PHYSICAL PROPERTIES OF Lyα EMITTERS AT z ∼ 0.3 FROM UV-TO-FIR MEASUREMENTS

In recent years, a substantial number of Lyα-emitting galaxies have been discovered over a wide range of redshifts, from the local universe up to z ∼ 7, and even beyond. At z ≳ 2.0, the Lyα emission is located in the optical or near-IR and LAEs are mostly found using the narrowband technique, where a combination of narrow- and broadband filters are employed to sample the Lyα emission and constrain its nearby continuum, respectively (e.g., Guaita et al. 2010; Bongiovanni et al. 2010; Ouchi et al. 2008, 2010; Cowie & Hu 1998; Gawiser et al. 2006; Gronwall et al. 2007; Shioya et al. 2009; Murayama et al. 2007; Nakamura et al. 2011; Sobral et al. 2009). At z ≲ 2, the Lyα line is in the UV and, so far, LAEs can only be found via Galaxy Evolution Explorer (GALEX; Martin et al. 2005) grism spectroscopy or other space-borne UV observatories. In this case, the selection technique is based on searching for Lyα emission in the UV spectrum of objects with a measured UV continuum (Deharveng et al. 2008; Cowie et al. 2010, 2011).

The physical properties of star-forming (SF) LAEs have been classically studied by fitting (Bruzual & Charlot 2003, hereafter BC03) templates to their observed spectral energy distribution (SED, e.g., Ono et al. 2010; Lai et al. 2007, 2008; Finkelstein et al. 2008; Nilsson et al. 2007, 2009, 2011; Pirzkal et al. 2007; Guaita et al. 2011). With this method, the stellar mass and age of galaxies can be reasonably constrained with good sampling of the rest-frame UV to mid-IR SEDs, but other physical properties, such as dust attenuation, star formation rate (SFR), and metallicity, tend to suffer from large uncertainties: metallicities can be obtained only by using rest-frame optical emission lines, and information about the dust emission in the FIR is essential for obtaining accurate values of dust attenuation and SFR.

At z ∼ 0.3, Cowie et al. (2010), Finkelstein et al. (2011), and Cowie et al. (2011) agree in their optical spectroscopic studies that LAEs are metal-poor galaxies with low dust attenuation. Cowie et al. (2011) also report that LAEs with higher EW(Hα) have bluer colors, lower metallicities, and less extinction, consistent with their being at a primeval evolutionary stage. Employing an SED-fitting procedure, Cowie et al. (2011) and Finkelstein et al. (2011) report that LAEs are mainly young galaxies with median ages of 100 and 300 Myr, respectively.

With the launch of the Herschel Space Telescope (Pilbratt et al. 2010) and the data taken with the Photodetector Array Camera and Spectrometer (PACS; Poglitsch et al. 2010), we are in possession of deep FIR data that allow us to constrain the FIR SED of LAEs. Unfortunately, few FIR counterparts for LAEs at z ≳ 2.0 have been reported so far; therefore, a deep analysis has not yet been possible. In fact, in Oteo et al. (2012) we only found four SF LAEs at 2.0 ≲ z ≲ 3.5 with FIR counterparts out of a sample of 140, only 2 of them having a Lyα rest-frame equivalent width (Lyα EW_rest-frame) above 20 Å, the typical minimum value for LAEs found via narrowband imaging. Despite this low number, the detection of some LAEs in the FIR reveals that some of them are red and dusty objects, thus dust and Lyα emission are not mutually exclusive (see also Chapman et al. 2005).

At z ∼ 0.3, the typical FIR-observed fluxes of LAEs make them probably detectable under the limiting fluxes of PACS and MIPS-24 μm observations. In Oteo et al. (2011), we report PACS-FIR detections and study dust attenuation in a sample of 12 spectroscopically GALEX-selected LAEs at z ∼ 0.3 and ∼1.0. In this paper, we expand the work started in Oteo et al. (2011) about the physical properties of LAEs at z ∼ 0.3. Specifically, we obtain UV and total IR luminosities, dust attenuation, SFR, and size and examine the morphology of 23 IR-detected LAEs. We also compare the results with those obtained for a sample of galaxies which, under the same limiting UV and IR luminosities, do not show Lyα emission in their rest-frame UV spectra.

The structure of this paper is as follows. In Section 2, we present the LAE data sample, reporting their PACS and MIPS-24 μm counterparts in Section 3. The comparison sample is presented in Section 4. Section 5 gives the results of our study and in Section 6 we show the main conclusions of this work.

Throughout this paper, we assume a flat universe with (Ω_m, Ω_Λ, h₀) = (0.3, 0.7, 0.7), and all magnitudes are listed in the AB system (Oke & Gunn 1983).

In this study, we use a sample of GALEX-discovered LAEs at z ∼ 0.3 built in Cowie et al. (2010) by looking for an emission line in the FUV spectra of objects with a measured UV continuum. Since the FUV spectra become very noisy at the edges of the spectral range, only those objects with Lyα emission within 1452.5–1750 Å were selected. In the redshift space, this wavelength range implies that LAEs are distributed within z = 0.195–0.44, with a median value of z ∼ 0.3. In Cowie et al. (2010), LAEs were classified as SF or active galactic nuclei (AGNs) via rest-frame UV emission-line diagnostics (shape and width of the Lyα line and the presence/absence of AGN ionization lines) and their X-ray measurements. In this study, we consider only those LAEs with an SF nature, ruling out the AGNs. Among the fields studied in Cowie et al. (2010), we focus on COSMOS, ECDF-South, ELAIS-S1, and Lockman. These fields were selected by their wealth of FIR data from PACS-Herschel or MIPS-Spitzer (see Section 3 for more details).

