A Close Comparison between Observed and Modeled Ly Lines
for Lyman Alpha Emitters
We present the results of a Ly profile analysis of 12 Ly emitters (LAEs) at with high-resolution Ly spectra. We find that all 12 objects have a Ly profile with the main peak redward of the systemic redshift defined by nebular lines, and five have a weak, secondary peak blueward of the systemic redshift (blue bump). The average velocity offset of the red main peak (the blue bump, if any) with respect to the systemic redshift is km s ( km s), which is smaller than (comparable to) that of Lyman-break galaxies (LBGs). The outflow velocities inferred from metal absorption lines in three individual and one stacked spectra are comparable to those of LBGs. The uniform expanding shell model constructed by Verhamme et al. (2006) reproduces not only the Ly profiles but also other observed quantities including the outflow velocity and the FWHM of nebular lines for the non-blue bump objects. On the other hand, the model predicts too high FWHMs of nebular lines for the blue bump objects, although this discrepancy may disappear if we introduce additional Ly photons produced by gravitational cooling. We show that the small values of our sample can be explained by low neutral-hydrogen column densities of log() = 18.9 cm on average. This value is more than one order of magnitude lower than those of LBGs but is consistent with recent findings that LAEs have high ionization parameters and low Hi gas masses. This result suggests that low values, giving reduced numbers of resonant scattering of Ly photons, are the key to the strong Ly emission of LAEs.
Subject headings:galaxies: high-redshift — galaxies: ISM — line: profiles — radiative transfer
Ly emitters (LAEs) are objects commonly seen in both the local and high- universes with large Ly equivalent widths, EW(Ly) Å(local: Deharveng et al. 2008; Cowie et al. 2011, high-: Hu & McMahon 1996; Rhoads & Malhotra 2001; Ouchi et al. 2008, 2010). Previous studies based on Spectral Energy Distributions (SEDs) have revealed that typical LAEs are young, low-mass galaxies with a small dust content (e.g., Nilsson et al. 2007; Gawiser et al. 2007; Guaita et al. 2011; Nakajima et al. 2012; Kusakabe et al. 2015), although there are some evolved LAEs with a moderate mass and dust (Ono et al. 2010b; Hagen et al. 2014). Morphological studies of their UV continuum have shown that the galactic counterparts of LAEs to be typically compact (e.g., Bond et al. 2009) and their typical size does not evolve with redshift (Malhotra et al. 2012). Furthermore, clustering analyses have revealed that LAEs have the lowest dark matter halo masses at every redshift (Ouchi et al. 2010; Guaita et al. 2010). These properties suggest that LAE is an important galaxy population as the building block candidates in the CDM model (Rauch et al. 2008).
Given their importance in galaxy evolution, the Ly escape mechanism in LAEs is still poorly understood. Resonant scattering strongly extends the path-length of Ly photons through galactic gas and renders them prone to absorption by dust grains. On one hand, some observational studies at the local universe have proposed that outflows facilitate the escape of Ly photons from galaxies (e.g., Kunth et al. 1998) as they reduce the number of scattering. Likewise, others (e.g., Kornei et al. 2010; Atek et al. 2014) have shown that the dust content correlates with Ly emissivity. While these effects would certainly be at work, there has been no decisive conclusion (cf., Cassata et al. 2015). On the other hand, theoretical studies have computed the Ly radiation transfer (RT) through idealized spherically symmetric shells of homogeneous and isothermal neutral hydrogen gas, especially in a form of an expanding shell (e.g., Zheng & Miralda-Escudé 2002; Verhamme et al. 2006; Dijkstra & Loeb 2009; Kollmeier et al. 2010). They have investigated how properties of the interstellar medium (ISM) affect the Ly escape and emergent Ly profiles. The result is that the Ly RT is a complicated process altered by galactic outflows/inflows, the neutral hydrogen column density and dust content of the ISM, and the inclination of the galaxy disk (e.g., Verhamme et al. 2012; Zheng & Wallace 2014; Behrens & Braun 2014). One of the goals in these theoretical studies is to aid understanding the galaxy properties from observed Ly lines, and to identify the key factor for the Ly escape.
To study the Ly RT and escape through close comparisons of observed and modeled Ly lines, it is important to obtain spectral lines other than the Ly line. The central wavelength and the width of nebular lines (e.g., H and [Oiii]) tell us the galaxy’s systemic redshift and internal velocity. The blue-shift of interstellar (IS) absorption lines with respect to the systemic redshift gives the galactic outflow velocity, and the width of the IS lines can be interpreted as the sum of thermal and macroscopic (rotation and turbulence) velocities of the outflowing gas. These lines can help us to disentangle the complicated Ly RT and understand the Ly escape.
However, due to the typical faintness of LAEs, it is only recently that these additional lines have been successfully detected in narrow-band selected LAEs (nebular lines: e.g., McLinden et al. 2011; Finkelstein et al. 2011; Hashimoto et al. 2013, IS absorption lines: Hashimoto et al. 2013; Shibuya et al. 2014b). Thus, in contrast to LBGs whose Ly profiles have been closely compared with Ly RT models (e.g., Verhamme et al. 2008; Kulas et al. 2012; Christensen et al. 2012), there are only a few studies that have performed Ly profile comparisons of LAEs (e.g., Chonis et al. 2013). Recent simultaneous detections of Ly and nebular emission lines have statistically confirmed that the Ly profiles of LAEs are asymmetric with a red main peak redshifted with respect to the systemic , 0 km s (e.g., Shibuya et al. 2014b; Song et al. 2014; Erb et al. 2014). Likewise, IS absorption studies in LAEs have shown that they are blue-shifted with respect to the systemic by km s (Hashimoto et al. 2013; Shibuya et al. 2014b), which is comparable to those of LBGs (e.g., Pettini et al. 2001; Shapley et al. 2003; Steidel et al. 2010; Kulas et al. 2012). These results suggest that LAEs do have outflows and motivate us to apply expanding shell models to LAEs.
