are challenging, and parameters such as L* and α are not nearly so well measured as they are in the nearby universe. For comparing Lyman-α luminosities, we take the average of six different measurements of L*_Ly−α values from Ouchi et al. (2008) that were fit using samples of LAEs at z ∼ 3.7 and z ∼ 5.7, assuming three possible values of the faint end slope of the luminosity function, α. For the purpose of comparing continuum luminosities, we take L*_UV as the average of the two L*₁₅₀₀ best-fit values for LAEs at z ∼ 3.7 and z ∼ 5.7 from Ouchi et al. (2008). SGAS J091541+382655 is (L*_Ly−α, L*_UV) ⩽ (0.6 L*_Ly−α, 2 L*_UV) and SGAS J134331+415455 is (L*_Ly−α, L*_UV) = (0.5 L*_Ly−α, 0.9 L*_UV). Both of these galaxies live at ≲L* on the luminosity function for similar galaxies at comparable redshifts, which makes them interesting targets for studying the properties of typical LAEs at z ∼ 5 on an individual basis (Figure 6).

Zoom In Zoom Out Reset image size

We also attempt to measure the morphologies of both lensed LAEs in our Gemini/GMOS imaging and place constraints on the intrinsic sizes of these two galaxies. Neither source is detected at S/N ≳ 12, which limits our ability to make detailed morphological measurements or shape fits, but we can check the object profiles against the image PSFs and use our estimates of the magnification due to lensing to place some rough constraints on the physical sizes of these galaxies. Studies of LAE morphologies at z = 3.1 using HST imaging find that the Lyman-α and rest-frame UV emission from these sources are spatially coincident and have half-light radii ≲1.5 kpc (Bond et al. 2009, 2010). Bond et al. (2009) argue that S/N ≳30 is necessary to make robust measurements of half-light radii for LAEs and find that it is difficult to distinguish between resolved and unresolved compact cores at S/N ≲30. The two lensed LAEs are best detected in our Gemini/GMOS i-band imaging, so these are the data we use for our morphological measurements.

SGAS J134331+315455 is slightly elongated in the tangential direction with respect to the center of the cluster. We measure its profile to be consistent with a Gaussian along the tangential axis with an FWHM =11 and consistent with a Gaussian of FWHM =057 along the radial axis. The FWHM of the PSF measured from stars in the image is ∼057, suggesting that the LAE is unresolved along one axis and resolved along the other. This kind of morphology is natural for a strongly lensed background source where the magnification is generally much greater along one axis than the other; a source located on the tangential caustic will be highly magnified in the tangential direction with respect to the center of the lensing potential and magnified very little or not at all in the radial direction. We conclude that SGAS J134331+415455 is unresolved in the radial direction and barely resolved in the tangential direction. Assuming that all of the magnification factor of m = 12 is applied in the tangential direction, we convert an FWHM of 11 into a physical linear size of ∼0.6 kpc, which indicates that most of the emission from this galaxy originates from a region very compact in size. This constraint must be taken with the caveat that the data are lower S/N than would be optimal, but the significant linear magnification of these sources due to gravitational lensing does enable us to probe very physically interesting scales even with ground-based imaging. We also measure the locations of SGAS J134331+415455 to be identical in the Gemini/GMOS i- and z-band images to the accuracy of our astrometry, ∼02, corresponding to physical scales of ∼100 pc. Based on the observer frame equivalent width of this source, the flux in the i-band filter is ∼60% Lyman-α line emission, whereas the z-band filter measures only rest-frame UV continuum emission. Given the spatial coincidence of the LAE in the i and z bands, we conclude that the UV continuum and Lyman-α emission cannot originate from regions separated by more than of order ∼100 pc projected onto the sky. We have no measurement of the relative velocities of the UV continuum and the Lyman-α emission and therefore cannot constrain a possible separation in velocity.

SGAS J091541+382655 is similar to SGAS J134331+415455 in that the LAE is elongated on the tangential axis relative to the center of the lensing cluster. Using the GALFIT subtracted i-band image, we measure the LAE to be resolved in the tangential direction, having a shape consistent with a Gaussian of FWHM = 182, which equates to a constraint on the physical scale of ⩽1.2 kpc, where the lower limit on the magnification of this source translates to an upper limit in the physical size (and again assuming that the magnification is entirely along the tangential direction). We also measure the LAE to have a Gaussian shape with FWHM = 075 in the radial direction, which—given the FWHM ∼073 measured from the PSFs of stars in the image—is consistent with the source being unresolved in the radial direction. In the observer frame the equivalent width of the Lyman-α emission for this source is EW_Ly−α = 157 Å, from which we estimate that Lyman-α photons are contributing less than 20% of the i-band flux, so that we cannot make any attempt to measure a spatial offset between the Lyman-α and UV continuum emission. Our size constraints for both sources are in good agreement with the literature, in that we find the sizes of the LAEs to be consistent with compact sources of size ≲1.5 kpc.

