SPATIALLY RESOLVED HST GRISM SPECTROSCOPY OF A LENSED EMISSION LINE GALAXY AT z ∼ 1^*

The central portion of A1689 is observed with the G800L grism with ACS on HST along with the broadband (F475W, F625W, F775W, and F850LP) exposures presented in Broadhurst et al. (2005). A single pointing is used at one position angle in a 7.1 ks exposure (three orbits). The resulting dispersed image covers a wavelength range of 5700–9800 Å at a resolution of R ∼ 90. The analysis of these images is performed using software discussed in detail by Meurer et al. (2007, hereafter M07). M07 present G800L grism observation of a well-studied unlensed field, the Hubble Deep Field-North (HDF-N), to a similar depth as our observations. They reduce and analyze this data set in two different ways: Method A, aXe selection and reduction (Kümmel et al. 2009), which is similar to the GRAPES team pipeline and Method B, "blind" EL source detection. Method A yields a similar set of sources as Method B, with the latter additionally yielding >50% more sources and more cases of multiple EL sources (i.e., star-forming regions) per object. A salient feature of Method B is its emphasis on finding ELs in the two-dimensional dispersed frames. This approach enables lone ELs with weak galaxy continuum flux to be detected against the high background that is characteristic of grism images. For our aims to do spatially resolved spectroscopy and to maximize the number of EL sources, we perform the reductions using Method B. The primary steps of the reduction algorithm are discussed below.

The initial data reduction is performed with the Space Telescope Science Institute (STScI) ACS CCD reduction pipeline CALACS (Hack 1999), from which we use the cosmic-ray-rejected "CRJ" G800L images along with the individual flat-fielded "FLT" broadband images (F475W, F625W, F775W, and F850LP). Since CALACS does not flatfield G800L exposures we apply our own corrections for pixel-to-pixel variations using a standard F814W flat from the STScI archive. To remove small dithers between exposures and to correct for geometric distortion we employ the ACS team pipeline Apsis (Blakeslee et al. 2003). These initial reduction steps produce geometrically corrected and aligned grism images, which are processed further as described below. Similarly, Apsis is used to align and process the set of broadband images; the weighted sum of these is what we designate as the "direct" image.

Galaxies with strong ELs should be the easiest sources to identify in slitless images like our G800L data. Method B of M07 requires no prior knowledge of the location of the emitting source, although the direct image is needed to find the precise position of the emission. Moving forward from the initial reductions discussed above, we next subtract off a 13 × 3 median smoothed version of the dispersed image, revealing the ELs. We also subtract off a smoothed version of the direct image from itself. The result of this high-pass filter is to remove slowly varying background levels, including galaxy continuum, starlight from galaxy neighbors, and other sources. As detailed in M07 we have determined a geometric distortion solution relating positions in the direct image to the corresponding positions of the zeroth order in the grism image. This solution along with the measured flux scaling between direct and zeroth-order images is used to mask out all significant flux from zeroth-order images in the grism frame. SExtractor (Bertin & Arnouts 1996) is used to find the EL sources in this masked and filtered grism frame. A five-row segment of both of the filtered dispersed and direct image is extracted, collapsed to 1D spectra and cross-correlated. The peak in the cross-correlation gives the location of the emitting source and an estimated line wavelength. The aXe package is then used to extract a 1D spectrum from the dispersed image (prior to any filtering) using the location we found as the adopted source position. Gaussian fits to the lines in this spectrum are used to determine their final wavelength and flux values. A portion of the dispersed image that includes the spatially resolved line emission for ELG F12_ELG1 at z = 0.79 appears in Figure 1. We present the 1D spectrum of F12_ELG1 in Sections 4 and 6. Figure 2 shows 1D spectra for other representative ELGs in our sample. Some line pairs can be distinguished which assist with line identifications. Notably, nebular lines [O iii] and Hβ are marginally resolved in the new galaxy ELG 5158, and the close line pair Hα and [S ii] can both be identified in the new galaxy ELG 4298.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We refer to an emission feature in the dispersed image as an EL (typically Hα, Hβ, [O iii], or [O ii]). The corresponding position in the direct image is called an emission line source (ELS). An ELS is typically an H ii region or galaxy nucleus, and in some cases there is more than one EL per ELS (e.g., [O ii] and [O iii]). The galaxy containing the ELS(s) is an emission line galaxy (ELG). If the galaxy is spatially resolved, there may be multiple ELSs per ELG. We identify a total of 43 galaxies (ELGs), 52 ELSs, and 66 ELs. Somewhat surprisingly, three-quarters of the ELGs are new in this well-studied cluster field. The spectroscopic results appear in Table 1. We identify nine ELGs in the cluster, 30 ELGs in the background, and five ELGs in the foreground. As the number of ELSs is larger than the number of ELGs, some ELGs have multiple H ii regions spatially resolved by the grism. In Table 2, we separate out the ELGs with two or more ELSs. The details concerning our line list and EL properties sorted by species are given in the Appendix.

