Free Access
Volume 547, November 2012
Article Number L1
Number of page(s) 4
Section Letters
Published online 22 October 2012

© ESO, 2012

1. Introduction

Studying the distribution of neutral gas in and around galaxies at different cosmological times provides a wealth of information about the formation and evolution of galaxies. The 21-cm hyperfine emission of atomic hydrogen has been used to trace the neutral gas in nearby galaxies and estimate their total H i mass. Given the sensitivity of current radio telescopes, this technique remains limited to z ≤ 0.2 (e.g. Lah et al. 2007). At high redshift, neutral gas is revealed by the damped Lyman-α absorption systems (DLAs) that it imprints in the optical spectra of bright background sources such as quasars. Because the detection of DLAs is only cross-section dependent, it is possible to statistically derive the amount of neutral gas and the corresponding column density distribution at different redshifts independently of the nature of the absorbers (see Wolfe et al. 2005).

The most recent contributions to the census of DLAs have used data mining of thousands of quasar spectra from the SDSS (SDSS, York et al. 2000) by Prochaska et al. (2005), Prochaska et al. (2009), and Noterdaeme et al. (2009, hereafter N09. These studies indicate that the N(H   i) distribution function (f(NHI), where χ is the absorption distance, see Lanzetta et al. 1991) steepens at log N(H   i) > 21 and that the cosmological density of neutral gas contained in DLAs () decreases significantly with time between z ~ 3.5 and z = 2.2.

Several explanations for the steepening of f(NHI) have been discussed in the literature, including conversion from atomic to molecular hydrogen (Schaye 2001; Zwaan & Prochaska 2006), small-scale turbulence, or stellar feedback (Erkal et al. 2012). Selection effects such as dust-reddening (e.g. Vladilo & Péroux 2005) could also alter the slope of f(NHI) in magnitude-limited surveys. However, the slope of the frequency distribution itself has not yet been well constrained at the high-column-density end due to fewer statistics. Similarly, the evolution of has long been discussed in the literature. Values at z ~ 1 (Rao et al. 2006) have been considered uncomfortably high when compared to that at z = 0 from Zwaan et al. (2005) and z ~ 2 from Prochaska et al. (2005). However, N09 corrected the value at z ~ 2 upwards. Since then, the value at z = 0 has also been corrected upwards by Braun (2012) and could indicate a flatter evolution over 0 < z < 2.

In this letter we present a search for DLAs in quasars observed in the course of the Baryonic Oscillation Spectroscopic Survey (BOSS, Dawson et al. 2012), one of the legacy surveys in the third stage of the SDSS (Eisenstein et al. 2011). We use the same formalism as described in N09 and adopt a ΛCDM cosmology with ΩΛ = 0.73, Ωm = 0.27, and H0 = 70 km   s-1 Mpc-1 (Komatsu et al. 2011).

2. Method

BOSS is a five-year programme using improved spectrographs (Smee et al. 2012) on the SDSS telescope (Gunn et al. 2006) to obtain spectra of 1.5 million galaxies and over 150 000 z > 2.15 quasars reaching up to 1 mag deeper than SDSS-II. The survey is mainly designed to measure the characteristic scale imprinted by baryon acoustic oscillations (BAOs) in the early Universe from the spatial distribution of luminous galaxies at z ~ 0.7 and the large-scale correlation of H i absorption lines in the intergalactic medium at z ~ 2.5 (Dawson et al. 2012). BOSS uses the same imaging data as in SDSS-I and II with an extension in the south galactic cap (see Aihara et al. 2011). The SDSS-DR9 (Ahn et al. 2012) makes publicly available the spectra of 87 822 quasars over an area of 3275 deg2, 65 205 having z > 2 (Pâris et al. 2012). The quasar target selection is described in Ross et al. (2012; see also Bovy et al. 2011).

2.1. Detection of DLAs

Intervening DLAs were automatically searched for in quasar spectra following the method described in N09. We briefly summarise here the main steps. For the purpose of collecting the largest number of DLAs1, we search the full line-of-sight to each quasar starting where the spectral signal-to-noise ratio per pixel reaches two (defining zmin) and up to the quasar redshift. We avoid sightlines with broad absorption lines of balnicity index BI  >  1000 km   s-1 (Pâris et al. 2012).