To examine Ly escape mechanisms in LAEs through detailed Ly modeling, we focus on the small of LAEs, 200 km s, compared to those of LBGs, 400 km s (e.g., Steidel et al. 2010; Kulas et al. 2012), with similar physical quantities such as stellar mass, star formation rate (SFR), and velocity dispersion (Hashimoto et al. 2013; Shibuya et al. 2014b; Song et al. 2014; Erb et al. 2014). Hashimoto et al. (2013) and Shibuya et al. (2014b) have also shown that LAEs have comparable outflow velocities, measured from IS absorption lines, to those of LBGs. These results imply that a definitive difference between LAEs and LBGs in velocity properties is . In addition, Hashimoto et al. (2013) have demonstrated that EW(Ly) anti-correlates with using a large sample of LAEs and LBGs (see also Shibuya et al. 2014b; Erb et al. 2014). Therefore, understanding the reason why LAEs have small , through detailed Ly modeling should shed light on the Ly RT and Ly escape mechanisms in LAEs.
According to the theoretical studies, there are several possible explanations for a small : a high outflow velocity (e.g., Verhamme et al. 2006), a very low neutral hydrogen column density () of the ISM (e.g., Verhamme et al. 2006, 2015), an inhomogeneous ISM with a covering fraction () below unity, where is defined as the fraction of sightlines which are optically thick to the Ly radiation, i.e., gas with holes (e.g., Behrens et al. 2014; Verhamme et al. 2015), and a clumpy ISM with a low covering factor, , which is defined as the average number of clouds intersected by a random line of sight (e.g., Hansen & Oh 2006; Dijkstra & Kramer 2012; Laursen et al. 2013)
In this work, we focus on applying the uniform expanding shell model based on a 3D Ly RT constructed by Verhamme et al. (2006) and Schaerer et al. (2011), to 12 LAEs whose Ly and nebular emission lines (e.g., H, Oiii) have been observed at a high spectral resolution (Hashimoto et al. 2013; Nakajima et al. 2013; Shibuya et al. 2014b). With the systemic redshifts and the full width half maximums (FWHM) determined from nebular emission lines, the stellar dust extinction derived from SED fitting, and the galactic outflow velocities inferred from LIS absorption lines, we first statistically examine how well the model can reproduce the Ly profiles and other observables (cf., Verhamme et al. 2008; Kulas et al. 2012; Chonis et al. 2013). After demonstrating the validity of the model, we securely derive physical quantities such as and discuss the origin of the small and implications for the Ly escape in LAEs. Possible other scenarios mentioned above for the small are also qualitatively discussed.
This paper is organized as follows. We describe our spectroscopy observations in Section 2, and discuss profiles of Ly and nebular emission lines in Section 3. We apply the uniform expanding shell model to our data and show comparisons with observables in Section 4. Discussion on the blue bumps as well as the origin of the small Ly velocity offsets are given in Section 5, followed by conclusions in Section 6.
Throughout this paper, magnitudes are given in the AB system (Oke & Gunn 1983), and we assume a CDM cosmology with , and km s Mpc.
2. Data and Observations
Our initial sample of objects are taken from large LAE samples in the COSMOS field, the Chandra Deep Field South (CDFS), and the Subaru/XMM-Newton Deep Survey (SXDS) (Nakajima et al. 2012, 2013; Nakajima et al. in prep.). These LAE samples are all based on Subaru/Suprime-Cam imaging observations with our custom made narrow band filter, NB387 ( = 3870Å and FWHM = 94Å). The LAEs have been selected by color criteria of NB387 and NB387, satisfying the condition that the rest frame photometric Ly EW (EW(Ly)) be larger than 30Å. From these, we only use 12 LAEs whose Ly and nebular emission lines (e.g., H and [Oiii]) are both spectroscopically confirmed. Among the 12 objects, 11 have been presented in Hashimoto et al. (2013) and Shibuya et al. (2014b). We add one new LAE with EW(Ly) 280Å whose detailed properties will be discussed in Hashimoto et al. in prep.
2.1. Near-Infrared Specctroscopy
In order to detect nebular emission lines, we performed three near-infrared observations with Magellan/MMIRS (PI: M. Ouchi), Keck/NIRSPEC (PI: K. Nakajima), and Subaru/FMOS (PI: K. Nakajima). Canonical spectral resolutions for our observation settings are 1120, 1500, and 2200 for MMIRS, NIRSPEC, and FMOS, respectively.
Details of the observation and data reduction procedures for MMIRS and NIRSPEC have been presented in Hashimoto et al. (2013) and Nakajima et al. (2013), respectively. Briefly, two CDFS objects, CDFS-3865 and CDFS-6482, were observed with MMIRS using the HK grism covering m, resulting in successful H and [Oiii] detections. A follow-up observation was carried out for CDFS-3865 with NIRSPEC. The [Oii] line was additionally detected with the band ( m). Four COSMOS objects, COSMOS-08501, COSMOS-13636, COSMOS-30679, and COSMOS-43982, were observed with NIRSPEC and its band ( m), resulting in H line detections. The [Oiii] line was also detected from COSMOS-30679 using the band ( m).
The data from FMOS will be presented in Nakajima et al. (2015, in prep). Its spectral coverage is m. We detected [Oiii] line(s) in eight objects: COSMOS-08357, COSMOS-12805, COSMOS-13138, COSMOS-13636, COSMOS-38380, COSMOS-43982, SXDS-10600, and SXDS-10942.
