The ISM and stellar populations of a LBG at z=3.8

Medium-resolution spectroscopy of FORJ0332-3557: Probing the interstellar medium and stellar populations of a lensed Lyman-break galaxy at z=3.771


We recently reported the discovery of FORJ0332-3557, a lensed Lyman-break galaxy at in a remarkable example of strong galaxy-galaxy gravitational lensing. We present here a medium-resolution rest-frame UV spectrum of the source, which appears to be similar to the well-known Lyman-break galaxy MS1512-cB58 at . The spectral energy distribution is consistent with a stellar population of less than 30 Ma, with an extinction of mag and an extinction-corrected star formation rate of 200–300. The Lyman- line exhibits a damped profile in absorption produced by a column density of about  cm, superimposed on an emission line shifted both spatially (05 with respect to the UV continuum source) and in velocity space ( with respect to the low-ionisation absorption lines from its interstellar medium), a clear signature of outflows with an expansion velocity of about 270 . A strong emission line from Heii nm indicates the presence of Wolf-Rayet stars and reinforces the interpretation of a very young starburst. The metallic lines indicate sub-solar abundances of elements Si, Al, and C in the ionised gas phase.

galaxies: high-redshift – galaxies: starburst – galaxies: ISM – galaxies: abundances – galaxies: evolution – gravitational lensing – galaxies: stellar content

1 Introduction

Very few high-redshift galaxies () have been observed spectroscopically at medium and high resolution until now because they are generally too faint for 8/10-m class telescopes equipped with medium resolution spectrographs. Over the past years the most successful technique to probe these high-redshift galaxies consisted in building composite spectra of low signal-to-noise samples (Shapley et al., 2001, 2003; Noll et al., 2004) and redshift-selected sub-samples (Steidel et al., 2001; Ando et al., 2004). Very recently, high-redshift gamma-ray bursts have also led to a wealth of low-resolution spectral data on the interstellar medium of young star-forming regions in high-redshift galaxies (Berger et al., 2006; Fynbo et al., 2006; Vreeswijk et al., 2004, 2006). From these composite samples a picture emerges of very young and massive stellar populations of 1/10th- to 1/4th- solar metallicities, dominated by on-going star formation, with strong outflows, and a dusty component within abundant neutral gas.

As of today, only a handful of galaxies are bright enough to be studied spectroscopically at medium and high resolution, and among them the most well-known example is MS 1512-cB58 (cB58) (Pettini et al., 2000, 2002; Savaglio et al., 2002, ; look-back time=11.4 Ga). A more recent example is the 6 starburst (Dow-Hygelund et al., 2005) discovered in the field of a galaxy cluster. The lensed galaxy FORJ0332-3557 (Cabanac et al., 2005, , look-back time=12.1 Ga), which is the object of the present study, is the most recent addition to this select few8. Because the galaxy is magnified 13 times by a foreground lens at , it is bright enough to be observed at medium to high spectral resolution. We present here a 9h-deep medium-resolution spectrum of the lensed galaxy taken with VLT/FORS2. The observations and data reduction are presented in Section 2 while Section 3 outlines the properties of the emission and absorption lines and in Section 4 we analyse the spectral energy distribution and the underlying stellar populations.

Throughout this paper we assume a flat FRW metric with , and a Hubble constant normalised at  km s Mpc. Following the recommendations adopted by the IAU, the letter a is used for the non-SI unit of year.

2 Observations and data reduction

The observations were carried out with FORS2 on the European Southern Observatory VLT, under ESO programs 74.A-0536 and 78.A-0240. A total exposure time of 8.71 hours, split into 61495 s (1slit) and 161400 s (08 slit), was obtained (see Table 1). We used the holographic grism (0.163 nm pixel) together with the order sorting filter, which results in resolving powers of 1200 and 1000 for the 08 the 1 slits, respectively. Since the 08 the 1 slits are slightly shifted with respect to each other in the focal plane, the resulting spectral coverage, 529.7-862.8 nm for the 1slit and 501.1-832.5 nm for the 08 slit, differ slightly. The detector was the upgraded MIT 240962048 mosaic, which with the standard resolution collimator yields a physical scale of 0126 pixel. We oriented the slits North-South and centered them on the brightest part of the main arc (see Figure 1). We used the standard 22 binned mode, which results in a final physical scale of 0252 pixel and a sampling of 4 and 3.2 pixels for the 1 and 08 slits, respectively. We placed the source on CCD1 (the upper northern chip), and used a 20nod-on-slit strategy that allowed us to remove most sky artefacts.

The science data were taken during seven nights (2 nights for 1 slit, 5 nights for 08). The ESO standard calibration scheme delivers a set of bias, flatfield and wavelength calibration (from an He-Ar arc) frames for each night. We, then, subtracted master biases and divided by the 2-D normalised master flatfields. We evaluated the fringing contamination to be ca. 1% peak-to-valley in the most sensitive part of the frames.

We performed two independent reductions on the 1 slit dataset. Firstly, we reduced separately each science exposure and recombined the resulting one-dimensional spectra after all calibrations. Secondly, we combined all science exposures using the FITS header World Coordinate info (WCS keywords) and removed the instrumental and sky artefacts on the combined 2-D image before extracting and calibrating the source spectrum.

