THE LYMAN ALPHA REFERENCE SAMPLE. III. PROPERTIES OF THE NEUTRAL ISM FROM GBT AND VLA OBSERVATIONS

The Lyman-alpha emission line (Lyα) at 1216 Å fulfills several extremely important roles in observations of the high-redshift (z) universe. As recombination nebulae re-process one-third of the raw ionizing power of hot stars into a single spectral emission feature, it acts as a luminous spectral beacon by which to identify galaxies at the highest redshifts (see, e.g., Hayes et al. 2010 and references therein). Further, at the highest redshift it provides the potential for the all-important spectroscopic confirmation of distant galaxies selected by other methods, transforming the status of candidates to objects with secure redshifts (e.g., Stark et al. 2010).

Lyα is a resonance line, and photons scatter wherever they encounter neutral hydrogen atoms. This scattering may occur in the interstellar medium (ISM) of the galaxies themselves, or in the circumgalactic medium (CGM) that immediately surrounds them. Scattering in the CGM can lead to extended Lyα structures such as those found in high-z Lyα emitters (LAE) by Steidel et al. (2011) and Zheng et al. (2011) among others. Inside the ISM, the radiative transport of Lyα depends strongly on dust extinction (Atek et al. 2014). This dependence, however, shows large scatter due to the complicated resonant scattering of Lyα, in which the visibility of the line is influenced by a large number of factors, including dust content (Charlot & Fall 1993; Verhamme et al. 2008; Atek et al. 2009; Hayes et al. 2010), dust geometry (Scarlata et al. 2009), neutral gas content and kinematics (Kunth et al. 1998; Mas-Hesse et al. 2003; Cannon et al. 2004), and gas geometry (Neufeld 1991; Giavalisco et al. 1996; Hansen & Oh 2006; Laursen et al. 2013; Duval et al. 2014). Taken together, these myriad factors indicate that the true total escape fraction of Lyα photons ($f_\mathrm{esc}^{\mathrm{Ly}\alpha }$; defined as the ratio of observed to intrinsic Lyα luminosity; Hayes et al. 2005, 2013), is still poorly understood. Thus, photometric measurements of Lyα will reflect the underlying properties of galaxies only in the very broadest statistical sense.

The only way to remedy this complicated situation is to study the Lyα emission line in a sample of star-forming galaxies on a spatially resolved basis. From the UV perspective, obtaining Lyα images in the nearby universe is expensive and requires space-based platforms. To this end, our team has acquired data from the Hubble Space Telescope (HST) to produce Lyα images of a statistically selected sample of star-forming galaxies in the low-z universe (see G. Östlin et al., in preparation); details about the image processing methods can be found in Hayes et al. (2009). This program, called the "Lyman Alpha Reference Sample" ("LARS"), includes HST Cycle 18 imaging and Cycle 19 spectroscopic observations of 14 galaxies. These HST data sets form the backbone of a comprehensive multi-wavelength observational campaign that will allow great strides forward in our understanding of Lyα radiative transport. Since the LARS sample is selected on UV luminosity and Hα equivalent width, and not on Lyα characteristics, the composite sample was designed to overlap with both high-z and low-z star-forming galaxies. LARS will form a critical local interpretive benchmark for studies of high-z galaxies for decades to come, and is well placed to take advantage of the remaining HST lifetime and motivate future observations with JWST.

The LARS program (including sample selection) is described in detail in Östlin et al. (2014) and Hayes et al. (2014), hereafter referred to as Papers I and II respectively. Preliminary results derived from the Lyα images of the LARS galaxies are presented in Hayes et al. (2013). That work showed that in many LARS galaxies, Lyα emission is prominent on physical scales that exceed those of both the massive stellar populations and the star formation regions that give rise to the Lyα photons; this trend is very prominent in the panels of Figures 1–3, which show optical and UV images of the LARS galaxies. This may point towards Lyα photons being resonantly scattered to large radii in most of the LARS galaxies, although other possibilities have been proposed, such as ionization caused by hot plasma (Otí-Floranes et al. 2012). It is likely that multiple of the aforementioned physical properties (dust content and geometry, neutral gas content, kinematics, geometry) are facilitating this efficient resonant scattering.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In order to decouple the effects of dust and neutral gas kinematics and densities, the Lyα images must be compared with spatially resolved information about the neutral ISM. While UV and optical absorption line spectroscopy offers one avenue to address this issue, such data are limited to those regions in the foreground of only the brightest continuum sources (usually the emission line nebulae themselves). While our team is pursuing such analyses (see T. Rivera-Thorsen et al., in preparation, and Lyα profile modeling from I. Orlitova et al., in preparation), these avenues should be complemented by direct probes of the neutral interstellar medium in all regions of the LARS galaxies. To this end, in this manuscript we begin a detailed exploration of the neutral hydrogen in the LARS galaxies by presenting new single-dish H i spectroscopy and spatially resolved interferometric H i imaging. We use these data to derive global properties and to study the properties of the H i gas on bulk scales.

Previous spatially resolved H i observations of galaxies in which Lyα radiative transport has been studied in detail are few. While the sample of systems with spectroscopic information about Lyα is robust, there are comparatively few systems that have direct Lyα imaging (see the discussion in Hayes et al. 2005, 2009; Östlin et al. 2009 and references therein). For those systems that do, there has, until the present work, been no systematic investigation of the H i on a spatially resolved basis; only a few selected systems have detailed H i studies in the literature.