2.2. Optical Spectroscopy
In order to detect Ly and metal absorption lines, we carried out several observations with Magellan/MagE (PI: M. Rauch) and Keck/LRIS (PI: M. Ouchi). The spectral resolutions for our observations were and for MagE and LRIS, respectively. The slit was positioned on the Ly centroids in the NB387 images.
Details of the observation and data reduction procedures for MagE and LRIS have been presented in Hashimoto et al. (2013) and Shibuya et al. (2014b), respectively, except for COSMOS-08501. First, we describe this new object in detail (§2.2.1) and then giver a brief summary for the rest of the sample (§2.2.2).
Optical Spectroscopy for COSMOS-08501
The MagE observations were carried out for COSMOS-08501 on 2012 February and 2013 December. We obtained s and s exposure times during each run, resulting in a 12000 s total integration time. Spectroscopic standard stars, dome flats, and Xenon flash lamp flats, were obtained on each night for calibrations. On these nights, the typical seeing sizes were . The slit width was for both runs, corresponding to . The spectra were reduced with based pipeline, , constructed by G. Becker (see also Kelson 2003). This pipeline processes raw frames, performing wavelength calibration and optimal sky subtraction, and extracts 1D spectra. Each of these reduced frames was then combined to form our final calibrated spectrum. From this, the Ly line was identified above the 3 noise of the local continuum.
Optical Spectroscopy for the Rest of the Sample
CDFS-3865, CDFS-6482, and COSMOS-30679 were observed with MagE; COSMOS-08357, COSMOS-12085, COSMOS-13138, COSMOS-38380, SXDS-10600, and SXDS-10942 were observed with LRIS; and finally, COSMOS-13636 and COSMOS-43982 were observed with both spectrographs. We identified the Ly line in all objects. In addition, we detected several metal absorption lines (e.g., Si ii and Civ lines) in a stacked MagE spectrum of CDFS-3865, CDFS-6482, COSMOS-13636, and COSMOS-30679 (Hashimoto et al. 2013) as well as in individual LRIS spectra of COSMOS-12805, COSMOS-13636, and SXDS-10600 (Shibuya et al. 2014b).
A summary of our observations is listed in Table 1, and our Ly and nebular emission line profiles are shown in Figure 1.
|Object||(J2000)||(J2000)||EW(Ly)||NIR obs.||opt. obs.||Source|
|(Å)||(10 erg s)|
|CDFS-3865||03:32:32.31||-28:00:52.20||NIRSPEC (J)||MagE||H13, N13|
|CDFS-6482||03:32:49.34||-27:59:52.35||MMIRS (HK)||MagE||H13, N13|
|COSMOS-30679||10:00:29.81||+02:18:49.00||NIRSPEC (H and K)||MagE||H13, N13|
|COSMOS-13636||09:59:59.38||+02:08:38.36||FMOS (H)||MagE and LRIS||H13, N13, S14|
|COSMOS-43982||09:59:54.39||+02:26:29.96||MMIRS (HK)||MagE and LRIS||H13, N13, S14|
|COSMOS-08357||09:59:59.07||+02:05:31.60||FMOS (H)||LRIS||S14, N15|
|COSMOS-12805||10:00:15.29||+02:08:07.50||FMOS (H)||LRIS||S14, N15|
|COSMOS-13138||10:00:02.61||+02:08:24.50||FMOS (H)||LRIS||S14, N15|
|COSMOS-38380||09:59:40.94||+02:23:04.20||FMOS (H)||LRIS||S14, N15|
|SXDS-10600||02:17:46.09||-06:57:05.00||FMOS (H)||LRIS||S14, N15|
|SXDS-10942||02:17:59.54||-06:57:25.60||FMOS (H)||LRIS||S14, N15|
(1) Object ID;
(2), (3) Right Ascension and Declination;
(4), (5) rest-frame Ly EW and luminosity derived from narrow- and broadband photometry;
(6) Instruments and filters used for the NIR observations;
(7) Instruments used for the optical observations;
(8) Source of the information
2.3. AGNs in the Sample
In short, for the MagE objects, we inspected it in three ways. We first compared the sky coordinates of the objects with those in very deep archival X-ray and radio catalogs. Then we checked for the presence of high ionization state lines such as Civ 1549 and He ii lines in the spectra. Finally, we applied the BPT diagnostic diagram (Baldwin et al. 1981) to the objects. No AGN activity is seen except for COSMOS-43982 whose high [Nii] /H line ratio is consistent with that of an AGN.
On the other hand, due to the lack of H or [Nii] data, we were only able to use the two forms of investigation for the LRIS objects. Of these, only COSMOS-43982 showed clear detection of the Civ 1549 line in its optical spectrum.
In summary, we have ruled out AGN activity in all but COSMOS-43982.