The quasar continuum is modelled over the Ly-α forest by fitting a modified power-law with a smoothly changing index plus Moffat profiles on top of the emission lines. Whenever the Ly-α emission line is severely absorbed (>30%), we use the predicted unabsorbed emission from principal component analysis (see Pâris et al. 2011) as a proxy for the true Ly-α emission before fitting the continuum. We then use the median continuum-to-noise ratio (CNR) as an estimate of the quality of the spectrum, independent of the presence of a DLA. Spectra with median CNR  <  2 over the Ly-α forest were not further considered.

Damped absorption lines are recognised through their characteristic shape by correlating the data against synthetic profiles of increasing column density. In short, (N(H   i),z) pairs with Spearman’s correlation above 0.5 (and significance >3σ) are recorded. To constrain the strength of the absorption, we also require that the absorbed flux is consistent with the presence of a DLA combined with possible Ly-α forest absorptions. The pairs are then grouped into individual DLA candidates (a gap of  >1000 km   s-1 indicates separate absorption systems), and the first guess for N(H   i) is taken from the pair with the highest correlation. The DLA redshift measurement is then improved whenever possible by cross-correlating the quasar spectrum on the red side of the Lyman-α emission line with a mask representing metal absorption lines. Finally, N(H   i) is obtained by fitting a Voigt profile to the damped Lyman-α line.

This approach provides us with an overall sample of 12 081 DLA candidates with log N(H   i) ≥ 20, out of which 6839 have log N(H   i) ≥ 20.3 (Table 3). We also provide values of (or limits on) the equivalent widths of associated metal lines redwards of the Ly-α emission line.

2.2. Statistical sample

We subsequently define a statistical sub-sample that is used to derive the N(H   i) distribution function and the integrated cosmological mass density of neutral gas. First of all, we conservatively reject all quasars with even moderate balnicity (BI  > 0 or flagged visually, Pâris et al. 2012) and apply a more stringent threshold to the data quality, keeping only spectra with CNR  > 3. We then restrict the redshift range as follows: i) N09 showed that the presence of a DLA near the blue end of the spectrum can bias the definition of zmin and proposed an application of a systematic 10 000 km   s-1 velocity shift to zmin that we also apply here; ii) we consider only the region between 3000 km   s-1 redwards of the Ly-β emission line and 5000 km   s-1 bluewards of the Ly-α emission line. The first cut ensures that we consider only the Ly-α forest and avoid the Ly-β+O vi region where associated broad O vi absorption can occur (even if no broad C iv absorption is seen) and be mistaken as DLAs. The second cut avoids DLAs located in the vicinity of the quasar (e.g. Ellison et al. 2002). Finally, we restrict our study to the range z ∈  [2,3.5] . This avoids the very blue end of the spectra (below 3650 Å) where reduction problems have been identified (Pâris et al. 2012). We set the upper limit because the increasing density of the Ly-α forest can introduce a significant fraction of false DLA identifications owing to strong blending of the Ly-α forest lines at the SDSS resolution (Rafelski et al. 2012). The value of 3.5 corresponds to the redshift out to which this systematic can be reliably estimated based on an analysis of mock spectra.

These cuts leave us with 37 503 lines-of-sight bearing 5428 systems with log N(H   i) ≥ 20 (3408 bona-fide DLAs with log N(H   i) ≥ 20.3). We present in Fig. 1 the sensitivity functions g(z) (i.e. the number of lines-of-sight covering a given redshift) for the full and statistical samples. Our statistical DR9 sample is more than three times larger than the overall DR7 sample and it also extends to lower redshifts, thanks to the improved blue coverage of the SDSS spectrograph. The total absorption path length probed by our statistical sample over z = 2–3.5 is Δχ ≈ 45 000 with an average redshift ⟨z⟩ = 2.5.

thumbnail Fig. 1

Redshift sensitivity function g(z) of our full DR9 sample (dotted) and statistical sample (black) compared to that of DR7 (N09, grey).

