A GLIMPSE AT QUASAR HOST GALAXY FAR-UV EMISSION USING DAMPED Lyα's AS NATURAL CORONAGRAPHS

DLAs are characterized by a wide flat trough with zero transmitted flux (i.e., dark trough or dark core), resulting from the natural broadening of the absorption at high H i column densities. In this study, we define the DLA dark trough as a region where the flux is negligibly small; specifically, regions where the ratio of the expected flux to the continuum flux is <1 × 10⁻⁴. Intervening DLAs can act as natural coronagraphs, because the quasar continuum, emitted from the centrally bright nucleus, is completely absorbed within the DLA dark trough. However, it is quite likely that regions of the extended emission from the quasar hosts are unobstructed by the intervening clumpy H i absorbers, and the extended stellar light thus leaks through and can be detected within the dark trough of the foreground DLAs.

Our sample is selected from the BOSS (Dawson et al. 2013; Ahn et al. 2014). BOSS is a spectroscopic survey that will measure redshifts of 1.5 million luminous red galaxies and Lyα absorption toward 160,000 high redshift quasars (Eisenstein et al. 2011). One thousand optical fibers with 2'' diameters are plugged into the 1000 holes in an aluminum plate to receive light from 160–200 quasars, 560–630 galaxies, and 100 ancillary science targets and standard stars (e.g., Dawson et al. 2013). Also, each plate contains at least 80 sky bers that are placed on the blank area of the sky to model the background for all the science fibers. The distribution of the sky fibers is constrained to cover the entire focal plane. This design allows the sampling of the varying sky background over the focal plane of the spectrograph (e.g., Dawson et al. 2013). The BOSS spectra have a moderate resolution of R ∼ 2000 from 3700 Å to 10,400 Å. With a typical exposure time of 1 hr for each plate, the BOSS spectra have a modest median signal-to-noise ratio (S/N) of ∼2–3 per resolution element for quasars at a rest-frame wavelength of λ = 1041–1185 Å (Lee et al. 2012).

The Data Release 10 Quasar (DR10Q) catalog (Pâris et al. 2014) includes 166,798 quasars detected over 6000 deg², of which 117,774 are at z > 2.15. A combination of target selection methods (Ross et al. 2012; Bovy et al. 2011) achieved this high surface density of the high-redshift quasars. Using a fully automated procedure based on profile recognition using correlation analysis, Noterdaeme et al. (2012) constructed a catalog of DLAs and sub-DLAs with N_H i > 10^20.0 cm⁻² detected from an automatic search along DR9Q lines-of-sight. Using a similar technique, Noterdaeme et al. (2014) generated the DLA catalog for Data Release 10 (DR10).

The DLA detection technique (Noterdaeme et al. 2012, 2009) pays special attention to the low-ionization metal lines in order to remove possible blends or misidentifications. The redshift measurement is mainly based on the profile fit to the DLA. Some of the DLA redshift measurements have been refined based on the low-ionization metal lines. The equivalent width of the metal lines is obtained automatically by locally normalizing the continuum around each line of interest and then modeling the absorption lines with a Voigt profile. For non-detections, upper limits are obtained from the noise around the expected line positions.

The final BOSS DR10 DLA catalog includes 11,030 DLAs with N_H i > 10^20.3 cm⁻², continuum-to-noise ratio (CNR) >2, and absorption redshift z_abs > 2.15. We avoided proximate DLAs at velocities less than 5000 km s⁻¹ from the quasar as these have a high probability of being physically associated with the local quasar environment. Strong proximate DLAs from BOSS are studied in Finley et al. (2013).

In our investigation, the DLA purity is crucial. Guided by this consideration, we include only DLAs satisfying the following stringent criteria: (1) spectra with median CNR >4.0; (2) >3σ low-ionization metal line detections; (3) column densities of N_H i > 10^20.6 cm⁻², which requires the width of the DLA trough to be greater than four times the spectral resolution; and (4) Spearman's correlation coefficient of the Voigt profile fitting that is greater than 0.6, which ensures an accurate estimate of the DLA column density (see details in Noterdaeme et al. 2009). Criterion (2) ensures that it is a DLA that is responsible for absorption rather than an unresolved blend of multiple narrow absorption features, while criteria (1), (3), and (4) enable an accurate determination of the column density and redshift. ∼2300 out of 11,030 DLAs in the catalog satisfy these four criteria. Furthermore, we visually inspect the 2300 systems to confirm that they are DLAs. We eliminate ∼40 systems close to the quasar O vi+Lyβ region. Also, visual inspection enables us to find a small number of DLAs with inaccurate redshift measurements, and we pay careful attention in order to refine the redshifts for some of the DLAs according to Voigt profile fitting and metal lines if necessary. We also drop a number of DLAs with Lyman limit systems blending or strong Lyα forest absorption in both DLA wings, which make the measurements of the DLA column density and redshift difficult. A sample of 2138 DLAs pass the visual inspection. In addition, we constrain the redshift of the DLAs to the moderate redshift range of 2.2 < z < 3.2. A sample of 1940 DLAs within the redshift range are finally selected and defined. In the remainder of the paper, we refer to these 1940 DLAs as the clean DLA sample (CDLA sample). In Figure 1, the right panel shows the redshift distribution of these 1940 DLAs with different quasar luminosity bins. The orange histogram represents the highest luminosity bins, while the red histogram represents the lowest luminosity bin. The left panel shows the redshift distribution for the corresponding background quasars. The redshift distribution of DLAs in different luminosity bins is similar within the redshift range of 2.2 < z < 3.2.