3. Observational Results
3.1. Line Center and FWHM Measurements for Nebular Emission Lines
Line center (i.e., redshift) and FWHM measurements of nebular emission lines are crucial for a detailed modeling of the Ly line, since they encode information on the intrinsic (i.e., before being affected by radiative transfer) Ly redshift and FWHM. In order to obtain these parameters and their uncertainties, we apply a Monte Carlo technique as follows. First, for each line of each object, we measure the 1 noise of the local continuum. Then we create 10 fake spectra by perturbing the flux at each wavelength of the true spectrum by the measured 1 error (Kulas et al. 2012; Chonis et al. 2013). For each fake spectrum, the wavelength at the highest flux peak is adopted as the line center, and the wavelength range encompassing half the maximum flux is adopted as the FWHM. The standard deviation of the distribution of measurements from the 10 artificial spectra is adopted as the error on the line center and FWHM. When multiple lines are detected, we adopt a weighted mean value of them. A summary of the measurements are listed in the columns 2 and 3 of Table 2. All redshift (FWHM) values are corrected for the LSR motion (instrumental resolution). When the line is unresolved, the instrumental resolution is given as an upper limit. The mean FWHM value for a sample of eight objects with a measurable velocity dispersion is FWHM(neb) km s, which is smaller than that of LBGs, FWHM(neb) km s (Pettini et al. 2001; Erb et al. 2006a; Kulas et al. 2012). This is consistent with the recent results by Erb et al. (2014), who have found that the median FWHM(neb) of 36 LAEs is 127 km s. These results indicate that LAEs have smaller dynamical masses than LBGs.
|(km s)||(km s)||(km s)||(km s)||(Å)|
Note. – The symbol “-” indicates we have no measurement. (1) Object ID; (2) Systemic redshift derived from the weighted mean of the nebular emission redshifts; (3) Weighted mean FWHM of nebular emission line; (4) Presence of a blue bump emission in the Ly profile; (5) Velocity offset of the Ly main red peak with respect to ; (6) Velocity offset of the Ly blue-bump with respect to ; (7) Separation between and ; (8) Rest-frame Ly EW derived from spectroscopy; and (9) Weighted skewness of the Ly line.
3.2. Two Component [Oiii] Profiles
Among the nebular emission lines we have obtained, while most objects show normal symmetric Gaussian profiles, COSMOS-13138 and SXDS-10600 show an asymmetric [Oiii] profile with a secondary blueshifted and redshifted component, respectively (see Figure 2). Such a profile has been reported in various objects: both local and high- star-forming galaxies and ULIRGs (e.g., Shapiro et al. 2009; Genzel et al. 2011; Newman et al. 2012; Soto et al. 2012), a high- Oxygen-Two Blob ([Oii] blob) (Harikane et al. 2014), and a few Lyman-Alpha Blobs (LABs) (Yang et al. 2014). However in LAEs, there has been no study which reports its presence.
Aforementioned studies apply a two Gaussian components fit with a narrow and broad components to the line. To examine the presence of two components, we also perform a fit with two Gaussians. We have six parameters: fluxes, line centers, and FWHMs for both components. We require that the widths of both components are larger than the spectral resolution, and that the broad component has a larger FWHM than the narrow component. Best fit parameters are determined through minimum realizations, and the parameter range satisfying + 1 is adopted as the error, where denotes the minimum value. The results are listed in Table 3. For each object, both components are significantly detected with , demonstrating that some fraction of LAEs have two-component line profiles.
The velocity offsets of the two components are km s (COSMOS-13138) and km s (SXDS-10600), respectively.
The FWHM values of the broad component after correction for instrumental resolution are and km s. These are much smaller than those of the star forming galaxies at (FWHM km s, Genzel et al. 2011), and slightly smaller than those of the [Oii] blob (FWHM km s) of Harikane et al. (2014) and the LABs (FWHM km s) of Yang et al. (2014). Our small values exclude the possibility of the broad component originating from an AGN activity (cf., Osterbrock & Ferland 2006) or a powerful outflow driven by a starburst (e.g., Shapiro et al. 2009; Genzel et al. 2011; Newman et al. 2012) because in these cases, the FWHM of the broad component should be as large as km s. It is possible that the two component lines originate from two large star-forming regions (e.g., Harikane et al. 2014) or mergers. As discussed in Harikane et al. (2014), the velocity offset of the two components, km s, may be due to a rotation of the objects.
|(10 erg s cm)||(10 erg s cm)||(km s)||(km s)|
Note. – (1) Object ID; (2), (3) Fluxes of the narrow () and the broad () components. Note that the values are not corrected for the slit loss; (4), (5) Redshifts of the narrow () and the broad () components; (6), (7) FWHM measurements of the narrow (FWHM) and the broad (FWHM) components.
3.3. Ly Profile with a Blue Bump
While the majority of Ly profiles are single-peaked (e.g., Shapley et al. 2003; Steidel et al. 2010), a fraction of Ly profiles are known to be multiple-peaked (e.g., Rauch et al. 2008; Yamada et al. 2012; Kulas et al. 2012). In particular, we shall refer to a secondary small peak blueward of the systemic redshift as “the bluebump” (see the case 2 profile in Figure 12 in Verhamme et al. 2006). Theoretical studies have shown that the blue bump is a natural outcome of the radiative transfer in a low speed galactic outflow (e.g., Zheng & Miralda-Escudé 2002).
We consider a blue bump to be detected if there exists an excess emission blueward of the systemic redshift above noise of the local continuum. We detect a blue bump of five objects; the MagE ones of CDFS-3865 and COSMOS-43982, and the LRIS ones of COSMOS-12805, COSMOS-13138, COSMOS-43982, and SXDS-10942 (the column 4 of Table 2). The position of the blue bump is designated by a blue arrow in Figure 1.
The frequency of blue-bump objects in the sample is (5/12). There are four LAEs in the literature that have a blue bump: one among the two LAEs studied in McLinden et al. (2011) and all three LAEs studied in Chonis et al. (2013). For the total sample of 17 LAEs, the frequency is calculated to be (9/17). Note that this is a lower limit due to the limited spectral resolution. On the other hand, Kulas et al. (2012) have studied 18 LBGs with measurements which are preselected to have multiple-peaked Ly profiles. They have argued that of the parent sample are multiple-peaked and that 11 out of the 18 objects have a blue bump, indicating that the blue bump frequency in LBGs is (). These results imply that the blue bump feature is slightly more common in LAEs than in LBGs although a larger sample observed at higher spectral resolution is needed for a definite conclusion.