thumbnail Fig. 2

Column density distribution functions from synthetic (left) and real data (centre) at ⟨z⟩ = 2.5. Horizontal bars represent the bin over which f(NHI) is calculated and vertical error bars represent Poissonian uncertainty. The difference between output and input mock distributions is shown at the bottom of panel a). The double power-law and Γ-function fits to the DR7 distribution (N09, ⟨z⟩ = 2.9) are shown as red dashed lines. f(NHI)(z = 0) are taken from Braun (2012, purple) and Zwaan et al. (2005, green). Right: the contribution of DLAs in a given N(H   i) range to the total mass census of neutral gas. DR9 values are corrected for systematics.

2.3. Mocks

The BOSS collaboration is constantly developing mocks that simulate the Lyman-α forest seen towards BOSS quasars (e.g. Font-Ribera et al. 2012). While mocks were principally designed for BAO studies, the important point for this study is that simulated spectra are produced with the same noise and flux distributions as in the actual DR9 data (Bailey et al., in prep.). In addition, DLAs have been introduced to those mocks with a known distribution (Font-Ribera & Miralda-Escudé 2012). We also applied our DLA-searching algorithm to 33 realisations of 3861 mocks representative of the DR9 data with the same cuts as in real data. From this exercise, the completeness and purity (1 minus the fraction of false identifications) in the statistical sample are both found to be above 95% for log N(H   i) ≥ 20.3 (and higher when restricting to higher N(H   i) systems). Overall, the automatic procedure systematically overestimates N(H   i) by 0.03 dex. This is much lower than the dispersion (0.20 dex) that corresponds to the typical 1σ error on log N(H   i).

3. The column density distribution at ⟨z⟩ = 2.5

In Fig. 2a, we compare the simulated input distribution of H i column densities (in the range N(H   i) = 1020–4 × 1021 cm-2) at z = 2–3 with the one recovered from mocks by our procedure over the same redshift range. We can see that the overall agreement is excellent, although our procedure slightly overestimates f(NHI), particularly at the low column density end. We use the difference between the input and output distributions as the correction to apply to the observed distribution from real data.

To ascertain the properties of the high-column-density end of f(NHI), where statistics are much smaller, we have visually checked all DLA candidates with log N(H   i) ≥ 21.6. In this regime, blind correction using mocks could be more uncertain since the corresponding H i fits are only based on Ly-α, while metals are systematically detected in the real data. Indeed, we found a few cases where two closely-spaced DLAs were mistaken for a higher column density one. Disentangling such blends was possible thanks to the presence of metal lines. For each DLA candidate with log N(H   i) ≥ 21.6, the absorption profile was carefully refitted manually, improving the continuum determination and using metal lines to determine a precise redshift of the absorber. The resulting f(NHI) at z = 2−3 is shown in Fig. 2b with values given in Table 1. It is apparent that the distribution extends beyond 1022 cm-2 with five systems with log N(H   i) ≥ 22 in the statistical sample (8 in the full sample). Extrapolating this function, we might expect to detect DLAs reaching log N(H   i) = 23 at the completion of BOSS.

Following N09, we measure the total amount of neutral gas in DLAs at ⟨z⟩ = 2.5 to be . Figure 2c represents the contribution to the total amount of neutral gas as a function of N(H   i). We confirm the N09 result that the largest contribution comes from systems with N(H   i) ~ 1021 cm-2. However, it is interesting that the systems with N(H   i) in excess of 5    ×    1021 cm-2 contribute a non-negligible fraction of (~10%), although they are rarely represented in most surveys.

Table 1

N(H   i) distribution function at ⟨z⟩ = 2.5.