With the aim of comparing with narrowband-selected high-redshift LAEs, we only include in the sample those Lyα-emitting galaxies whose Lyα EW_rest-frame are above 20 Å, which is the typical threshold in narrowband searches. In this sense, we define LAEs as those Lyα-emitting galaxies whose Lyα EW_rest-frame are above 20 Å, using this definition throughout the study. All the previous considerations produce a sample of 30 SF LAEs. This sample is nearly complete up to m_NUV ∼ 21.5 mag, which represents the UV limiting magnitude of the sample. This implies a UV limiting magnitude of log (L_UV/L_☉) = 9.9 at z ∼ 0.3, which is similar to that in the samples of high-redshift LAEs found via the narrowband technique.

All our LAEs have Lyα luminosities below 10⁴³ erg s⁻¹, the typical median value for LAEs at z ≳ 2.0. Indeed, the lower values of the Lyα luminosities for LAEs at z ∼ 0.3 are an indication of an evolution in the physical properties of LAEs between that redshift and z ≳ 2.0 (see Cowie et al. 2011).

The PACS-FIR observations used in this study were taken within the framework of PACS Evolutionary Probe project (PEP, PI: D. Lutz). PEP is the Herschel Guaranteed Time Key Project to obtain the best profit from studying FIR galaxy evolution with Herschel instrumentation (Lutz et al. 2011). Among the fields where LAEs cataloged in Cowie et al. (2010) are located, COSMOS, GOODS-South, and ECDF-South have been already observed in PACS-100 μm and PACS-160 μm bands.¹³ In this study, we use PACS fluxes extracted by using MIPS-24 μm priors, with limiting (100 μm, 160 μm) fluxes in COSMOS, GOODS-South, and ECDF-South of (5.0, 11.0) mJy, (1.1, 2.0) mJy, and (4.2, 8.2) mJy, respectively. The PACS-FIR counterparts of the LAEs at z ∼ 0.3 in COSMOS and GOODS-South have already been reported in Oteo et al. (2011), by looking for a PACS detection within 2'' around the location of each LAE, which is the typical astrometric uncertainty in the position of the FIR sources. In this study, we add the PACS counterparts in ECDF-South to the previous set, following the same matching criterion. We find one extra LAE detected in both PACS-100 μm and PACS-160 μm, assembling, in total, a sample of six PACS-detected SF LAEs. All these FIR counterparts represent a direct measurement of the FIR dust emission in LAEs at z ∼ 0.3. As an example, we show in Figure 1 the FIR SED of the three LAEs with counterparts both in PACS-100 μm and PACS-160 μm. Both observational FIR data points and best Chary & Elbaz (2001) and Polletta et al. (2007) fitted templates are represented.

Zoom In Zoom Out Reset image size

MIPS-24 μm measurements allow us to constrain the FIR SED of PACS-undetected but MIPS-24 μm-detected LAEs. In Oteo et al. (2011), we report MIPS-24 μm counterparts for LAEs in the COSMOS field by using data taken from the S-COSMOS survey (Sanders et al. 2007). We found that 90% of LAEs at z ∼ 0.3 were detected in that band. Here, we extend the MIPS-24 μm counterparts with data coming from the Spitzer Wide-area InfraRed Extragalactic survey (SWIRE; Lonsdale et al. 2003) in the ECDF-South, Lockman, and ELAIS-South fields, using a matching radius of 2'' as well. We find 13 extra MIPS detections.

Summarizing, among the 30 LAEs with MIPS/PACS coverage, 23 are detected in MIPS/PACS. Nine of them belong to the sample in Oteo et al. (2011). Therefore, we find that more than 75% of LAEs at z ∼ 0.3 are detected in the mid-IR/FIR. Figure 2 shows the detection fraction in MIPS-24 μm and PACS bands of a sample of UV-selected galaxies at z ∼ 0.3 (see Section 4). It can clearly be seen that UV-brighter galaxies are more likely to be detected in the IR. This tendency is similar to that found for MIPS-24 μm detections in high-redshift galaxies (Reddy et al. 2010). There is a decrease in the detection fraction at the highest UV luminosities but in those cases the number of galaxies in each luminosity bin is low and the error bars are too large to draw any conclusion. As commented above, the limiting UV luminosity associated with the limiting magnitude of the LAEs in the sample studied is about 9.9 at z ∼ 0.3. From Figure 2, it can be deduced that UV luminosities above that value correspond to a high detection fraction in the IR, which is compatible with the large percentage of IR-detected LAEs found in this study.

Zoom In Zoom Out Reset image size

According to the limiting fluxes of the MIPS-24 μm observations, MIPS-detected LAEs at z ∼ 0.3 have log (L_IR/L_☉) ≳  10.0 in COSMOS and ≳ 10.40 in the SWIRE fields. Regarding PACS observations, the limiting luminosities are log (L_IR/L_☉)  ∼  10.5, ∼9.76, and ∼10.4 in COSMOS, GOODS-South, and ECDF-South, respectively. In this sense, we adopt a limiting luminosity of log (L_IR/L_☉)  ∼  10.4 for the whole sample. Note that, although PACS observations in GOODS-South and MIPS observations in COSMOS are deeper than that threshold, there is only one LAE with L_IR below 10^10.4 L_☉ and it is due to its lower redshift, z  ∼  0.2. The adopted IR limiting luminosity represents an IR star formation rate of SFR_IR ∼ 4.5 M_☉ yr⁻¹, according to the Kennicutt (1998) calibration.