3.4. Ly Velocity Properties
We derive three velocity offsets related to the Ly line: the velocity offset of the main red peak of the Ly line with respect to the systemic redshift,
that of the blue bump of the Ly line with respect to the systemic redshift, if any,
and that of the two peaks,
where , , and represent the systemic redshift, the Ly redshift of the main red peak, and that of the blue bump, respectively.
Ly Main Red Peak Velocity Offsets,
We estimate the value using a Monte Carlo technique in a similar manner to that in §3.1. First, for each object, we measure the error in the Ly spectrum set by the continuum level at the wavelength longer than Å. Then we create 10 fake spectra converted to velocity space by simultaneously perturbing the flux at each wavelength and the systemic redshift listed in Table 2 by their errors. Finally, we measure the velocity at the highest flux peak. The mean and the standard deviation value of the distribution of 10 measurements are adopted as the and its error, respectively. The derived values are listed in the column 5 of Table 2, ranging from km s to km s with a mean value of km s. In most cases, these values are consistent with those measured in Hashimoto et al. (2013) and Shibuya et al. (2014b) within , however, they are not for COSMOS-08357 and COSMOS-12805. This is due to the fact that these studies have applied a symmetric/asymmetric profile fit to the Ly line. In Figure 1, we show the two values derived from the Monte Carlo and the profile fit technique as the orange and green line segments, respectively. For the sake of consistency in the definition of the in the shell model (Verhamme et al. 2006; Schaerer et al. 2011), we adopt here the new measurements. We note that our discussion is unchanged even if we adopt the previous values.
The value has been measured in more than 60 LAEs (McLinden et al. 2011; Finkelstein et al. 2011; Hashimoto et al. 2013; Guaita et al. 2013; Chonis et al. 2013; Shibuya et al. 2014b; Song et al. 2014; Erb et al. 2014). These studies have shown that LAEs at have a mean of km s, which is significantly smaller than that of LBGs at a similar redshift, km s (e.g., Steidel et al. 2010; Rakic et al. 2011; Kulas et al. 2012). The left panel of Figure 3 represents the histogram of for the 12 LAEs (14 spectra) studied in this study and 18 LBGs given by Kulas et al. (2012). We carry out the Kolmogorov-Smirnov (K-S) test for the two populations. The resultant probability is , indicating that is definitively different between LAEs and LBGs.
Ly Blue Bump Velocity Offsets,
For each detected blue bump in §3.3, we measure value in the same manner as for . We obtain km s (CDFS-3865), km s (MagE-COSMOS-43982), km s (COSMOS-12805), km s (COSMOS-13138), km s (LRIS-COSMOS-43982), and km s (SXDS-10942) as listed in the column 6 of Table 2. We have obtained two different measurements for COSMOS-43982 due to the spectral resolution effect, however, they are consistent with each other within (see also §4.3.3). We combine our measurements with those in the four aforementioned LAEs with a blue bump to construct a large sample of LAEs with a blue bump consisting of 9 objects (10 spectra): one from McLinden et al. (2011) with km s and three from Chonis et al. (2013) with , and km s. The mean value of the large sample is km s, which is consistent with that of 11 LBGs with a blue bump, km s (Kulas et al. 2012). We calculate the K-S probability to be 0.3901, indicating that LAEs’ values are comparable to LBGs’. The middle panel of Figure 3 shows the distribution for the LAE and LBG samples.
We check if our conclusion remains unchanged even if the spectral resolution effect is taken into account. The sample by Kulas et al. (2012) has been obtained with three settings: 300-line grating, 400-, and 600-line grisms, corresponding to a spectral resolution of and , respectively. We compare the mean value of our four LAEs taken by LRIS () and that of six LBGs with a blue bump obtained at a similar resolution (). The resultant mean values for LAEs and LBGs are and km s, respectively, and the K-S probability is 0.9238. Thus, we obtain the same conclusion.
Velocity Offsets Between the Main Red Peak and the Blue Bump,
Finally, for each of the spectra with a blue bump, we measure the velocity offset between the red and blue peaks: = km s (CDFS-3865), km s (MagE-COSMOS-43982), km s (COSMOS-12805), km s (COSMOS-13138), km s (LRIS-COSMOS-43982), and km s (SXDS-10942), as listed in the column 7 of Table 2. In order to make a large sample with measured, we utilize again the four LAEs with the blue bump from the literature: one LAE studied in McLinden et al. (2011) with km s and three LAEs studied in Chonis et al. (2013) with , and km s. The mean value of the nine objects (ten spectra) is km s, which is significantly smaller than the value derived for 11 LBGs with a blue bump, km s (Group I in Kulas et al. 2012). The K-S probability is calculated to be , indicating that LAEs and LBGs have distinctive values. See the right panel of Figure 3 for their distributions.
We examine the spectral resolution effect exactly the same manner as in §3.4.2. The mean value of the four LAEs taken by LRIS () and that of the six LBGs with a blue bump obtained at a similar spectral resolution () are and km s, respectively. In conjunction with the K-S probability, 0.09524, we conclude that LAEs have a significantly smaller value than that of LBG even at the same spectral resolution. Our finding is recently supported by Henry et al. (2015) and Yang et al. (2015), who have examined Ly velocity properties and their relations to the Ly escape fraction for local galaxies called “Green Peas” galaxies (Cardamone et al. 2009). They have found that the Ly escape fraction is higher for objects with smaller .
In summary, we have derived three Ly velocity offsets, , , and . While we need a larger sample of objects with a blue bump for a definite conclusion, we find that LAEs have a smaller (comparable) () value relative to LBGs, which makes their value also smaller than that of LBGs.