4. Cosmological mass density of neutral gas

Figure 3 (see also Table 2) shows the evolution of the cosmological mass density in DLAs as a function of redshift. Using mock spectra, we estimate a correction for systematics (over/under-estimate of N(H   i), incompleteness, and contribution of false positives) as a function of redshift. At high redshift, the correction is mostly because of N(H   i) overestimation caused by the denser Ly-α forest together with increasing false positive identifications. At z < 2.3, the correction is upwards owing to higher incompleteness and a slight underestimation of N(H   i). We note that the zero-point photometric calibration can be in error by about 5% below 4000 Å (Pâris et al. 2012), which could affect the detection of DLAs and N(H   i)-measurements in different ways in mocks and real data. This problem will be addressed in forthcoming versions of the pipeline (Bolton et al. 2012).

We observe a decrease in from z = 3.5 to z = 2.3 as in N09 and Prochaska & Wolfe (2009), although with higher values at z < 3.2. This can be explained by the  ~10% contribution of very large column density systems in DR9 and better knowledge of systematics. It is unclear which value of should be used at z = 0. The measurement by Braun (2012) is based on only three galaxies, and although the high-column density end of f(NHI) seems to be well constrained, this may not be true at low N(H i) (Zwaan et al. 2005). Measurements at z ~ 1 (Rao et al. 2006) are still indirect and, while direct searches for DLAs at low redshift are possible (Meiring et al. 2011), they are still quite limited in terms of sample size. It appears that systematics are a larger source of error than statistical uncertainties across most of the redshift range. Keeping this in mind, we can still conclude that evolves only mildly over the past 12 Gyr.

thumbnail Fig. 3

Cosmological mass density of neutral gas in DLAs as a function of redshift (Z05: Zwaan et al. 2005; B12: Braun 2012; R06: Rao et al. 2006; PW09: Prochaska & Wolfe 2009; DR9: this work).

Table 2

and DLA incidence (dN/dz) in different redshift bins.

5. Conclusion

We have presented the first results of our ongoing survey for DLAs in the SDSS-III BOSS, Data Release 9. This represents by far the largest sample of DLAs to date (with  ~12 000 systems with log N(H   i) ≥ 20) and should allow numerous follow-up studies. We expect the sample to be increased by a factor larger than two at the completion of BOSS. Using a well defined sub-sample, and controlling systematics (which dominate the error budget) with synthetic spectra, we derive the H i column density distribution at ⟨z⟩ = 2.5 in the range 1020 − 2 × 1022 cm-2 and characterised the evolution of the cosmological mass density of neutral gas in DLAs at 2 < z < 3.5. This study should help to constrain models of galaxy formation and evolution by measuring the amount of neutral gas immediately available to fuel star formation through cosmic history.


DLAs are contaminants for the study of the Ly-α forest correlation function (Font-Ribera & Miralda-Escudé 2012).


We thank the anonymous referee for helpful comments and suggestions. The French participation group to SDSS-III was supported by the Agence Nationale de la Recherche under grant ANR-08-BLAN-0222. Funding for SDSS-III has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, and the US Department of Energy Office of Science. The SDSS-III web site is SDSS-III is managed by the Astrophysical Research Consortium for the Participating Institutions of the SDSS-III Collaboration including the University of Arizona, the Brazilian Participation Group, Brookhaven National Laboratory, University of Cambridge, Carnegie Mellon University, University of Florida, the French Participation Group, the German Participation Group, Harvard University, the Instituto de Astrofísica de Canarias, the Michigan State/Notre Dame/JINA Participation Group, Johns Hopkins University, Lawrence Berkeley National Laboratory, Max Planck Institute for Astrophysics, Max Planck Institute for Extraterrestrial Physics, New Mexico State University, New York University, Ohio State University, Pennsylvania State University, University of Portsmouth, Princeton University, the Spanish Participation Group, University of Tokyo, University of Utah, Vanderbilt University, University of Virginia, University of Washington, and Yale University.