In the first method we used standard NOAO/IRAF routines in order to extract the trace with a 4th-degree Legendre polynomial in the dispersion direction, weighted by the spectrum variance perpendicular to the dispersion direction. We removed the sky emission from 1-D modelling of the sky emission as measured on both sides of the trace. We then calibrated each extracted trace with the wavelength calibration frames (1st order cubic spline) respecting the source spectrum slit position. FORS2 flexures are very small and the final rms of 0.02 nm derived for the wavelength frame was assumed to apply to the science exposures as well. All wavelengths were shifted to the heliocentric reference frame. Finally the combination of the resulting 6 1-D spectra (1 slit data; Table 1) was done using a median algorithm rejecting the minimum value and two maxima.

In the second method the sky substraction was done using an optimal fitting while combining the 2-D spectra. The wavelength calibration frames proved to be stable between the two nights (1 slit data; Table 1). Figure 2 shows the resulting 2-D frame after sky substraction.

Both methods lead to a similar final S/N and wavelength calibration accuracies across the useful wavelength range. The second method seems to lead to a slightly larger convolution of the observed line widths than the first method. We reduced the 08 slit data using the second method only.

An important calibration step is the instrument response correction. The response curve of FORS2 was computed from spectrophotometric standard observations (LTT1788 and LTT2415 using a MOS 5slit). Because the 08 and 1slits are shifted with regards to the MOS slit, the wavelength range of the standard covers a slightly different range. We extrapolated the response curve linearly in the missing regions.

A final step was required to remove the contribution of the elliptical galaxy that produces the gravitational lens effect on the 2-D spectrum of the source. We used two methods, which gave similar results. First, we subtracted the contribution of the lens measured on the southern profile at a symmetrical angular distance. Second, we measured the dominant IR contribution of the lens over the arc and subtracted the best fit SED normalized to that contribution (cf. section 4).

Figure 1: VLT/FORS2 acquisition frame (in the band) and 1-slit mask position on FORJ0332-3557: the central object is the lens at and the top arc is the lensed Lyman break galaxy at .
Date UT time exp. time airmass seeing slit
y-m-d h:m:s s -
2004-11-16 01:02:19 1495 1.446 0.56 1.0
2004-11-16 01:28:01 1495 1.326 0.55 1.0
2004-11-16 01:54:29 1495 1.231 0.65 1.0
2004-11-16 02:20:09 1495 1.161 0.76 1.0
2005-01-30 01:20:45 1495 1.103 0.50 1.0
2005-01-30 01:46:26 1495 1.154 0.54 1.0
2006-10-16 06:02:29 1400 1.028 0.66 0.8
2006-10-16 06:26:57 1400 1.020 0.53 0.8
2006-10-16 06:51:33 1400 1.021 0.61 0.8
2006-10-16 07:15:52 1400 1.031 0.50 0.8
2006-10-16 07:55:52 1400 1.068 0.73 0.8
2006-10-16 08:20:31 1400 1.104 0.61 0.8
2006-10-21 07:37:59 1400 1.070 0.48 0.8
2006-10-21 08:02:09 1400 1.106 0.61 0.8
2006-11-21 07:03:26 1400 1.266 0.85 0.8
2006-11-22 04:33:11 1400 1.023 0.85 0.8
2006-11-22 04:57:20 1400 1.036 0.89 0.8
2006-11-28 04:24:07 1400 1.030 0.54 0.8
2006-11-28 04:48:36 1400 1.048 0.63 0.8
2006-11-28 05:20:59 1400 1.087 0.50 0.8
2006-11-28 05:45:14 1400 1.130 0.57 0.8
2007-01-25 04:30:28 1400 1.916 0.67 0.8
Table 1: Description of spectroscopic data with VLT/FORS2.
Figure 2: Continuum-normalised rest-frame ultraviolet spectrum of FORJ0332-3557 (black solid line), superimposed on its error spectrum (blue solid line, offset for clarity) and the spectrum of the lensed galaxy cB58 (red solid line). The 2-D spectrum of the source, part of a compact lensing configuration, is also shown on the top.

3 Emission and absorption lines

3.1 Source redshift

In the low-resolution discovery spectrum (Cabanac et al., 2005) we estimated a source redshift of from a few strong interstellar lines. The improved signal-to-noise obtained here (continuum peak S/N15) allows us to measure here a revised systemic redshift of , using 11 stronger detections of 5 atomic species : C, O, Al, Si, and Fe, and excluding uncertain identifications and blended lines (see Table 2). We marginally detect two important photospheric lines O iv) and S v (see Fig. 2). Hence we can only make a very tentative estimate of a possible velocity offset between the photospheric lines and the low-ionisation ISM lines of about . This is about half the value inferred in cB58 of (Pettini et al., 2000, 2002).

