WHAT IS THE PHYSICAL ORIGIN OF STRONG Lyα EMISSION? II. GAS KINEMATICS AND DISTRIBUTION OF Lyα EMITTERS^*

We have carried out optical spectroscopy for our z = 2.2 LAE sample with the LRIS (Oke et al. 1995; Steidel et al. 2004) on the Keck I telescope in order to detect their redshifted Lyα emission lines. We used six multi-object slit (MOS) masks for LAEs selected in the NB387 imaging observations in the COSMOS, HDFN, HUDF, SSA22, and SXDS fields. The mask for the objects in the HUDF includes two LAEs whose nebular emission lines were detected in the 3D-HST survey (H. Atek et al. in preparation). The total number of LAEs observed with these LRIS masks is 83. The observations were conducted on 2012 March 19–21 and November 14–15 (UST) with seeing sizes of 07–16. Spectrophotometric standard stars were observed on each night for flux calibrations. The spectral resolution is R ~ 1000. The number of observed LAEs, grisms, central wavelengths, and observing time in each slit-masks are summarized in Table 1.

To calculate systemic redshifts of our LAEs from their nebular emission lines, we use NIR spectroscopic data obtained from observations with the FMOS (Kimura et al. 2010) on the Subaru telescope on 2012 December 22, 23, and 24 (UST). All LAEs in the SXDS and COSMOS fields are observed with the J- and H-band filters of FMOS. Details of the FMOS observation and reduction are shown in K. Nakajima et al. (in preparation). The systemic redshifts for objects were derived by simultaneously fitting to Hβ and [O iii] λ4958, 5007 emission lines by using their vacuum wavelengths in rest frame.

Our LRIS spectra in each MOS mask are reduced with the public Low-Redux (XIDL) pipeline,⁹ for long-slit and multi-slit data from the spectrographs on the Keck, Gemini, MMT, and Lick telescopes. We reduce the spectra of LAEs with this software in the following manner. First, we create flats, calibrate wavelengths with the arc data, and reject sources illuminated by the cosmic ray injections for two-dimensional (2D) spectra in the MOS masks. Next, we automatically identify emission lines and continua, and trace them in each slit in individual one-frame masks. After the source identification, we subtract the sky background, and correct for the distortion of the 2D MOS mask images using sky lines. According to the information on the source identifications, we extract one-dimensional (1D) spectra from each slit in individual mask images. Finally, we stack the extracted 1D spectra.

The public XIDL software extracts 1D spectra from each one-exposure frame before combining these 2D mask images. This process makes it difficult to detect faint emission lines and continua that are undetectable in individual one-exposure images. Then, we additionally search for faint emission lines from stacked 2D images by visual inspection after combining one-exposure frames.

In total, the Lyα emission lines are detected from 26 objects in the LRIS spectroscopy. Figure 1 shows the spectroscopic success rate in the detection of Lyα emission. The success rate is ~70% for bright objects with NB387  24.5. However, low detection and/or selection completeness at NB387  24.5 reduces largely the success rate (~20%). The photometric and spectroscopic properties of these Lyα-detected objects are listed in Table 2. Among these LRIS spectra, we identify eight LAEs with detections of Lyα and nebular emission lines excluding active galactic nucleus (AGN) like objects.