We also aim at comparing the derived physical properties of our IR-detected SF LAEs with those of other IR-detected SF galaxies in the same redshift range and with the same UV and IR limiting luminosities but which do not exhibit Lyα emission in their UV spectrum. To do that, we define a control sample focusing on the COSMOS field and using data from GALEX (Guillaume et al. 2006; Zamojski et al. 2007) and PACS observations plus the photometric redshifts, z_phot, of the COSMOS photometric catalog (Capak et al. 2007). At z ∼ 0.3, z_phot are quite reliable and can be used instead of spectroscopic surveys, which contain many fewer objects. In this way, we select all the sources with z_phot between 0.2 and 0.4 with measurements both in GALEX and PACS and whose UV luminosities are higher than the UV limiting luminosity of LAEs at z ∼ 0.3. In order to avoid possible contamination from AGNs, we rule out from the sample those samples which are detected in X-rays with Chandra. From now on, we refer to those galaxies as non-LAEs, and they will be the main source of comparison to study the differentiating characteristic of LAEs. The sample contains 135 galaxies. This sample also contains Lyα-emitting galaxies whose Lyα EW_rest-frame are below 20 Å; therefore, although they exhibit a Lyα line, it is not strong enough to be selected in narrowband searches.

Furthermore, we also retain those galaxies with z_phot between 0.2 and 0.4, which are detected in GALEX and PACS but have UV luminosities fainter than the UV limiting luminosity of LAEs. We call them UV-faint SF galaxies. These sources will not be directly compared with LAEs because of their different selection criterion in the UV, but will be used to place LAEs within a more general scenario of SF galaxies at the same redshift. Note that, owing to GALEX spectroscopic limitations, we do not have UV spectra for these sources and we therefore do not know whether they exhibit Lyα emission.

It should be noted that, for the same reasons as pointed out in Section 3, there is no bias in the IR selection between LAEs and the galaxies in the control sample, all of them being limited to the same IR luminosity, log (L_IR/L_☉) ∼ 10.4.

To summarize, we have a control sample formed by two kinds of galaxies: (1) non-LAEs, which are galaxies that do not show Lyα emission with Lyα EW_rest-frame above 20 Å, are in the same redshift range of our LAEs, and have the same UV and IR limiting luminosities; (2) UV-faint SF galaxies: formed by galaxies which are at the same redshift as our LAEs, have the same IR limiting luminosity, but are fainter in the UV than our LAEs and non-LAEs. Only the non-LAEs will be used directly to compare the properties of LAEs with those of other galaxies without Lyα emission. The UV-faint galaxies will be used only to place LAEs and non-LAEs within a more general framework of SF galaxies at z ∼ 0.3.

The UV and mid-IR/FIR measurements, which represent the emitted light from young populations and the re-emission of the light absorbed by dust in the UV, respectively, can be used to analyze the physical properties of our galaxies. The UV/IR combination allows an accurate determination of dust attenuation and total SFR, SFR_total. Furthermore, FIR detections themselves enable us to examine the IR nature of galaxies.

5.1. UV and IR Luminosities

L_UV, expressed as νL_ν, is obtained for each galaxy from its observed magnitude in the NUV band and using the assumed cosmology. We choose the NUV band because at z ∼ 0.3, it covers the continuum near the Lyα line and is not contaminated by the Lyα emission. On the IR side, L_IR is defined as the integrated IR luminosity between 8 and 1000 μm in the rest frame and is obtained by using different calibrations depending on whether an object is PACS-detected or PACS-undetected but MIPS-24 μm-detected. In the first case, L_IR is calculated using calibrations between PACS bands and total IR luminosities (Equations (1) and (2)):

$$\begin{equation} \log {L_{\rm IR}} = 0.99 \log {L_{100\,\mu {\rm m}}} + (0.44 \pm 0.25) \end{equation} \tag{ 1 }$$

$$\begin{equation} \log {L_{\rm IR}} = 0.96 \log {L_{160\,\mu {\rm m}}} + (0.77 \pm 0.21), \end{equation} \tag{ 2 }$$

where all the luminosities are in solar units and L_100 μm and L_160 μm are defined as νL_ν. The calibration employed for each galaxy is that associated with the reddest PACS band where it is detected. This would represent the nearest measurement to the dust emission peak, ensuring a better determination of L_IR. To derive these calibrations, we focus on the COSMOS field and select all the PACS-detected objects both at 100 μm and 160 μm, which are spectroscopically cataloged to be at z ≲ 0.5 in the zCOSMOS survey (Lilly et al. 2007). This condition enables us to carry out accurate FIR SED fits with Chary & Elbaz (2001, hereafter CE01) templates since at that z ≲ 0.5 the dust emission peak is well sampled with those FIR bands. The fits are performed with the Zurich Extragalactic Bayesian Redshift Analyzer (Feldmann et al. 2006) in the maximum likelihood mode and the L_IR for each galaxy is obtained by integrating its best-fitted template between 8 and 1000 μm in the rest frame. The calibrations were built by comparing the L_IR of each source with that in the PACS bands and the errors are twice the standard deviation in the fittings. In this process, we ruled out AGNs via X-ray emission diagnosis.

For the PACS-undetected but MIPS-24 μm-detected objects, we convert MIPS-24 μm fluxes into L_IR by fitting the fluxes to CE01 templates. These templates are built in such a way that for a given flux and redshift a unique solution for L_IR exists. The determination of L_IR from single FIR band extrapolations has also been employed in other studies (Elbaz et al. 2010, 2011; Nordon et al. 2010). Elbaz et al. (2010) analyze the applicability of this procedure, finding that MIPS-24 μm extrapolations to L_IR are valid for galaxies up to z ∼ 1.5 and which fall below the ultraluminous infrared galaxy (ULIRG) limit. Both conditions are met in our case, both in LAEs and in the galaxies belonging to the control sample.