Zoom In Zoom Out Reset image size

DLAs with N_H i > 10^20.6 cm⁻² have a dark trough with a rest-frame velocity width extending at least ±340 km s⁻¹ from the center, much larger than the DLA redshift uncertainties. In addition to refining the DLA redshifts from Voigt profile fitting on the Lyα part, we use multiple low-ionization metal lines to fit the redshifts (z_ion) of ∼500 DLAs. We find that the largest difference between the redshift fitted by low-ionization metal lines and the redshift fitted by the Lyα Voigt profile (c × Δz/(1 + z)) is ∼150 km s⁻¹. We also checked the DLA trough of the composite DLA spectra (see Sections 3.1 and 3.3, and also the inset in Figure 3), and the central ±250 km s⁻¹ is consistent with a flat absorption trough within 1σ. All of this evidence supports that the DLA redshift uncertainties are much smaller than the width of DLA dark trough; and we conservatively define the dark trough in the composite spectra as the region within ±150 km s⁻¹ about the DLA center.

The Sloan Digital Sky Survey (SDSS) spectra are shifted to the rest frame of the DLA while conserving the flux, i.e., f_i(λ_res) = (1 + z_DLA) × f_{i, sdss}(λ_obs), where λ_res is the DLA rest-frame wavelength, λ_obs is the wavelength in the SDSS spectra (λ_obs = λ_res × (1 + z_DLA)), f_i(λ) is the flux density of an individual spectrum redshifted to the rest frame, and f_{i, sdss}(λ) is the flux density of an individual SDSS spectrum in the observed frame. In the dark trough, the flux density in each pixel (wavelength) of the composite spectra is calculated by two methods: (1) the 3σ-clipped mean of all the spectra at the same wavelength and (2) the median of all the spectra at the same wavelength. Outside the dark trough with λ < 1213 Å or λ > 1218 Å, we take the median value to stack the spectra. We calculate the error at each pixel (wavelength) in the stacked spectrum by propagating the errors of the pixels at the same position in every individual spectrum $(\sigma _{\rm {stacked, i}}= {1}/{n}\times \sqrt{{{{\sum _{i=0}^n \sigma ^2_{i}}}}})$. Note that 3σ-clipping mainly removes the large outliers due to noise, without clipping away the Lyα emission from DLA galaxy. On the one hand, even assuming that most of DLA host galaxies are as bright as a L* galaxy at z = 2–3 (e.g., Shapley et al. 2003), the Lyα emission from DLA host is still submerged within the noise of the individual SDSS-III spectrum. The detailed discussion is included in Section 4.1. On the other hand, from visual inspection of the entire CDLA sample, we do not find DLAs with strong Lyα emission in the dark trough.

In the stacking process, we do not introduce any additional scaling. It is true that different quasars are different in luminosity, quasar host flux, and DLA host emission. However, in the DLA dark trough, the quasar continuum is completely blocked in the DLA dark trough. Also, for SDSS-III spectra, the average flux densities of the DLA hosts and quasar hosts are generally about or more than one order of magnitude lower than the noise level (see more details in Sections 4.1 and 4.2). For the SDSS-III spectrum, the noise is mainly due to the sky background and the CCD read noise (e.g., Dawson et al. 2013), and in the dark trough of the DLAs at z = 2.6 ± 0.3, different systems generally have similar noise level. Therefore, following previous similar studies, we simply stacked the spectra without any additional scaling (e.g., Rahmani et al. 2010; Noterdaeme et al. 2014).

After the initial stacking, we find that the dark trough of the stacked DLA shows a positive offset. We calculate the mean flux density by averaging the flux density in the dark trough region. Following the discussions in Section 2.2, we conservatively define the stacked dark trough region within ±150 km s⁻¹ about the center. For the stacking of the 3σ-clipped mean, the average dark trough flux density is $\bar{F}_{\rm {dark}}=6.5 \,{\pm}\, 0.6 \times 10^{-19}$ erg cm⁻² A⁻¹ s⁻¹; and for the median stacking, the average dark trough flux density is $\bar{F}_{\rm {dark}}=7.0 \,{\pm}\, 0.6 \times 10^{-19}$ erg cm⁻² A⁻¹ s⁻¹. A general flux residual in the stacked DLA dark trough has been documented in previous work with much smaller sample size (Rahmani et al. 2010; Pâris et al. 2012). The large DLA database of SDSS-III/BOSS DR10 enables us, for the first time, to carefully study and test the origin of the positive residual flux in the dark core. This residual flux can arise from three sources: (1) sky subtraction residual (systematic sky subtraction errors), (2) Lyα emission around λ ∼ 1216 Å from the DLA galaxies, and (3) FUV continua emission at (1 + z_DLA)/(1 + z_QSO) × 1216 Å ∼1100 Å from the quasar host galaxies that is not blocked by the foreground DLAs.