Cannon et al. (2004) presented H i observations of two of the most well-studied and nearby Lyα-emitting galaxies in the local universe, ESO 338–IG 004 (Tol 1924–416) and IRAS 08339+6517. In each galaxy, extended neutral gas connects the target and nearby neighbors, suggesting that interactions have played an important role in triggering the massive starbursts in ESO 338−IG 004 and IRAS 08339+6517. The close proximity of the companions suggests that the interactions were recent. Since both primary systems are Lyα emitters, these data support two possible interpretations of Lyα escape from starburst galaxies: either (1) the bulk ISM kinematics provides the means of escape for Lyα photons by shifting H i atoms out of resonance, or (2) the H i is sufficiently clumpy on scales below our resolution to allow for efficient Lyα escape. These interpretations are further supported by H i (Bergvall & Jorsater 1988) and Lyα observations (Kunth et al. 1998) of ESO400−G43.

In this work, we begin to extend this type of analysis to the entire LARS sample. Basic properties of the LARS galaxies are summarized in Table 1. Figures 1–3 show optical and UV images of the 14 LARS galaxies, with the left and right columns displaying the color HST images and the SDSS color images respectively. The right column includes the immediate surroundings, while the left column has smaller fields of view, and are scaled to show detail in the galaxies.

We organize this manuscript as follows. In Section 2 we describe the data acquisition, reduction, and analysis. Section 3 presents the single-dish H i profiles of the 14 LARS galaxies, while Section 4 presents spatially resolved H i images of 5 of the 14 LARS systems. We discuss each system individually in Section 5, and discuss various correlations between galaxy global properties in Section 6. We draw our conclusions in Section 7. Throughout this paper we adopt the value of H₀ = 70 ± 1.4 km s⁻¹ Mpc⁻¹ from the averaged WMAP 7 yr data as presented in Komatsu et al. (2011).

2.1. GBT Data

We obtained data from the National Radio Astronomy Observatory 100 m Robert C. Byrd Green Bank Telescope (GBT¹³) under program GBT/11A-057 (Legacy ID QO13; PI: Östlin). Data were acquired in four observing sessions in 2011 March and April. Position switching observations were performed for all sources; 32,768 channels over the 12.5 MHz of total bandwidth produce a spectral resolution of 381.469 Hz (0.08 km s⁻¹ ch⁻¹).

All reductions were performed in the IDL environment,¹⁴ using the GBTIDL package designed at NRAO. We first imported the raw data for each galaxy into GBTIDL using standard averaging techniques, with zenith opacity coming from atmospheric data from the NRAO CLEO weather system. The data was moderately contaminated by terrestrial radio frequency interference (RFI); bad data were identified by examining individual spectra, blanked, and interpolated.

The standard position switching algorithms in GBTIDL were used to produce a combined spectrum for each galaxy. To increase the signal-to-noise ratio (S/N) of the observations we first averaged the reference spectrum by 16 channels. This procedure can have a negative effect on the baseline shape, increasing the order of the polynomial (often third order) that we used to fit and subtract the continuum (see Table 2 for the order of the fit for each galaxy). The S/N was quite low on most of the LARS galaxies and we thus smoothed our spectrum by 256 channels resulting in a velocity resolution of ∼20 km s⁻¹. Flux and width measurements were taken on this smoothed profile. Two regions were selected by hand for each galaxy: the first around the profile, and the second in an area free from RFI (used as an rms noise region).

Table 2. Observed and Derived GBT H i Properties

Galaxy	Vsysa	W50b,c	W20b	Velocity Offset	SH id	MH i	SNR	RMS	Baselinee	Typef
	(km s−1)	(km s−1)	(km s−1)	(km s−1)	(Jy km s−1)	(109 M☉)	(Sum, Peak)	(mJy)	Fit Order
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
LARS 01	8339 ± 5	160 ± 10	180 ± 16	−55 ± 5	0.70 ± 0.07	2.5 ± 0.3	(3.1, 6.5)	1.1	2	D
LARS 02	8940 ± 8	140 ± 17	150 ± 26	3 ± 8	0.70 ± 0.07	2.8 ± 0.3	(2.2, 2.9)	1.8	3	D
LARS 03	9421 ± 20	310 ± 39	380 ± 62	230 ± 20	1.6 ± 0.2	6.9 ± 0.7	(3.1, 4.9)	1.2	3	D
LARS 04	9740 ± 16	150 ± 31	260 ± 49	−3 ± 16	1.6 ± 0.2	7.7 ± 0.8	(5.1, 11.)	1.0	3	S
LARS 05	10117 ± 14	160 ± 28	170 ± 44	−10 ± 14	<0.55	<2.9	(1.3, 1.5)	1.1	1	I
LARS 06	10378 ± 12	370 ± 3	390 ± 5	155 ± 2	4.4 ± 0.4	23. ± 2	(9.5, 18.)	1.1	3	D
LARS 07	11315 ± 16	100 ± 32	170 ± 50	−10 ± 16	0.47 ± 0.05	3.1 ± 0.3	(3.3, 6.2)	0.86	3	S
LARS 08	11468 ± 23	310 ± 47	470 ± 73	20 ± 23	3.4 ± 0.3	22. ± 2	(5.7, 11.)	0.91	3	I
LARS 09	14020 ± 46	270 ± 92	490 ± 140	−130 ± 46	1.2 ± 0.1	13. ± 1	(2.8, 7.3)	0.73	3	S
LARS 10	17284 ± 69	280 ± 140	380 ± 220	70 ± 69	0.29 ± 0.03	4.5 ± 0.5	(1.1, 2.1)	0.69	3	C
LARS 11	25366 ± 56	260 ± 11	290 ± 18	65 ± 6	0.75 ± 0.08	26. ± 3	(2.9, 6.4)	0.89	3	D
LARS 12	(30609)	(264)	(290)	N/A	<2.9	<150	N/A	3.4	3	N
LARS 13	(43980)	(264)	(290)	N/A	<33	<3800	N/A	38.	5	N
LARS 14	(54172)	(264)	(290)	N/A	<1.7	<310	N/A	2.0	1	N