Zoom In Zoom Out Reset image size

Table 2. Summary of the Lyα-detected Objects in the LRIS Spectroscopy

Slit Mask	Object	R.A.	Decl.	U	NB387	B	λobs	zLyα	f(Lyα)	L(Lyα)	EW(Lyα)	zsys	ΔvLyα	Fblue/Ftot
							(Å)		(10−17erg s−1 cm−2)	(1042erg s−1)	(Å)		(km s−1)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)
COSMOS	12027	149.9343976	+2.1285326	24.2	23.4	24.5	3878.72 ± 0.24	2.1906	5.6 ± 0.3	2.0 ± 0.1	$73.69^{+6.91}_{-6.44}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	12805a	150.0637013	+2.1354116	23.7	23.3	23.8	3843.27 ± 0.65	2.16144	7.4 ± 0.7	2.6 ± 0.3	$33.73^{+5.98}_{-5.52}$	2.15872	258 ± 51	0.24
	13138	150.0108585	+2.1401388	24.9	24.6	25.0	3866.73 ± 0.89	2.18074	1.2 ± 0.2	0.43 ± 0.07	$40.36^{+9.64}_{-8.44}$	2.17921	144 ± 69	0.40
	13636a,b	149.9974498	+2.1439906	23.9	23.0	24.1	3844.68 ± 1.30	2.1626	9.6 ± 0.5	3.3 ± 0.2	$86.80^{+8.31}_{-7.76}$	2.16052	197 ± 102c	0.13
	14212a	149.9585714	+2.1482830	24.0	23.3	24.0	3879.99 ± 0.52	2.19165	6.7 ± 0.6	2.4 ± 0.2	$54.98^{+5.56}_{-5.18}$	2.18955	188 ± 40	0.23
	08357d	149.9961405	+2.0921070	24.8	24.4	24.9	3868.79 ± 0.86	2.18243	1.4 ± 0.4	0.50 ± 0.14	$46.68^{+8.60}_{-7.69}$	2.18044	205 ± 66	0.13
	13820d	149.9554179	+2.1470628	25.6	25.1	25.9	3820.20 ± 1.01	2.14246	2.1 ± 0.2	0.72 ± 0.07	$98.86^{+31.30}_{-26.40}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	14135d	149.9770609	+2.1508410	27.0	25.9	26.8	3893.06 ± 1.08	2.2024	0.62 ± 0.1	2.3 ± 0.4	$100.02^{+30.70}_{-24.26}$	⋅⋅⋅	⋅⋅⋅
HDFN1	18325e	189.0973399	+62.1014179	22.9	21.6	23.2	3858.78 ± 0.25	2.1742	30.0 ± 0.9	10.6 ± 0.3	$151.61^{+3.52}_{-3.45}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	20042d	189.0293966	+62.1176510	25.4	24.7	25.6	3864.87 ± 2.12	2.17921	1.5 ± 0.4	0.53 ± 0.14	$108.42^{+12.48}_{-11.36}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
HDFN2	31902	189.3127706	+62.2091548	25.4	23.9	24.9	3865.31 ± 0.31	2.17957	1.6 ± 0.2	0.6 ± 0.08	$146.30^{+13.35}_{-12.31}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	43408	189.4532215	+62.2639356	26.7	25.4	26.3	3886.59 ± 0.47	2.19708	5.6 ± 0.2	2.0 ± 0.1	$72.09^{+6.29}_{-5.89}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	42659d	189.4575590	+62.2917868	25.9	23.8	25.9	3882.07 ± 0.61	2.19336	1.4 ± 0.1	0.51 ± 0.04	$661.41^{132.72}_{-104.76}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
COSMOS3B	38380	149.9205873	+2.3844960	24.4	23.4	24.5	3909.79 ± 0.47	2.21616	6.9 ± 0.7	2.6 ± 0.3	$137.19^{+14.80}_{-13.48}$	2.21256	336 ± 36	0.01
	43982e,b	149.9766453	+2.4416582	24.3	23.2	24.6	3883.01 ± 0.73	2.19413	6.9 ± 0.6	2.6 ± 0.2	$130.06^{+12.35}_{-11.35}$	2.19333	75 ± 56	0.41
	46597	149.9415665	+2.4688913	24.7	23.8	24.7	3857.99 ± 0.74	2.17355	3.5 ± 0.6	1.2 ± 0.2	$58.75^{+7.07}_{-6.53}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	38019d	149.9020418	+2.3815371	25.9	25.0	25.7	3900.36 ± 0.49	2.2084	1.5 ± 0.2	5.5 ± 0.7	$146.34^{+31.88}_{-26.40}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	40792d	149.9444266	+2.4094991	26.7	25.5	27.4	3901.43 ± 0.53	2.20928	2.3 ± 0.4	0.85 ± 0.14	$394.30^{+121.34}_{-88.39}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	41547d	149.9246216	+2.4166699	26.0	24.9	26.6	3832.08 ± 0.65	2.15224	2.5 ± 0.6	0.86 ± 0.21	$298.88^{+77.85}_{-61.81}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	44993d	149.9744788	+2.4530529	26.5	25.0	26.5	3907.65 ± 0.80	2.2144	2.1 ± 0.4	0.78 ± 0.15	$240.61^{+53.56}_{-43.25}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
HUDF_maB	17001f	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	3705.44 ± 0.73	2.04806	22.0 ± 4.0	6.7 ± 1.2	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	31000f	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	4969.63 ± 0.47	3.08798	3.7 ± 0.6	3.1 ± 0.5	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
SXDS495B	06713	34.4224906	−5.1136338	24.5	23.5	24.6	3894.18 ± 1.02	2.20332	3.9 ± 0.2	1.4 ± 0.1	$118.77^{+5.24}_{-5.04}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅
	10600a	34.4420541	−5.0486039	23.7	23.0	23.6	3903.62 ± 0.69	2.21109	5.2 ± 0.2	1.9 ± 0.1	$58.19^{+2.62}_{-2.54}$	2.20915	181 ± 53	0.03
	10942	34.4980945	−5.0428800	25.6	24.2	25.6	3887.51 ± 0.55	2.19783	0.77 ± 0.05	0.28 ± 0.02	$134.94^{+9.24}_{-8.68}$	2.19557	212 ± 42	0.12
	10535d	34.4246768	−5.0488535	25.9	24.8	26.2	3905.50 ± 0.36	2.21263	3.4 ± 0.2	1.3 ± 0.1	$223.09^{+22.04}_{-19.96}$	⋅⋅⋅	⋅⋅⋅	⋅⋅⋅

Notes. Columns: (1) Slit mask. (2) Object ID. (3) Right ascension. (4) Declination. (5)–(7) U, NB387, and B-band magnitudes. (8) Observed wavelength of Lyα measured by the asymmetric Gaussian fitting. (9) Redshift of Lyα corrected for the heliocentric motion. (10) Lyα flux uncorrected for slit loss in LRIS spectroscopy. (11) Lyα luminosity. (12) Lyα equivalent width estimated from the NB387 magnitudes. Lyα position in the transmission curve is taken into account from the spectroscopic redshift of Lyα. (13) Redshift of nebular emission lines corrected for the heliocentric motion (K. Nakajima et al. in preparation). (14) Lyα velocity offset relative to nebular emission lines. (15) Ratio of Lyα flux in the bluer side relative to the systemic redshift to total Lyα flux. ^aUV continuum-detected LAEs. ^bThese objects have also been observed with Magellan/MagE in Hashimoto et al. (2013). ^cHashimoto et al. (2013) have reported that LAE 13636 has a Δv_Lyα of $99^{+16}_{-16}$ km s⁻¹. However, the Hα line profile would have been affected by a residual of a neighboring OH line due to the low spectral resolution of their Keck-II/NIRSPEC observation (R ~ 1500), making it difficult to determine accurately the systemic redshift. Our FMOS spectroscopy with R ~ 2200 would securely detect nebular emission less affected by OH lines. ^dThese objects are reduced without the Keck/LRIS public pipeline. ^eAGN-like objects. ^fK. Nakajima et al. (in preparation).

Download table as:  ASCIITypeset image

We measure the Lyα velocity offset for the eight LAEs with detections of Lyα and nebular lines:

$$\begin{equation} \Delta v_{\rm Ly\alpha } = c \frac{z_{\rm Ly\alpha } - z_{\rm sys}}{1+z_{\rm sys}}, \end{equation} \tag{ 1 }$$

where c, z_Lyα, and z_sys, are the speed of light, and the Lyα and systemic redshifts, respectively. The systemic redshift is determined from nebular emission lines obtained with FMOS.