In order to calculate the sky subtraction residual, we select a group of 150 DLAs at z > 3.8 with a median redshift 〈z〉 = 4.0, and examine the flux in the DLA Lyman limit region ranging from rest-frame λ = 800–900  Å, where the optical depth τ(λ) is expected to be over 1000. For the sample of z > 3.8 DLAs, the Lyman limit region at rest-frame λ = 820–900 Å corresponds to an observed wavelength from 4100 Å to 4500 Å, which covers the average wavelength of DLA trough in our CDLA sample. The expected flux of quasar continua should be negligibly small in the DLA Lyman limit region because of the large optical depth. The flux in this region can only be contributed by (1) sky-subtraction residuals, (2) unobscured escaping ionizing photons with λ ≲ 900 Å from DLA galaxies, and (3) photons from quasar hosts at even shorter wavelengths. Therefore, the average flux in DLA Lyman limit regions can be regarded as an upper limit for the sky-subtraction residual. A number of studies have found that the escape fraction is small, ≲5%–10% relative to photons escaping at 1500 Å (Malkan et al. 2003; Siana et al. 2007; Bridge et al. 2010). This small escape fraction suggests that the ionizing photons could only have a minor or even negligible contribution to the average flux in the DLA Lyman limit region.

The 150 DLAs at z > 3.8 we have selected from the DLA catalog all have C ii and/or Si ii detections and passed the visual inspection to ensure the DLA nature. We use the median to stack the spectra at rest-frame λ = 820 –900 Å in the stacked spectra. Then, we average the flux from λ = 820 Å–900 Å. The average flux density is equal to 3.502  ±  0.008 × 10⁻¹⁹ erg cm⁻² A⁻¹ s⁻¹, and this value can be regarded as an upper limit of the sky-subtraction (see red dot in Figure 2). We double checked our results using the average sky subtraction residual as a function of wavelength, which is determined by the SDSS pipeline group (Bolton et al. 2012; D. J. Schlegel et al. 2014, in preparation; D. J. Schlegel et al. 2014, private communication). The sky-subtraction residual is ∼3 × 10⁻¹⁹ erg cm⁻² A⁻¹ s⁻¹ at ∼4300 Å, consistent with the upper limit on the sky-subtraction residual we have derived. In Figure 2, the inset shows a zoom-in of the DLA absorption for the composite spectrum of the group of z ⩾ 3.8 DLAs. Also, in Figure 2, we present the median flux in the composite DLA dark trough (blue dot with error bar) of the CDLA sample is 6.5  ±  0.6 × 10⁻¹⁹ erg cm⁻² Å⁻¹ s⁻¹, which is significantly larger than the SDSS sky-subtraction residual (red dot in Figure 2). The observed wavelength of the blue dot in Figure 2 is determined by the center of the DLA trough at the median redshift of the CDLA sample.

Zoom In Zoom Out Reset image size

In Section 3.1, we initially stacked all the spectra of the CDLA sample, and find that the average dark trough flux density is $\bar{F}_{\rm {dark}}= 6.5 \,{\pm}\, 0.6 \times 10^{-19}$ erg cm⁻² A⁻¹ s⁻¹, significantly greater than the upper limit of the sky subtraction residual determined in Section 3.2 (Figure 2). This dark trough flux density indicates that in addition to the sky subtraction residual, flux must be contributed by other sources. In this section, we will examine and test the possible origin of the flux residual in the DLA dark trough.

In order to examine the origin of the flux residual in the DLA dark core, we will group the quasars according to their luminosities, and check if the residual intensities in the dark trough correlate with the quasar luminosities. The motivation for this experiment is that if the intensity in the stacked dark trough is mainly contributed by the Lyman continua of the foreground DLA, then this intensity should be uncorrelated with the background quasar luminosity, and we expect the three luminosity bins to yield similar intensities in the dark trough. We divide the quasar into three bins according to their luminosity bins: (1) the highest luminosity bin contains quasars with the luminosity density at 1450 Å (L₁₄₅₀) > 9 × 10⁴² erg s⁻¹ Å⁻¹; (2) middle luminosity bin with 5 × 10⁴² erg s⁻¹ Å⁻¹ < L₁₄₅₀ < 9 × 10⁴² erg s⁻¹ Å⁻¹; and (3) low luminosity bin with L₁₄₅₀ < 5.0 × 10⁴² erg s⁻¹ Å⁻¹. Each luminosity bin contains a similar number of quasars.