Extensive ancillary data exist to support the grism observations. Images of the central portion of A1689 were taken in several bands, as follows: U (DuPont Telescope, Las Campanas), B (NOT, La Palma), V (Keck II LRIS), I (Keck II LRIS), g₄₇₅r₆₂₅i₇₇₅ and z₈₅₀ (HST ACS), and P_3.6 and P_4.5 (Spitzer IRAC). Spectroscopy of hundreds of arclets has been acquired at several large ground-based observatories. The existing photometry and spectroscopy are discussed in detail elsewhere (Broadhurst et al. 2005; Frye et al. 2002, 2007, 2008).

Additional spectroscopy was obtained for F12_ELG1 at z = 0.79 in 2010 May with the DEep Imaging Multi-Object Spectrograph (DEIMOS) on Keck II (Faber et al. 2003). The observations were made through a 1 arcsec slit width with the 1200 lines mm⁻¹ grating blazed at 1016 and set to a central wavelength of 7800 Å. A combination of 12 slitlets was placed together to construct the long-slit mask "Long1.0B." Two 300 s exposures were taken during dusk twilight. The seeing was 0.5–06 FWHM. Long1.0B provided a resolution of R = 5870 measured from an isolated night sky line at 7571.75 Å. The data were reduced with the DEEP2 DEIMOS pipeline (Cooper et al. 2012; Newman et al. 2012). Observations were also acquired on the Magellan Telescope I (Walter Baade) in 2009 March and 2010 March on the Inamori Magellan Areal Camera and Spectrograph (IMACS). One multislit mask was used with 35 targets using the 600 lines mm⁻¹ grism at a blaze angle of 14.67 Å and a central wavelength of 8410 Å as a part of a different program (PIs: Malhotra and Rhoads). The grism provided a resolution of R = 2300. Eight 1800 s exposures were taken in a single position.

The image and spatially resolved G800L spectra of F12_ELG1 with a visual extent of >8 arcsec and a magnification provided by the cluster of a factor of 4.5 appear in Figure 5. This is one of a handful of the brightest star-forming galaxies at z ∼ 1 (M_B = −21.3), allowing for a study of one ELG in detail at intermediate redshift. The HST grism spectroscopy shows three separate ELSs as ELS 20002 ("A"), ELS 10638 ("B"), and ELS 10640 ("C"). Each ELS has at least two emission lines: [O ii] λλ3726, 3729 and [O iii] λλ4959, 5007, with the central knot "B" also showing Hβ line emission. The line fluxes and rest equivalent widths are given in Table 1. A fit to the line centroids of the emission features in Component A yields a redshift of z = 0.790, confirming the redshift in Duc et al. (2002). Knots A and B have a separation at the source of 1.5 kpc, while knots B and C are separated by only 0.5 kpc. There appear to be significant redshift differences between knots A, B, and C. These are likely to be owing to misidentification of the precise x position of the ELSs, which will translate into a wavelength error.

Zoom In Zoom Out Reset image size

In our Keck and Magellan spectra of F12_ELG1 we see line emission from star-forming regions across a continuous 4 From the 1D spectroscopy of this extended line emission we recover all the emission features of the HST grism spectrum and also detect additional ELs. We identify [O ii] λ3726, [O ii] λ3729, [O iii] λ4959, [O iii] λ5007, Hβ and Hγ, with Hδ, H, H8, and [Ne iii] λ3869, and weak [C iii]/[C iv] λ4650 (in the Magellan spectrum). While the results are similar for our two ground-based data sets, we will focus primarily on the Keck spectrum herein with its higher spectral resolution and flux calibration taken with a standard star at the time of the observations. The line fluxes for the strong ELs for the Keck spectroscopy appear in Table 3.

A Gaussian fit to [O ii] and [O iii] in our Keck spectrum yields a new systemic redshift of z = 0.7895. The line width is estimated from the fit to the [O iii] λ5007 line and is found to be Δv = 500 km s⁻¹ after subtraction of the instrumental resolution. This is the line width set for all other emission features. To account for the slightly asymmetric line profile of the [O iii] λ5007 line owing to an adjacent sky line, the flux value is taken from the Gaussian fit rather than from the data values. The [O iii] λ4959 line also suffers from its unfortunate placement relative to a sky line. In this case the well-defined ratio of [O iii] λ5007 to [O iii] λ4959 of 3:1 is used for any calculations involving the sum of the fluxes or equivalent widths of these two lines. For all other line features, the values are measured directly from the data. The Balmer transition lines Hβ through H8 are shown in Figure 6. The velocity range over our 500 km s⁻¹ measurement width is shown in blue and varies from −250 km s⁻¹ < v < +250 km s⁻¹ with reference to the systemic redshift of z = 0.7895. This large family of Balmer features all in emission is rare for an extragalactic source and signals the early stages of a major starburst.