In Figure 3, we present the L_UV and L_IR for LAEs at z ∼ 0.3 and those for the control sample, both non-LAEs and UV-faint galaxies. On the UV side, it can be seen that the histogram for LAEs includes larger values than that for non-LAEs, indicating that LAEs are UV-brighter than non-LAEs. Median log (L_UV/L_☉) for LAEs and non-LAEs are 10.1 and 9.9, respectively. On the IR side, L_IR values for LAEs and non-LAEs are in the same range, mainly spanning from log (L_IR/L_☉) = 10.4 to 11.2. Even for UV-faint galaxies, the histogram of L_IR is similar to that for LAEs and non-LAEs, despite their difference in the UV selection.

Zoom In Zoom Out Reset image size

5.1.1. IR Emission and ULIRG Fraction Evolution

According to their L_IR, galaxies can be classified into three types: normal SF galaxies: L_IR < 10¹¹ L_☉, luminous infrared galaxies (LIRGs): 10¹¹ < L_IR [L_☉] < 10¹², and ULIRGs: L_IR > 10¹² L_☉. As can be seen in Figure 3, most of our IR-detected LAEs are in the normal SF regime. Due to the correlation between MIPS-24 μm luminosity and L_IR, IR-undetected LAEs with FIR coverage have L_IR less than the IR limiting luminosity (∼10^10.4 L_☉), being normal SF galaxies too. In this way, considering the 30 LAEs with IR coverage (both IR-detected and IR-undetected), we find that at z ∼ 0.3 more than 80% of LAEs are SF galaxies. Only six LAEs have 10¹¹ < L_IR[L_☉] < 10^11.5, and none of them have L_IR > 10^11.5 nor fall in the ULIRG class. Note that, considering both IR-detected and -undetected LAEs, we work with a nearly complete sample of LAEs up to m_NUV ∼ 21.5 mag (Cowie et al. 2010), and therefore, the non-existence of ULIRGs in our sample of LAEs is an unbiased result under that limit. If there were LAEs with a ULIRG nature at z ∼ 0.3, they would have to be fainter in the UV than those analyzed in the present work. No ULIRG is found in the control sample (non-LAEs and UV-faint galaxies) either, most galaxies being in the normal SF regime as well. Indeed, most recent studies of the FIR properties of galaxies with Herschel do not find many ULIRGs at z ∼ 0.3, but they begin appearing from z ∼ 1.0 (Elbaz et al. 2011, 2010; Buat et al. 2010; Lutz et al. 2011).

At z ≳ 2, Chapman et al. (2005) found Lyα emission in the optical spectra of a sample of 850 μm detected SF submillimeter galaxies with a ULIRG nature. Nilsson & Møller (2009) and Nilsson et al. (2011) also suggest that some LAEs at z ∼ 2 have a ULIRG nature. In this way, the discovery of Lyα emission in ULIRGs at z ≳ 2 and the fact that most of our LAEs at z ∼ 0.3 have log (L_IR/L_☉) ≲ 11.2 suggests that the IR nature of the galaxies found via their Lyα emission is changing from z ∼ 2 to z ∼ 0.3: there is a population of red and dusty LAEs with a ULIRG nature at z ≳ 2, which is not seen at z ∼ 0.3. This trend is similar to that found in previous studies for general galaxies detected in MIPS-24 μm and to the trend of cosmic star formation (Hogg et al. 1998; Hopkins 2004; Hopkins & Beacom 2006; Pérez-González et al. 2005; Le Floc'h et al. 2005). Cowie et al. (2011) find a dramatic evolution in the maximum of the Lyα luminosity function between z ∼ 0.3 and z ≳ 1.0. This result, together with the evolution in the IR emission of LAEs suggested above, indicates that either the properties of galaxies selected via their Lyα emission are evolving over cosmic time or the Lyα selection technique is not tracing the same kind of objects at different redshifts.

5.2. Dust Attenuation

The ratio between L_IR and L_UV is a good tracer of dust attenuation in galaxies. Here, we adopt the calibration between the IR/UV ratio and dust attenuation found in Buat et al. (2005) (Equation (3)) to obtain dust attenuation for our IR-detected LAEs at z ∼ 0.3 and for the galaxies in the control sample, both non-LAEs and UV-faint galaxies:

$$\begin{equation} A_{\rm NUV} = -0.0495x^3 + 0.4718x^2 + 0.8998x + 0.169, \end{equation} \tag{ 3 }$$

where A_NUV is the dust attenuation in the NUV band and x = log (L_IR/L_UV). The conversion from A_NUV to the dust attenuation at 1200 Å, A_1200 Å, is made by using the Calzetti et al. (2000) reddening law. As in Buat et al. (2007), we do not apply K-correction to the UV luminosities given that the L_ν spectrum of galaxies is quite flat in the UV range. In this way, we assume that the NUV fluxes observed at z ∼ 0.3 are the same as those observed at z ∼ 0 in the NUV band.

In Figure 4, we represent the dust attenuation of our LAEs versus their L_IR along with the dust attenuation distribution, which has a median value of A_1200 Å ∼ 1.5 mag. We also plot in that figure the dust attenuation for non-LAEs and for the UV-faint galaxies. It can be seen that, for each value of L_IR, LAEs are slightly less dusty than non-LAEs. In fact, the median value of the dust attenuation distribution for non-LAEs is A_1200 Å ∼ 2.0 mag. Furthermore, for both LAEs and non-LAEs there is a trend toward brighter galaxies in the IR being dustier.

Zoom In Zoom Out Reset image size

Cowie et al. (2011) find, comparing the UV spectral index to the Hα/Hβ flux ratio, that the dust attenuation in most LAEs at z ∼ 0.3 can be also described by Cardelli et al. (1989) or Fitzpatrick (1999) laws, i.e., the stellar extinction is better represented as a uniform screen rather than a patchy distribution. The inclusion of those laws here would change the final numbers in the sense that the dust attenuation would be lower than that obtained with the Calzetti et al. (2000) law, although the relation between the dust attenuation in different populations would remain the same.