3.2 The Lyman- lines

Figure 2 shows the 2-D spectrum and the associated 1-D spectrum normalised to the continuum. Bluewards of Ly the spectrum is poorly defined mostly because of foreground hydrogen absorption, and the continuum has been assumed to be located at the peaks of the spectrum. We also added the low-resolution VLT/FORS2 spectrum at wavelengths shorter than 110 nm (Cabanac et al., 2005). The Ly, Ly, and Ly absorption features are conspicuous, and unresolved Ly absorption is detected as well.

Figure 3: Zoom on the 2-D spectrum around Ly, showing the spatial offset of pixels ( arcsec) and the velocity shift of , indicative of outflows (121.567 nm position is the narrow green vertical line). The lower trace is the spectrum of the elliptical galaxy that produces the gravitational lensing effect.

Figure 3 zooms on the region around Ly. A faint emission peak at  km/s of the Ly absorption centroid and pixels () north of the continuum emission is clearly visible. This feature can be interpreted either as a very bright Hii region in the periphery of the main galaxy, or, more probably, as the signature of an expanding outflow. Assuming that the results of detailed 3-D Monte Carlo Ly radiative transfer codes Verhamme et al. (2006) (where the Ly  emission is produced by an expanding shell that is being ionised) can be applied to this case, we can infer that the expansion velocity of the shell that gives rise to the Ly emission is constrained to be between 800 (low column density case) and 270 (large column density case). It is therefore essential to measure the Hi  column density in this context.

If we assume that the Ly absorption can be fitted by a damped Voigt profile9, we derive a column density of cm. Figure 4 shows the best-fit damped Voigt profile and errors, along with the emission profile in velocity space obtained when the absorption profile is subtracted. The column density is remarkably large, about 4 times larger than in MS1215-cB58 (Pettini et al., 2002; Savaglio et al., 2002). If we adopt the model of an expanding shell at constant velocity (Verhamme et al., 2006), this column density still implies a large expansion velocity of about km/s, similar to the one found in cB58 (Schaerer & Verhamme, 2008), and at the upper end of the Lyman-break galaxies (Shapley et al., 2003) if they do have large Hi  column densities. It is remarkable that in spite of the very large column densities, such strong Lyman- emissions are detected, confirming that the interstellar medium is very clumpy, yielding smaller effective optical depths which are very similar to the ones inferred in the low-redshift starbursts (Valls-Gabaud, 1993), and allowing many Ly photons to escape.

Figure 4: The best-fit Voigt profile corresponding to a column density of  cm (solid and red dotted lines) is shown on the rest-frame spectrum (top frame). The bottom frames show the same fit in velocity space . On the panel at the bottom right the model is subtracted out of the spectrum, showing a conspicuous emission feature peaking at  km/s. This feature is also offset spatially by 2 pixels (ca. 05) from the main UV-continuum emitting source (see Fig. 3).

The neutral gas content derived with Voigt profile fitting is fully consistent with the neutral gas content derived independently from the analysis of stellar populations (Section 4) with the gas-to-dust ratio taking Milky Way or LMC values, but is 5–10 times smaller than the values observed in the SMC.

3.3 The Helium ii 164.04nm emission line: the signature of very massive stars

A broad emission feature at 164.04 nm is seen in Fig. 2 at a redshift of consistent with other faint photospheric features. We associate this feature with the well-known He ii emission produced by Wolf-Rayet stars. Although not present in cB58, this feature has already been detected in absorption in young LBGs (Shapley et al., 2001, ; cf. Fig. 15 and 16), and in emission in the composite LGB spectrum of Shapley et al. (2003) but the origin of the feature cannot be asserted by the authors because the composite spectrum mixes a wide range of stellar populations and ages. If FORJ0332-3557 is similar to other LBGs, it belongs to the strongly absorbed Ly quartile, and the case for a very young stellar population (5 Ma) of massive stars is strong. The He ii feature of Fig. 2 shows the typical asymmetric profile produced by strong stellar outflows of the evolved descendants of O stars more massive than –-. The equivalent width yields the ratio of WR to O stars following Schaerer & Vacca (1998),


which for nm  implies a ratio O/WR. In turn, this value sets a strong age limit on the stellar population of 6 Ma rather independently of metallicity (Cerviño & Mas-Hesse, 1994; Schaerer, 2003). Stochastic effects due to the sampling of the stellar mass function (e.g. Cerviño & Valls-Gabaud (2003); Cerviño et al. (2003)) imply that only the most recent burst can be dated through this short-lived feature.

3.4 Interstellar absorption lines

The spectrum of FORJ0332-3557 shows characteristic absorption lines from starburst galaxies. They are summarised in Table 2. The medium-resolution spectrum of the LBG cB58 Pettini et al. (2000) is shown in Fig. 2 in red, superimposed on our source, for a direct comparison of the insterstellar features. It is immediately obvious that the FORJ0332-3557 source is qualitatively similar to cB58.