Notes. ^aValues from nondetections are from the SDSS optical redshifts. ^bThese values are derived from fits to the slopes on either side of the profile. See discussion in Section 2. ^cAssumed linewidths for nondetections are the average linewidth of detected sources. ^dUpper limits are 1.5σ above the local rms noise over the average W₂₀ value (290 km s⁻¹). ^eOrder of polynomial that was fit to the GBT baseline. ^fThe classification of the profile: "S" for single-horned, "D" for double-horned, "C" for confusion caused by other galaxies, "I" for an irregular profile, "N" for no detection.

Download table as:  ASCIITypeset image

The redshifts of the LARS galaxies (see Table 1) place some of them in proximity to a bandpass filter in the 1.2–1.3 GHz region. This bandpass filter is designed to shield against a local radar signal, but contributes higher than usual system temperature (T_sys) values (and therefore system noise) to calibrated data in this frequency range. Further, the T_sys values fluctuated markedly from one spectral scan to the next. The affected galaxies were all nondetections (LARS 12, LARS 13, and LARS 14) and have markedly higher T_sys values (in particular LARS 13) than would otherwise be expected. It should also be noted that these three galaxies are at the largest redshifts in our sample.

Systematically measuring the flux and width of irregular, low S/N, H i spectra can be challenging. To do so, we have modified previously well known measurement techniques and find that they are robust over the range of S/N and profile shapes in our data. Our measurement routines take as inputs the smoothed spectrum and the region suspected of containing the galaxy. In our initial tests of archival data for NGC 5291 (available in the distribution of GBTIDL), we found that hand selecting regions of the profile could add additional bias, causing the width and flux measurements to vary by 2%–5%. Instead, we opted to use χ² minimization techniques to fit Gaussian components (a single component fit for galaxies with a single peak and a multi component fit for double-horned profiles) to the smoothed spectrum. These components were used to guide later analysis and ensured repeatability in our results even with low S/N. These Gaussian fits serve only to guide analysis, however, and the flux and width measurements follow the methodology described in Springob et al. (2005). This method fits lines to the profile in the region over which it varies the most (here between 15% and 85% of the peak value) and measures width properties from these fits. Springob et al. (2005) found that this reduced the dependence on noise on their measurements versus direct measurements. We modified this method to respond better to low S/N values and to handle single peaked data more accurately. We consider this method to be the "best representation" of the width measurements for these spectra, although it is not the only method (e.g., the busy function; Westmeier et al. 2014). We recorded two S/N based on the peak-to-rms ratio and on the integrated flux to integrated rms ratio.

The flux accuracy of GBT data using standard reduction techniques is quoted as 10% in the GBTIDL calibration guide. The primary error terms are the rms fluctuations of the spectra in the region around the target and the subjectivity of selecting the baseline regions and rms/profile regions. In testing it was seen that increasing the order of the baseline generally led to increasing error. We thus adopt a 10% flux error for "good" profiles and a conservative 20% for "irregular" profiles.

Special attention was paid to the error terms presented in Springob et al. (2005) as modified from Schneider et al. (1990). The error term given there adds an error associated with the quality of the baseline fit. Because we have assumed a separate baseline error, we remove this term to get a more or less standard error term:

$$\begin{equation} \epsilon _S^{{\rm stat}}=\,\mathrm{rms}\sqrt{W \Delta V}, \end{equation} \tag{ 1 }$$

where $\epsilon _S^{{\rm stat}}$ is the noise in units of Jy km s⁻¹, and W is the area of integration for the profile. The width at 50% of the maximum value (W₅₀) results reported are highly dependent on the accuracy of the method described in Springob et al. (2005) which, as mentioned in the above discussion, breaks down at lower S/N.

The error for the width and systemic velocity measurements is based on the equations found in Schneider et al. (1990)

$$\begin{equation} \sigma V = 1.5(W_{20}-W_{50}) ({\rm S/N})^{-1}. \end{equation} \tag{ 2 }$$

This equation is doubled to give the error of the width measurements W₂₀ and W₅₀, which refer to the width at 20% and 50% of the maximum, respectively.

2.2. VLA Data

2.3. Lyα Data

We use global Lyα data for all 14 LARS galaxies as presented in Paper I and Paper II. The HST data were obtained with the Wide Field Camera 3 and Advanced Camera for Surveys using Hα and Hβ filters, and a combination of U, B, and I bands for FUV continuum imaging. We also acquired data with the Solar Blind Channel using a bandpass designed to isolate Lyα emission.

We reduced imaging data using MULTIDRIZZLE¹⁸ in the standard HST pipelines and used the Lyman-alpha extraction Software (LaXs; Hayes et al. 2009) to continuum subtract the Lyα UV data. We refer the reader to Papers I and II for details of this process. The final global properties (see Paper I and Table 5) are measured within apertures that correspond to twice the Petrosian radius (η = 0.2; Petrosian 1976). This reduces the effect of noise on the measurements at large radii.