LAEs with a fainter UV continuum than the limiting magnitude are not included in the sample and they could be dustier than those presented in this study. Therefore, there could be some LAEs at z ∼ 0.3 which are dustier and have larger rest-frame Lyα equivalent widths than those presented here. The finding of a Lyα emission in the UV spectra of a dusty object at z ∼ 0.3 could indicate the presence of a clumpy interstellar medium (Neufeld 1991). The possible existence of this and other kinds of geometries has been reported to be able to recover the observed Lyα/Hα and Hα/Hβ ratios and the observed SEDs of LAEs at different redshifts (Finkelstein et al. 2008, 2011; Scarlata et al. 2009; Guaita et al. 2011).

5.2.1. Dust Attenuation and UV Continuum Slope

Dust attenuates the rest-frame UV light of a galaxy in a wavelength-dependent way, this attenuation being larger for shorter UV wavelengths. Therefore, dust produces an increase in the rest-frame UV continuum slope, β, from negative to less negative or even positive values. On the other hand, part of the absorbed light in the UV is in turn re-emitted in the FIR regime. In this way, dust attenuation and β are expected to be related. Indeed, many studies employ different relations between dust attenuation and β to obtain the dust attenuation in galaxies that are not detected in the FIR.

In this section, we compare dust attenuation and β for the IR-detected LAEs, non-LAEs, and UV-faint SF galaxies. Shown in Figure 5 is the IRX-β diagram for the three kinds of galaxies. β values are obtained by employing the Kong et al. (2004) recipe, as in Buat et al. (2010). In this way, we assume that the rest-frame UV continuum can be described by a power law in the form f_λ ∼ λ^β and obtain β from GALEX photometry in the FUV and NUV channels. It can be clearly seen that LAEs have UV continuum slopes much bluer than non-LAEs, which is compatible with their being among the least dusty galaxies at their redshift (see Section 5.2). The finding of larger (redder) values of β for the UV-faint galaxies than for LAEs and non-LAEs is due to their faintness in the UV.

Zoom In Zoom Out Reset image size

Shown in Figure 5 are also the curves of Kong et al. (2004) and Boissier et al. (2007) corresponding to normal SF and starburst (SB) galaxies, respectively. It can be seen that although many LAEs are distributed around the curve corresponding to the normal SF galaxies, some of them occupy a different location in the IRX-β diagram, quite near to the SB relation. Some studies use a relation between dust attenuation and β to obtain the dust attenuation from the values of the UV continuum slope for objects which are not detected in the FIR. The reason for this procedure is that, whereas the UV continuum slope is relatively easy to obtain from broadband photometry over a wide range of redshifts, the detection rate in the FIR of certain kinds of galaxies is very low, mainly at the highest redshifts. This is typically the case for LAEs (Oteo et al. 2012). But it should be noted that the reliability of this technique is based on the idea that the relation assumed between dust attenuation and β applies for the population of galaxies under study. What we find here is that this is not the case for LAEs at z ∼ 0.3, which are distributed between the SB and the SF relations so that a unique relation cannot be applied to the whole population to recover the dust attenuation from β. Furthermore, it has been reported that the location of galaxies in the IRX-β diagram also depends on the L_IR and age. For example, at z ∼ 2, galaxies with ages below 100 Myr tend to be less reddened at a given UV continuum slope than the values predicted by the SB relation (Reddy et al. 2006, 2010, 2012). This tendency might have a significant influence on LAEs, since most of them have been reported to be young galaxies in a wide range of redshifts.

On the other hand, non-LAEs are mostly distributed around the normal SF curve, although in a zone associated with higher β values than those for LAEs. There is a very low percentage of non-LAEs near the SB curve. From another point of view, for each value of dust attenuation, LAEs tend to have bluer UV continuum slopes than non-LAEs. This could indicate a different star formation mode between LAEs and non-LAEs.

5.2.2. Dust Attenuation and Redshift

Two open questions in the study of LAEs are their dust attenuation and its possible evolution with redshift. In the previous sections, we provided additional clues to answer the former question for LAEs at z ∼ 0.3 by using UV and mid-IR/FIR measurements. Now, in an attempt to answer the latter question, we compare our values with those at different redshifts reported in previous studies (Finkelstein et al. 2008, 2009, 2011; Cowie et al. 2011; Guaita et al. 2011; Nilsson et al. 2007, 2011; Gawiser et al. 2006; Lai et al. 2008; Ono et al. 2010; Pirzkal et al. 2007). Figure 6 shows the representation of the results reported in those works, along with the curve of the redshift evolution of a general population of galaxies found in Hayes et al. (2011). We also include the line associated with a dust attenuation of A_1200 Å = 1.0 mag at all redshifts in order to clarify the differences in dust attenuation between LAEs at different redshifts. It can be seen that most LAEs at z ≳ 2.5 have dust attenuation below 1.0 mag in 1200 Å. Actually, most results indicate that the SEDs of LAEs at those redshifts are compatible with dust attenuation below 0.5 mag at 1200 Å (Lai et al. 2008; Gawiser et al. 2006, 2007).

Zoom In Zoom Out Reset image size

At z ∼ 0.3, we find a dust attenuation distribution centered in A_1200 Å = 1.5 mag. This is compatible with the results reported by Finkelstein et al. (2011) and Cowie et al. (2011) even though the results were obtained with rest-frame optical spectroscopy and SED fitting and consequently different methodologies.