Prior to the measurement of Δv_Lyα, we measure the wavelength of Lyα in the following line-fitting procedures. We use the peak wavelength of the best-fit asymmetric Gaussian profile for measurements of the Lyα wavelength. We first automatically search for an emission line in a wavelength range of 3500–4000 Å in each spectrum. This range includes the wavelength range of the NB387 filter. Next, we fit an asymmetric Gaussian profile to the detected lines. The asymmetric Gaussian profile is expressed as

$$\begin{equation} f(\lambda) = A \exp \left(\frac{-(\lambda -\lambda _0^{\rm asym})^2}{2 \sigma _{\rm asym}^2}\right) + f_0, \end{equation} \tag{ 2 }$$

where A, λ₀, and f₀ are the amplitude, peak wavelength of the emission line, and continuum level, respectively. The asymmetric dispersion, σ_asym, is represented by $\sigma _{\rm asym} = a_{\rm asym} (\lambda -\lambda _0^{\rm asym})+d$, where a_asym and d are the asymmetric parameter and typical width of the line, respectively. An object with a positive (negative) a_asym value has a skewed line profile with a red (blue) wing. The fitting with the asymmetric Gaussian profile is efficient for Lyα line from high-z galaxies affected by complex kinematic structure of infalling and/or outflowing gas and IGM absorption. For fitting, we use data points over the wavelength range where the flux drops to 10% of its peak value at the redder and bluer sides of the emission line. We use the peak flux, the wavelength of the line peak, 0.4, 1.0 × 10⁻¹⁷, and 2.0 as the initial parameters of A, $\lambda _0^{\rm asym}$, a_asym,  f₀, and d for the line-fitting. The last two are typical values of our spectra. If profile fitting does not converge to the minimum in χ², we search for the best fit by changing the initial value of a_asym.

We show the best-fit asymmetric Gaussian profile for an example spectrum in Figure 2. We also fit a symmetric Gaussian profile to the emission lines in addition to asymmetric one. For the symmetric Gaussian fitting, we adopt two wavelength ranges where the flux drops to 70% and 10% of its peak value, and denote the corresponding peak wavelengths by $\lambda _0^{\rm gauss}$ and $\lambda _0^{\rm cent}$, respectively. The fitting procedure in the former narrow range is similar as in Hashimoto et al. (2013) in terms of avoiding systematic effects due to asymmetric line profile. As shown in Figure 2, the best-fit $\lambda _0^{\rm asym}$ is broadly equal to $\lambda _0^{\rm cent}$ for the example line. The wavelength difference is ~ + 0.1 Å (~ + 10 km s⁻¹ at z = 2.2). In contrast, $\lambda _0^{\rm gauss}$ differs from $\lambda _0^{\rm asym}$ by ~ + 0.4 Å which corresponds to a velocity difference of ~ + 30 km s⁻¹ at z = 2.2. This is likely to be caused by the sharp drop on the blue side and the extended red tail which cannot be fit well with symmetric profiles.

Zoom In Zoom Out Reset image size

This trend is more clearly shown in Figure 3 which exhibits the wavelength difference of $\lambda _0^{\rm gauss}$ and $\lambda _0^{\rm cent}$ from $\lambda _0^{\rm asym}$ as a function of the asymmetric parameter, a_asym. The wavelengths of individual profiles are in good agreement for almost symmetric lines with a_asym ~ 0. Even for moderately asymmetric profiles with |a_asym| 0.2, $\lambda _0^{\rm cent}$ tends to correct for the systematic effects of skewed lines compared to $\lambda _0^{\rm gauss}$. However, both $\lambda _0^{\rm gauss}$ and $\lambda _0^{\rm cent}$ are redshifted (blueshifted) from $\lambda _0^{\rm asym}$ by ~0.5–1.0 Å for highly asymmetric lines with a_asym ~ +0.4 (−0.4).

Zoom In Zoom Out Reset image size

After correcting for the heliocentric motion of the Earth for the redshifts of Lyα and nebular lines,¹⁰ we calculate Δv_Lyα following Equation (1). Table 2 lists the z_Lyα, z_sys, and Δv_Lyα for the 26 Lyα-detected objects observed with LRIS. Figure 4 presents Lyα spectra as a function of velocity for LAEs with detections of nebular emission lines. In Table 3, we also list these quantities of the four LAEs with detections of Lyα and nebular lines obtained by previous Magellan/IMACS observations (Nakajima et al. 2012). Almost all objects observed with LRIS have a Δv_Lyα of ~200 km s⁻¹ which is consistent with values in previous studies (e.g., Hashimoto et al. 2013). The values of Δv_Lyα in the IMACS sample are calculated to be smaller than the LRIS results. This could be caused by large uncertainties due to the IMACS spectroscopy with a lower spectral resolution than LRIS.

Zoom In Zoom Out Reset image size

Table 3. Summary of the Lyα-detected Objects in the IMACS Spectroscopy

Slit Mask	Object ID	λobs	zLyα	EW(Lyα)	zsys	ΔvLyα
		(Å)		(Å)		(km s−1)
(1)	(2)	(3)	(4)	(5)	(6)	(7)
IMACS-SXDS	04640	3865.08 ± 0.37	2.17938	$164.75^{+4.63}_{-4.51}$	2.17822	110 ± 138
	08204	3895.11 ± 5.59	2.20408	$88.61^{+5.87}_{-5.56}$	2.20329	74 ± 505
	09219	3890.71 ± 6.08	2.20047	$29.62^{+2.38}_{-2.29}$	2.20004	40 ± 508
	11135	3882.27 ± 0.56	2.19352	$111.96^{+4.78}_{-4.60}$	2.19238	107 ± 151

Notes. Columns: (1) Slit mask. (2) Object ID. (3) Observed wavelength of Lyα measured by the asymmetric Gaussian fitting. (4) Redshift of Lyα corrected for the heliocentric motion. (5) Lyα equivalent width. (6) Redshift of nebular emission lines corrected for the heliocentric motion (K. Nakajima et al. in preparation). (7) Velocity offset of Lyα relative to nebular emission lines.

Download table as:  ASCIITypeset image

Additionally, we provide a consistency check for our measurement of z_Lyα by using the same object as in Hashimoto et al. (2013), COSMOS-13636. The object has been observed with both of LRIS in this work and Magellan/MagE in a previous work. The redshift of Lyα estimated from the LRIS spectrum ($z_{\rm Ly\alpha }^{\rm LRIS}=2.1626\, {\pm }\, 0.00073$) is in good agreement with that of MagE ($z_{\rm Ly\alpha }^{\rm MagE}=2.16229\, {\pm }\, 0.00008$) within a 1σ fitting error. The difference in velocity is 30 ± 70 km s⁻¹. The large error in $z_{\rm Ly\alpha }^{\rm LRIS}$ is likely to be due to the lower spectral resolution of LRIS (R ~ 1000) than that of MagE (R ~ 4100).

