Profiles of Lyman $\alpha$ Emission Lines of the Emitters at $z=3.1$

A large number of the Ly$\alpha$ emitters at high redshift have been detected in sensitive search using the narrow-band filters (e.g., Hu & McMahon 1996; Steidel et al. 2000; Rhoads et al. 2001; 2003; Kodaira et al. 2003; Palunas et al. 2004; Hayashino et al. 2004; Matsuda et al. 2004; Taniguchi et al. 2005; Gronwall et al. 2006; Iye et al. 2006, Ouchi et al. 2008). From the photometric study, they are generally considered to be young star-forming objects and found to be widely distributed in their luminosity, size, and equivalent width (e.g., Gawiser et al. 2007; Finkelstein et al. 2007).

While the Ly$\alpha$ emission is one of the fundamental tools to study the galaxy formation at high redshift, it is, however, still difficult to understand what the dominant origins of the Ly$\alpha$ emission are, and how the Ly$\alpha$ photons escape from the galaxy. Resonance scattering and extinction by the neutral hydrogen or dust in galaxies may easily modify the observed flux and line profile. There are at least three different physical origins of the Ly$\alpha$ lines from high-redshift star-forming galaxies; photo-ionization by hot massive stars, cooling radiation from the gas heated either by the shock during to the gravitational collapse or by the shock due to the gas outflow driven by the thermal energy of frequent supernovae or active galactic nuclei. The Ly$\alpha$ photons produced by any of these process are affected by the neutral hydrogen gas mixed with or surrounding the emission-line regions. Recent results suggest that statistical average of the Ly$\alpha$ escape fraction of a sample of the star-forming galaxies at redshift $z\sim 2$ is only $\sim 5$$\%$ (Hayes et al. 2009).

It is important to study their spectroscopic properties, especially their line profiles to study how the lines are produced and modified. If the Ly$\alpha$ emission is largely reprocessed by the neutral gas with large bulk motion, we may observe the signatures in their line profiles. For example, Dijikstra et al. (2006) studied the expected Ly$\alpha$ profile of optically thick, spherically symmetric collapsing gas clouds. The expected profiles are generally blueshifted and show double peaked profile possibly dominated by the blue peak due to the combination of the infalling gas motion and line radiation transfer effects. On the other hand, Verhamme et al. (2006, 2008) modeled the line profiles expected for the expanding neutral hydrogen shell. Although the expected profiles are also double-peaked, in many cases they are redshifted and the red peak appears stronger due to the scattered components behind the original Ly$\alpha$ emission-line region in our line of sight. It is, however, not easy to discriminate these two cases from the observation of the individual Ly$\alpha$ line itself if we do not know the systemic velocity.

It is known that high-redshift star-forming galaxies typically show gas outflow from the sources. Pettini et al. (2001) and Shapley et al. (2003) compared the redshifts of the interstellar absorption line with the rest-frame optical emission lines which are nearly at the systemic velocity of the photo-ionized region and found that the inter-stellar absorption gas is generally blueshifted for typically 200 km s${}^{-1}$ indicating that the outflow gas motion is common among the Lyman Break Galaxies. Ly$\alpha$ lines in these objects also show P Cyg-type profiles, namely, a narrow emission line with a blueshifted absorption line. Verhamme et al. (2008) applied their models to the Ly$\alpha$ line profile of the Lyman Break Galaxies and found that they can be explained by the scattering in the simple expanding shell models. Steidel et al. (2010) also showed that the cool gas around $z=2$-3 Lyman Break galaxies typically exhibit outward motion.

On the other hand, the models of cold accretion (Dekel et al. 2009; Goerdt et al. 2009) also suggest that the observed luminosity, surface brightness, and size distributions of the Ly$\alpha$ emitters, especially for the extended Ly$\alpha$ Blobs (Steidel et al. 2000; Matsuda et al. 2004) can be explained by the gas heated by the collisional excitation in the cold gas stream.

Spectroscopic observations to study the line profiles of the Ly$\alpha$ emitters are essential to discriminate these cases although only a small sample of spectroscopy with enough high spectral resolution is available so far. Matsuda et al. (2006) studied the line profiles of the extended Ly$\alpha$ emitters, or Ly$\alpha$ Blobs as well as more ’ordinary’ Ly$\alpha$ emitters, and found that the Ly$\alpha$ lines of the blobs have more structure and the whole velocity width is positively correlated with their size.