Table 5. Global UV Properties of the LARS Sample Galaxies^a

Galaxy	LLyα	LHα	$f_{{\rm esc}}^{\mathrm{Ly}\alpha \ }$	SFR$^{{\rm FUV}}_{corr.}$	Metallicity	WLyα	ξLyα	M*
	(1042 cgs.)	(1042 cgs.)		(M☉ yr−1)	(12+log(O/H))	(Å)		(109 M☉)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
LARS 01	0.85 ± 0.09	0.63 ± 0.019	0.119 ± 0.012	3.14 ± 0.09	8.26 ± 0.04	33.0 ± 3.44	$3.37_{-0.90}^{+1.1}$	6.10 ± 0.25
LARS 02	0.81 ± 0.02	0.18 ± 0.005	0.521 ± 0.015	1.01 ± 0.03	8.23 ± 0.04	81.7 ± 2.36	$2.27_{-0.57}^{+0.83}$	2.35 ± 0.20
LARS 03	0.10 ± 0.02	0.59 ± 0.018	0.003 ± 0.000	2.41 ± 0.07	8.38 ± 0.20	16.3 ± 2.71	$0.77_{-0.15}^{+0.10}$	20.1 ± 1.6
LARS 04	0.00	0.60 ± 0.018	0.000	3.37 ± 0.10	8.20 ± 0.04	0.00	...	12.9 ± 0.85
LARS 05	1.11 ± 0.13	0.51 ± 0.015	0.174 ± 0.021	3.23 ± 0.10	8.13 ± 0.04	35.9 ± 4.32	$2.61_{-0.60}^{+0.77}$	4.27 ± 0.10
LARS 06	0.00	0.08 ± 0.002	0.000	0.62 ± 0.02	8.08 ± 0.06	0.00	...	2.09 ± 0.15
LARS 07	1.01 ± 0.01	0.52 ± 0.016	0.100 ± 0.001	2.61 ± 0.08	8.36 ± 0.05	40.9 ± 0.41	$3.37_{-0.46}^{+0.77}$	4.75 ± 0.11
LARS 08	1.00 ± 0.07	1.50 ± 0.045	0.025 ± 0.002	8.81 ± 0.26	8.50 ± 0.15	22.3 ± 1.59	$1.12_{-0.12}^{+0.19}$	93.3 ± 11
LARS 09	0.33 ± 0.03	2.61 ± 0.078	0.007 ± 0.001	15.0 ± 0.45	8.40 ± 0.05	3.31 ± 0.28	>2.85	51.0 ± 1.4
LARS 10	0.16 ± 0.05	0.34 ± 0.010	0.026 ± 0.008	2.29 ± 0.07	8.51 ± 0.14	8.90 ± 2.82	$2.08_{-0.38}^{+0.42}$	21.5 ± 3.2
LARS 11	1.20 ± 0.20	1.66 ± 0.050	0.036 ± 0.006	22.3 ± 0.67	8.43 ± 0.30	7.38 ± 1.22	$2.27_{-0.94}^{+0.65}$	121 ± 2.6
LARS 12	0.93 ± 0.10	1.96 ± 0.059	0.009 ± 0.001	13.3 ± 0.40	8.35 ± 0.05	8.49 ± 0.87	$3.48_{-0.33}^{+0.46}$	7.41 ± 1.5
LARS 13	0.72 ± 0.08	2.46 ± 0.074	0.010 ± 0.001	18.8 ± 0.56	8.50 ± 0.12	6.06 ± 0.68	$1.74_{-0.13}^{+0.06}$	59.2 ± 1.8
LARS 14	4.46 ± 0.43	1.99 ± 0.060	0.163 ± 0.016	11.0 ± 0.33	8.15 ± 0.05	39.4 ± 3.83	$3.62_{-0.22}^{+0.82}$	1.75 ± 0.10

Note. ^aAll values from Hayes et al. (2013), G. Östlin et al. (2014, in preparation) and Hayes et al. (2014).

Download table as:  ASCIITypeset image

Presented in Figure 4 are the fully smoothed and reduced GBT H i profiles of the LARS galaxies. Plotted over these are the Gaussian fits in gray, solid lines in red showing the sides of the profile used in the peak W₅₀ calculations, and the peak W₅₀ line itself shown as a horizontal blue dotted line. A discussion of each galaxy is included below.

Zoom In Zoom Out Reset image size

The observed properties from our GBT observations of each galaxy can be found in Table 2 along with properties derived from these using methods described below. The distances of each galaxy are calculated from the standard LARS luminosity distances. The H i mass in units of M_☉ is calculated via M_H i = 2.36 × 10⁵ D² S_H i where D is the distance in Mpc and S_H i is the H i flux integral in units of Jy km s⁻¹. We will discuss the mass value for each galaxy in Section 5. Overall we find that our detections and mass estimates show a strong dependence on distance.

It is important to emphasize that the 8' primary beam of the GBT at 21 cm is sensitive to all H i-bearing objects in the observed frequency range. Even for the closest LARS galaxy (LARS 01; D = 120 ± 2.4 Mpc), this 8' beam subtends a projected circular area that is nearly 280 kpc in diameter; the potential for contamination in the GBT beam is significant. We thus examined SDSS images of the area surrounding each LARS galaxy for nearby objects. Objects with spectroscopic data available from SDSS that might have contributed to the GBT flux are noted below. These objects have velocities within 300 km s⁻¹ of the target LARS galaxy and are within the 8' primary beamsize of the GBT. Optical images of the fields that contain possible confusing sources (LARS 06 and LARS 11) are shown in Figures 5 and 6; these two sources are discussed in more detail in Section 5. We used the SDSS g band to convert to luminosity L_g, which we compared with a log(M_H i/L_B) value of −0.5 for late type galaxies found in (Haynes & Giovanelli 1984). The masses of potential contamination galaxies are computed from this ratio. We also use the 22.2 mag limit for the SDSS to determine the possible contributions from optically dim sources of contamination. For our closest nine galaxies this is roughly 10⁶M_☉ of H i; this increases to 10⁸ for our furthest two systems. This information is also presented in Table 2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The profiles in Figure 4 show a variety of H i structure. Some galaxies are simple single peak profiles (e.g., LARS 04, LARS 09), while others show complex and asymmetric profiles (e.g., LARS 01, LARS 03). Still other systems have H i spectra that are almost certainly confused; the classic double-horn profile of LARS 06 is likely due to the contribution of UGC 10028, a large spiral galaxy located only 1' to the southeast of LARS 06 (see Figure 5).