All common interstellar absorptions are found in FORJ0332-3557. The strong absorption features include low-ionisation lines associated with neutral gas (Si ii 126.04 130.47 152.67, C ii 133.45, O i 130.22, Al ii 167.08, Fe ii 160.84) and high-ionisation lines associated with a hot gas phase (Si iv 139.38 140.28, C iv154.82 155.08). Table 2 lists the ion line identification, vacuum rest-frame wavelength , observed wavelength , redshift , rest-frame equivalent width , oscillator strengths , column density, ion abundance with respect to solar [] and comments. Additional uncertain identifications are question-marked, the lines noted i? might belong to interlopers at unknown redshift(s). We emphasize that the derived are very sensitive to both sky subtraction and continuum normalisation, hence the systematic errors caused by the continuum normalisation have tentatively been computed and are shown to be close to photon counting errors, while the sky subtraction error is much more difficult to quantify. As a sanity check, we computed the equivalent widths of absorption lines in the spectrum of cB58 and found our measurements to be fully consistent with those published by Pettini et al. (2002).

No nebular emission lines are detected in the present spectrum other than He ii 164.04nm (§ 3.3). A weak detection of C iii] 190.87 seen on a previous spectrum (Cabanac et al., 2005) suggests that contamination of the high-ionisation lines by nebular emission is present but small. P-Cygni profiles are visible on C ii 133.45, and C iv155.08.

There are several ways to derive the abundances in the interstellar medium of distant galaxies (Spitzer, 1978; Pettini et al., 2002; Savaglio et al., 2002). Ideally one should build a curve of growth by fitting Voigt profiles and Doppler parameters for all ions independently. Because the resolution of our observed spectrum is just under the resolution one needs for Voigt profile fitting, and is penalised by a low signal-to-noise ratio, most of the strong lines appear saturated, and most weak lines are dominated by noise.

A careful analysis of the ISM metallicity goes beyond the present paper and will be done elsewhere. Here we present only qualitative arguments on the curve of growth, and Doppler parameters . Assuming that the interstellar medium in FORJ0332-3557 is optically thin, one can infer lower limits to column densities, log([cm]), and abundances (given in Table 2) by taking the optically thin approximation


where is the line oscillator strength. The equivalent width, , and the wavelength, , are in nm. A tentative curve of growth indicates that the ion abundances could be 2-3 dex larger for a Doppler parameter of . In this context, the most constraining line, besides Siii153.3 which may be blended, is Feii160.8, which appears unsaturated and whose small equivalent width is similar to the one measured in cB58 and would yield  , similar to the reported in cB58 (Pettini et al., 2002).

Compared to cB58, FORJ0332-3557 are lower by factors of 2-3 (C iv155.08 155.08, Al ii167.08, O i130.22) to a factor of 1-1.2 (Si iv139.38 140.28, Fe ii160.84). More detailed analyses on elemental abundances and depletion in the interstellar medium of FORJ0332-3557 will be presented elsewhere.

Ion redshift 10 log11 []12 Comments
nm nm nm cm
H i Ly 97.254 - - 0.0290 - -
H i Ly 102.57 - - 0.07912 - -
H i Ly 1215.7 - - - 0.4164 -
He ii 164.04 783.09 3.7738 - - - WR feature
C ii 133.45 637.01 3.7732 0.1278 blended with C ii*133.57
C iv 154.82 738.74 3.7716 0.1908
C iv 155.08 740.00 3.7718 0.09522 sky contamination
N i? 572.70 3.7718 0.04023 triplet N i 119.95 120.02 120.07
N iii? 132.43 632.24 3.7741 - - - - blended with C ii132.39
O i 130.22 621.31 3.7714 0.04887 blended with Si ii130.44
O iv 134.34 641.27 3.7735 - - - photospheric
Al ii 167.08 797.41 3.7726 1.833 sky contamination
S i/i? 138.16 659.29 3.7719 - - -
S ii/i? 125.38 598.57 3.7740 0.01088
S ii/i? 125.95 601.31 3.7742 0.01624 - - blended with Si ii 126.04
S v? 150.18 716.39 3.7741 0.00545 - - photospheric
S ii/i? 151.12 721.43 3.7740 - - - blended with Si ii151.21
Si ii 126.04 601.84 3.7749 1.007 blended with S ii125.95
Si ii 130.47 622.49 3.7710 0.094 blended with O i130.22
Si ii 152.67 728.39 3.7710 0.130
Si ii*? 153.34 731.40 3.7697 0.132
Si iv 139.38 665.09 3.7718 0.5140
Si iv 140.28 669.38 3.7718 0.2553 blended with S i140.15?
Fe ii 160.84 767.68 3.7728 0.058
Table 2: Interstellar absorption lines

4 Stellar populations

After analysing the absorption and emission lines here we study the properties of the continuum, using recent stellar synthesis codes, which, for the first time, predict high-resolution spectra of starburst populations. We used both starburst99 (Leitherer et al., 1999; Vázquez & Leitherer, 2005) and sed@ (Buzzoni, 1989; Cerviño et al., 2002; Cerviño & Luridiana, 2004) in an attempt to derive plausible combinations of ages, metallicities and extinctions by dust from the FORJ0332-3557 ultraviolet spectrum (Fig. 2).

