SMALL-SCALE PROPERTIES OF ATOMIC GAS IN EXTENDED DISKS OF GALAXIES

We carried out follow-up VLA observations of three galaxies with radio-bright background sources. Properties of the target galaxies such as their redshifts, stellar and H i masses, central metallicities, H i velocity centroids, and widths are provided in Table 1. The GBT spectrum of the three galaxies obtained by B11 shows strong H i emission. The measured H i masses indicate that these galaxies have more H i than stars. Based on the H i mass to disk size relationship derived by Swaters et al. (2002), these galaxies are expected to have extended H i disks well beyond their optical disks.

These three galaxies are probed by six sightlines corresponding to six radio-bright background sources. Information on each of the sightlines including their IDs from the study by B11, position of the background radio sources, their 20 cm fluxes, and impact parameters is provided in Table 2. Three of the sightlines are QSOs with optical spectroscopic data from the Sloan Digital Sky Survey (SDSS). The non-detection of any Hα in emission at the redshift of the foreground galaxy in the QSO spectra provides us with an upper limit on the star formation rate (SFR) surface density (Σ_SFR).

Since all the galaxies were found to have strong 21 cm H i emission in the GBT spectra, it is impossible to rule out absorption being filled in by emission from a different spatial location. Hence, we carried out high spatial resolution VLA observations to confirm the non-detections at the location of the background radio sources. The VLA in B-configuration is most appropriate for such a study as the size of the synthesized beam (∼8'' × 6'') matches the extent of the background continuum sources and is hence most conducive to detecting absorption features. For example, the spatial resolution achieved in our imaging (Column 6 of Table 3) is better than that prescribed by Dickey et al. (2000, resolution of <10'') for extragalactic sources.

We obtained UV spectra for two of the QSO sightlines from our original sample that had confirmed 21 cm absorbers. The purpose of these observations was to detect and measure the Lyα (λ1215.670) transition associated with the 21 cm absorbers and hence provide a measure of the H i column density independent of any assumption regarding the spin temperature of the gas. The combination of H i absorption properties from the 21 cm hyperfine transition and the Lyα transition can be used to measure the column density and spin temperature of the absorber simultaneously. The QSOs were observed with the Cosmic Origins Spectrograph (COS) aboard the HST under programs 12467 and 12486. One of the sightlines, SDSS J122115.22+454843.2, is the optical/UV counterpart of the radio sightline #2 (see Table 1). The other sightline is toward SDSS J104257.58+074850.5 probing galaxy GQ1042+0747 at an impact parameter of 1.7 kpc (Borthakur et al. 2010). The setups for the HST observations are presented in Table 4.

Unfortunately, the COS spectra for both the targets had insufficient flux at the rest-frame wavelength of Lyα for the targeted galaxies. As a result no measurements could be made. The drop in UV flux of the QSOs is consistent with the presence of Lyman limit systems at higher redshift toward these sightlines that have consumed most of the flux at shorter wavelengths. Therefore, for the remainder of the paper, we concentrate on the analysis and results from the 21 cm follow-up observations.

We carried out a total of 18 hr of observations with the VLA in B-configuration under program 10C-120. We observed the field toward SDSS J122115.22+454843.2 (UCG 7408 hereafter) for 2 hr and the fields toward SDSS J141629.25+372120.4 (J141629 hereafter) and SDSS J160659.13+271642.6 (J160659 hereafter) for 8 hr each. The observations were carried out in dual polarization mode with a total bandwidth of 1, 2, and 4 MHz, respectively. This setup was chosen to yield velocity coverage of 211, 436, and 883 km s⁻¹, respectively. The VLA correlator, WIDAR, was set up to deliver 256 spectral channels for each of the used bandwidths, resulting in a velocity width of 0.8, 1.7, and 3.5 km s⁻¹ per channel, respectively. WIDAR being an XF correlator,⁸ the frequency resolution of the raw (unbinned and unsmoothed) spectra is 1.2 times the channel width (Rots 1982).