Various observed properties of each GBT H i profile, and the quantities derived from these, are tabulated in Table 2: Column 1 identifies the galaxy number; Column 2 is the systemic velocity in km s⁻¹; Column 3 is the linewidth at 50% in km s⁻¹; Column 4 is the linewidth at 20% in km s⁻¹; Column 5 is the velocity offset between the optical velocity and the H i central velocity; Column 6 is the single-dish integrated flux in Jy km s⁻¹; Column 7 is the mass from the single-dish measurements in units of 10⁹ M_☉; Column 8 is the S/N from the integrated flux and the peak; Column 9 is the rms noise value of the GBT in units of mJy; Column 10 gives the baseline polynomial fit used in the calibration of the GBT spectrum; Column 11 is the classification of the profile (''S" for single-horned, "D" for double-horned, "N" for no detection, "I" for highly irregular shapes, or the presence of RFI near the signal region, and "C" for confusion due to known nearby galaxies.)

The H i moment zero (representing H i mass surface density) and moment one (representing intensity-weighted velocity) images of the five LARS galaxies observed in program 13A-181 are presented in Figures 7–11. In panel (a) of each figure, a Digitized Sky Survey image is overlaid with contours of H i mass surface density (individual contour levels are provided in the caption to each figure). The beam size is shown by a blue circle, while a red square shows the location of the HST pointing for which we have UV imaging (see discussion in Papers I and II). In panel (b) of each figure, the H i moment one image (representing intensity weighted velocity) is presented in a rainbow color format; the velocity scale of each galaxy is shown by a colorbar.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Based on the distances and optical angular sizes of the LARS galaxies (of order 1' or less; see Figures 1–3), one would expect the H i to be distributed on similar angular scales. At the resolution of these data (beam sizes between 59'' and 72''), the LARS galaxies would thus appear as unresolved sources in the H i data cubes. As we discuss below, three of the LARS galaxies have H i morphologies consistent with this interpretation. Surprisingly, two of the sources have H i distributions that largely exceed the beam size; we discuss these sources in detail below.

It is important to stress that even in cases where galaxies are unresolved by an H i beam, one can still extract bulk characteristics of the neutral ISM because of the spectrally resolved nature of the data. Detailed kinematic analyses (e.g., rotation curve work or spatially resolved position–velocity diagrams) will be unavailable in such cases. However, within the spatial and spectral resolution elements of the data, one can derive bulk constraints on the motions of the H i gas.

We use the VLA H i images of the five LARS galaxies presented here to constrain two critical properties. First, the H i data allow us to localize the H i associated with the LARS galaxy. This in turn allows us to accurately measure the total H i mass of each system. These interferometric measurements often recover less of the H i flux, and therefore mass, than the GBT measurements. We attribute this to the low S/N of H i in these galaxies, the possible presence of low surface brightness H i features missed by the VLA in extended regions or velocity space, and flux contamination from other galaxies, although we suspect that this is only relevant in the cases marked as confused.

It is interesting to note that at the distances of the LARS galaxies, interferometric observations are perhaps better suited to determine the H i mass than single-dish observations (for example, the 60'' characteristic resolution of the present data are about a factor of three better than the H i beam of Arecibo, the largest single-dish radio telescope in the world, and a factor of eight better than our GBT resolution), simply because of the potential for contamination in the single-dish beam. Second, the present data are sensitive to large-scale distributions of H i gas, some of which is expected to be tidal in origin based on the optical morphologies of the LARS galaxies alone (see, e.g., Figures 1–3).

In Table 4 we present VLA H i quantities for our five targeted galaxies derived in the same method as described above for the GBT data. Column 1 identifies the galaxy number; Column 2 is the systemic velocity in km s⁻¹; Column 3 is the linewidth at 50% in km s⁻¹; Column 4 is the linewidth at 20% in km s⁻¹; Column 5 is the velocity offset between the optical velocity and the H i central velocity; Column 6 is the interferometric integrated flux in Jy km s⁻¹; Column 7 is the mass in units of 10⁹ M_☉; Column 8 is the rms noise value in units of mJy Beam⁻¹; Column 9 is the median global column density of H i from a column density map convolved with the beam size given in 10¹⁹ cm⁻².

We now present discussion of each LARS target based on our new H i data and on the HST imaging presented in Hayes et al. (2013), Paper I, and Paper II.

5.1. LARS 01

5.2. LARS 02

5.3. LARS 03

5.4. LARS 04

5.5. LARS 05

5.6. LARS 06

5.7. LARS 07

5.8. LARS 08

5.9. LARS 09

5.10. LARS 10

5.11. LARS 11

Due to resonant scattering with neutral hydrogen atoms, the radiative transport of Lyα photons may well have its strongest dependence on the characteristics of the H i that surrounds the star forming regions in which the photons are produced. With the present H i spectroscopy and imaging, we now have a first order understanding of the neutral gas contents of many of the LARS galaxies. Comparing the qualities of the global H i reservoirs with the detailed characteristics of each system derived from HST imaging allows us to probe what roles H i kinematics and density play in governing Lyα radiative transport.