Zoom In Zoom Out Reset image size

Our Keck spectrum has at best only a slight continuum break at rest-frame 4000 Å. Balogh et al. (1999) define this break D_n(4000) as the ratio of the average flux density in the wavelength band 4000–4100 Å to the 3850–3950 Å band. We measure a break index of D_n(4000) = 0.96–1.0 after first masking out the strong [Ne iii] λ3869 and H8 ELs. This near lack of a continuum depression indicates a dominant population of hot young stars. The population synthesis models of Bruzual & Charlot (2003, hereafter BC03) provide a value of D_n(4000) for each spectral energy distribution (SED; see Section 4.3 for a description of our model). For a reasonable subset of models over four metallicities (0.2 Z_☉, 0.4 Z_☉, Z_☉, and 2.5 Z_☉) and a range of star formation histories, we have set an initial constraint on the age of the dominant stellar population of the galaxy to a range of 6 Myr ⩽t ⩽ 100 Myr. Note that any contribution from AGN continuum light would operate to raise the value for D_n(4000), and hence lead to an underestimate in the age of the galaxy (Kauffmann et al. 2003). We infer from spatially resolved analysis that this object is most likely dominated by star formation but also shows evidence of a harder ionizing source (see Section 6).

The [O ii] λλ3726, 3729 doublet is spectrally resolved in our Keck DEIMOS data, and this is useful as these are density-sensitive lines (Figure 7). The intensity ratio is measured to be 1.44, which for a typical temperature of T = 10,000–20,000 K equates to N_e = 980–1280 cm⁻³. This is a factor of ∼3 higher than the star-forming galaxies in the Kobulnicky & Kewley (2004) sample with 0.3<z < 1.0 and is consistent with ELG samples at z ∼ 2 (Lehnert et al. 2009; Hainline et al. 2009). These, and all other fluxes in this paper, are corrected for underlying stellar absorption, extinction by dust, and cluster magnification as described in Sections 4.3 and 5.

Zoom In Zoom Out Reset image size

The hydrogen Balmer features consist of both a nebular emission and a stellar absorption term. The ELs tend to weaken toward higher level transitions owing to the sum of a rapidly decreasing nebular EL strength with decreasing wavelength and an equivalent width of the stellar absorption component that at best increases only modestly with wavelength. To make a correction for the underlying stellar absorption we compute an intrinsic SED model template and then subtract it from the galaxy spectrum in a manner similar to that of Tremonti et al. (2004). We note that as our object is rare for showing Hβ through H8 all in emission, the underlying stellar absorption is not expected to be a major contaminant.

Our optical through infrared photometry is used to construct the matching SED model. We delens the photometry using a magnification factor of 4.5. The observed (unlensed) photometry of F12_ELG1 is as follows: U = 21.05 ± 0.16, B = 21.22 ± 0.17, V = 21.08 ± 0.15, g' = 21.06 ± 0.08, r' = 20.87 ± 0.01, i' = 20.56 ± 0.01, z' = 20.44 ± 0.01, P_3.6 = 20.11 ± 0.07, and P_4.5 = 20.46 ± 40.08. Our measurements of the Balmer decrement allow a bracketed range of model stellar ages 6 Myr<t <100 Myr (see Section 4.2). This constraint is in keeping with our detection of higher level Balmer ELs, as population synthesis models for star-forming galaxies predict for these features an increase with evolution up to 500 Myr (González Delgado et al. 1999). Our model-fitting approach allows for a range of metallicities and assumes a single starburst model with a range of decay rates τ and a star formation rate that depends exponentially on τ as follows: SFR(t)∝exp (t/τ). A summary of our parameter ranges is as follows:

• 1.  
  τ = 0.1, 0.2, 0.3, 0.5, 1, and 5 Gyr;
• 2.  
  t = 6–100 Myr;
• 3.  
  Z = 0.2 Z_☉, 0.4 Z_☉, 1 Z_☉, and 2.5 Z_☉.

For each fixed τ, t, and Z, two parameters are fitted to the data: color excess E(B − V) and stellar mass M_*. A suite of models is constructed over the allowable parameter space, and each model is corrected for dust extinction. We compute synthetic photometry for comparison with the observed photometry until the lowest value of reduced χ² is obtained, as is described in detail in Frye et al. (2008). Our best-fit model estimate yields E(B − V) = 0.45 and M_* = 2 × 10⁹ M_☉ for a young stellar age of t = 8 Myr, Z = 0.4 Z_☉, and τ = 5 Gyr. After redshifting and binning to the correct spectral resolution, we measure a correction in rest-frame equivalent width of W = 1.5 Å for the Hβ line. All Balmer ELs include this correction for underlying stellar absorption. Our measured EL rest equivalent widths are too low to affect the SED fits, in contrast to the large equivalent widths found in Atek et al. (2011). Note although t is less than the estimated galaxy crossing time by a factor of ∼2, we find the BC03 models to provide an adequate fit to the data for our purposes of determining estimates for stellar absorption.