Figure 5 shows the Hubble Space Telescope (HST)/ACS I₈₁₄-band images of LAEs with a z_sys measurement in the COSMOS field. Unfortunately, the LAEs in the SXDS field are not covered by the CANDELS project. Several LAEs have multiple components, which could be mergers. The merger fraction of LAEs and its Lyα dependence are discussed in Shibuya et al. (2014).

Zoom In Zoom Out Reset image size

We measure the IS velocity offset of IS absorption lines for our LAEs. Due to the faintness of their UV continuum emission, it is difficult to detect IS absorption lines from high Lyα EW galaxies with EW 50 Å in individual spectra. However, owing to the high sensitivity of Keck/LRIS, the rest-frame UV continuum emission is clearly detected from four individual LAEs, LAE 12805, 13636, and 14212 in COSMOS, and LAE 10600 in SXDS, among the 26 Lyα-detected objects.

We first fit a power-law curve to the UV continuum emission in four individual objects in order to normalize the continuum level, and derive the properties of IS absorption lines. The normalized continuum emission in the rest frame is shown in Figure 6. Next, we fit the symmetric Gaussian profile to each IS absorption line in a wavelength range of ±5 Å around the expected line center. We summarize the best-fit peak wavelength, line depth, width, and EW in Table 4. The noise in each line is estimated from spectra at 1250–1700 Å avoiding regions close to the absorption lines. Most absorption lines are found to be detected at the >5σ levels except for several low ionization interstellar (LIS) lines.

Zoom In Zoom Out Reset image size

Table 4. Absorption-line Features of the UV-continuum Detected LAEs

Object	Ion	λrest	$\lambda _{\rm rest}^{\rm sys}$	I/I0	σ	EW(IS)	ΔvIS
		(Å)	(Å)		(Å)	(Å)	(km s−1)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
12805	Si ii	1260.4221	1259.64 ± 0.43	−0.69 ± 0.56	0.67 ± 0.58	−1.15 ± 0.21	−186 ± 101
EW(Lyα)=33.73 [Å]	O i	1302.1685	1301.02 ± 0.61	−0.65 ± 0.33	0.97 ± 0.54	−1.56 ± 0.21	[−518 ± 139]
ΔvLyα = 258 [km s−1]	Si ii	1304.3702	1301.02 ± 1.52	−0.65 ± 0.33	0.97 ± 0.54	−1.56 ± 0.21	[−518 ± 139]
mB = 23.8	C ii	1334.5323	1333.57 ± 0.72	−0.58 ± 0.33	0.90 ± 0.50	−1.32 ± 0.21	−215 ± 162
	Si iv	1393.76018	1392.39 ± 1.40	−0.22 ± 0.27	0.64 ± 0.76	−0.36 ± 0.21	−295 ± 307
	Si ii	1526.70698	1525.91 ± 0.36	−0.60 ± 0.17	0.63 ± 0.17	−0.94 ± 0.21	−156 ± 71
	C iv	1548.204	1545.60 ± 1.05	−0.27 ± 0.29	0.70 ± 0.70	−0.48 ± 0.21	[−750 ± 203]
	C iv	1550.781	1545.60 ± 1.05	−0.27 ± 0.29	0.70 ± 0.70	−0.48 ± 0.21	[−750 ± 203]
	Fe ii	1608.45085	1607.75 ± 1.11	−0.37 ± 0.48	0.49 ± 0.47	−0.46 ± 0.19	−131 ± 207
	Al ii	1670.7886	1670.90 ± 2.13	−0.35 ± 0.37	1.63 ± 1.9	−1.43 ± 0.22	21 ± 382
13636	Si ii	1260.4221	1260.23 ± 0.70	−0.21 ± 0.068	1.76 ± 0.65	−0.91 ± 0.19	−46 ± 165
EW(Lyα) = 86.80 [Å]	O i	1302.1685	1301.25 ± 1.20	−0.33 ± 0.26	1.26 ± 1.07	−1.05 ± 0.19	[−465 ± 280]
ΔvLyα = 197 [km s−1]	Si ii	1304.3702	1301.25 ± 1.20	−0.33 ± 0.26	1.26 ± 1.07	−1.05 ± 0.19	[−465 ± 280]
mB = 24.1	Si iv	1393.76018	1392.80 ± 0.64	−0.38 ± 0.15	1.27 ± 0.56	−1.19 ± 0.19	−206 ± 138
	Si iv	1402.77291	1401.92 ± 0.62	−0.50 ± 0.74	0.35 ± 0.95	−0.44 ± 0.20	−182 ± 134
	C iv	1548.204	1546.80 ± 0.83	−0.42 ± 0.14	2.00 ± 0.85	−2.09 ± 0.19	[−518 ± 161]
	C iv	1550.781	1546.80 ± 0.83	−0.42 ± 0.14	2.00 ± 0.85	−2.09 ± 0.19	[−518 ± 161]
	Al ii	1670.7886	1670.35 ± 0.62	−0.73 ± 0.62	0.39 ± 0.36	−0.71 ± 0.20	−96 ± 112
14212	Si ii	1260.4221	1258.98 ± 0.56	−0.65 ± 0.31	0.93 ± 0.48	−1.52 ± 0.24	−343 ± 133
EW(Lyα)=54.98 [Å]	O i	1302.1685	1301.56 ± 0.74	−0.52 ± 0.28	1.06 ± 0.62	−1.37 ± 0.25	[−394 ± 172]
ΔvLyα = 188 [km s−1]	Si ii	1304.3702	1301.56 ± 0.74	−0.52 ± 0.28	1.06 ± 0.62	−1.37 ± 0.25	[−394 ± 172]
mB = 24.0	C ii	1334.5323	1333.42 ± 0.55	−0.69 ± 0.16	1.93 ± 0.55	−3.34 ± 0.24	−249 ± 123
	Si iv	1402.77291	1402.54 ± 0.49	−0.49 ± 0.15	1.30 ± 0.43	−1.59 ± 0.24	−50 ± 105
	Si ii	1526.70698	1525.49 ± 0.58	−0.86 ± 0.38	0.96 ± 0.42	−2.06 ± 0.24	−238 ± 113
	C iv	1548.204	1545.01 ± 0.60	−0.73 ± 0.17	2.32 ± 0.66	−4.27 ± 0.30	[−864 ± 123]
	C iv	1550.781	1545.01 ± 0.60	−0.73 ± 0.17	2.32 ± 0.66	−4.27 ± 0.30	[−864 ± 123]
	Fe ii	1608.45085	1606.52 ± 1.00	−0.32 ± 0.23	0.49 ± 0.40	−0.40 ± 0.25	−360 ± 185
10600	Si ii	1260.4221	1258.90 ± 1.07	−0.28 ± 0.11	2.34 ± 1.18	−1.65 ± 0.19	−363 ± 254
EW(Lyα)=58.19 [Å]	O i	1302.1685	1298.58 ± 0.39	−0.61 ± 0.47	0.28 ± 0.18	−0.42 ± 0.19	[−1079 ± 90]a
ΔvLyα = 181 [km s−1]	Si ii	1304.3702	1298.58 ± 0.39	−0.61 ± 0.47	0.28 ± 0.18	−0.42 ± 0.19	[−1079 ± 90]a
mB = 23.6	C ii	1334.5323	1333.38 ± 0.45	−0.54 ± 0.28	0.54 ± 0.26	−0.74 ± 0.16	−258 ± 100
	Si iv	1393.76018	1392.57 ± 0.38	−0.50 ± 0.18	0.74 ± 0.26	−0.92 ± 0.16	−256 ± 82
	Si iv	1402.77291	1401.64 ± 0.46	−0.74 ± 0.35	0.32 ± 0.13	−0.60 ± 0.19	−242 ± 97
	Si ii	1526.70698	1525.24 ± 0.65	−0.30 ± 0.14	1.00 ± 0.50	−0.75 ± 0.18	−289 ± 127
	C iv	1548.204	1546.75 ± 0.36	−0.92 ± 0.32	0.69 ± 0.22	−1.59 ± 0.19	[−528 ± 71]b
	C iv	1550.781	1546.75 ± 0.36	−0.92 ± 0.32	0.69 ± 0.22	−1.59 ± 0.16	[−528 ± 71]b