We focus on four global properties derived from the present H i observations of the 11 lowest-redshift LARS galaxies (those for which the GBT spectra provide meaningful measurements or limits); for the galaxies where it is available (LARS 02, 03, 04, 08, 09) we use the total fluxes and masses VLA images preferentially over the GBT data. In general the VLA data recovers 30% less flux than the GBT data, which contributes to 6% shorter linewidths. As mentioned in Section 4, this is most likely due to a combination of decreased surface brightness sensitivity in the VLA data and decreased contamination from other sources.

First, the H i linewidth (specifically, the width at 50% of the maximum W₅₀) is a distance-independent variable that has been shown to correlate with absolute magnitude and thus to serve as a proxy for mass in star-forming disk galaxies (Tully & Fisher 1977). For rotationally dominated galaxies, a larger H i linewidth can often be interpreted as an indicator of a more massive galaxy. To know with certainty, a detailed study of the dynamics and complete censuses of the baryonic components of individual galaxies is needed. In our case, the galaxies with irregular morphology may have linewidths that are significantly increased due to kinematic disturbances and merging activity.

The second global parameter we examine is the total H i mass of each LARS galaxy. This parameter is of course related to the H i linewidth, but not in a one to one sense. The LARS galaxies span a range of morphological types, including violent interactions (e.g., LARS 03), disk-dominated spirals (e.g., LARS 05 and LARS 11), and irregulars (e.g., LARS 04, LARS 06). While for disk dominated systems the H i linewidth and H i mass should be closely correlated, for the other types of systems the H i line profile may not only be indicative of rotation; H i gas from multiple components may create irregular single-dish profiles. Examples of this are clearly seen in the GBT spectra (Figure 4) and in the VLA images (see, e.g., Figure 8).

Third, we examine the offset in velocity between the centroid of the H i profile and the known systemic velocity of each LARS galaxy as derived from optical spectroscopy. Assuming that such offsets are caused by large-scale kinematic deviations from regularly rotating galaxy disks (e.g., tidal interaction, gas outflow or infall), this offset parameter can be considered a rough proxy for the presence or absence of large-scale motions of neutral gas. As discussed in Cannon et al. (2004) and above for LARS 03, we find evidence for extended tidal structure in some Lyα-emitting galaxies. However, we also find systems that are strong Lyα emitters that do not show such extended neutral gas components (e.g., LARS 02; see Section 5.2 above).

Finally, we examine the gas mass fraction defined as the H i mass divided by the stellar mass. Numerous works have shown that normal, star-forming disk galaxies populate a robust trend of decreasing gas fraction with increasing stellar mass (see recent results in Huang et al. 2012; Papastergis et al. 2012; Peeples et al. 2014 and various references therein). In Figure 12 we show the gas fraction as a function of total stellar mass (M_*) derived from two-component SED modeling to the HST data (see Paper II and Hayes et al. 2009 for details). Except for LARS 06, whose H i flux integral is likely strongly contaminated by the nearby field spiral UGC 10028 (see discussion in Section 5.6 and Figure 5), the LARS galaxies closely follow the derived relations for nearby galaxies as found in Papastergis et al. (2012) and Peeples et al. (2014).

Zoom In Zoom Out Reset image size

It is important to remain cognizant of the possible confusion within the GBT beam for some of the sources without VLA data. In particular, the H i properties of LARS 06 are almost certainly contaminated by UGC 10028, and those of LARS 11 may be contaminated by SDSS J140401.00+062901.7 and/or SDSS J140353.36+062504.8. If these sources are in fact contaminated, then the H i linewidths and masses will be overestimated; the velocity offsets may or may not change.

We examine the relationships of these four H i-based quantities with seven global properties derived from HST data in Paper II: the Lyα/Hα flux ratio, the Lyα luminosity, the Hα/Hβ flux ratio, the Lyα escape fraction, the Lyα equivalent width, the Lyα extension parameter (Lyα ξ), and the SFR per unit area within the Petrosian radius (η = 0.2; see Paper II). As discussed in detail in Hayes et al. (2014), there is a complex interrelation between these (and other) properties in the process of Lyα radiative transport; the total Lyα luminosity of a given galaxy does not correlate in a statistically significant way with any individual quantity. Further, there is significant variation in these values as a function of position within an individual galaxy; some galaxies appear as net absorbers within the disk and as net emitters when the diffuse Lyα halo is included. Nevertheless, we seek statistical comparisons with many of the global quantities of both Lyα and H i. These comparisons most closely match detections of higher-redshift Lyα emitters and will constrain one of the many poorly understood aspects of Lyα propagation and detection in the distant universe.

From a simplified interpretative standpoint, these seven global quantities can be related as follows. The Hα/Hβ ratio indicates extinction; deviations above the intrinsic ratio of 2.86 (Osterbrock 1989) indicate nonzero E(B − V) values. These E(B − V) values can in turn be used to derive the intrinsic Lyα luminosity of a system based on its observed (and extinction corrected) Hα luminosity and also to estimate the Lyα escape fraction. The Lyα/Hα ratio has a (Case B) recombination value of 8.7. Lower values can indicate stronger suppression of Lyα versus Hα (e.g., by attenuation from dust, or by resonant scattering away from the production site); super-recombination values are seen in a few localized regions of the LARS galaxies, as well as in some of the large scattered Lyα halos surrounding some of the systems (see Paper I and Paper II). The Lyα equivalent width is related to the recent star formation history of the galaxy or region; global values of ∼100 Å indicate constant star formation over ∼100 Myr timescales, while values exceeding 250 Å occur only in the very youngest burst episodes (Schaerer et al. 2003; Raiter et al. 2010.)