At z ∼ 2.0, Guaita et al. (2011) and Nilsson et al. (2011) report high dust attenuation values, higher than those reported at z ∼ 0.3 and z ≳ 2.5. Although these median values are high, the dust attenuation distributions found in both studies are quite wide, which makes it difficult to establish a clear tendency between z ∼ 2.0 and other redshifts. However, the high median values themselves are an indication that there should exist a population of dusty LAEs at z ∼ 2.0 that are not seen at other redshifts. Furthermore, the redshift distribution of the red and dusty objects selected via their submillimeter emission is centered around z ∼ 2.3 (Chapman et al. 2005). Some of those galaxies are SF ULIRGs exhibiting Lyα emission in the spectra, which reinforces the idea that, at those redshifts, the Lyα selection technique can also segregate dusty galaxies, in contrast with the classical idea of LAEs as galaxies with low dust attenuation. Blanc et al. (2011) find the emergence of a small population of red galaxies at z < 3 (in terms of their UV continuum slope) in their study of integral-field spectroscopically selected LAEs at 1.9 < z < 3.8. Moreover, in Oteo et al. (2012) we report the detection in PACS-160 μm of a sample of LAEs at 2.0 ≲ z ≲ 3.5 and derive their dust attenuation by employing the IR/UV ratio (Buat et al. 2005). As result, we obtain values of A_1200 Å ≳ 4.5 mag, higher than the median values reported in Guaita et al. (2011) and Nilsson et al. (2011) but compatible with the width of the distribution in Nilsson et al. (2011). Note that the values of dust attenuation reported in Oteo et al. (2012) represent the upper limit on LAEs at that redshift, since we only work with FIR-detected sources.

At z ∼ 3.1, most LAEs are found to be nearly dust-free objects (Gawiser et al. 2006, 2007; Lai et al. 2008), representing a great difference from z ∼ 3.1 to z ∼ 2.1–2.3. However, it should be considered that at z ≳ 2.5, the dust attenuation in most studies has been obtained by using a stacking analysis of the sample. Nilsson et al. (2011) compared the dust attenuation obtained for LAEs at z ∼ 2.3 when performing a stacking analysis with those found when considering individual objects. They found that dust attenuation is considerably reduced when stacking the total sample and a subsample of old LAEs, indicating that dust attenuation is very sensitive to stacking.

Despite the uncertainties of stacking, the evolution of dust attenuation between z ≳ 3 and z ∼ 2 could be related to the evolution in other properties of LAEs. Ciardullo et al. (2012) report that there is a significant evolution in the LAE luminosity function between z ∼ 3.1 and z ∼ 2.1. Bond et al. (2011) claim for an evolution in the size of LAEs from z ∼ 3.1 to z ∼ 2.1, raising the median half-light radius from 1.0 kpc at z ∼ 3.1 to 1.4 kpc at z ∼ 2.1. Nilsson et al. (2009) also found evolution in the physical properties of LAEs between z ∼ 3 and ∼2, in the sense that at z ∼ 2 LAEs have redder SEDs, which indicates that they are more evolved (dustier, older, and more massive) than those at z ∼ 3.

At z ≳ 4, the samples studied contain very few objects and the results are not as statistically significant as at lower redshifts; therefore, although most studies have reported that the SED of LAEs is compatible with some levels of dust, more work is needed to confirm that behavior.

Therefore, the general trend is that LAEs at z ≳ 2.5 tend to be slightly less dusty than their low-redshift analogs. Furthermore, it can be seen that LAEs at z ≳ 2.0 tend to follow the dust evolution of galaxies with redshift, while at z ∼ 0.3, they deviate from that behavior.

Apart from indicating a possible evolution of dust attenuation with redshift for LAEs, Figure 6 also shows the lack of knowledge that we still have about the physical properties of those galaxies. While the best method of determining dust attenuation is the combination of direct UV and FIR measurement used in this study, different procedures have been used to derive that parameter. However, it has not been checked yet whether they provide consistent results at a given redshift. Furthermore, at the highest redshifts, either the number of objects studied is low, or stacking analysis had to be done due to the non-detection of LAEs in many photometric bands.

5.3. Star Formation Rates

The combination of UV and IR data also provides the most accurate determination of SFRs in galaxies. Adopting the Kennicutt (1998) calibrations, the SFRs associated with the observed UV and IR luminosities are given by the expressions

$$\begin{equation} {\rm SFR}_{\rm UV,uncorrected}[M_{\odot }{\rm \,yr}^{-1}] = 1.4\times 10^{-28}L_{\nu,\rm observed} \end{equation} \tag{ 4 }$$

$$\begin{equation} {\rm SFR}_{\rm IR}[M_{\odot }{\rm \,yr}^{-1}] = 4.5\times 10^{-44}L_{{\rm IR}}, \end{equation} \tag{ 5 }$$

where L_{ν, observed} is in units of erg s⁻¹ Hz⁻¹ and L_IR is defined in the same way as in Section 5.1. In order to obtain the SFR_total for our galaxies, we assume that all the light absorbed in the UV is in turn re-radiated in the FIR. In this scenario, SFR_total can be calculated as the sum of a dust-uncorrected component, SFR_{UV, uncorrected}, and the correction term shown in Equation (5). Thus, it can be written

$$\begin{equation} {\rm SFR}_{\rm total} = {\rm SFR}_{\rm UV, uncorrected} + {\rm SFR}_{\rm IR}. \end{equation} \tag{ 6 }$$

The distributions of the SFR_total for IR-detected LAEs and the galaxies in the control sample are shown in Figure 7. It can be seen that IR-detected LAEs have SFR_total ranging mainly from 10 to 40 M_☉ yr⁻¹, with a median of 18 M_☉ yr⁻¹, and peaking around 13 M_☉ yr⁻¹. The median value of SFR_total for non-LAEs is 15 M_☉ yr⁻¹, with a distribution peaking around 8 M_☉ yr⁻¹, which tends to be shifted to lower values of SFR_total than that for LAEs. The median value found in this study is larger than that reported in Cowie et al. (2011) by using extinction-corrected Hα luminosities, SFR_{Hα-corrected} ∼ 6 M_☉ yr⁻¹.