The detailed stacking procedure is as follows. We first calculate the quasar intensity for each quasar using the following expression:

$$\begin{equation} I_{i, \rm {dark}}(\lambda _{\rm {res}})= (f_{\rm {i,sdss}}(\lambda _{\rm {obs}})- f_{\rm {sky}}) \times 4\pi {D_L}^2(z_{\rm {QSO}})\times (1+z_{\rm {DLA}}), \end{equation} \tag{ 1 }$$

where λ_obs = λ_res × (1 + z_QSO), and f_{i, sdss} is the flux density for an individual spectrum, f_sky = 3.5 × 10⁻¹⁹ erg s⁻¹ cm⁻2, which is determined in Section 3.2 and further supported by SDSS calibration group (D. J. Schlegel et al. 2014, in preparation). Following the stacking method described in Section 3.1, we stack the quasar intensity I_{i, dark}(λ_res) for each luminosity bin by taking the 3σ-clipped mean and median value in the dark trough. Outside the dark trough, we just take the median value to stack the spectra.

Note that the 3σ-clipped mean is only a legitimate method when we stack the spectra within the dark trough. In the dark trough, the residual flux, if exists, is only in a very low level. It is necessary to first rule out the large outliers due to the noise before taking the average. However, outside the dark trough, the different quasar continua have large scatter in luminosity. The 3σ-clipping will reject a large number of bins in the continua simply because the quasars differ in luminosity. Therefore, the 3σ-clipped mean is no longer a legitimate way to stack the quasar continua outside the dark trough. Figure 3 further demonstrates this point. In Figure 3, one can see that in the DLA dark trough, the number of rejected pixels is small (only ∼0.5% of the sample size). However, for the quasar continua at rest-frame wavelength λ < 1214 Å or λ > 1218 Å, the number of the rejected pixels for each wavelength becomes significantly larger than 1% of the sample size. This is because outside the dark trough, the 3σ-clipping mean rejects bins in the continua because quasars differ in luminosity. We have checked that, for the stacked quasar bolometric luminosity, the median value is 10% smaller than the 3σ-clipped mean, 25% lower than the average without sigma-clipping. Outside the DLA dark trough, we take the median to stack the quasar continua. In the next Section (Section 4.2), we will be consistent by only using the median composite spectrum to compare the quasar bolometric luminosity with the dark trough intensities (see details in Section 4.2).

Zoom In Zoom Out Reset image size

We summarize our results in Table 1 and Figure 4. The upper panel in Figure 4 presents the composite DLA spectra in three quasar luminosity bins using the CDLA sample. The middle panel presents the 3σ-clipped mean of the DLA dark trough after removing the sky-subtraction residual (Equation (1)). The stacked dark trough, which lies between the two vertical dashed lines, is defined within ±150 km s⁻¹ from the center (see Section 2.2). The lowest panel presents the median value of the flux residual in the DLA dark trough. The zoom-in insets in the middle and lowest panel presents the 3σ-clipped mean and median value of the overall CDLA sample, centered on the DLA absorption to show the non-detection of Lyα emission from the DLA hosts. The middle and lower panels demonstrate that the dark trough flux density is non-zero for the two higher luminosity bins, and moreover, the dark core intensity increases for the higher quasar luminosity bin. We have compared the results derived from 3σ-clipped mean and median value. We found the residual intensities derived using 3σ-clipped mean are consistent with the values obtained from the median stacking (Table 1).

Zoom In Zoom Out Reset image size

Figure 5 shows that the correlation between the dark trough intensity and quasar luminosity still holds for DLAs with the largest dark troughs in the CDLA sample. DLAs with N_H i > 10^20.9 cm⁻² have an expected dark trough width within ±580 km s⁻¹ about the center. Again, we conservatively define the dark trough region to be ±250 km s⁻¹ about the center. The upper panel of Figure 5 shows the composite DLA (N_H i > 10^20.9 cm⁻²) spectra for the same three luminosity bins with that in Figure 4. The middle panel shows the 3σ-clipped mean of the expanded DLA dark core region after subtracting the sky-calibration residual, and the lowest panel shows the median value for the same dark trough region. The insets in the middle and lower panel present the composite spectra of the N_H i > 10^20.9 cm⁻² CDLA sample, centered on the composite DLA absorption to show the non-detection of the Lyα emission from the DLA host galaxies. The sample size of the N_H i > 10^20.9 cm⁻² CDLA sub-sample is only half of the overall CDLA sample, yet it is evident that the higher quasar luminosity bin still corresponds to a higher dark trough intensity, i.e., the dark trough intensity with highest quasar luminosity (yellow in Figure 5) is 3σ higher than that with the lowest quasar luminosity (red in Figure 5).

Zoom In Zoom Out Reset image size

The relation between the quasar luminosities and the intensities in the stacked dark trough is summarized in Figure 6. The CDLA sample with N_H i>10^20.6 cm⁻² is shown in red, and the luminosity distribution is shown in the lower panel in red. A sub-group of DLAs in the CDLA sample with multiple metal lines are shown in black, where the redshifts are obtained from the low ionization lines. The blue points represent DLAs in the CDLA sample with N_H i > 10^20.9 cm⁻². The comparison between N_H i > 10^20.6 cm⁻² and N_H i > 10^20.9 cm⁻² DLAs demonstrates that the dark core intensities are not correlated with the DLA column densities. We also compare DLAs in two bins of equivalent width, which are denoted by triangles in red and yellow in Figure 6. We find that the DLA dark core intensities do not correlate with the DLA metal line equivalent width. Figure 6 clearly suggests a >3σ detection of the correlation between the mean DLA dark core intensity and quasar luminosity. The histogram in the lower panel shows the distribution of the quasar luminosities for CDLA samples.