In this paper, we present the results of our new spectroscopic observations for the 91 Ly$\alpha$ emitters to show their Ly$\alpha$ line profiles. The photometric data for the sample selection is similar as in Matsuda et al. (2006) but the number of the spectra for normal Ly$\alpha$ emitters are significantly increased. We describe our observations and data reduction in Section 2, the observed line profiles in Section 3, and the discussion in Section 4.

The sample of the Ly$\alpha$ emitters are selected based on the same data presented in Hayashino et al. (2004) at the field of SSA22-Sb1 where one of the largest overdensity of Lyman Break Galaxies as well as Ly$\alpha$ emitters at $z$=3.1 is known to exist (Steidel et al. 1998; 2000). The selection criteria for the emitters are, however, slightly modified to include the objects with relatively low equivalent width. We also did not use the $B-V$ color criteria, which was needed to make the very robust imaging sample of Ly$\alpha$ emitters to exclude the contamination by the foreground objects. We here applied only the following criteria.

or

where $BV$ is the effectively averaged $AB$ magnitude of the $B$ and the $V$-band and $NB497$ represents the $AB$ magnitude in the NB497 narrow-band filter (Hayashino et al. 2004).

The observations of the Ly$\alpha$ profiles of the emitters are done by the Faint Object Camera And Spectrograph (FOCAS) equipped with the Subaru 8.2m telescope in Jul 2004- Aug 2005. The VPH grating 600_450 was used to obtain the spectral dispersion of 0.37 Å per pixel. Using slits with a width of 1 arcsec and spectral resolution, $\lambda$/$\delta \lambda$ $\approx 1700$ was achieved as measured from the widths of the arc lines. In order to observe as many targets as we can in one exposure, we used the custom-made intermediate band filter with $\delta \lambda$ $\sim 200$Å to limit the spectral length on the detectors. By this method, we obtain the range of the spectra only around the emission lines, but the spectral resolution is high enough to clearly discriminate the [OII] 3727 Å doublet, which is the only significant contaminants, from the $z=3.1$ Ly$\alpha$ lines. The data was reduced in the standard manner using IRAF. The accuracy of the wavelength calibration is higher than 0.2 Å . The one-dimensional spectra are then extracted by averaging the three spatial pixels, each 0.$"$2. We also smoothed the spectra by three pixels in the dispersion direction.

As a result, we obtained the Ly$\alpha$ emission-line spectra for the 91 objects. The sample contains the 12 objects classified as Ly$\alpha$ Blobs by Matsuda et al. (2004) whose isophotal area above the surface brightness of $\approx 7\times {10}^{-17}$ erg s${}^{-1}$ cm${}^{-2}$ arcsec${}^{-2}$ is larger than 16 arcsec${}^2$ or 900 kpc${}^2$ at $z=3.1$. Matsuda et al. (2004; 2006) showed that the photometric or the spectroscopic properties of the Ly$\alpha$ Blobs are not quite discriminated from other objects but rather continuously distributed in the whole sample of the emitters.

As described in Matsuda et al. (2005) and Matsuda et al. (2006), the sample selected by the criteria in Hayashino et al. (2004) is almost free for contamination ($<1\%$). For the objects with lower equivalent width (i.e., $0.5<BV-NB497<1.2$), the fraction of the [OII] emitter at $z=0.33$ increases, as expected, but still less than $\sim 5$% . We identified in total of 91 spectroscopic Ly$\alpha$ emitters in the 11 masks located in the SSA22-Sb1 field.

Fig.1a-1f show the observed Ly$\alpha$ spectrum for all the 91 objects. Two of them (LAE J221709.6+001801 and LAE J221720.2+002019) clearly show a very large line width ($\gtrsim$ 1000 km s${}^{-1}$) and possibly host active galactic nuclei (AGN). Table 1 summarizes their observed properties.

In Fig.1, we first notice that many of the objects show the characteristic profiles, namely the double-peaked spectra with a strong, asymmetrical red peak and a much weaker blue peak. While the most conspicuous cases are the objects LAE J221720.3+002438, LAE J221745.3+002006, and LAE J221759.2+002254 (LAB28 in Matsuda et al. 2004), many other cases can be recognized in Fig.1. As a large fraction of the observed Ly$\alpha$ spectra shows the similar profiles, it may represent the common physical properties of the emitters.