We present scatter plot comparisons of the four global H i-derived properties and seven global HST-derived properties in Figures 13, 14, 15, and 16. We restrict the Lyα properties to positive values, which has the effect of setting EW and luminosity measurements equal to zero. We mark this region on the plots with a hashed region. We use the Spearman ρ_s correlation coefficient to quantify possible monotonic correlations between these properties; a perfect correlation or anti-correlation between two properties will have ρ_s values of +1 or −1, respectively. Each panel of Figures 13 through 16 shows the corresponding ρ_s value. To estimate the errors on these correlation coefficients we resample the properties 1000 times with a random realization of the associated errors and measure, recording the correlation coefficients each time. For each correlation we show the value from the sample of uncontaminated detections as the primary result, and include the value for the entire sample including upper limits and confused detections below.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Overall the results do not show strong evidence of correlation between properties (ρ_s < −0.6 or ρ_s > +0.6). We present all correlations in Table 6, along with the dispersion and discuss the correlations for each H i property below.

Table 6. Correlations between H i and Global Lyα Properties^a

Lyα	W50	MH i	Velocity Offset	fgas
Property	ρsa	ρsa	ρsa	ρsa
(1)	(2)	(3)	(5)	(4)
Lyα/Hα	−0.52 [0.2]	−0.53 [0.06]	−0.083 [0.08]	0.47 [0.4]
Lum(Lyα)	−0.27 [0.1]	0.05 [0.1]	0.17 [0.04]	0.00 [0.1]
Hα/Hβ	0.67 [0.1]	0.75 [0.1]	0.50 [0.05]	−0.45 [0.3]
f$^{\mathrm{Ly}\alpha }_{{\rm escp}}$	−0.53 [0.2]	−0.63 [0.1]	−0.15 [0.1]	0.40 [0.4]
Lyα EW	−0.33 [0.2]	−0.60 [0.1]	0.10 [0.1]	0.45 [0.3]
ξLyα	−0.81 [0.1]	−0.45 [0.1]	−0.44 [0.2]	0.57 [0.1]
SFR area−1	−0.15 [0.1]	−0.40 [0.2]	−0.067 [0.04]	0.43 [0.2]

Notes. ^aValues are for all galaxies with a positive, unconfused detection. Standard deviation values are given in square brackets.

Download table as:  ASCIITypeset image

6.1. Linewidth

6.2. Mass

6.3. Gas Fraction

6.4. Velocity Offset

We have presented new GBT spectroscopy and VLA H i spectral line imaging of the galaxies in the "Lyman Alpha Reference Sample" ("LARS"). LARS is a comprehensive, multi-wavelength study of 14 UV-selected star-forming galaxies that functions as a local benchmark for studies of Lyα emission and absorption at higher redshifts. The centerpiece is a suite of HST images that allows us to study Lyα emission and related quantities on a spatially resolved basis. The preliminary results presented in Hayes et al. (2013) showed that most of the LARS galaxies have large, diffuse halos of Lyα emission that exceed the sizes of the stellar populations; this is strong evidence for the importance of resonant scattering in the neutral hydrogen gas component. The H i structure and content are important pieces for understanding the origin of the Lyα halos and their different sizes. Our current H i angular resolution probes the large-scale global properties rather than those in the immediate vicinity of the Lyα halo regions imaged with HST. Nevertheless, these provide important insights into the galaxy structure and environment. The LARS survey products are described in Paper I of this series (Östlin et al. 2014); the integrated properties of the LARS galaxies as derived from the HST imaging are presented in Paper II of this series (Hayes et al. 2014).

In this manuscript we present a first exploration of the H i properties of the LARS galaxies. Using data from the GBT 100 m telescope, we have presented direct measurements of the neutral gas contents of 11 of the 14 galaxies; we place limits on the three most distant (z > 0.1) LARS galaxies. These profiles show a wide variety of structure. Some systems harbor multi-component profiles that are indicative of interaction or the presence of nearby companions; others show more simple single H i peaks. Two sources (LARS 06 and LARS 11) are likely confused with nearby galaxies. We fit each of these profiles in order to derive global line properties, including systemic velocity, linewidths, H i flux integrals, and H i masses.

For a subset of five of the LARS galaxies, we also present new, low-resolution H i spectral line imaging obtained with the VLA. These H i images allow us to localize the H i associated with each galaxy. Three of the systems (LARS 02, LARS 08, and LARS 09) are unresolved at this angular resolution, while LARS 04 is resolved by a few beams. LARS 03 is highly resolved by the present data; we have discovered an enormous tidal structure that contains more than 10⁹M_☉ of H i, and that extends more than 160 kpc from the main interacting galaxies in the LARS 03 system. Based on the recovered flux integrals from these VLA data, each of the five systems presented here would be amenable to higher spatial resolution observations. The VLA C and B configurations would offer a factor of roughly 4 and 10 improvement in resolution respectively, but at the cost of longer integration times.

In particular LARS 03 presents an interesting case where we have unequivocal evidence of kinematic disturbances on large scales, yet no accompanying boost relative to nondisturbed galaxies in Lyα escape fraction. A high-resolution follow up with the VLA and future radio telescopes, especially coupled with larger maps of the Lyα flux, would disentangle the local versus global kinematic effects, and probe the Lyα escape on physical scales larger than the ∼10 kpc already probed in the HST imaging.