Notes. Columns: (1) Object ID. (2) Ion. (3) Wavelength in rest frame. (4) Observed wavelength of the line. (5) Amplitude of the emission line. (6) Width of the absorption line uncorrected for the instrumental broadening. (7) Equivalent width of the line. (8) Velocity offset of emission line relative to nebular emission lines. ^aThe value of Δv assumes that the rest wavelength of the blend is 1303.2694 Å. ^bThe value of Δv assumes that the rest wavelength of the blend is 1549.479 Å.

Download table as:  ASCIITypeset image

We calculate Δv_IS in a similar manner as for Lyα in Section 6.1. Several pairs of absorption lines such as O i λ1302-S i λ1304, and C iv λ1548-C iv λ1550 are likely to be blended at the resolution of our spectroscopy. For this reason, we define the wavelengths of the line pairs as central values between the pairs. We also derive the properties of fine-structure emission lines such as Si ii* as summarized in Table 5. We find that the velocity offsets of these ion lines from z_sys are almost zero, indicating that the fine-structure emission lines also trace the systemic redshift of galaxies. This is because these emission lines come from nebular regions photoionized by radiation from massive stars (e.g., Shapley et al. 2003).

Table 5. Emission Line Features of UV-continuum Detected LAEs

Object	Ion	λrest	$\lambda _{\rm rest}^{\rm sys}$	I/I0	σ	EW	ΔvIS
		(Å)	(Å)		(Å)	(Å)	(km s−1)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
12805	Si ii*	1309.276	1309.26 ± 0.44	0.78 ± 0.37	0.82 ± 0.45	1.59 ± 0.21	5 ± 100
	O iii]	1660.809	1660.58 ± 0.51	0.64 ± 0.49	0.41 ± 0.31	0.66 ± 0.19	−42 ± 93
	O iii]	1666.150	1665.34 ± 0.45	0.89 ± 0.76	0.46 ± 0.29	1.02 ± 0.19	−147 ± 81
13636	Si ii*	1533.431	1532.97 ± 0.62	0.48 ± 0.40	0.44 ± 0.39	0.52 ± 0.20	−90 ± 121
	O iii]	1666.150	1665.92 ± 1.4	0.39 ± 0.30	1.47 ± 1.23	1.45 ± 0.19	−42 ± 257
14212	Si ii*	1264.738	1265.06 ± 0.49	0.80 ± 0.54	0.65 ± 0.45	1.30 ± 0.24	76 ± 117
10600	Si ii*	1264.738	1264.94 ± 0.61	0.45 ± 0.32	0.52 ± 0.34	0.59 ± 0.16	47 ± 144
	Si ii*	1533.431	1534.03 ± 0.44	0.73 ± 0.48	0.22 ± 0.13	0.41 ± 0.18	116 ± 85
	O iii]	1666.150	1665.52 ± 0.97	0.37 ± 0.27	0.94 ± 0.69	0.86 ± 0.20	−113 ± 175

Notes. Columns: (1) Object ID. (2) Ion. (3) Wavelength in rest frame. (4) Observed wavelength of the line. (5) Amplitude of the emission line. (6) Width of the absorption line uncorrected for the instrumental broadening. (7) Equivalent width of the line. (8) Velocity offset of emission line relative to nebular emission lines.