Zoom In Zoom Out Reset image size

More than hundreds of DLAs at z > 2 have been identified over the last 20 yr. However, only a handful of DLAs have been confirmed to have Lyα emission originated from the DLA host galaxies. Here, we first give a brief summary of such studies. Møller & Warren (1993) confirmed an Lyα emission line associated with a DLA with column density N_col = 10²¹ cm⁻² at z = 2.8. The luminosity of Lyα emission is about 2 × 10⁴² erg s⁻¹ (∼0.4 × L* at z = 3; Ciardullo et al. 2012). The velocity offset of the Lyα emission is about +50 ± 100 km s⁻¹. Djorgovski et al. (1996) confirmed an Lyα emission associated with a sub-DLA of column density N_col = 10^20.0 cm⁻² at z = 3.1. The luminosity of the Lyα emission is 5 ×10⁴² erg s⁻¹ (∼L*), and the velocity offset is about +200 km s⁻¹. Leibundgut & Robertson (1999) confirmed an Lyα emission from galaxy associated with a DLA with N_col = 10^20.85 cm⁻² at z = 3.1. The Lyα luminosity is about L*, and the velocity offset is about 490 km s⁻¹. Møller et al. (2002) presented a gamma-ray burst DLA with a ∼0.5 L* Lyα emission of DLA galaxy at z = 2.3, and the Lyα emission is redshifted by 530 km s⁻¹. Møller et al. (2004) detect a ∼0.5 L* Lyα emission in the dark trough of a DLA with large column density N_col ∼ 10^21.8 cm⁻² at z = 2.0. The velocity offset of this Lyα emission is about +10 ± 150 km s⁻¹. Kulkarni et al. (2012) presents a super-DLA with column density of 10^22.0 cm⁻² associated with a L* Lyα emission at z = 2.2. In the high S/N spectrum, the Lyα emission is consistent with a velocity offset of +300 km s⁻¹. By stacking a sample of 99 DLAs with N_H i > 10^21.7 cm⁻², Noterdaeme et al. (2014) present the detection of an Lyα emission line with a luminosity of 0.65 × 10⁴² erg s⁻¹ in the composite DLA trough, corresponding to 0.1 × L* galaxies at z = 2–3, and this composite Lyα emission has a velocity offset of 130 km s⁻¹.

After stacking the spectra using the DLAs in the CDLA sample, we did not detect Lyα emission in the dark trough of the composite spectra (e.g., see the insets in the middle and lower panels of Figures 4 and 5).

In Figure 7, we present the composite spectra of 2000 simulated DLAs with Lyα emission being added, and the stacking method is 3σ-clipped mean. The simulated DLAs have same column density distribution with that of our CDLA sample. The noise has been included according to the realistic BOSS spectra. The Lyα emission is assumed to have the same luminosity, FWHM (750 km s⁻¹), and velocity offset (445 km s⁻¹) with the composite spectrum of L* Lyman break galaxies (LBGs) at z = 2–3 (Shapley et al. 2003). Brown line shows the 3σ-clipped mean of simulated DLAs by assuming that 100% DLA host galaxies have Lyα emission similar to that of L* LBGs, and all the Lyα emission entered the SDSS fibers. The blue line assumes that 50% of the simulated DLA host galaxies have L* Lyα emission. The red line assumes that 10% of the simulated DLA hosts have such strong Lyα emission. The yellow line indicates that the Lyα emission from the DLA hosts do not contribute to the residual flux in the DLA dark trough, either because the emission from DLA host galaxies is below the detection limit (see first probability in Section 4.1), or because the impact parameter between the DLA clouds and Lyα emitting region is bigger than 10 kpc (see third probability in Section 4.1). From the region between two vertical dashed lines, one can see that the strong Lyα emission will significantly affect the profile of the stacked DLA trough. In the region of the stacked dark trough, with the present of strong Lyα emission, the stacked flux will increase with the wavelength, yet our observed flux (black curve) does not. The observed results are generally consistent with the non-detection of the strong Lyα emission from DLA galaxies. Also, this figure demonstrates that 3σ-clipped stacking does not clip away Lyα emission from DLA host galaxies.