We therefore tried to characterize the profile and to evaluate the fraction of the objects with the similar profile among those observed in more objective way. We first fitted the visually isolated peaks in each spectrum with the Gaussian profiles. These procedures are rather formal ones but useful in the following discussions. Fig.2 and Fig.3 show the example of the fitting for the two objects with high and moderate signal-to-noise ratio, respectively. Fig.2a and 3a are for the multiple Gaussian fitting. The green lines show the each component and the blue ones the total. The red dashed lines in the figures show the noise level measured at the same wavelength by using the blank part of the slits. Bottom panels show the residual. On the other hand, Fig.2b and 3b show the examples of the single Gaussian fitting similar with those tried in Matsuda et al. (2006). For the objects with multiple peaks, the data points between the peaks are not used for the fitting (thin crosses in Fig.2b and 3b). The center of the wavelength in each panel in Fig.1 is the central wavelength of the single-component Gaussian fitting. The central wavelengths of the primary red and secondary blue peaks (for the objects with the feature) as well as that of the single-component fitting (for all the objects) are also presented in Table 1.

Of the 89 objects except for the two broad-line AGN, 50 objects have the visually identified strong blue and weak red peaks. 4 other objects (221728.3+001212, 221740.3+001129, 221740.9+001125 , and 221812.5+001433) do not show the weak blue peaks but have notable components to the red of the strongest peak. Three of these four objects, 221728.3+001212, and closely located objects 221740.3+001129 and 221740.9+001125 are associated with the Ly$\alpha$ Blobs LAB33 and LAB7 in Matsuda et al. (2004), respectively.

For these 50 candidates of the “strong red and weak blue” profiles, we then evaluated the significance of the weak peaks relative to the noise level. We evaluated the r.m.s. noise in the blank-sky spectra (the red dashed lines in Fig.2 and Fig.3 for the example) at the line core ($\pm 2$${\sigma }_{\lambda }$ of the Gaussian) and then selected only the objects with the weak peaks which are more significant than the 3$\sigma$ level as the final sample of the “strong red and weak blue” profiles. For the cases between 3$\sigma$ and 4$\sigma$, we further visually checked the spectra and rejected a few cases. After all, 39 objects among the 89 (44$\%$) show the significant characteristic profile. The wavelength of the primary and the secondary peaks are shown by the vertical dotted lines in Fig.1. This is a large fraction, which implies that the “strong red and weak blue” profile is a common characteristic property of the observed Ly$\alpha$ emitters. The fraction should be considered even as the lower limit because it is more difficult to detect the secondary weaker peaks significantly if the objects are faint or their emission-lines are weak. Indeed, if we limit the objects with $NB497<25$ and further with $BV-NB497>1.0$, the fraction is even larger, 53 and 55$\%$, respectively. Table 2 summarizes the number of the objects with the characteristic profile. We also listed the wavelength of the secondary component for five marginal objects whose blue peaks were visually identified but not found to be significant in Table 1 with ’:’.

While it is difficult, only from such formal fitting procedure as shown in Fig.2 and Fig.3, to tell whether the peaks are the multiple emission lines with different central velocities or the single line modified by the absorption and/or scattering by neutral gas in vicinity, the latter interpretation is plausible as such large fraction of the sample always show much stronger peak at the long wavelength. We confirm this further by the following two different methods.

First, we investigated the asymmetry of the strongest peaks of the sample. We identify the pixel with the largest flux and obtained the ratio of the flux integrated over the 4.5Å (about 1.5 times the spectral resolution) toward longer wavelength to those toward shorter wavelength. Fig.4 shows the result for the objects with $NB497<25.0$ and $BV-NB407>1.0$, avoiding the objects with relatively low signal to noise ratio. The symmetry index (the ratio) of those with the characteristic “strong red and weak blue” profile is strongly concentrated around 0.5 while those of other objects are distributed around the unity (The median is 0.83) with the r.m.s. scatter of $\sim 0.25$. 12 objects with characteristic profile has the value below 0.6 and they indeed show the conspicuous “strong red and weak blue” profiles. We applied the Kolmogorov-Smirnov Test to the symmetry distribution of the objects with and without the characteristic profile and the probability that they are drawn from the same population is 7%. We also changed the magnitude or color range as well as the wavelength region for the index and found that the trend is always seen.

Next, for those objects with the characteristic profile, we compared the observed width of the stronger red peak and the separation of the peaks obtained in the formal Gaussian fitting. Fig.5 shows the result; the clear correlation between these two quantities is seen. This strongly support that the two peaks are indeed the parts of a single feature. If they are multiple different components, such trend as seen in Fig.5 is not necessary to be observed. Fig.6 shows the distribution of the observed FWHM of the most prominent peak of the each object. Those with the characteristic profile tend to show the smaller values.