Download table as:  ASCIITypeset image

We estimate the covering fraction, f_c, of surrounding H i gas from the depth of low ionization IS absorption lines for our four continuum-detected LAEs. If the H i gas is distributed in a spherical shell, the depth of the lines may be related to f_c. The covering fraction of any ion is estimated from

$$\begin{equation} \frac{I}{I_0} = 1 - f_c (1 - e^{-\tau }), \end{equation} \tag{ 3 }$$

where τ, I, and I₀ are optical depth of an absorption line, its residual intensity, and the continuum level, respectively. The optical depth is linked to the column density as

$$\begin{equation} \tau = f \lambda \frac{\pi e^2}{m_e c} N = f\lambda \frac{N}{3.768\times 10^{14}}, \end{equation} \tag{ 4 }$$

where f, λ, and N are the ion oscillator strength, the wavelength of the absorption line in Å, and the column density of the ion in cm⁻² (km s⁻¹)⁻¹, respectively. Jones et al. (2013) use Si ii λ1260, 1304, and 1526 lines in order to solve the above two equations and estimate f_c for gravitationally lensed LBGs at z ~ 4. They find best-fit values of N and f_c by fitting observed the Si ii line profiles with the intensity as a function of N and f_c, I(N, f_c), derived from the above equations. In addition to the fitting to Si ii lines, they use several strong absorption lines, Si ii λ1260, O i λ1302, Si ii λ1304, C ii λ1334, and Si ii λ1526 to put a lower limit on f_c via

$$\begin{equation} f_c = 1 - \frac{I}{I_0}, \end{equation} \tag{ 5 }$$

which is a simplified case of Equation (3) when τ 1. For our LAEs, we estimate f_c in the latter method for the following reasons. (1) It is relatively difficult to fit our Si ii line profiles with a low signal-to-noise ratio (S/N) due to the faintness of UV-continuum emission and low resolution of our spectroscopy. (2) Jones et al. (2013) use mainly the f_c value derived in the latter method in their discussion. We would like to compare f_c for LAEs with that for LBGs in the same manner.

We derive the average absorption line profile of these strong transitions as a function of velocity. In the calculation, we do not use O i λ1302 and Si ii λ1304 transitions, since they could be heavily blended owing to the low spectral resolution. The derived average line profiles are shown in Figure 7. The covering fractions are estimated to be ~0.7 for LAE 12805, ~0.3 for 13636, ~0.9 for 14212, and ~0.4 for 10600 from the residual intensity in the core of the absorption line profiles. We additionally calculate the average depth of each best-fit Gaussian profile derived in Section 4.3. This alternative is helpful to adequately estimate the depth of a profile with a low S/N. The values of f_c are comparable to those derived from the average line profiles, with the exception of LAE 14212. The difference for LAE 14212 is because the f_c of the average line profile is affected by a singular count of C ii λ1334 line profile.

Zoom In Zoom Out Reset image size

The spectral resolution of our LRIS spectroscopy is ~4 times lower than that of Jones et al. (2013), preventing us from making a fair comparison between our LAEs and LBGs. We alternatively estimate EW of strong LIS absorption lines, EW(LIS), for our UV-continuum detected LAEs, and compare with results of composite LBG spectra in Section 6.3.

In this section, we statistically investigate the difference in Δv_Lyα between LAEs and LBGs in a compilation of LAEs with a Δv_Lyα measurement in the previous studies including our 12 LAEs. The Lyα velocity offsets have previously been estimated for two objects in McLinden et al. (2011), three in the HETDEX survey (Finkelstein et al. 2011; Chonis et al. 2013), four from Hashimoto et al. (2013), and two LAEs in the MUSYC project (Guaita et al. 2013) at z ~ 2–3. Table 7 summarizes EW(Lyα) and Δv_Lyα of the LAEs in these studies. Among the objects in previous studies, COSMOS 13636 from (Hashimoto et al. 2013) is included in our sample of the 12 LAEs. We combine these 11 LAEs with our new 11, and construct a large sample consisting of 22 objects, which doubles the number of LAEs with Δv_Lyα measurements. Figure 9 shows the histogram of Δv_Lyα using the newly enlarged sample. This is the updated version of Figure 6 in Hashimoto et al. (2013). Similar to Hashimoto et al. (2013), we confirm that Δv_Lyα of LAEs is systematically smaller than the values of LBGs. We carry out the non-parametric Kolmogorov–Smirnov (K-S) test for the two populations. The K-S probability is calculated to be ~10⁻⁷, indicating that Δv_Lyα is definitively different between LBGs and LAEs. The weighted mean of the 22 objects is Δv_Lyα = 234 ± 9 km s⁻¹.

Zoom In Zoom Out Reset image size

Table 7. Properties of the NB-selected Galaxies with Detections of Lyα and Nebular Emission Lines in the Previous Studies

Object	zsys	EW(Lyα)	ΔvLyα	SFR	E(B − V)	log M*	Comments
		(Å)	(km s−1)	(M☉ yr−1)		(M☉)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
McLinden et al. (2011)
LAE27878	3.11879	$118^{+34}_{-40}$	125 ± 17.3	$32^{+9}_{-7}$	⋅⋅⋅	$9.97^{+0.378}_{-0.395}$a	[O iii]λ5007
LAE40844	3.11170	$78^{+8}_{-8}$	342 ± 18.3	$113^{+120}_{-60}$	⋅⋅⋅	$9.80^{+0.734}_{-0.363}$a
Finkelstein et al. (2011) and Chonis et al. (2013)
HPS 194	2.28628	114 ± 13	303 ± 28	>29.3b	0.09 ± 0.06	$10.2^{+0.08}_{-0.14}$	HETDEX sample
HPS 256	2.49024	206 ± 65	$177^{+52}_{-68}$	>35.4b	0.10 ± 0.09	$8.28^{+0}_{-0.02}$	Hβ, [O iii]λ4959, [O iii]λ5007, Hα
HPS 251	2.28490	140 ± 43	$146^{+116}_{-156}$	>9.9b	0.07 ± 0.08	$9.04^{+0.73}_{-0.04}$
Hashimoto et al. (2013)
CDFS-3865	2.17210	$64^{+29}_{-29}$	$281^{+99}_{-25}$	$190^{+13}_{-13}$b	$0.185^{+0.009}_{-0.009}$	$9.50^{+0.028}_{-0.018}$	Subaru NB387 sample
CDFS-6482	2.20443	$76^{+52}_{-52}$	$156^{+52}_{-25}$	$48^{+10}_{-9}$b	$0.185^{+0.026}_{-0.018}$	$9.72^{+0.087}_{-0.071}$	[O iii]λ5007, Hα
COSMOS-13636	2.16125c	$73^{+5}_{-5}$	$99^{+16}_{-16}$c	$18^{+3}_{-3}$b	$0.273^{+0.018}_{-0.079}$	$9.30^{+0.078}_{-0.330}$
COSMOS-30679	2.19776	$87^{+7}_{-7}$	$253^{+26}_{-26}$	$45^{+5}_{-5}$b	$0.528^{+0.026}_{-0.026}$	$10.3^{+0.124}_{-0.151}$
Guaita et al. (2013)
LAE27	3.0830	25.7	167.8 ± 105.3	⋅⋅⋅	0.1	$9.95^{+0.13}_{-0.17}$	MUSYC sample
z3LAE2	3.1118	23.8	221.8 ± 90.0	⋅⋅⋅	$0.32^{+0.06}_{-0.23}$	$9.95^{+0.13}_{-0.17}$	Hβ, [O iii]λ4959, [O iii]λ5007