With several photometric measurements in hand for each of our sources, it becomes interesting to investigate the spectral energy distributions of the LAEs to try and recover the properties (e.g., stellar mass, age, SFR, and dust content) of the underlying stellar populations in a way that does not rely on as many problematic assumptions as, for example, the SFR estimated from the rest-frame UV continuum flux that we calculate in Section 3.1 above. We compare our photometry for these sources against modified stellar population synthesis models. More details can be found in Wuyts et al. (2010); we only summarize the main procedure here. We use the revised templates of Bruzual and Charlot (CB07, based on Bruzual & Charlot 2003) with solar metallicity, a Chabrier initial mass function (Chabrier 2003), and Calzetti (2000) dust extinction law. We investigate a range of exponentially declining star formation histories of the form SFR(t) ∼ exp(t/τ), with e−folding times τ= 0.01, 0.05, 0.1, 0.2, 0.5, 1, 2, and 5 Gyr, as well as single bursts (SSP) and continuous star formation models (CSF). An updated version of the code Hyperz (Bolzonella 2000) is used to obtain the best-fit SED at the fixed spectroscopic redshift of the source via a maximum likelihood procedure.

The best-fit SED for SGAS J091541+382655 is shown in Figure 7 and corresponds to a dust-free single burst with an age of 1.4 Myr. It is important to supplement the best-fit stellar population parameters with confidence intervals allowed by the photometric uncertainties. We create 1000 fake realizations of the observed SED by perturbing each broadband magnitude measurement in a manner consistent with its error bars. This set of fake SEDs is fit in exactly the same manner as described above for the observed SED; bad fits with χ²>3 are excluded. This procedure results in a very young (age ⩽5 Myr) and dust-free (E(B − V) ⩽ 0.01) stellar populations. Though we can place constraints on the age of the observed stellar population, these data are not sufficient for the SED fits to simultaneously constrain the star formation history and consequently the current star formation rate for this galaxy. After correcting for a lensing magnification factor of ≳10, we find a Chabrier stellar mass of M_stars ⩽ 7.9^+3.7_−2.5 × 10⁷ M_☉ h⁻¹_0.7.

Zoom In Zoom Out Reset image size

SGAS J134331+415455 is detected in only two bands, which is insufficient to produce a robust best-fit stellar population. Instead, we compare the photometry to single burst, dust-free CB07 models of different ages to obtain some constraints on the age and stellar mass. The results are shown in Figure 8. The age of the observed stellar population lies in the range between 100 Myr and 400 Myr, which makes SGAS J134331+415455 significantly older than the very young stellar population seen in SGAS J091541+382655. After correcting for a lensing magnification factor of 12, we constrain the corresponding stellar masses to be within the range 2 × 10⁸ M_☉ h⁻¹_0.7< M_stars < 6 × 10⁹ M_☉ h⁻¹_0.7. All of our photometric data for both of these sources sample light blueward of the 4000 Å break in the source frame so that the stellar population parameters obtained from the SED fitting procedures correspond to the most recent episode of star formation in each LAE; we have no power to constrain stellar mass that may be present in an underlying population of older stars.

Zoom In Zoom Out Reset image size

In addition to SED modeling of the photometry of both LAEs, we also examine the modest continuum signal that is detected for SGAS J091541+382655. Cross-correlation of our GMOS spectrum for this source over a wavelength range of 1230–1350 Å (source frame) against the composite LBG spectrum from Shapley et al. (2003) results in a peak at z ∼ 5.2, in agreement with the redshift measured from the Lyman-α emission. We explicitly exclude the Lyman-α emission portion of the spectrum in this analysis so that the cross-correlation signal that we detect is entirely due to low-significance continuum features redward of Lyman-α. The composite LBG from Shapley et al. (2003) is plotted with our GMOS spectra for both targets in Figures 2 and 3, and visual comparison of the continuum against the LBG composite spectrum suggests the presence of several strong UV metal absorption lines that are observed in well-studied LBGs at lower redshifts (e.g., Pettini et al. 2000). The features are at too low a significance in our data to claim a robust detection of individual absorption lines, but the cross-correlation signal alone is encouraging given our low-resolution spectra and limited integration time. Both of the sources presented here are excellent candidates for a more aggressive spectroscopic follow-up effort to explore the gas properties and stellar metallicity in representative low-mass star-forming galaxies at z ∼ 5.