4.1 Spectra predicted by starburst99

We build two series of starburst99 spectra using (1) an IMF with a Salpeter slope from 1-100 M13, (2) a self-consistent association of low metallicity tracks with LMC/SMC atmospheric stellar templates, and (3) solar metallicity tracks with galactic atmospheric stellar templates, from ages 0.01 Ma to 50 Ma, for metallicities of 0.001 (1/20), 0.004 (1/5), 0.008 (2/5), and 0.020 (). All starburst99 models were fitted to the spectrum of Fig. 2, pre-filtered with a Savitsky-Golay filter (conserving the equivalent-width, FWHM and line profiles), to the same resolution as the starburst99 synthetic templates. The region below 123 nm was excluded from the fits, because the source is heavily contaminated by foreground interlopers. For solar metallicities, starburst99 high-resolution atmospheric templates go up to 180 nm, but only 160 nm for subsolar SMC/LMC metallicities. Only the highly ionised absorption lines were considered for the fitting, to avoid excessive interstellar contamination. An additional range of wavelengths (near 137 and 144 nm) was excluded due to telluric contamination. The regions included in the fit are shown as solid lines in Fig. 5. Different sets of constraints can be derived by fitting the line equivalent widths of the normalized spectrum and by fitting the shape of the spectrum continuum. Fig. 5 compares the best fit models with the normalised spectrum for two extreme scenarios of star formation: (i) a single burst of star formation (bottom) and (ii) a continuous SFR (top). The best fit models for =0.004 have ages of 8 Ma (single burst) and 29 Ma (constant SFR). The age of the best fit models is found to be insensitive to metallicity in the case of single bursts. In the case of a constant SFR, increasing the metallicity tends to increase the age of the best fit model : 29 Ma at =0.001,  Ma at =0.008, and more than 50 Ma at =0.02. We explored solar metallicity models with ages of 200 Ma and although the reduced decreased slightly the highly-ionised absorption lines did not get deeper significantly. The reduced shows that constant SFR models become insensitive to age above 20 Ma, whereas single burst models appear to isolate a well-defined small region (6-9 Ma). However, neither of these scenarios can reproduce the observed depth of the absorption lines, which probably argues for interstellar origin for the highly ionised lines. Finally, the single burst scenario seems to produce slightly deeper absorption lines, whereas the constant SFR scenario better fits the P-Cygni profiles of C iv155.08 155.08 nm and the red side of Ly. We also tested different stellar wind models without any significant differences.

Figure 5: The best fit starburst99 models (red lines) are overlaid on the normalised spectrum of FORJ0332-3557 source (blue line) for two extreme scenarios of star formations with a Salpeter IMF from 1-100 M, at a metallicity =0.004 (1/5 ). The parts of the spectrum excluded of the fit are shown as dotted lines (see text). The top frame shows a constant SFR for 29 Ma. Ages older than 20 Ma are strongly favoured. The bottom frame shows an 8 Ma-old instantaneous burst of star formation. No model (at any available metallicities (1/20–1 ) can reproduce the observed depths of the lines. For a constant SFR, larger metallicities produce older ages (29 Ma at = 0.001, 49 Ma at = 0.008). Instantaneous burst models tend to produce deeper lines while constant SFR models tend to better fit the observed P-Cygni profiles of C iv155.08 nm.

4.2 Spectra predicted by sed@

Figure 6: Reduced contours are given for the best fit =0.008 sed@ models for a grid of ages and extinctions. The contour levels are arbitrarily chosen to be 1, 3 and 10% of the reduced minimum, to outline the trends. The left panel shows the contours derived for a continuum-normalised set of deep absorption lines alone (cf. text), the center panel shows the constraints derived from the photometric colours alone, and the right panel shows the combined contours of the two independent sets taken together. The optimal parameter set is age=20 Ma, .