Notes. Columns: (1) Object ID. (2) Systemic redshift. (3) Lyα equivalent width. (4) Lyα velocity offset. (5) SFR. (6) Dust extinction. (7) Stellar mass. (8) Comments. ^aEstimated in Rhoads et al. (2014). ^bBased on Hα flux. ^cThe Δv_Lyα of this object is calculated to be 197 ± 102 km s⁻¹ in our FMOS observation with higher spectral resolution than that of the Keck-II/NIRSPEC spectrosocpy in Hashimoto et al. (2013) (see Table 2).

Download table as:  ASCIITypeset image

We plot EW(Lyα) and Δv_Lyα of our new sample in the right panel of Figure 10. We confirm the anti-correlation between EW(Lyα) and Δv_Lyα in high-z star-forming galaxies including objects with a high Lyα EW (Hashimoto relation) suggested by Hashimoto et al. (2013). The larger sample clarifies that Δv_Lyα decreases with increasing EW(Lyα). A similar trend has been found for UV-continuum selected galaxies at z ~ 2–3. Shapley et al. (2003) have calculated velocity offsets between Lyα emission and IS absorption (Δv_em–abs) for composite spectra of LBGs, and have investigated relation between Lyα EW and Δv_em–abs. Four LBG subsamples divided according to their Lyα EW reveal a trend that Lyα EW increases with decreasing Δv_em–abs in the EW range of −15 to +50 Å. The Δv_em–abs of our UV-continuum detected LAEs is very consistent with the trend of Shapley et al. (2003), as shown in the top panel of Figure 11. Nevertheless, Lyα and IS velocity offsets from z_sys would be capable of distinguishing effects of Lyα resonant scattering and galactic outflow on Δv_em–abs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We additionally examine a possible difference in Δv_IS between LAEs and LBGs. The weighted means of Δv_IS of the absorption lines are calculated to be −134 ± 67, −261 ± 48, −216 ± 56, and −169 ± 52 km s⁻¹ for LAE 13636, 10600, 14212, and 12805, respectively. In the calculation of the average Δv_IS for each object, we exclude several line pairs with a large Δv_IS of − 500 km s⁻¹ which are not reliably determined due to a line blending. As shown in Table 4, we find that almost all IS absorption lines are blueshifted with respect to z_sys by − 200 km s⁻¹, which indicates that gaseous outflows are present in the continuum-detected LAEs.

The left panel in Figure 10 represents the relation between EW(Lyα) and Δv_IS. The average of the four is Δv_IS = −204 ± 27 km s⁻¹, which is comparable to that of LBGs (e.g., Erb et al. 2006b; Steidel et al. 2010) in contrast to Δv_Lyα, although the current small sample of LAEs with a Δv_IS is insufficient to provide a definitive conclusion for Δv_IS of LAEs and LBGs.

We compare the H i covering fraction f_c of LAEs derived in Section 4.4 with that of z ~ 2–3 LBGs in Jones et al. (2013). Note that here we place lower limits on f_c when τ 1. Figure 12 displays the relation between f_c and Lyα EW, indicating a tentative trend that f_c decreases with Lyα EW. This trend has already been found in Jones et al. (2013) using an LBG sample. We find that the trend continues in objects with a higher Lyα EW. Our slope of the trend is slightly steeper than that in Jones et al. (2013), which would result from the wider dynamic range in Lyα EW. However, this trend could arise from the difference in the spectral resolution, although this tendency may marginally be found for LAEs alone. In addition to f_c, we compare EW(LIS) between LAEs and z = 3–4 LBGs in the bottom panel of Figure 11. Shapley et al. (2003) have found that EW(LIS) decreases with increasing EW(Lyα) with composite spectra of LBGs. Our LAEs with EW(Lyα) = 30–90 Å follow the trend between EW(Lyα) and EW(LIS), which might be indicative of a low velocity dispersion and/or low f_c, as suggested by Shapley et al. (2003). These results related to the low f_c imply the need for modeling Lyα line profiles emitted from a non-spherical shell of neutral gas (e.g., Zheng & Wallace 2013; Behrens et al. 2014).

Zoom In Zoom Out Reset image size

As described in the previous sections, we definitely confirm the anti-correlation between Δv_Lyα and Lyα EW by using a larger LAE sample than previously available. In this section, we explore the physical origin of the small Δv_Lyα in high Lyα EW galaxies.

Models predict that the redshift of the Lyα emission line should increase with either outflow velocity or neutral hydrogen column density (N_H i) (Verhamme et al. 2006, 2008). We have shown that the outflow velocities of LAE are comparable to those of LBGs, so the smaller Δv_Lyα for LAE is likely to be due to lower column densities in these objects.

In order to address the origin of the small Δv_Lyα in LAEs, we introduce the velocity offset ratio,

$$\begin{equation} R^{\rm Ly\alpha }_{\rm IS} \equiv \frac{\Delta v_{\rm Ly\alpha }}{\Delta v_{\rm IS}}. \end{equation} \tag{ 6 }$$

The value of $R^{\rm Ly\alpha }_{\rm IS}$ could trace purely physical properties such as N_H i and the dust amount by excluding the kinematic effect of a bulk outflow, since the quantity is normalized by the outflowing velocity, as suggested in (Verhamme et al. 2006). Hashimoto et al. (2013) infer the average value of Δv_IS for LAEs from a stacked spectrum of four LAEs with a z_sys, and compare $R^{\rm Ly\alpha }_{\rm IS}$ between LAEs and LBGs. In the stacking analysis, $R^{\rm Ly\alpha }_{\rm IS}$ is found to be ~1 for LAEs which is slightly smaller than that of LBGs, but the uncertainties are large.