The data were reduced and calibrated following the standard VLA calibration and imaging procedures using the Common Astronomy Software Applications (CASA) package. The data were imaged using a "natural" weighting scheme for spectral line images. The natural weighting was preferred as it offered the highest possible signal-to-noise ratio in the images. Details of the observations and data products are provided in Table 3.

We also present existing VLA D-configuration data that were obtained under legacy ID AY190 and were first published by B11. The data were obtained in 2008 in dual polarization mode with a bandwidth of 1.5 MHz and 128 spectral channels using the old VLA correlator. These correspond to a velocity coverage of 330 km s⁻¹ and a channel width of 2.6 km s⁻¹, respectively. The data were reduced and calibrated following the standard calibration and imaging procedures using NRAO's Astronomical Image Processing Software. The spatial resolution achieved with uniform weighting of the data is 5217×3811, which is not enough to resolve the background radio source into three individual sources. And hence, detailed spatial information on the associated H i absorption could not be made with these data, thus requiring the need for higher spatial and spectral resolution B-configuration data.

We present the H i map from the channel corresponding to the velocity of the absorption feature (444.5 km s⁻¹) seen toward sightline #2 in Figure 6. The background radio source is shown in black contours. The map is color coded to show absorption in blue, noise (±1σ) in green, and emission in yellow/red. However, unlike emission, H i absorption can only be measured in the region covered by the background radio source.

Zoom In Zoom Out Reset image size

Absorption associated with sightline #2 can be seen in blue in Figure 6, covering the northeast quadrant of the background source. The physical size (area) of the H i cloud responsible for the absorption is ∼0.04 kpc². The implied covering fraction, f, of the background source by the cloud is ≈25%–30%.

A spectrum extracted from the position of the absorber (identified as a white oval of same size as the beam) is shown on the right. We measured the peak depth in flux density for the feature to be 10.41 mJy. For comparison, we also present in Figure 6 the D-configuration absorption spectrum that was obtained by subtracting the H i emission model from the spectrum in Figure 5. The region of the galaxy from which the D-configuration spectrum was extracted is shown as the white dashed rectangle on the image in the left panel of Figure 6. As noted earlier, the absorption feature seen in the B-configuration data is deeper and narrower than that seen in the VLA D-configuration data. The B-configuration data have higher spectral and spatial resolution; however, they have lower signal-to-noise ratio than the D-configuration data. Nevertheless, the feature was confirmed at >5σ at that channel or at >3σ_average, the average noise over all channel.

From our B-configuration data, we estimated the peak optical depth of this absorber to 1.37 using the following relation:

$$\begin{equation} \tau = - ln\left(\frac{I_o-I_{{\rm abs}}}{I_o}\right), \end{equation} \tag{ 1 }$$

where I_o is the flux density of the background quasar at the position of the absorber of 13.98 mJy and I_abs is the peak strength of the absorption feature of 10.41 mJy (see Figure 6, left panel). If we assume the width of the line is due to collisional excitation of the hydrogen atoms, we can relate the kinetic temperature of the gas to the width of the line as

$$\begin{equation} T_{k}\le 21.855 (\Delta v)^{2}, \end{equation} \tag{ 2 }$$

where Δv is the FWHM of the line in km s⁻¹. In this case, Δv =1.1 km s⁻¹ and implies a kinetic temperature, T_k ⩽ 26.4 K. This indicates that the temperature of this absorber is similar to that observed in cold neutral medium (CNM) of the LMC and SMC (Dickey et al. 1994, 2000).