Note. – The symbol “-” indicates no measurement. (1) Object ID; (2) Mean velocity offset of LIS absorption lines with respect to ; (3) Stellar mass estimated from SED fitting; (4) Stellar dust extinction estimated from SED fitting; (5) Presence of merger examined via close-pair method and system studied in Shibuya et al. (2014a); (6) Ellipticity defined as , where and are the major and minor axes, respectively.
3.5. Other Physical Quantities
In this section, we describe other physical quantities related to with this work. We describe metal absorption line properties in §3.5.1, SED fitting properties in §3.5.2, and morphological properties in §3.5.3.
Metal Absorption Line Properties
Low ionization state (LIS) metal absorption lines encode information on cold neutral gas in galaxies. The mean blueshift of LIS absorption lines with respect to the systemic velocity, , gives the average speed of the galactic outflow (e.g., Pettini et al. 2001; Shapley et al. 2003; Martin 2005). In the following sections, we compare the values of our LAE sample with the results from Ly radiation transfer fitting.
Shibuya et al. (2014b) have detected several LIS absorption lines in a few narrowband-selected LAEs on the individual basis. The derived mean blue shifts are km s (COSMOS-13636), km s (COSMOS-12805), and km s (SXDS-10600). Additionally, Hashimoto et al. (2013) have detected several LIS absorption lines in a stacked spectrum of four LAEs: CDFS-3865, CDFS-6482, COSMOS-13636, and COSMOS-30679. The mean blueshift of the LIS metal absorption lines is km s. These values are listed in the column 2 of Table 4.
SED Fitting Properties
In this study, we utilize SED fitting results of the sample, in particular, stellar dust extinction, , and stellar mass, . In the following sections, we compare the values with the results from Ly radiation transfer fitting, and investigate the correlation between the Ly profile trends and
SED fitting results for the MagE (LRIS) objects have been presented in Hashimoto et al. (2013) and Nakajima et al. (2013) (Shibuya et al. 2014b). For the detail procedure of the fitting, we refer the reader to Ono et al. (2010a, b). The derived and values are listed in the columns 3 and 4 in Table 4. The former range from to with a mean value of , and the latter from log = 7.7 to 10.8 with a mean of log = 9.3, respectively.
In the following sections, we use three morphological properties studied for LAEs in Shibuya et al. (2014a): the presence of a merger, the spatial offset between Ly and stellar-continuum emission peaks, , and the ellipticity. Shibuya et al. (2014a) have utilized and data taken with ACS and WFC3 on to examine rest-frame UV and optical morphologies, respectively. Among the objects presented in this study, the rest-frame UV images of the eight COSMOS objects have been investigated in Shibuya et al. (2014a).
The presence of a merger has been examined with two methods: the close-pair method(e.g., Le Fèvre et al. 2000; Law et al. 2012) and the morphological index method, especially system (Abraham et al. 1996; Conselice et al. 2000). In Shibuya et al. (2014a), the former method has been applied to objects with , which is the case for all the COSMOS objects presented in this study except for COSMOS-13138. The result is that two objects, COSMOS-13636 and COSMOS-12805, are mergers, while the remaining seven are not. On the other hand, the latter method has been done for objects with and a half light radius, , larger than . The reason why Shibuya et al. (2014a) have limited the sample for the latter method is to obtain reliable values of the indices. This is the case for three COSMOS objects presented in this study, COSMOS-13636, COSMOS-43982, and COSMOS-38380. The result is that none of the three is a merger. The two results for COSMOS-13636 are not consistent with each other because we have used two different methods. Thus, among the eight COSMOS objects, COSMOS-13636 and COSMOS-12805 may be a merger (the column 5 of Table 4).
The Ly spatial offset, , has been examined by performing source detections with SExtractor for Subaru NB387 and images. While compact objects with a symmetric UV light profile tend to have a small value, objects with an asymmetric, disturbed UV light profile likely to have a large value (e.g., Jiang et al. 2013; Shibuya et al. 2014a). Thus, this quantity could be a useful tracer of the Hi gas stability around the galaxy. The value is reliably obtained for the objects with and NB387 , where the typical positional error in (NB387) is less than (). For the eight COSMOS objects in this study, none has a significant Ly spatial offset larger than the typical error of the , .
The ellipticity, , where and are the major and minor axes, is a useful indicator of the galactic disk inclination. In Shibuya et al. (2014a), this has been measured using GALFIT software (Peng et al. 2002) for the objects with 25.0 and larger than the typical PSF size. The former criterion, corresponding to detection, is needed for the reliable ellipticity measurements (e.g.,Mosleh et al. 2012; Ono et al. 2013). Only three objects, COSMOS-30679, COSMOS-38380, and COSMOS-43982, satisfy these criteria. The resultant ellipticity values are 0.24 (COSMOS-30679), 0.34 (COSMOS-38380), and 0.49 (AGN-COSMOS-43982), respectively (the column 6 of Table 4).
4. Ly radiative transfer model and fitting procedure
4.1. A Library of Synthetic Spectra
The library of synthetic Ly spectra used in this study has been described in Schaerer et al. (2011). Ly radiation transfer has been computed with McLya (Verhamme et al. 2006) through spherically symmetric expanding shells of homogeneous and isothermal neutral hydrogen gas. The shell is describe by four parameters:
the radial expansion velocity, ,
the neutral hydrogen column density along the line of sight, ,
the Doppler parameter, , describing the thermal and turbulent motion in the shell,
The Ly source is located at the center of the shell. The intrinsic (i.e., before being affected by the radiative transfer effect) spectrum is a Gaussian Ly line plus a flat continuum, and is characterized by two parameters :
the Ly equivalent width, EW(Ly),
and the full width at half maximum, FWHM(Ly).