The attenuation of the intrinsic light due to dust is calculated from a standard curve, which for a starbursting galaxy is given in Calzetti et al. (2000), their Equation (4). Dust extinction is calculated using the Balmer decrement method, whereby a pair of ELs with a well-defined intrinsic ratio from atomic theory such as Balmer lines is compared with the data. We use Hβ and Hγ from our Keck data set, and attribute any deviation from the intrinsic value to dust. After correcting for the underlying stellar absorption, we measure f(Hγ)/f(Hβ) = 0.321. For an intrinsic ratio of f(Hγ)/f(Hβ) = 0.469 given by Osterbrock (1989), we compute E(B − V)_gas ≈ 0.78. This value is different from the color excess measured by SED fitting of E(B − V) = 0.45. Some of this discrepancy is due to incomplete areal coverage of the galaxy image. We adopt the more general value measured from the SED modeling for this study.

The spectral classification of ELGs at low redshift can be determined from EL diagnostic diagrams such as the classical Baldwin, Phillips, & Terlevich (BPT) diagram (Baldwin et al. 1981). The BPT diagram is based in part on Hα and [N ii] lines which are redshifted out of the optical passband for z ⩾ 0.4. Classification systems based on bluer ELs are also well developed (Lamareille et al. 2004, 2009; Pérez-Montero et al. 2009; Lamareille 2010; Rola et al. 1997; Marocco et al. 2011). Marocco et al. (2011) revise the blue EL scheme of Lamareille (2010). Their samples are derived from the Sloan Digital Sky Survey (SDSS) and do not include a correction for dust extinction.

We classify F12_ELG1 for our ACS grism (G800L) data set for which we have complete coverage of the emitting line region. We correct all fluxes for underlying stellar absorption and dust extinction and sum up the flux over all three ELSs. Under the Marocco et al. (2011) scheme, which involves the line ratios f ([O iii])/f ([Hβ]) and f ([O ii])/f ([Hβ]), F12_ELG1 is situated in the intermediate region that includes both star-forming galaxies (SFGs) and Seyfert 2 objects. In a different classification scheme, Pérez-Montero et al. (2009) use also the VVDS samples but instead consider the ratios of f ([O ii])/f (Hβ) and f ([Ne iii])/f (Hβ) after accounting for dust extinction. With this diagnostic set, F12_ELG1 is placed in a region that includes SFGs and the uncertainty domain between SFGs and AGNs.

The detection of additional spectral features similarly suggests that F12_ELG1 is in a position intermediate between star-forming and AGN source types. We detect [C iii]/[C iv] λ4650 commonly seen in AGNs and measure a flux ratio of [Ne iii]/Hβ ≈ 0.6, a value consistent with excitation by a hard ionizing source (Osterbrock 1989). At the same time we fail to detect [Ne v] λ3426 and [He iiλ4686], two ELs typically associated with AGNs, although the signal-to-noise is poor at the expected position of [Ne v] λ3426 owing to its placement on top of a strong sky line. Note the AGN diagnostic line [Ne v]λ3346 is blueward of our passband. As for the common nebular lines, the flux ratio f ([O iii])/f (Hβ) is low in the center, a trend that runs contrary to the expected behavior of a central AGN (Figure 5). Interestingly, the Balmer decrement is at best only marginally detected, which when used as an additional diagnostic places this object as an SFG for any value of [O ii] and [Ne iii] (Marocco et al. 2011). We conclude that there is at least a strong star-forming component to F12_ELG1, and that the AGN interpretation cannot be ruled out. We will use spatially resolved spectroscopy to address the possibility that this object supports AGN-like activity in Section 4.7.

The gas-phase oxygen abundance is estimated from the metallicity-sensitive rest-frame optical ELs. The direct measurement from the [O iii] λ4363 line typically seen in metal-poor galaxies is not detected in any of our spectroscopy. The indirect measurement using EL ratios involving Hα is redshifted out of our passband. From the available ELs we can compute R₂₃ = ( f ([O ii] λ3727)+f ([O iii] λλ4959,5007))/f (Hβ) and O₃₂ = ( f ([O iii] λ4959)+f ([O iii] λ5007))/f ([O ii]). R₂₃ = 4.6^+2.3_−1.5 and O₃₂ = 0.96^+0.39_−0.28 for the G800L data set and after first correcting all line fluxes for underlying stellar absorption, reddening, and cluster magnification. The uncertainties reflect 1σ errors in the noise and continuum placement.

The calculation of gas-phase oxygen abundances from R₂₃ is complicated by its double-valued behavior, with a given value representing either the metal-poor lower branch, the metal-rich upper branch, or a transition at 12 + log (O/H) ≈ 8.4. We measure values for several empirical calibrations. Our value for R₂₃ yields an abundance of 12 + log (O/H) = 8.1 (lower branch) and 8.9 (upper branch; Kobulnicky & Kewley 2004). For Tremonti et al. (2004) as adapted from the method of Charlot & Longhetti (2001) we obtain 12 + log (O/H) = 8.8 (upper branch). For Zaritsky et al. (1994) we measure 12 + log (O/H) = 8.8 (upper branch). Finally, we compute 12 + log (O/H) = 7.8 (lower branch) and 8.7 (upper branch) for McGaugh (1991) as given by Kobulnicky & Johnson (1999). F12_ELG1 is consistent with being on either branch.