Zoom In Zoom Out Reset image size

Figure 8 shows the ratio between SFR_IR and SFR_total against L_UV for LAEs, non-LAEs, and UV-faint SF galaxies. Considering the galaxies in the control sample, it can be seen that the ratio has a strong dependence on NUV luminosity: the IR contribution is lower with increasing L_UV. In Section 3, we found that UV-bright galaxies at z ∼ 0.3 are more like those detected in the FIR than UV-faint ones. Now, we have found that, for IR-detected galaxies, UV-bright ones have a lower IR contribution to SFR_total (are more transparent) than UV-faint ones. There is a limit in the NUV up to which SFR_IR is a good estimator of SFR_total. Considering ratios of 80% and 90%, SFR_IR is a good indicator of SFR_total for galaxies at z ∼ 0.3 with L_NUV ≲ 9 L_☉ and ≲ 9.5 L_☉, respectively. IR-detected LAEs at z ∼ 0.3 are all above the mentioned thresholds; therefore, both the UV and IR light contribute significantly to SFR_total. Despite this, the FIR contribution to SFR_total for most LAEs is greater than 60%, indicating that rest-frame UV-based methods would underestimate SFR_total by a factor greater than two. This is an indication of the great importance of FIR measurements when calculating the SFR: even in galaxies with low dust attenuation, such as LAEs, FIR emission makes a significant contribution; therefore, FIR data must be used to obtain accurate results.

Zoom In Zoom Out Reset image size

5.4. Morphology

In the previous sections, we have analyzed the physical properties of LAEs with direct IR and UV measurements. We now examine whether there is a difference between LAEs and non-LAEs regarding their morphology in the optical side of their SED. In order to carry out a precise morphological analysis, high spatial resolution images (i.e., Hubble Space Telescope (HST)/Advanced Camera for Surveys (ACS)) are required. In the case of non-LAEs, given that they are all located in the COSMOS field, there is ACS information for almost all of them, 116 (about 85%) to be precise. However, the IR-detected LAEs are distributed in different fields, some of which do not have available HST images. This limits the morphological study to eight IR-detected LAEs.

We take 12'' × 12'' I-band ACS cutouts of our sources (LAEs and control sample) and follow both an analytical and a visual procedure. The analytical one is aimed at obtaining the physical sizes of the galaxies, whereas the visual one has the objective of classifying the galaxies within the different types of the Hubble sequence. Note that we use the images of the galaxies in the same passband and, since they are all in the same redshift range, we are analyzing the same zone of their SED, with no need for morphological K-correction. In order to obtain the sizes of our galaxies, we fit Sersic profiles (Sersic 1968) to their light distributions by using GALFIT (Peng et al. 2010). Figure 9 shows the half-light radii of LAEs, non-LAEs, and UV-faint SF galaxies. It can clearly be seen that there is a remarkable difference: LAEs tend to be significantly smaller than non-LAEs. The median values of R_eff for LAEs, non-LAEs, and UV-faint galaxies are 1.5, 4.1, and 3.4 kpc, respectively. Note that non-LAEs are bigger than UV-faint galaxies owing to the difference in their UV brightness. The difference in size between LAEs and other UV-selected galaxies is also present at higher redshifts. Malhotra et al. (2011) report that LAEs tend to be smaller than Lyman break galaxies (LBGs) at the same redshift from z ∼ 2 up to z ∼ 4, while from z ≳ 5, LAEs and LBGs show similar sizes and have similar properties. At z ∼ 3–4, Vanzella et al. (2009) find that galaxies with strong Lyα absorption tend to be more diffuse than galaxies with Lyα in emission.

Zoom In Zoom Out Reset image size

At z ≳ 2, narrowband-selected LAEs have been reported to be compact (Bond et al. 2009, 2011; Malhotra et al. 2011). Bond et al. (2009) and Bond et al. (2011) report median values for R_eff of 1.0 and 1.4 kpc for their LAEs at z ∼ 3.1 and ∼2.1, respectively. The median value for LAEs that we obtain in this study is not significantly different from the values reported at higher redshifts; hence, a clear evolution in the median value of the physical size of LAEs is not seen from z ∼ 2 down to z ∼ 0.3.

Figures 10 and 11 show the ACS cutouts of LAEs and non-LAEs, respectively, with available ACS information. It can be seen that the vast majority of non-LAEs have clear disk-like morphologies, most of them belonging to the Sb, Sc, or Sd classes. However, LAEs tend to depart from that tendency. Although two of them seem to be disk-like galaxies, the other six LAEs have compact or interacting/merging morphologies. As was pointed out before, we only have eight LAEs with available ACS information, so the results are not statistically significant. However, despite the scarcity, the morphology of LAEs seems to be more heterogeneous than that of non-LAEs and this is an indication of a possible morphological differentiation.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Cowie et al. (2011) also studied the sizes and morphologies of their LAEs and UV-selected galaxies at z ∼ 0.3, examining the dependence of size upon the Hα equivalent width (EW(Hα)) as well. They report that LAEs with higher EW(Hα) are generally unresolved at the 1'' resolution of the Canada–France–Hawaii Telescope (CFHT) MegaPrime U-band images, while the galaxies with lower EW(Hα) are generally extended. Furthermore, at the same EW(Hα), they found that LAEs are significantly smaller than galaxies without Lyα emission. This agrees with the result we find here for our IR-detected LAEs, despite the results being obtained with optical observations in different bands. Cowie et al. (2010) analyzed the morphology of a sample of LAEs and other galaxies without Lyα emission by using the i'-band ground-based data from the CFHT MegaPipe database. In agreement with our result, they found that the LAE sample contains a much larger fraction of mergers and compact galaxies then the NUV-continuum-selected sample.