The stellar population synthesis code sed@ is included in the Legacy Tool project of the Violent Star Formation Network14 (Buzzoni, 1989; Cerviño et al., 2002; Cerviño & Luridiana, 2004). It computes UV spectra with the highest resolution currently available, consistently computing the errors introduced by a limited number of bright stars in a small burst (this effect can probably be neglected in our case unless the lens amplifies a single star-forming region within the source). We use a series of instantaneous burst models from 0 to 10 Ga, a Salpeter IMF, and Geneva stellar tracks at 4 metallicities: of 0.001 (1/20), 0.004 (1/5), 0.008 (2/5), and 0.020 (). As in the previous section, we treat independently the constraints coming from the normalised UV absorption features, and the overall photometric SED of the source. For the UV-spectrum fitting, we isolate the strongly-ionised absorption features containing most of the signal and exclude the weakly-ionised features dominated by interstellar features, using the wavelength intervals 120–125, 127–129, 131–133, 134–143, 145–152, and 153–158 nm. We normalise each interval to the continuum by dividing each wavelength bin by the linear interpolation between the two extreme values of the interval. We compute the reduced between the templates and the source UV spectrum for a grid of extinctions (), and ages (1-1000 Ma). The photometric SED of the templates are integrated over the instrumental transmission curves (filter+telescope+CCD detector) given by VLT/FORS1 (, and bands) and VLT/ISAAC (, , bands) manuals, normalised to VEGA magnitude using Kurucz’s solar model (Castelli & Kurucz, 1994). Figure 6 shows the resulting reduced contours. Ages are along the horizontal axis (steps of 1 Ma from 0 to 10 Ma, steps of 5 Ma from 10 to 100, steps of 100 Ma from 100 Ma and older ages). The extinction is given on the vertical axis. The left panel shows the contours of values obtained in the fitting of the UV features described above (the contour levels are given at 1, 3 and 5% above the value of the minimum ). The central panel shows the contours obtained in the fitting of the photometric SED (contour levels at 1.5, 3, and 5 times the minimum ), while the right panel shows the combined constraints from both the overall SED and the UV features (contour levels at 1  3, and 10% above the minimum ). The left frame shows a complete degeneracy in extinction, as expected since we normalise the continuum and the UV is no longer sensitive to extinction. The central panel shows the well-known degeneracy between age and extinction but clearly shows a tendency to prefer young ages with higher extinction, rather than older ages with small extinctions. It should be noted that the of the UV fitting alone shows a minimum at =0.004, whereas the of SED-fitting alone favours the smallest metallicity models =0.001. The combined set leads to =0.008 (Fig. 6) but it is clearly not a very constraining result due the degeneracies involved15. Because the two sets of constraints are independent and almost orthogonal, the combined shown in the right panel of Fig. 6 (each is given the same weight) gives a tighter constraint on the age- parameter space. The combined best fit parameters (, age [Ma], [mag]) are (0.001, 15, 0.62), (0.004, 20, 0.54), (0.008, 20, 0.54), (0.020, 10, 0.64), (0.040, 15, 0.52) for reduced values of 0.853, 0.800, 0.788, 0.792, 0.789.

In conclusion, both stellar synthesis codes give a consistent picture of the global properties of the source galaxy in FORJ0332-3557, although they marginally disagree on details. They both find young stellar populations of  Ma and the presence of some dust extinction. Neither the metallicity, nor the IMF can be robustly constrained although sed@ marginally favours =0.008. We emphasize that none of the models, at any metallicity and age, can fit the depth of the observed strongly-ionized absorption lines. starburst99 points to a young stellar population of 8–10 Ma, in the case of an instantaneous burst, and a few tens of Ma, in case of a constant SFR, whereas sed@ models seem to favour slightly older ages (10–20 Ma) for an instantaneous burst. Lastly, starburst99 tends to give a better fit of the absorption features than sed@, including interstellar features of course because sed@ uses synthetic stellar spectra whereas starburst99 uses observed stellar templates where the Milky Way interstellar absorption features have not been removed. Finally, sed@ over predicts an absorption line Si ii 126.5 nm and a P-Cygni profile of O i130.22 nm. Finally, the ages derived by stellar synthesis are older than the presence of the Helium ii emission line seems to suggest. This is not contradictory but rather suggests that the two analyses are not sensitive to the same features, which probably come from different regions of the galaxy and may point at a range of ages and multiple star formation episodes.

4.3 Integrated spectral energy distribution and dust extinction

The photometry in optical bands , , , , and in infrared bands , , and is described in Cabanac et al. (2005); Valls-Gabaud et al. (2006). Because the deflector is compact in the optical range, the source and magnitudes were simply extracted in an annulus of inner/outer radii of 15/3. The source is faint in and bands, and we can only put an upper limit to its flux in these bands. The IR magnitudes are more difficult to extract because the foreground deflector covers part of the source. We subtracted the central lens with galfit (Peng et al., 2002), using the best fit parameters of Cabanac et al. (2005). The derived IR magnitudes are sensitive to the lens substraction and we estimate that the systematic errors are of the order of 0.3 mag. We obtain the integrated Spectral Energy Distribution (SED, in Vega magnitudes) as shown in Figure 7.

We derived dust extinctions , and ages of the stellar populations using HyperZ (Bolzonella et al., 2000), for the known redshift of 3.7723, an extinction law similar to the Small Magellanic Cloud and several libraries of HyperZ templates, spanning single burst to constant SFR. The stellar Initial Mass Function (IMF) is that of Miller & Scalo (1979) and the metallicity of HyperZ templates is fixed at the solar value. Although it is likely that FORJ0332-3557 has a subsolar metallicity at a look-back time of  Ga, at face value solar metallicities are as good a guess as sub-solar metallicities. In order to estimate the effect of metallicity on , we also fed HyperZ with the best fit starburst99 templates (cf. previous section) of = 0.004.

Both Single-burst and Irregular HyperZ templates yield the same dust extinction  mag (for , ) for the best fit 23 Ma-old single-burst template and the best fit 128 Ma constant SFR template (Figure 7). The ages of the stellar populations are not well constrained by stellar synthesis, and Fig. 7 illustrates this point. One template is 100 Ma older than the other and yet they are indistinguishable with broadband data only. The inferred colour excess of is similar to the value found by Shapley et al. (2003) () for the Group 1 (Ly deficient) LGBs, using a similar extinction law and photometric bands, but a 300-Ma-old constant star formation rate. FORJ0332-3557 also falls in the middle of the extinction distribution of the LBG sample Steidel et al. (1999).