Of the available emission features, the secondary indicator [Ne iii] can break the metallicity degeneracy (Nagao et al. 2006). We compute f ([Ne iii])/f ([O ii]) = 0.05^+0.01_−0.07, which has a best-fit polynomial correspondence value of 12 + log(O/H) = 8.8 ± 0.2. Our value is close to the solar value of 12 + log (O/H) = 8.66 (Allende Prieto et al. 2002; Asplund et al. 2004). It also overlaps with the range of values obtained from current metallicity history studies at z = 1–2 of 12 + log(O/H) = 8.3–9.0 (Lamareille et al. 2006; Liu et al. 2008; Hainline et al. 2009). In turn, all values are higher than those of direct measurements from strongly lensed high-z galaxies (Yuan & Kewley 2009; Rigby et al. 2011).

We estimate the star formation rate in two ways: (1) by extrapolating the intrinsic Hβ line flux into an estimate of the Hα flux and (2) from the intrinsic (corrected for reddening and lensing magnification) [O ii] line flux. Although our observations do not cover the most reliable tracer of star formation, Hα, one can infer the Hα line flux from the E(B − V) value and the intrinsic flux ratio between Hβ and Hα (Osterbrock 1989). We estimate the intrinsic fluxes for Hα to be f (Hα) = 1.4^+0.53_−0.61 × 10⁻¹⁵ erg s⁻¹ cm⁻² and f (Hα) = 2.1^+0.38_−0.33 × 10⁻¹⁶ erg s⁻¹ cm⁻² for the Keck and HST G800L observations, respectively. The SFR can be measured using the relation starting from Kennicutt (1998): SFR(M_☉ yr⁻¹) = 7.9 × 10⁻⁴²f(Hα)*4πD²_L erg s⁻¹. For the Keck observations we measure SFR = 31⁺¹¹₋₁₃ M_☉ yr⁻¹. For the HST G800L observations there is insufficient spectral resolution to detect Hβ in every ELS, making our value of SFR ≳ 6.5 M_☉ yr⁻¹ a lower limit. We will adopt the value for the SFR from the Keck observations for this study.

Calibration of the star formation rate using the intrinsic [O ii] λ3727 line is more challenging, as it is more sensitive to reddening, metallicity, and ionization parameter. Kewley et al. (2004) present an algorithm that takes these dependencies into account, especially for high gas-phase oxygen abundances which apply to our case (log (O/H) + 12 > 8.4). Their Equation (14) for SFR is based on ionization parameter and gas-phase oxygen abundance. Using this equation yields SFR_[O ii] ≈ 3 M_☉ yr⁻¹ for the G800L grism data set, a value that is small compared to estimates based on Hβ flux. We will adopt the value for the SFR from the Hβ flux for this study as its relation to Hα is better understood, and plan to measure SFR directly by Hα in grism observations with VLT SINFONI in an upcoming paper. Finally, we compute sSFR ≈20 Gyr⁻¹.

The image of this giant arc is noticeably bisected into a broad, line-emitting clump and a long extended tail. Line emission is detected only in the clump. We have spatially resolved spectroscopy with sufficient angular resolution to subdivide the line-emitting clump into seven contiguous bins of 06 each, or equivalently 1 kpc each after correcting for a magnification factor of 4.5. Each bin is assigned a different color. The bins have the following assignments: Bin 1 (purple), Bin 2 (rose), Bin 3 (lavender), Bin 4 (light green), Bin 5 (dark green), Bin 6 (olive), and Bin 7 (orange). We extract the seven spectra and compute line fluxes, line centroids, and velocity widths for common ELs. A sample stack plot centered on the position of Hβ in particular is shown in Figure 8, and an image of F12_ELG1 with the seven bins overlaid appears in the top panel of Figure 9.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The velocity dispersion σ is derived from the FWHM of the Hβ line and based on the assumption that the line width is set by the spread of velocities of the line emitting gas about the local central mass. The instrumental profile is subtracted off of the FWHM in quadrature, and then σ is computed as follows: FWHM/2.355*c/λ_obs. There should be a peak at the galaxy's center of mass if the object is supported by a central mass pulling on the stars. For F12_ELG1 the peak is at Bin 2, making this location the likely location for the galaxy nucleus (Figure 9). The uncertainties in σ come from continuum fitting and are less than the size of the plotting symbols. It is interesting that the Hβ line centroid blueshifts relative to the mean value along the galaxy, from Δv = +35 ± 5 km s⁻¹ for Bin 1 to Δv = −13 ± 5 km s⁻¹ in Bin 7. These results are similar to what is also seen across the lensed galaxy Lens22.3 at z = 1.7 (Yuan & Kewley 2009). From the velocity dispersion one can compute a rough value for the dynamical mass of ∼2 × 10⁹ M_☉. This value is consistent with our estimated stellar galaxy mass from SED fitting. We measure M_B = −22.2 and in turn obtain an estimate of the mass-to-light ratio of M/L ≈1. Our low measured M/L and stellar mass are consistent also with our stellar age by SED fitting of t = 8 Myr. This is similar to the work of van der Wel et al. (2005) who find a low M/L and low stellar mass at z ∼ 1 to be correlated with a young stellar age.