Zoom In Zoom Out Reset image size

Figure 8 presents a similar plot to Figure 7. In this figure, we assume that the Lyα emission from DLA host galaxies have the same luminosity as 0.1 × L* star-forming galaxies at z = 2–3 (Noterdaeme et al. 2014). The luminosity, FWHM (300 km s⁻¹), and velocity offset (130 km s⁻¹) of the Lyα emission are followed by that described in Noterdaeme et al. (2014). Brown presents the 3σ-clipped mean of simulated DLAs by assuming that 100% of the simulated DLA host galaxies have Lyα emission with a luminosity of 0.1 L*, and all the Lyα emission entered the SDSS fibers. The blue (red) line assumes that 50% (10%) of the simulated DLA host galaxies have 0.1 × L* Lyα emission entering the fibers. The yellow line indicates that the flux from the DLA hosts did not contribute to the residual flux in the DLA dark trough (see Section 4.1). From the observed spectra (black line), we did not detect the Lyα emission from DLA host galaxies. Our results are consistent with the above discussions: (1) DLAs are generally hosted by faint galaxies in relatively low-mass halos; and (2) for column density of N_col > 10^20.6 cm⁻², most of the DLAs have their impact parameters between the DLA clouds and Lyα emitting regions larger than 10 kpc.

Zoom In Zoom Out Reset image size

Figure 9 shows a similar plot to Figure 8. In this figure, we assume that the Lyα emission from DLA host galaxies have a typical stellar mass of 10^8.5 M_☉ (Møller et al. 2013). Assuming that the DLA hosts follow the galaxy main sequence, the typical luminosity of DLA host galaxies should be about 0.03  × L* galaxies at z = 2–3. In this figure, we further assume that the DLA host galaxies have a luminosity of 0.03  × L*. Brown presents the 3σ-clipped mean of simulated DLAs by assuming that 100% of the simulated DLA host galaxies have Lyα emission with a luminosity of 0.03 L*. The blue (red) line assumes that 50% (10%) of the simulated DLA host galaxies have 0.03  × L* Lyα emission. The yellow line indicates that the flux from the DLA hosts did not contribute to the residual flux in the DLA dark trough. From the figure, if only 10% of the Lyα emission similar to 0.03  × L*, the stacked Lyα emission will fall below our current detection limit. The black line shows that we did not detect the Lyα emission from DLA host galaxies, which is consistent with the red and yellow lines.

Zoom In Zoom Out Reset image size

We estimate the 3σ upper limit (F_3σ (Lyα)) by using the Gaussian fitting errors, assuming that the rest-frame FWHM of the expected Lyα emission is 400 km s⁻¹. The 3σ upper limit of the Lyα emission line flux is F_3σ(Lyα) = 7.5 × 10⁻¹⁹ erg s⁻¹ cm⁻². Considering the 2'' SDSS-III fiber diameter, this limit translates to a limiting 1σ surface brightness limit of 0.8  × 10⁻¹⁹ erg s⁻¹ cm⁻² arcsec⁻². This depth is about a factor of two deeper than Rahmani et al. (2010) and comparable to the depth of the long-slit spectroscopic observations of Rauch et al. (2008). At the mean redshift (〈z〉 = 2.6), the 3σ limit on the Lyα emission corresponds to an Lyα luminosity of 4.0  × 10⁴⁰ erg s⁻¹. This upper limit on the Lyα luminosity reaches a depth of ∼0.01 L*, where L* ∼ 4 × 10⁴² erg s⁻¹ at z = 3 (Cassata et al. 2011; Ciardullo et al. 2012). This corresponds to a 3σ upper limit on the Lyα-based SFR (SFR_Lyα) of 0.04 M_☉ yr⁻¹ (Kennicutt 1998; Dijkstra & Westra 2010).

The non-detection of the Lyα emission from the DLA galaxies in our CDLA sample can be explained by the following possibilities: (1) DLAs are mostly harbored by faint (sub-L*) galaxies in relatively low-mass dark matter halos; (2) A large fraction of DLA clouds have impact parameters much larger than the stellar half light radius of the DLA host galaxy and exceed the 1'' fiber radius, corresponding to ∼6 kpc at 〈z〉 = 2.65, such that a large number of the fibers are not exposed to the DLA galaxy stellar light; or (3) systemic offsets and redshift uncertainties of the Lyα emission lines in the DLA galaxies.

The first possibility is supported by an analysis of the mass–metallicity relation in a sample of 110 DLAs from z = 0.1–5.1 conducted by Møller et al. (2013). They suggest that the typical DLA has a stellar mass of log(M_*/M_☉) ∼8.5, in agreement with earlier results that most massive DLA galaxies only correspond to the least massive LBGs (Fynbo et al. 1999). Also, the low-mass DLA halos have been suggested by the simulations (e.g., Mo et al. 1999; Haehnelt et al. 2000; Nagamine et al. 2004). However, a few works favor more massive DLA dark matter halos on the order of 10¹² h ⁻¹ M_☉ (e.g., Font-Ribera et al. 2012).