Following Pettini et al. (2000), we infer the extinction at 150 nm, and the Hi  content, Hi . For an SMC like extinction law, . Assuming gas-to-dust ratios ofHi cmmag for the Milky Way, cmmag for LMC, and cmmag for the SMC, we find cmHi cm (MW), cmHi cm (LMC), and cmHi cm (SMC). The column density inferred from the Ly absorption line is compatible with the first two values and would be inconsistent with an SMC gas-to-dust ratio.

Figure 7: The SED of FORJ0332-3557 (red crosses, the vertical bars indicating systematic errors and the horizontal ones giving the spectral bandwith of each filter). The top frame shows HyperZ best fit solar metallicity templates: single-burst template of age 23 Ma (blue solid line) and constant SFR followed by passive evolution template of age 128 Ma (green dotted line). Both templates are the best fit with an extinction  mag. The bottom frame shows HyperZ best fit with starburst99 =0.004 templates: instantaneous burst of age 8 Ma (blue solid line) and constant SFR of age 35 Ma (green dotted line), both with a best fit extinction of  mag. The bottom frame shows a clear difference between a constant SFR and an instantaneous burst near the 400 nm break. We can favour an instantaneous burst scenario based on band data.

4.4 Dust-corrected star formation rate

The UV (150 nm) luminosity is a standard indicator of (massive) star formation. Without correcting for dust extinction and assuming a constant SFR, Cabanac et al. (2005) derived a, where is the gravitational amplification produced by the lens. Using the extinction measured in the previous section, , we get a. Using the same method on cB58, Pettini et al. (2000) derives  a (for  km s Mpc, ). This puts FORJ0332-3557 in a very dynamic star forming episode, similar to what is observed in SDSS J11470250 (Bentz et al., 2004).

5 Conclusions

We have presented a medium-resolution spectrum of FORJ0332-3557, a lensed starburst galaxy, magnified 13 times, very similar to both cB58 and the composite spectrum of Lyman-break galaxies at .

We find spectral signatures of outflows of , commonly found in starburst galaxies, through a distinct residual Ly emission off-centered by 05 and peaking at ca. . This offset emission is most probably produced by the outflow rather than by an isolated, bright Hii region within the galaxy.

By combining visible-to-infrared colours and spectral features, we derive consistent ages of  Ma for single bursts and 20-40 Ma for constant SFRs and an extinction of using two independent stellar populations synthesis codes. Young stellar ages are also supported by the detection of the Wolf-Rayet Heii 164.04 nm line, indicator of stellar ages in the range 4-6 Ma. There is a clear trend to sub-solar metallicities in both cases. The ongoing rate of star formation, computed from the 150 nm continuum is  a.

We also derive preliminary lower limits to the abundances of some the low-ionisation interstellar lines suggesting sub-solar metallicities in the ionised gas phase.


We are grateful to Max Pettini for providing the spectrum of MS1512-cB58 and important suggestions, to Miguel Cerviño for computing a complete set of SED@ templates, to Lise Christensen for useful discussions, and to the anonymous referee for stimulating comments.


  1. thanks: Based on observations made at the ESO VLT under programs 74.A-0536 and 78.A-0240
  2. pagerange: Medium-resolution spectroscopy of FORJ0332-3557: Probing the interstellar medium and stellar populations of a lensed Lyman-break galaxy at z=3.77thanks: Based on observations made at the ESO VLT under programs 74.A-0536 and 78.A-0240References
  3. thanks: Based on observations made at the ESO VLT under programs 74.A-0536 and 78.A-0240
  4. pubyear: 2008
  5. pagerange: Medium-resolution spectroscopy of FORJ0332-3557: Probing the interstellar medium and stellar populations of a lensed Lyman-break galaxy at z=3.77thanks: Based on observations made at the ESO VLT under programs 74.A-0536 and 78.A-0240References
  6. thanks: Based on observations made at the ESO VLT under programs 74.A-0536 and 78.A-0240
  7. pubyear: 2008
  8. The advent of systematic surveys on wide fields is yielding many more potential candidates, see e.g. Bolton et al. (2006); Cabanac et al. (2007).
  9. This is clearly a rough approximation since the spectrum comes from the integrated light of the source and there are photons that are scattered into the line of sight. Hence the profile cannot be a pure absorption profile, and the inferred column density is just indicative.
  10. Errors are given as statistical (photon noise) and systematic (systematic uncertainty in location of the continuum)
  11. Lower limits based on the assumption of an optically thin medium
  12. Assuming solar abundances from Asplund et al. (2005)
  13. Tests where the upper stellar mass cutoff of the IMF is is reduced to 30 M give very similar best-fit ages.
  14. See the sed@ online Reference Manual at for more information.
  15. See for example Schaerer & Pelló (2005) for other examples of degeneracies in the SED fitting of high- galaxies.