The trends of various metallicity-dependent line ratios across the galaxy are also shown in Figure 9. We retain the same bin coloring scheme and apply the same modest correction for stellar absorption and extinction to each bin as computed in Section 4.3. While the peak in the FWHM is at Bin 2, there is a temptation to find a peak over the metal line ratios at Bin 3, and a drop off to the left-hand side (Bins 1 and 2). We derive below a number of interesting results by a comparison of values in Bin 3 with the mean of Bins 1 and 2 (Bins 1–2). By Marocco et al. (2011) Bin 3 encompasses the composite star-forming galaxy (SFG)/Seyfert 2 and Seyfert 2-only regions of their f ([O iii]/Hβ) versus f ([O ii]/Hβ) H ii excitation diagram. Meanwhile Bins 1–2 yield a different result and reside entirely in the SFG/Seyfert 2 region. By the approach of Pérez-Montero et al. (2009), which includes [Ne iii], Bin 3 plus uncertainty region lies in the transition region between SFGs and AGNs, while Bins 1–2 plus uncertainty region lies entirely inside the SFG region. We conclude that the disky Bins 1–2 is consistent with an SFG while Bin 3 exhibits AGN-like behavior. These strong metal line ratios that are driven by a hard ionizing source and are offset by ∼1 kpc from the peak of the velocity dispersion appear to be best explained by shocks in the direction of the galaxy tail (Bins 4–7).

The values for [O iii]/Hβ are large all across the galaxy and are also higher in Bin 3 compared to Bins 1–2. In Bin 3 we measure log( f ([O iii])/f (Hβ)) = 0.49 ± 0.04. There is only a hint of a trend in the behavior of [Ne iii]/Hβ, but it is worth noting that [Ne iii]/Hβ is high in the central three bins, and for Bin 3 in particular the value is consistent with 0.4, a value commonly associated with AGNs (Osterbrock 1989). It is tempting to say the disky Bins 4–7 also show a decline from Bin 3, and if so then the various line ratios are not falling off as rapidly as in Bins 1–2. We speculate that there is a moderately elevated metal fraction in Bins 4–7 that may be a result of increased star formation activity in the direction of the galaxy tail, indicating possible harassment in the galaxy's star formation history.

If Bins 4–7 indeed lead toward a galaxy tail, then one can ask the question: which are the likely culprits for past interactions? F12_ELG1 has neighbors roughly centered in redshift at the systemic redshift of z = 0.7895. There are seven galaxies with spectroscopic redshifts of 0.7625 <z < 0.8175 distributed over a field in the image plane of ≈3.5 arcmin on a side, which for a mean magnification of ∼4 × yields a physical size in the source plane of 0.8 Mpc. This structure or filament may contribute to the compound lensing effect at the 1'' level that is important for doing precision cosmology with clusters (Jullo et al. 2010).

In sum, with its high sSFR corresponding roughly to a cold gas fraction of ∼0.7 (Reddy et al. 2005), its low mass, and the presence of many Balmer series ELs, this strongly lensed but otherwise ordinary galaxy would appear to be caught at the beginning of a major burst of star formation. This is consistent with the picture of van der Wel et al. (2011) in which ∼1/2 the stars in a typical field galaxy are formed in only ∼2–3 bursts that produce the stellar population in an M_* = 10⁹ M_☉ by z ∼ 1. F12_ELG1 also looks similar to the objects in Kriek et al. (2009) that are AGN hosts of size ∼1 kpc.

F12_ELG1 has similar ionization properties compared to other z = 1–2 star-forming galaxies, and as a group such objects have elevated ionizations compared to the local galaxy population. It is thought that galaxy feedback must play a role in understanding these differences. One possible explanation is that many intermediate-redshift galaxies may harbor weak AGNs (Wright et al. 2010; Trump et al. 2011). F12_ELG1 has elevated metal-line ratios consistent with an AGN that are interestingly offset from the dynamical center of the galaxy. Given this extended tail in the same direction, we interpret the asymmetrically situated AGN-like region of this galaxy as shock excitation, possibly as a result of a past galaxy interaction. The line ratios decrease toward the outer disk on both sides of the peak in this one galaxy, similar to the spatially resolved spectroscopy of another giant arc at intermediate redshift, the "Clone" (Jones et al. 2010). By contrast, in a sample of 50 intermediate-redshift ELGs with integral field unit spectroscopy a significant fraction showed the opposite trend (Queyrel et al. 2012).