The second possibility is supported by several previous studies. The fibers may not be exposed to the DLA galaxy light due to the large impact parameter of the DLA clouds with respect to the center of the galaxy. Zwaan et al. (2005) present a comprehensive H i 21 cm absorption survey of nearby galaxies, which contains 40 H i absorption features with column densities N_H i > 10^20.0 cm⁻². Half of the DLAs were found to have an impact parameter b > 8 kpc from the center of the galaxy, indicating that the stellar emission from a large fraction of DLA-hosting galaixes would not enter the 2'' BOSS fiber, if the DLAs at z = 2–3 have the same impact parameter as that of the nearby galaxies. Krogager et al. (2012) studied all of the 10 known DLAs with the identified galaxy counterparts at z = 2–3, and found that DLAs with impact parameter ranges from 01 to 30. Pontzen et al. (2008) are able to reproduce several properties of DLAs in their simulations, and the impact parameters at z = 3 are expected to be b ≲ 30 kpc. Furthermore, Fynbo et al. (2008) supports that higher metallicity DLAs have larger disks. In our CDLA sample, we only selected DLAs with metal line detections, which biases our DLAs toward higher metallicity, which may have larger H i disks based on the simulations. It is true that diffuse halo Lyα emission (Steidel et al. 2011) may enter the fiber. However, the surface brightness of the diffuse Lyα halo emission significantly drops by a factor of >3 when the impact parameter is greater than 15 kpc.

The third possibility is also supported by numerous observations. Using a sample of 89 LBGs with 〈z〉 = 2.3  ±  0.3, Steidel et al. (2010) find that the Lyα lines have systemic velocity offsets of Δv_Lyα = 445  ±  27 km s⁻¹. Assuming that the Lyα emission of DLA galaxy hosts has the same systemic velocity offsets as LBGs, the Lyα emission would be located outside of the dark trough of DLAs in our CDLA sample. The three possibilities we have discussed likely work together to cause the non-detection of DLA Lyα emission.

By stacking a sample of 99 DLAs with N_H i > 10^21.7 cm⁻², Noterdaeme et al. (2014) present the detection of an Lyα emission line with a luminosity of 0.65 × 10⁴² erg s⁻¹ in the composite DLA trough, corresponding to 0.1 × L* galaxies at z = 2–3. The idea is that DLAs with large column densities are expected to have smaller galaxy impact parameters. This detection of Lyα emission from DLA galaxies does not contradict our result, because the very large DLAs with N_H i > 10^21.7 cm⁻² constitute only ∼2% of our CDLA sample. On the contrary, this further supports the conclusion that most DLAs with N_H i < 10^21.7 cm⁻² are generally not associated with close galaxy counterparts. Our non-detection of Lyα emission supports the hypothesis that the lower column density DLAs could have larger impact parameters (up to a few tens of kiloparsecs); thus a large number of DLA absorbers in our sample are located >10 kpc from the central galaxies.

Besides the contribution of Lyα emission from DLA galaxies, the dark core flux may also be contributed by the DLA Lyman continua at λ ∼ 1216 Å. However, if the dark core flux comes from the DLA Lyman continua, it is difficult to explain why the dark core flux correlates with quasar luminosity. Therefore, our results favor the interpretation that the flux residual in the dark trough is highly unlikely to be contributed by Lyman continua or Lyα emission from DLAs.

Theoretically, that the luminous quasar phase naturally coincides with intense star formation was been proposed early on, and has the natural interpretation that both processes rely on reservoirs of gas brought to the center by gas-rich mergers and disk instabilities (e.g., Sanders et al. 1988; Hopkins et al. 2006; Lutz et al. 2008). Several observational studies in the mid-infrared and submillimeter regime (e.g., Lutz et al. 2008; Serjeant & Hatziminaoglou 2009) confirmed that there exists a link between the quasar host SFR and the black hole accretion rate. Anglés-Alcázar et al. (2013) further propose that galaxy-scale torque-limited accretion (Hopkins & Quataert 2011) naturally and robustly yields black holes and galaxies evolving along the observed scaling relations, and that AGN feedback does not need to couple with galaxy-scale gas and regulate black hole growth. Therefore, FUV stellar light from the quasar hosts is the most plausible interpretation for the flux residual in the composite DLA dark trough.

It is possible that the DLA absorption, while blocking the quasar continuum region, only partially blocks the host galaxies. The SDSS fiber is exposed to the regions of the quasar host galaxy, which are not obscured by the intervening DLA and may contribute residual flux. This is consistent with models of clumpy H i distributions in DLA galaxies as suggested by recent observations (e.g., Kashikawa et al. 2014). Kanekar et al. (2009) used the Very Long Baseline Array to image 18 DLAs in redshifted H i 21 cm, and suggest that for quasar radio emission with source sizes >100pc, the median DLA H i covering factor for DLAs at z = 1.5–3.5 is f_med ∼ 0.6. If the 21 cm and Lyα (UV) absorption do arise in the same cloud complexes, this result indicates that the emission from quasar hosts is not entirely covered by the foreground DLAs. Recent simulations and observations also support that in the galactic and circumgalactic environment, the covering factor of optically thick systems is ≲ 40%–50% for ∼0.1 L* galaxies at z = 2–3 (Kanekar et al. 2009; Srianand et al. 2013). The half-light radius of quasar host galaxies at z = 2–4, typically 3–5 kpc (Peng et al. 2006), is fully covered by the SDSS fiber (∼12 kpc). The relation between the dark core residual flux and quasar luminosity supports a correlation between the SFR in the quasar hosts and the quasar luminosity.