Using these new GBT and VLA data, as well as the results derived from HST imaging and presented in Paper II, we compare various global H i and Lyα properties. While the sample is small, we find a few intriguing correlations in our robustly detected galaxies: The H i linewidths are strongly anti-correlated with ξ_Lyα (ρ_s = −0.81), while the total H i masses show anti-correlations with escape fraction and Lyα EW. Additionally, both the H i linewidths and the H i masses are each correlated with the H_α/H_β ratio. These intriguing but tentative (anti-)correlations are in general agreement with a significant dependence of Lyα propagation on the total mass and thus the H i mass of the source, and is in general agreement with previous work by Laursen et al. (2009) which found that the escape fraction decreases with increasing virial mass. These results also support the general trend that Lyα properties appear to correlate with total galaxy mass, as found in Paper II. That work showed that Lyα emission (quantified by Lyα equivalent width, escape fraction, etc.) is systematically larger in lower mass galaxies. Given the variety of properties that scale roughly with mass (e.g., metallicity, dust content, star formation rate, UV colors, etc.), a larger statistical sample of galaxies will be needed in order to better quantify these trends.

The famous metal-poor blue compact dwarf galaxy I Zw18 reminds us that the trend of lower mass galaxies having larger Lyα luminosities or escape fractions is not always true. While the system has an extremely low nebular oxygen abundance (Skillman & Kennicutt 1993), an appreciable star formation rate (Cannon et al. 2002), and a substantial UV luminosity (Grimes et al. 2009), Kunth et al. (1998) showed that I Zw 18 is in fact a Lyα absorber. This process was explained by numerous scattering events in the neutral, and, most importantly, static ISM even in the complete absence of dust (Atek et al. 2009). Future local Lyα studies should seek to expand the number of low-mass systems with analyses similar to the ones presented here for LARS.

Although it is tempting to view the increasing Lyα/Hα with decreasing mass as a result of less gas mass being available for scattering, this theory must be cast in the light of results that suggest that gas fraction increases with increasing redshift (e.g., Tacconi et al. 2013), and that lower mass galaxies tend to be more gas-rich (Saintonge et al. 2011) and less dusty (e.g., Blanc et al. 2011). In Figure 12 we see that the LARS galaxies fit the general trend with gas mass fraction.

While Lyα halos seem to appear in systems with a range of H i linewidths, large Lyα/H_α values only appear in galaxies with H i smaller linewidths. Large Lyα halos are also correlated with less massive galaxies, and with increasing velocity offsets. We interpret this to mean that low H i linewidths are necessary but not sufficient for Lyα escape, while larger linewidths contribute to the overall effects of Lyα destruction by dust, extinction, and scattering.

The bulk of the evidence presented by previous Lyα work shows the importance of H i properties, especially kinematic details, in determining the escape of photons (e.g., Mas-Hesse et al. 2003). This process is mostly governed by local effects, and it is unclear at this time what global H i properties are required for the escape and propagation of Lyα photons. The asymmetric GBT profiles and evidence of Lyα emission in the presence of large column densities of H i both point to interesting subresolution kinematic and density effects.

A larger sample of galaxies with data like those presented here will allow us to place these correlations on a more statistically robust footing. The HST eLARS program has been approved for 54 orbits and contains 28 more galaxies within our nominal sensitivity range of z∼ 0.08. In addition, higher resolution follow up observations will allow us to determine kinematic details on more applicable scales for a number of our galaxies. This program will likely be the largest statistical sample of both H i and Lyα detected galaxies until new advances of radio telescopes, such as the Square Kilometer Array,¹⁹ push H i detections to 20–100 kpc scales at z >1.

The authors thank the anonymous referee for insightful comments that improved the quality of this manuscript. J.M.C. and S.P. thank Sung Kyu Kim and Macalester College for research support. J.M.C. thanks the Instituto Nazionale di Astrofisica and the Osservatorio Astronomico di Padova for their hospitality during a productive sabbatical leave. G.Ö. is a Royal Swedish Academy of Science research fellow (supported by a grant from the Knut & Alice Wallenbergs foundation). M.H. acknowledges the support of the Swedish Research Council (Vetenskapsrådet) and the Swedish National Space Board (SBSB). This work was supported by the Swedish Research Council and the Swedish National Space Board. H.O.F. is funded by a postdoctoral UNAM grant. J.M.M-H is funded by Spanish MINECO grants AYA2010-21887-C04-02 and AYA2012-39362-C02-01. I.O. was supported by the Sciex fellowship of the Rectors' Conference of Swiss Universities. P.L. acknowledges support from the ERC-StG grant EGGS-278202. D.K is supported by CNES (Centre National d'études spatiales) for the HST project.

This investigation has made use of the NASA/IPAC Extragalactic Database (NED), which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration, and NASA's Astrophysics Data System. This investigation has made use of data from the Sloan Digitized Sky Survey (SDSS). Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, the Max Planck Society, and the Higher Education Funding Council for England. The SDSS Web site is http://www.sdss.org/.

The SDSS is managed by the Astrophysical Research Consortium for the Participating Institutions. The Participating Institutions are the American Museum of Natural History, Astrophysical Institute Potsdam, University of Basel, University of Cambridge, Case Western Reserve University, University of Chicago, Drexel University, Fermilab, the Institute for Advanced Study, the Japan Participation Group, Johns Hopkins University, the Joint Institute for Nuclear Astrophysics, the Kavli Institute for Particle Astrophysics and Cosmology, the Korean Scientist Group, the Chinese Academy of Sciences (LAMOST), Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New Mexico State University, Ohio State University, University of Pittsburgh, University of Portsmouth, Princeton University, the United States Naval Observatory, and the University of Washington.