For a comparison with the observed data, each rest-frame model has been shifted using the systemic redshift values listed in Table 2. To reflect the uncertainty, we have allowed the observed Ly spectra to shift relative to the velocity zero point within the error. Thus, combinations of seven free parameters are fitted to the data.
4.2. Fitting of Observed Spectra
To perform a statistical comparison between the observed and modeled Ly line profiles, we calculate the values for each of the possible combinations of the parameters for each galaxy (cf., Chonis et al. 2013). Since model spectra are normalized and at an infinite spectral resolution, two steps are needed before the calculation. First, we normalize the observed spectra using the continuum level estimated at wavelengths longer than 1216Å. Second, each model Ly spectrum has been convolved with a Gaussian whose FWHM corresponds to spectral resolutions:
where is the speed of light.
We note that our fitting technique gives exactly the same statistical weight to all data points of the continuum and the Ly line. Finally for the sake of consistency, for each object we calculate in the wavelength range from to around the Ly line center.
In Figure 4, we demonstrate how the best fit, and its associated errors, are found using values. To do this, examples of the fit to are shown for well and poorly constrained objects. In the left panels of this figure, one can see a broad range of values with low reduced for COSMOS-08357 whose Ly S/N ratio is , in comparison to CDFS-3865 with a Ly S/N of . To measure median and values, we convert values into probabilities using the formula, exp() for each five 2D parameter set ( vs. , vs. , vs. , vs. FWHM(Ly), and vs. EW(Ly)). After normalizing them so that the total probability is unity, we draw a probability (PDF) and a cumulative density function (CDF) as shown in the middle and the right panels, respectively. Finally, we adopt the values where the CDF value satisfying CDF and as the median and , respectively. Performing this for each five 2D parameter set results in five median and values. As can be seen, all the five median and values are consistent with each other for CDFS-3865, whereas those are not for COSMOS-08357. In the latter case, we adopt the average of the five median and values.
Figure 5 shows the best fit model spectra with the observed ones. All the Ly profiles are quite well reproduced by the model, which seems to differ from the previous studies by Kulas et al. (2012) and Chonis et al. (2013). These authors have had difficulty reproducing their Ly profiles, especially the position and the flux of the blue bump. This might be due to model differences. These two studies have utilized the uniform expanding shell model constructed by Zheng & Miralda-Escudé (2002) and Kollmeier et al. (2010). There are three major differences between the models (c.f., Chonis et al. 2013). First, in addition to the three common parameters, , , and , the model used in this study also includes an additional one for dust absorption. Second, the grid points and the physical range of parameters are different. The model by Zheng & Miralda-Escudé (2002) and Kollmeier et al. (2010) has four values for each parameter: = 50, 100, 200, 300 km s, log () = 17, 18, 19, 20.3 cm, and = 20, 40, 80, 120 km s, whereas the model used in this study has 12 , 13 , and 5 values spanning wider physical ranges. Finally, the intrinsic spectrum of the previous models is a monochromatic Ly line, while we model a Gaussian Ly plus a continuum. As we show in §4.3.2 and later sections, we infer that the key to better reproducing the blue bump is to assume the Ly profile to be a (broad) Gaussian.
The best fit parameters are summarized in Table 5. We describe the mean values of the derived parameters, and systematically compare them with those of LBGs modeled by the same code (Verhamme et al. 2008; Schaerer & Verhamme 2008; Dessauges-Zavadsky et al. 2010). For the parameter FWHM(Ly), we examine the mean values of two subsamples, objects with a blue bump and those without. We have checked that there is no significant difference between the two subsamples in the other parameters.
The mean value of the LAEs is km s, which is comparable to that of LBGs, km s. This strongly disfavors the hypothesis that the small of LAEs is due to their large outflow velocity.
The most interesting parameter, , ranges from log() = 16.0 to 19.7 cm, with a mean value of cm, which is more than one order of magnitude smaller than the typical log() value of LBGs, (cm).
The mean values of and are and km s, respectively, both of which are comparable to those of LBGs, and km s.
FWHM(Ly) values range from FWHM(Ly) = 50 to 847 km s. The mean values for the whole sample, the non blue bump sample, and the blue bump sample, are 354, 169, and 602 km s, respectively. This shows that the blue bump objects have significantly larger FWHM(Ly) than that found in the non blue bump objects. This trend is similar to Verhamme et al. (2008); They have found that most LBGs with a single peaked Ly profile are best fitted with moderate values of FWHM(Ly), km s, whereas the best fit FWHM(Ly) values for two LBGs with a blue bump are greater than km s. These results support our claim that large FWHM(Ly) helps fitting the blue bump. We investigate if there are any observational trends for the blue bump objects, and discuss possible mechanisms for the blue bump objects to have large FWHM(Ly) in §5.1.
Since starburst activities that produce Ly photons should be similar between LAEs and LBGs, we expect comparable EW(Ly) values for these two galaxy populations. The result is that the mean EW(Ly) value of LAEs, Å, is somewhat smaller than that of LBGs, Å.
In summary, the model parameter derived in LAEs is more than one order of magnitude smaller than that of LBGs, whereas the remaining parameters are consistent within between LAEs and LBGs.
Influence of Spectral Resolution on the Fitting Procedure
To investigate the influence of spectral resolution on the fitting results, we compare the best fit parameters of the two objects observed with the two spectrographs, COSMOS-13636 and COSMOS-43982. As can be seen in Table 5, the two fitting results of COSMOS-43982 are consistent with each other, whereas those of COSMOS-13636 are not, possibly owing to the large difference in the best-fit reduced , 1.1 and 6.2.