After correcting for the sky-subtraction residual in the highest luminosity bins in the CDLA sample, the median intensity in the dark core is ∼16.4  ±  2.1 × 10⁴⁰ erg s⁻¹ Å⁻¹. We take this value as the average intensity of a quasar host galaxy at a rest-frame wavelength λ ∼ 1100 Å. This value corresponds to an intensity of ∼7.5  ±  1.0 × 10⁴⁰ erg s⁻¹ Å⁻¹ at λ ∼ 1700 Å for a galaxy with a constant star formation history and an age of 0.5 Gyr, after correcting for intergalactic medium absorption (Bruzual & Charlot 2003; Schaye et al. 2003). At the mean redshift 〈z〉 = 3.1, this corresponds to 0.3 L*, where L* is determined from >2000 spectroscopic LBGs from z = 1.9–3.4 (Reddy et al. 2008). The inferred SFR based on the UV continua (SFR_UV) is ∼9 M_☉ yr⁻¹ (Madau et al. 1998). Following the same procedure, the middle-luminosity bin corresponds to SFR_UV= 5 M_☉ yr⁻¹, and the non-detection in the lowest luminosity bin translates to a 3σ upper limit of SFR_UV = 7.5 M_☉ yr⁻¹. If we assume that DLAs have a H i covering fraction of ∼0.5 as discussed in the last paragraph, the SFRs_UV for each luminosity bin will be a factor of two times higher than our measured value. We further derive the bolometric luminosities (L_bol) for the three luminosity bins from the absolute magnitudes at 1450 Å, assuming a bolometric conversion factor ζ_1450A = L_bol/νL_ν,1450 A = 4.4 (Richards et al. 2006). We obtain L_bol = 25.0, 13.0, 7.0 × 10¹² L_☉, for our three quasar bins, respectively. The properties of these three sub-samples are summarized in Table 1 based on our measurements.

Aretxaga et al. (1998) report that the SFR of quasar host galaxies is >100–200 M_☉ yr⁻¹ from a sample of three quasars at z = 2 using R and I band measurements. Jahnke et al. (2004) estimated an SFR ∼2–30 M_☉ yr⁻¹ for z = 1.8–2.6 quasar host galaxies using HST ACS F606W and F850LP bands. Villforth et al. (2008) found a moderate SFR of ∼33 M_☉ yr⁻¹ for quasar host galaxies using multi-band data from U to K. Our SFR_UV values from partially obscured quasar hosts are generally consistent with these previous observations.

Using the long-slit spectroscopy, Zafar et al. (2011) reported the detection of the host galaxy in the dark trough of the DLA in front of quasar Q0151+048A at z = 1.9. The detected positive residual flux matches the reported brightness of the host galaxy well (Fynbo et al. 2000), and the measured counts in the dark trough is about a factor of 27 less than the quasar continuum in the B  band. In Figure 10, we compare our results with the detection of Zafar et al. (2011). The Lyα mean optical depth at z = 2.6 (<τ > (z = 2.6)) is 0.3, and <τ > (z = 1.9) =0.1. After correcting the mean Lyα absorption due to IGM, the result in Zafar et al. (2011) is a factor of two times higher than our results. Note that Q0151+48A is one order of magnitude higher than BOSS quasars. Also, the detected residual flux for Q0151+48A is the Lyman continuum at λ ∼ 1216 Å, while the residual flux for our BOSS quasar is the Lyman continuum at <λ > ∼1100 Å. Overall, the difference between our results and that of Zafar et al. (2011) can be interpreted well by the large scatter of the covering fraction among different DLA clouds, and plus the uncertainties of the dust extinction among different quasar hosts.

Zoom In Zoom Out Reset image size

Extended scattered nuclear light from the surrounding nebulosity could also be considered as a source for the emission, in addition to light from star formation (Young et al. 2009). The signature of scattered nuclear light is suggested by a few polarimetric measurements (e.g., Smith et al. 2004; Letawe et al. 2007; Borguet et al. 2008). From Table 1 and Figures 4 and 5, the residual luminosity we detected is about 1.5% of quasar luminosity for the highest quasar luminosity bin. Observations of type 2 quasars suggest that scattering efficiencies can be at or above the 1% level (Zakamska et al. 2005). With the assumption that any broad Balmer line emission that is visible in the off-nuclear spectrum arises from scattered quasar light, Miller & Sheinis (2003) estimated the scattered light emission from four quasar host galaxies using Keck long-slit spectroscopy, and found that the fraction of the scattered nuclear light is generally small compared with the quasar host galaxies' stellar light (Miller & Sheinis 2003). However, Young et al. (2009) also report that based on their theoretical models, scattered light should be a concern for host characterization in high-redshift observations. From our current observations, any quantitative estimate of the quasar scattered light is difficult, and thus, extended scattered nuclear light remains a possibility for some of the residual flux. Thus the SFR_UV we derived could be regarded as a stringent upper limit for the quasar host galaxies.