We report new ELSs detected with the ACS G800L grism which have close angular separations. They are ELS 20004 "A" and ELS 11085 "B" (Figure 12). We designate this group of sources as "F12_ELG2." Sources A and B have an angular separation of ≈021 and are resolved despite the angular separation being slightly smaller than our predicted spatial resolution of R₀ = 025, set by our algorithm that extracts a minimum of five spatial pixels centered on an EL. The spectrum for component A shows two EL peaks with similar flux amplitudes. The spectrum for component B shows a single EL with an extended red tail that appears to be associated with at least some of the same emission sources as in component A (Figure 13).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Emission peaks B and G both have line emission at the same wavelength (see 2D spectrum in Figure 13). In addition, component G shows extended emission, making it the source likely to be most sensitive to the photometric redshift estimate. The photometric redshift for G is z_BPZ = 0.480 ± 0.145, from which we consider the peak "B+G" as [O iii] at z = 0.532. As it is unlikely for components B and G to have the same wavelength peak but be unrelated, we take this as the redshift for both components. The morphology of the galaxy nearest to the sources is an elliptical which is likely component G (see Figure 12). From profile fitting we do not find associated Hβ. Our double-peaked spectrum (A+G) has a peak-to-peak velocity separation of ∼9600 km s⁻¹. From this remarkably high value we can rule out the interpretation of this object as two separate ELSs close in space. Most likely, G and B are both at z = 0.532, while A is an unrelated source at z = 0.560. It is also possible that source A and/or B are situated behind the elliptical galaxy, which would alter the above redshift identifications. Additional spectroscopy is required at a competitive spatial resolution to identify these emission features and objects.

We detect the highest redshift object with a published spectrum in the A1689 field, ELG 10399 at z = 5.13 (Frye et al. 2002). Our grism spectrum is shown in the upper panel of Figure 14. This arclet is faint (i_AB = 25.85 ± 0.18) and extremely red (g − i > 4), yet is detected in our rather shallow grism survey owing to its large rest-frame equivalent width (W_r = 29.7^+13.29_−4.6), its high magnification, and its small size of 1'' (Figure 14). We compute a magnification of μ ≈ 4.5 and estimate an intrinsic size for its redshift of ≈1 kpc. There are no counterarcs predicted for this arclet. We identify the lone EL to be Lyα based on our higher resolution companion spectrum taken with Keck LRIS (blue dashed line) which shows also a Lyman-series break. With its high rest equivalent width of W_r > 20, this object would appear from the grism spectrum alone to be an Lyα emitter, but our Keck spectrum shows it to have a Lyman-break seen against stellar continuum.

Zoom In Zoom Out Reset image size

Of the six galaxies in the field of A1689 with z >2.5 (Table 4), interestingly only this one galaxy at z = 5.13 enters into our sample. The Sextet arcs at z = 3.038 are not detected, which is not surprising given that the arclet in this sextuply imaged system with the strongest line emission has an Lyα total of W_r = 4^+1.5_−5.0 Å, a value in the lowest single percentile of our sample. It is less well understood why the bright arclet at z = 4.868 is not detected. This arclet image has i₇₇₅ = 23.48 ± 0.03, an angular size of ∼13, an Lyα EL with W_r = 12.4^+8.83_−3.84 Å (the lowest 10% of our sample), and is situated in a relatively uncrowded location relative to the cluster members. Note we do not see any lone unassociated ELs of other faint, potentially high-redshift objects near to the locations of the tangential critical curve.

Previously published redshifts are the first resource for line identifications. For new ELGs, the objects with multiple ELs make the redshift determination straightforward, taking into account the cases of similar ratios of wavelengths between line pairs, such as the similar line ratios of λHα/λ[O iii] and λHβ/λ[O ii] (M07). The remainder of the catalog consists of single ELs. Single ELs present challenges as grism observations are not sensitive to small-scale changes that can serve as redshift-confirming features, such as strong absorption bands and the line shape of ELs closely separated in velocity space. The treatment of single ELs relies on the combination of photometric redshift, profile fitting, and other sanity checks such as the search for continuum depressions and the absence of conflicting emission features. The photometric redshift is derived from a Bayesian approach, described in detail in other papers (Benítez 2000; Benítez et al. 2004; Coe et al. 2006); it is measured including nine bands and found to be a robust redshift indicator (see Figure 9 in Frye et al. 2007). We fit profiles to distinguish secondary bumps in our single ELs to corroborate the likelihood of, for example, a bona fide [O iii]/Hβ detection. When one EL is observed, the photometric redshift is used to resolve the degenerate possible redshift solutions, resulting in a much more accurate redshift than from the photometry alone.

The largest source of uncertainty is in establishing the goodness of fit of the zero point of the wavelength solution. In our blind EL finding technique, ELs are identified and the offset of the position between the EL and the direct source is computed. The goodness of fit can be measured by comparing the grism and spectroscopic redshifts. From previous studies involving similar data sets and reduction techniques as described in M07, the scatter in grism and spectroscopic redshifts is equivalent to an error in position of ∼1 pixel, or Δv = 1650 km s⁻¹. In addition, the spectral resolution is low, with R = 90 at λ = 8500 Å, corresponding to Δv = 3300 km s⁻¹ or about 2.1 pixels. While in practice this means that we will have relatively little leverage on measuring kinematics of gas clouds within a disk, the grism has the advantage of obtaining spectroscopy that is spatially resolved on impressively small angular scales of ∼02

We compare line properties sorted by the three most common line species, Hα, [O iii], and [O ii] (Table 5), and compare these results with a recent HST ACS grism survey in an unlensed field to similar depth, the HDF grism survey (M07). By line species, Hα emitters are plentiful despite the small sample volume owing to the field selection of a massive galaxy cluster. The median redshift for our sample of HAEs of z = 0.22 is close to the cluster redshift of z = 0.187 and is dominated by galaxies in the cluster. We measure the median W_rs to be W_Hα = 29.3 Å, W_[O iii]= 67.2 Å, and W_[O ii] = 29.6 Å. These values are large, but typical of grism surveys.