  1. Ando M., Ohta K., Iwata I., Watanabe C., Tamura N., Akiyama M., Aoki K., 2004, ApJ, 610, 635
  2. Asplund M., Grevesse N., Sauval A. J., 2005, in Barnes T. G., Bash F. N., eds, ASP Conf. Ser. 336: Cosmic Abundances as Records of Stellar Evolution and Nucleosynthesis p. 25
  3. Bentz M. C., Osmer P. S., Weinberg D. H., 2004, ApJl, 600, L19
  4. Berger E., Penprase B. E., Cenko S. B., Kulkarni S. R., Fox D. B., Steidel C. C., Reddy N. A., 2006, ApJ, 642, 979
  5. Bolton A. S., Burles S., Koopmans L. V. E., Treu T., Moustakas L. A., 2006, ApJ, 638, 703
  6. Bolzonella M., Miralles J.-M., Pelló R., 2000, A&A, 363, 476
  7. Buzzoni A., 1989, ApJS, 71, 817
  8. Cabanac R. A., Alard C., Dantel-Fort M., Fort B., Gavazzi R., Gomez P., Kneib J. P., Le Fèvre O., Mellier Y., Pello R., Soucail G., Sygnet J. F., Valls-Gabaud D., 2007, A&A, 461, 813
  9. Cabanac R. A., Valls-Gabaud D., Jaunsen A. O., Lidman C., Jerjen H., 2005, A&A, 436, L21
  10. Castelli F., Kurucz R. L., 1994, A&A, 281, 817
  11. Cerviño M., Luridiana V., 2004, A&A, 413, 145
  12. Cerviño M., Luridiana V., Pérez E., Vílchez J. M., Valls-Gabaud D., 2003, A&A, 407, 177
  13. Cerviño M., Mas-Hesse J. M., 1994, A&A, 284, 749
  14. Cerviño M., Valls-Gabaud D., 2003, MNRAS, 338, 481
  15. Cerviño M., Valls-Gabaud D., Luridiana V., Mas-Hesse J. M., 2002, A&A, 381, 51
  16. Dow-Hygelund C. C., Holden B. P., Bouwens R. J., van der Wel A., Illingworth G. D., Zirm A., Franx M., Rosati P., Ford H., van Dokkum P. G., Stanford S. A., Eisenhardt P., Fazio G. G., 2005, ApJL, 630, L137
  17. Fynbo J. P. U., Starling R. L. C., Ledoux C., et al. 2006, ArXiv astro-ph/0602444
  18. Leitherer C., Schaerer D., Goldader J. D., Delgado R. M. G., Robert C., Kune D. F., de Mello D. F., Devost D., Heckman T. M., 1999, ApJS, 123, 3
  19. Miller G. E., Scalo J. M., 1979, ApJS, 41, 513
  20. Noll S., Mehlert D., Appenzeller I., Bender R., Böhm A., Gabasch A., Heidt J., Hopp U., Jäger K., Seitz S., Stahl O., Tapken C., Ziegler B. L., 2004, A&A, 418, 885
  21. Peng C. Y., Ho L. C., Impey C. D., Rix H., 2002, AJ, 124, 266
  22. Pettini M., Rix S. A., Steidel C. C., Adelberger K. L., Hunt M. P., Shapley A. E., 2002, ApJ, 569, 742
  23. Pettini M., Steidel C. C., Adelberger K. L., Dickinson M., Giavalisco M., 2000, ApJ, 528, 96
  24. Savaglio S., Panagia N., Padovani P., 2002, ApJ, 567, 702
  25. Schaerer D., 2003, A&A, 397, 527
  26. Schaerer D., Pelló R., 2005, MNRAS, 362, 1054
  27. Schaerer D., Vacca W. D., 1998, ApJ, 497, 618
  28. Schaerer D., Verhamme A., 2008, arXiv:astro-ph/0801.1187
  29. Shapley A. E., Steidel C. C., Adelberger K. L., Dickinson M., Giavalisco M., Pettini M., 2001, ApJ, 562, 95
  30. Shapley A. E., Steidel C. C., Pettini M., Adelberger K. L., 2003, ApJ, 588, 65
  31. Spitzer L., 1978, Physical processes in the interstellar medium. (New York : Wiley-Interscience)
  32. Steidel C. C., Adelberger K. L., Giavalisco M., Dickinson M., Pettini M., 1999, ApJ, 519, 1
  33. Steidel C. C., Pettini M., Adelberger K. L., 2001, ApJ, 546, 665
  34. Valls-Gabaud D., 1993, ApJ, 419, 7
  35. Valls-Gabaud D., Cabanac R., Lidman C., Diego J. M., Saha P., 2006, in Mamon G. A., Combes F., Deffayet C., Fort B., eds, EAS Publications Series Vol. 20 of EAS Publications Series. pp 149–152
  36. Vázquez G. A., Leitherer C., 2005, ApJ, 621, 695
  37. Verhamme A., Schaerer D., Maselli A., 2006, A&A, 460, 397
  38. Vreeswijk P. M., Ellison S. L., Ledoux C., et al. 2004, A&A, 419, 927
  39. Vreeswijk P. M., Smette A., Fruchter A. S., et al. 2006, A&A, 447, 145
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minumum 40 characters
Add comment
Loading ...
This is a comment super asjknd jkasnjk adsnkj
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test description