These significant morphological differences, along with the results found in previous sections, indicate that Lyα photons tend to escape preferentially from irregular or interacting galaxies of small size, low dust attenuation, and high SFRs.

In this study, we have obtained fundamental physical properties (UV and IR emission, dust attenuation, SFR, and morphology) of LAEs at z ∼ 0.3, defined as those Lyα-emitting galaxies with Lyα EW_rest-frame above 20 Å, the typical threshold in the narrowband searches at higher redshifts. Furthermore, we have compared LAEs with non-LAEs, defined as galaxies at the same redshift as LAEs which, with the same UV and IR limiting luminosities, show no Lyα emission with Lyα EW_rest-frame ≳ 20 Å in their spectrum. The main conclusions of our work are as follows.

• 1.  
  We find that a large percentage (∼75%) of our LAEs have MIPS/PACS IR counterparts under a limiting luminosity of log (L_IR/L_☉)  ∼  10.4. These mid-IR/FIR detections are a direct measurement of their dust emission.
• 2.  
  We find that 80% of the LAEs studied at z  ∼  0.3, both IR-detected and -undetected, are normal SF galaxies, L_IR  <  10¹¹ L_☉. We find only six LAEs with 10¹¹ ≲ L_IR[L_☉] ≲ 10^11.5 and none with L_IR > 10^11.6 L_☉ or with a ULIRG nature. The finding of a noticeable number of ULIRGs at z ∼ 2.5 exhibiting Lyα emission suggests that the IR nature of objects selected by means of their Lyα emission changes with redshift.
• 3.  
  For each value of L_IR, LAEs are among the least dusty galaxies at z ∼ 0.3. The distributions of the dust attenuation in 1200 Å of LAEs and non-LAEs are centered around 1.5 and 2.0, respectively. In this study, we have obtained dust attenuation by combining UV and FIR measurements without the uncertainties of rest-frame UV/optically based methods, which do not take into account the dust emission in the FIR.
• 4.  
  The dust attenuation of objects selected via their Lyα emission is evolving with redshift, from dust-free objects at z ∼ 3.0 to LAEs with high and low/moderate dust attenuation at z ∼ 2.3 and ∼0.3, respectively. However, it should be noted that the procedures followed in the different studies are not the same and that at z ≳ 3 very few LAEs have been individually studied without the uncertainties of stacking. The suggested evolution of dust attenuation of LAEs along with the finding that the UV and IR nature of LAEs is changing with redshift indicates that the physical properties of galaxies selected at different redshifts from their Lyα emission are not the same: either they are evolving or the technique is picking up galaxies of a different nature.
• 5.  
  LAEs have a much bluer UV continuum slope than non-LAEs, which is compatible with their being among the least dusty objects at their redshift. Furthermore, while most non-LAEs follow the trend of normal SF galaxies, some LAEs seem to be SB galaxies, indicating a possible difference in their mode of star formation. The distribution of LAEs between the SF and SB relations indicates that a unique relation between UV continuum slope and dust attenuation does not apply for the whole population. Therefore, the determination of dust attenuation from β fails for LAEs at z ∼ 0.3.
• 6.  
  The SFR_total for LAEs tends to be larger than for non-LAEs. We find a noticeable contribution of the IR emission to the SFR_total in LAEs and non-LAEs, about 60%. Therefore, although LAEs have a bright UV continuum and are among the least dusty galaxies at z ∼ 0.3, it is essential to take into consideration their dust emission in the FIR for obtaining accurate values of their SFR_total.
• 7.  
  The sizes of LAEs tend to be smaller than those of non-LAEs, with median values of 1.5 and 4.1 kpc, respectively. Despite the low number of LAEs with available ACS information, a visual inspection of their morphologies reveals that they tend to be compact, disk-like, or merging/interacting galaxies (i.e., there is a heterogeneity in morphology) in opposition to non-LAEs, which are mainly disk-like galaxies. These are the most noticeable differences between LAEs and non-LAEs.

We thank the anonymous referee for the comments provided, which have helped us to improve the manuscript. This work was supported by the Spanish Plan Nacional de Astrononomía y Astrofísica under grant AYA2008-06311-C02-01. Some of the data presented in this paper were obtained from the Multimission Archive at the Space Telescope Science Institute (MAST). STScI is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. Support for MAST for non-HST data is provided by the NASA Office of Space Science via grant NNX09AF08G and by other grants and contracts. Based on observations made with the European Southern Observatory telescopes obtained from the ESO/ST-ECF Science Archive Facility. Herschel is an ESA space observatory with science instruments provided by European-led Principal Investigator consortia and with important participation from NASA. The Herschel spacecraft was designed, built, tested, and launched under a contract to ESA managed by the Herschel/Planck Project team by an industrial consortium under the overall responsibility of the prime contractor Thales Alenia Space (Cannes), and including Astrium (Friedrichshafen) responsible for the payload module and for system testing at spacecraft level, Thales Alenia Space (Turin) responsible for the service module, and Astrium (Toulouse) responsible for the telescope, with in excess of a hundred subcontractors. PACS has been developed by a consortium of institutes led by MPE (Germany) and including UVIE (Austria); KUL, CSL, IMEC (Belgium); CEA, OAMP (France); MPIA (Germany); IFSI, OAP/AOT, OAA/CAISMI, LENS, SISSA (Italy); IAC (Spain). This development has been supported by the funding agencies BMVIT (Austria), ESA-PRODEX (Belgium), CEA/CNES (France), DLR (Germany), ASI (Italy), and CICYT/MICINN (Spain).