The largest difference between this sample and the HDF is in the values of the W_r. One-quarter of all of the emitters in our sample have W_r ⩾ 100 Å, compared to three-quarters for the comparison sample (HDF-N, M07). Also, the median W_r for each line species is smaller than the values in the comparison sample (Figure 16). The largest shift is found for the HAEs, which are dominated by galaxies in or near to this old and presumably relaxed cluster (z = 0.19). Interestingly, all of the HAEs in our sample that are brighter than the magnitude limit of the large EL survey of Balogh et al. (2002) of I = 19.3 are recovered in our census. Moreover, all of the HAEs in our sample that are fainter than their limiting magnitude are new to the literature. Thus, we find the slitless grism observing approach to be especially sensitive to the detection of low-luminosity (and largely unlensed) HAEs with low SFRs. On the other hand, the shift in the [O iii] and [O ii] emitters toward lower W_rs is due to lensing. Lensing boosts the brightnesses and sizes of the background objects, including the ELs, with the effect of improving the signal-to-noise of the grism spectra. This effect enables the identification of weaker ELs with smaller W_rs and lower SFRs than are found in the field. We compute rest-frame B-band magnitudes, M_B, by fitting a Bruzual–Charlot model to our optical ACS photometry, and then computing the k-correction onto a Johnson B-band filter template. The mean M_B for the 27 [O iii] emitters is fainter than for the HDF by 1.5 mag, yielding yet another indication of the gravitational lensing effect. The shift is not seen for the 21 Hα emitters or nine [O ii] emitters which is dominated by emitters inside the cluster lens and small number statistics, respectively.

Zoom In Zoom Out Reset image size

Seven ELs flagged by our purpose-built reduction code were later removed as bogus detections. In five cases, the lines were weak and other available spectroscopic data ruled out the grism features as ELs. In two cases, the continuum shapes mimicked an EL but turned out to be stellar continuum. There are three cases for which the EL redshifts are not consistent with published redshifts: ELS 1184, ELS 11324, and ELG 6621. We discuss these cases below.

• 1.  
  ELS 1184 is in a crowded environment and is separated by only 18 from ELS 1157 (see upper left of Figure 14). ELS 1157 has multiple emission features confirming the redshift of z = 0.195 for this previously identified object (Balogh et al. 2002; Duc et al. 2002). ELS 1184 has a single EL at 6917 Å that we do not identify with any strong emission feature at or near to z ∼ 0.195. We adopt the foreground redshift z = 0.054 for this compact and very bright ELS based on the absence of other expected ELs in the bandpass were the identifications to correspond to [O iii] or [O ii]. We measure a photometric redshift for the Sc galaxy of z_BPZ = 0.320 ± 0.129, but this measurement is taken at the position of ELS 1157.
• 2.  
  ELG 11324 is in the crowded outer regions of an extended face-on spiral galaxy with a photometric redshift of z_BPZ = 0.250^+0.260_0.122 and a spectroscopic redshift of z = 0.384 (J. Richard 2011, private communication). We detect a single EL at 6913.5 Å which we take to be [O iii] at the spectroscopic redshift.
• 3.  
  ELG 6621 is small and faint, with a magnified size of 03 and i₇₇₅ = 26.7. The grism spectrum shows a lone EL at 6712.4 Å with stellar continuum and no salient continuum features. This galaxy suffers from crowding by a second object that is confidently identified from our companion Magellan spectrum as ELG 6621a at z = 0.1868. ELG 6621a is a cluster member that is not detected in our grism data set and that has no feature at the position of the grism EL. The photometric redshift for ELG 6621 is z_BPZ = 3.780 ± 0.468. If the EL in ELG 6621 is Lyα then the redshift is z = 4.52, and we cannot rule out the possibility of high redshift at least according the redshift test available to us (Rhoads et al. 2009). The high-redshift scenario would seem unlikely given that the photometry does not correct for extinction at the cluster redshift, and thus we must attribute some of its "redness" to its being behind the foreground cluster member. Strictly, the EL can be also Hα, [O ii], or [O iii]. If the line is [O iii] then z = 0.346, in which case we do not detect the non-requisite Hβ EL. Based on the large observed equivalent width of 772.9 ± 497.9 we adopt a line identification of [O ii] at z = 0.800, and admit this as an uncertain redshift in our sample. Note ELG 6621a is detected only in our Magellan spectrum and so does not appear in Table 1.