Here, we estimate $R^{\rm Ly\alpha }_{\rm IS}$ for the four continuum-detected LAEs, and compare the quantities with those of z = 2–3 LBGs in Erb et al. (2006a, 2006b). In the comparison, we use LBGs with a negative Δv_IS value that indicates the outflow is present. The LAEs have a $R^{\rm Ly\alpha }_{\rm IS}$ of ~0.6–1.4, while LBGs have a wide variety of the quantity from 0 to ~10. Nevertheless, the average $R^{\rm Ly\alpha }_{\rm IS}$ for the LAEs is systematically smaller than that of LBGs. This indicates that LAEs tend to have a small N_H i compared to LBGs based on the expanding gas shell model of Verhamme et al. (2006). The small $R^{\rm Ly\alpha }_{\rm IS}$ in LAEs would be indicative of a small N_H i in LAEs.

Next, we examine possible correlations of Δv_Lyα and $R_{\rm IS}^{\rm Ly\alpha }$ with physical properties inferred from the SED fitting (Section 5). The physical properties of LAEs with Δv_Lyα in the literature are given in Table 7. In correlation tests, Δv_Lyα and $R_{\rm IS}^{\rm Ly\alpha }$ correlate most strongly with mass-related quantities and star formation rate (SFR), respectively. Figures 13 and 14 show the correlations of these quantities, respectively, including LAEs and LBGs with a z_sys in the literature. The SFR value of several LAEs is based on a Hα flux through the relation of Kennicutt (1998). The SFR based on a Hα flux is found to be comparable to the value inferred from SED fitting (Hashimoto et al. 2013). We conduct Spearman rank correlation tests in order to find the most related physical quantities to Δv_Lyα and $R_{\rm IS}^{\rm Ly\alpha }$ in the same manner as Steidel et al. (2010). Table 8 summarizes the results of the Spearman rank correlation tests.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 8. Correlations between Lyα Kinematics and Galaxy Properties

Quantity	ΔvLyα	$N_{\Delta v_{\rm Ly\alpha }}$	$R^{\rm Ly\alpha }_{\rm IS}$	$N_{R^{\rm Ly\alpha }_{\rm IS}}$
(1)	(2)	(3)	(4)	(5)
SFR	0.065	53	0.241	31
sSFR	−0.098	51	−0.852	31
E(B − V)	0.792	56	0.272	34
M*	0.001	52	0.810	31

Notes. Columns: (1) Physical quantity. (2) Probabilities satisfying the null hypothesis that the quantities are not correlated in Spearman rank correlation tests. A smaller absolute value of the probabilities implies that a physical property more correlates with a Lyα velocity offset. Negative values indicates anti-correlations. (3) Number of galaxies in the correlation test between Δv_Lyα and physical quantities. (4)–(5) Probabilities and galaxy numbers in the correlation tests for $R^{\rm Ly\alpha }_{\rm IS}$. Two LBGs with an extremely high $R^{\rm Ly\alpha }_{\rm IS}$ value of >25 are excluded in the correlation tests.

Download table as:  ASCIITypeset image

For Lyα velocity offsets, we find that the Δv_Lyα strongly correlates with SFR and stellar mass, which has not been observed previously in an LBG sample (Steidel et al. 2010). These correlations may have merged because our sample covers larger dynamic ranges of SFR and M_*. The correlation between Δv_Lyα and specific SFR (sSFR) may arise from the stellar mass.

As far as the velocity offset ratio is concerned, we do not find a notable correlation between $R_{\rm IS}^{\rm Ly\alpha }$ and the physical properties. Nonetheless, the correlation tests indicate that $R_{\rm IS}^{\rm Ly\alpha }$ most correlates with SFR among the four physical quantities. The correlation may reflect the connection between star formation and N_H i, if $R_{\rm IS}^{\rm Ly\alpha }$ is sensitive to N_H i. A larger sample of LAEs with a $R_{\rm IS}^{\rm Ly\alpha }$ measurement might reveal its physical connections with galactic properties.

With our larger sample of LAEs with a z_sys, we conclusively confirm that LAEs typically have a smaller Δv_Lyα than LBGs with a lower Lyα EW, while their outflowing velocities are similar in the two populations. These results yield a small $R^{\rm Ly\alpha }_{\rm IS}$ in LAEs, which indicates a small N_H i in galaxies with a high Lyα EW. The anti-correlations of f_c and EW(LIS) with Lyα EW in Figures 12 and 11 are consistent with the small N_H i in LAEs. The patchy H i gas clouds surrounding the central source would lead to a small flux-averaged N_H i corresponding to a small $R^{\rm Ly\alpha }_{\rm IS}$. In this condition, Lyα photons could easily escape less affected by resonant scattering in the clouds. The results of our kinematic analyses support the idea that the H i column density is a key quantity determining Lyα emissivity.

Moreover, recent NIR spectroscopy by Nakajima et al. (2013) has suggested that LAEs have a large [O iii]/[O ii] ratio, indicating these systems are highly ionized with density-bounded H ii regions. This tendency has been confirmed by a subsequent systematic study in Nakajima & Ouchi (2013). The large [O iii]/[O ii] ratio also indicates a low column density of H i gas. A stacked UV continuum spectrum of our eight LAEs shows that LIS absorption lines have a low EW, as shown in Figures 15 (see also Figure 11). The weak LIS absorption lines are consistent with a large [O iii]/[O ii] ratio in LAEs (e.g., Jones et al. 2012).

Zoom In Zoom Out Reset image size

In our first paper of the series investigating LAE structures, we find that LAEs with a high Lyα EW tend to be a non-merger, to show a small Lyα spatial offset between Lyα and stellar continuum emission δ_Lyα, and to have a small ellipticity by using a large sample of 426 LAEs (Shibuya et al. 2014). On the basis of these results on the gas distribution, the difference in H i column density explains the Lyα-EW dependences of the merger fraction, the Lyα spatial offset, and the galaxy inclination. For objects with density-bounded H ii regions, Lyα photons would directly escape from central ionizing sources, which produce a small δ_Lyα. The low H i abundance along the line of sight also induces the preferential escape of Lyα to the face-on direction.

All of the above results suggest that ionized regions with small amounts of H i gas dominate in galaxies with a high Lyα.