Assuming the kinetic temperature as a proxy for spin temperature (i.e., T_k = T_s = 26.4 K), we estimate the column density of H i in this absorber to be 6.3 × 10¹⁹ cm⁻² using the following expression:

$$\begin{equation} N({\rm H\,\scriptsize{I}}) = 1.823 \times 10^{18} \frac{T_{s}}{f}\int \tau (v) dv\, {\rm cm}^{-2}, \end{equation} \tag{ 3 }$$

where τ(v) is the 21 cm optical depth as a function of velocity in km s⁻¹. From the Gaussian fit to the absorber, we estimate ∫τ(v)dv = 1.3. In this case, the covering fraction is unity, f = 1, as we are measuring optical depth, τ(v), at the region of the background source where the absorber was detected.

The column density of cold gas as derived from the absorption feature is significantly lower than the total H i column density seen in emission. By comparing the two, we find the ratio of cold-to-total H i column density associated with the absorber to be ≈30%. Folding in the covering fraction, we find that only ≈10% of the total H i by mass exists in the cold phase. The non-detections of H i in absorption toward sightlines #1 and #3 provide upper limits on the column densities of 2.9 and 14.2 × 10¹⁹ cm⁻², respectively. The same exercise of comparing these column densities to that observed in emission implies that the cold-to-total H i toward sightlines #1 and #3 are ⪅20% and ⪅50%, respectively.

Our results suggest that the process of condensation is suppressed in UGC 7408. We found that a large fraction of atomic gas in the extended disk of this galaxy is warm with a temperature much larger than a few 100 K. Therefore, we conclude that most of the atomic gas in the extended disk of UCG 7408 failed to condense into the atomic gas clouds of 100–50 K (typically found in the Milky Way ISM). We did detect a small fraction of gas in the cold phase with temperature ∼25 K. This is much lower than the temperatures commonly seen in the CNM of the Milky Way (Heiles & Troland 2003; Strasser et al. 2007). In fact, at such low temperatures most of the gas in the Milky Way is in the molecular phase. The existence of such cold atomic gas suggests that the fraction of H i that was able to condense into the cold phase remains in the atomic state and avoids the transition to molecular phase.

The atomic hydrogen in the SMC also exhibits very similar properties. For instance, Dickey et al. (2000) found that less than 15% of the total H i in the SMC is in the cold phase. They also found the temperature of cold phase to be typically 40 K or less. They suggested that the difference in the properties of the CNM between the SMC and the Milky Way is a consequence of the difference in the metallicity in their ISM. Low metallicity implies lower radiative cooling. In particular for clouds where cooling is dominated by fine-structure line emission, the thermal equilibrium between the warm and the cold phase requires higher pressures with decreasing metal abundance (Wolfire et al. 1995). This limits the existence of cold clouds to regions of higher pressure, and hence the reduction in the cold gas fraction. Similar effects are observed in DLAs where the spin temperatures are inversely proportional to the metallicity of the DLA (Kanekar et al. 2009). Again, this emphasizes the inefficient cooling of gas at lower metallicities.

On the other hand, the lower metallicity would imply lower dust content and therefore, the fraction of atomic gas that is able to cool does not suffer from photoelectric heating by dust grains. However, the lack of dust grains is likely to impede the production of molecules. Dust plays a crucial role in shielding gas clouds against ionization thus aiding in atomic-to-molecular transition. Therefore, a drop in metallicity/dust content requires a much larger column of neutral gas to shield the clouds and form molecules (see Figure 1 in Krumholz et al. 2009). Observationally, this is evident from the variation in the minimum neutral gas column densities required to detect molecules in galaxies as a function of metallicity. For example, in the Milky Way the transitions occurs when the neutral gas column density is 10^20.7 cm⁻² or higher (Savage et al. 1977). The same for LMC and SMC was found to be at N(H i) >10^21.3 cm⁻² and >10^22.0 cm⁻², respectively (Tumlinson et al. 2002).

Therefore, it is likely that UCG 7408 has a very low molecular gas fraction as compared to its neutral gas content. This would explain the small physical extent of the stellar component in this galaxy unlike its extended H i distribution as well as the suppression in star formation (as indicated by the limiting SFR surface density in Table 2) in the vicinity of the absorber despite its low temperature.