Taking a closer look into these two fits, we see that the extremely small 1 noise in the flux of LRIS-COSMOS-13636 could be a key reason for its high value. On the other hand, the modeled spectrum seems to be over-smoothed, leading us to infer its Ly line resolution is under-estimated. Indeed, it is known that the spectral resolution for a given line can be higher than the canonical value. A combination of these factors would naturally cause the large resultant value, and the discrepancy between the different best-fit parameters at two resolutions.
|(km s)||(cm)||(km s)||(km s)||(Å)|
Note. – (1) Object ID; (2) Reduced value of the fitting calculated as , where and denote the number of data points and the degree of freedom, respectively; (3) (8) Best fit values of the radial expansion velocity, the column density of the neutral Hydrogen, the dust absorption optical depth, the Doppler parameter, the intrinsic Ly FWHM, and the intrinsic Ly EW, respectively.
4.4. Degeneracy among Parameters
In this subsection, we investigate degeneracies among the model parameters to understand how they affect our determination of the best fit parameters. First we describe possible degeneracies and then statistically examine them using 2D values.
It is possible that parameters and EW(Ly) are degenerated as an observed profile can be reproduced equivalently well either assuming a weak intrinsic line with low dust extinction, or a strong intrinsic line with high dust extinction. There would also be a degeneracy between and FWHM(Ly) in the sense that both broaden the line profile. Furthermore, when there is a blue bump in the profile, we need either a high or a low to reproduce it.
Figures 13 - 15 in the Appendix are 2D parameter grid maps for CDFS-3865 with the grey dots showing the entire grids. We use these maps and values to examine the actual degeneracies among the parameters. If there is a degeneracy between two parameters, the contour would be tilted and elongated. The blue grids in these figures show those satisfying above the raw minimum designated by the white dots, i.e., the 3 uncertainty in the parameter set (Press et al. 1992). Thanks to the number of data points given by high spectral resolutions, and the relatively coarse grids, even the 3 uncertainty is converged into one grid. This indicates that there is no degeneracy that affects our determination of the best fit. We have checked that this is also true for the rest of the sample in this study. Thus, we conclude that the systematic uncertainties among the parameters due to the degeneracies are small, and thus do not affect our discussions.
4.5. Comparison between Observation and Model
In order to examine if the best fit parameters are reasonable, we compare the derived parameters with the observables.
As stated in §3.5.1, several LIS absorption lines have been detected in individual spectra of COSMOS-12805, COSMOS-13636, and SXDS-10600 (Shibuya et al. 2014b), and in a stacked spectrum of four LAEs, CDFS-3865, CDFS-6482, COSMOS-13636, and COSMOS-30679 (Hashimoto et al. 2013). The measured blueshift of LIS absorption lines with respect to the systemic, , is listed in Table 4. Figure 6 shows a comparison between and the best-fit expansion velocity, . For the stacked spectrum, we plot the mean value of the four LAEs, km s. While there are only four data points, and are in excellent agreement with each other.
The stellar dust extinction values, , for the sample have been derived in previous studies (Hashimoto et al. 2013; Nakajima et al. 2013; Shibuya et al. 2014b) (see §3.5.1). Figure 7 compares them with gas dust extinction, , derived assuming the relation:
Dotted and dashed lines correspond to empirical relations (Erb et al. 2006a) and (Calzetti et al. 2000), respectively, for host galaxies. As Kashino et al. (2013) have shown, the difference between and becomes smaller for higher- galaxies.
In this study, we expect that data points are located below these relations. This is because obtained from Ly modeling is gas dust extinction for outflowing shells, which should be smaller than that for host galaxies. The figure shows that half of the sample roughly lie between the two lines, while the rest of the sample show low values. A similar trend has been found in Figure 12 of Verhamme et al. (2008) who have compared and for LBGs. They have assumed two different star formation histories (SFHs) in deriving : a constant SFH indicated by red triangles and an exponentially decreasing SFH indicated by blue open circles, the former of which is the same as that assumed in this study. Both our data and the red triangles in Verhamme et al. (2006) are similarly distributed in the sense that half of the sample has comparable extinction values and the rest has low values.
FWHM(neb) vs. FWHM(Ly)
Figure 8 plots the observed FWHM of nebular emission lines, FWHM(neb), versus modeled FWHM of the intrinsic (i.e., before being affected by the radiative transfer effect) Ly line, FWHM(Ly). Assuming that both Ly and nebular emission lines originate from Hii regions, the two FWHMs should be similar. However, FWHM(Ly) is systematically larger than FWHM(neb). Additional scattering of Ly photons in an H region due to residual H atoms in it may be at work. Assuming a static H region with a neutral hydrogen column density of log(N) cm, corresponding to an unity optical depth for ionizing photons, (cf., Verhamme et al. 2015), FWHM(Ly) can be broadened by km s compared to FWHM(neb). As can be seen from Figure 8, while this additional broadening would help explain the discrepancy for the non blue bump objects, it is still not enough for the blue bump objects. We discuss some interpretations for the huge FWHM(Ly) in the blue bump objects in §5.1.
We also perform Ly profile fitting of the blue bump objects with fixing FWHM(Ly) = FWHM(neb). As shown in Figure 9, the blue bumps are poorly reproduced compared to the fitting without fixing FWHM(Ly). We examine if the derived best-fit model parameters differ between the free and fixed FWHM(Ly) cases. While there is no systematic difference for and , we find that () becomes large (small) in the fixed FWHM(Ly) case. This would be related to the intrinsic degeneracy between them discussed in §4.4.