  1. Ahn, C. P., Alexandroff, R., Allende Prieto, C., et al. 2012, ApJS, accepted [arXiv:1207.7137] [Google Scholar]
  2. Aihara, H., Allen de Prieto, C., An, D., et al. 2011, ApJS, 193, 29 [NASA ADS] [CrossRef] [Google Scholar]
  3. Bolton, A. S., Schlegel, D. J., Aubourg, E., et al. 2012, AJ, accepted [arXiv:1207.7326] [Google Scholar]
  4. Bovy, J., Hennawi, J. F., Hogg, D. W., et al. 2011, ApJ, 729, 141 [NASA ADS] [CrossRef] [Google Scholar]
  5. Braun, R. 2012, ApJ, 749, 87 [NASA ADS] [CrossRef] [Google Scholar]
  6. Dawson, K. S., Schlegel, D. J., Ahn, C. P., et al. 2012, AJ, submitted [arXiv:1208.0022] [Google Scholar]
  7. Eisenstein, D. J., Weinberg, D. H., Agol, E., et al. 2011, AJ, 142, 72 [Google Scholar]
  8. Ellison, S. L., Yan, L., Hook, I. M., et al. 2002, A&A, 383, 91 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  9. Erkal, D., Gnedin, N. Y., & Kravtsov, A. V. 2012, ApJ, submitted [arXiv:1201.3653] [Google Scholar]
  10. Font-Ribera, A., & Miralda-Escudé, J. 2012, J. Cosmol. Astropart. Phys., 7, 28 [NASA ADS] [CrossRef] [Google Scholar]
  11. Font-Ribera, A., McDonald, P., & Miralda-Escudé, J. 2012, J. Cosmol. Astropart. Phys., 1, 1 [Google Scholar]
  12. Gunn, J. E., Siegmund, W. A., Mannery, E. J., et al. 2006, AJ, 131, 2332 [NASA ADS] [CrossRef] [Google Scholar]
  13. Komatsu, E., Smith, K. M., Dunkley, J., et al. 2011, ApJS, 192, 18 [NASA ADS] [CrossRef] [Google Scholar]
  14. Lah, P., Chengalur, J. N., Briggs, F. H., et al. 2007, MNRAS, 376, 1357 [NASA ADS] [CrossRef] [Google Scholar]
  15. Lanzetta, K. M., McMahon, R. G., Wolfe, A. M., et al. 1991, ApJS, 77, 1 [NASA ADS] [CrossRef] [Google Scholar]
  16. Meiring, J. D., Tripp, T. M., Prochaska, J. X., et al. 2011, ApJ, 732, 35 [NASA ADS] [CrossRef] [Google Scholar]
  17. Noterdaeme, P., Petitjean, P., Ledoux, C., & Srianand, R. 2009, A&A, 505, 1087 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  18. Pâris, I., Petitjean, P., Rollinde, E., et al. 2011, A&A, 530, A50 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  19. Pâris, I., Petitjean, P., Aubourg, E., et al. 2012, A&A, in press, DOI: 10.1051/0004-6361/201220142 [Google Scholar]
  20. Prochaska, J. X., & Wolfe, A. M. 2009, ApJ, 696, 1543 [NASA ADS] [CrossRef] [Google Scholar]
  21. Prochaska, J. X., Herbert-Fort, S., & Wolfe, A. M. 2005, ApJ, 635, 123 [NASA ADS] [CrossRef] [Google Scholar]
  22. Prochaska, J. X., Sheffer, Y., Perley, D. A., et al. 2009, ApJ, 691, L27 [Google Scholar]
  23. Rafelski, M., Wolfe, A. M., Prochaska, J. X., Neeleman, M., & Mendez, A. J. 2012, ApJ, 755, 89 [NASA ADS] [CrossRef] [Google Scholar]
  24. Rao, S. M., Turnshek, D. A., & Nestor, D. B. 2006, ApJ, 636, 610 [NASA ADS] [CrossRef] [Google Scholar]
  25. Ross, N. P., Myers, A. D., Sheldon, E. S., et al. 2012, ApJS, 199, 3 [NASA ADS] [CrossRef] [Google Scholar]
  26. Schaye, J. 2001, ApJ, 562, L95 [NASA ADS] [CrossRef] [Google Scholar]
  27. Smee, S., Gunn, J. E., Uomoto, A., et al. 2012, AJ, submitted [arXiv:1208.2233] [Google Scholar]
  28. Vladilo, G., & Péroux, C. 2005, A&A, 444, 461 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  29. Wolfe, A. M., Gawiser, E., & Prochaska, J. X. 2005, ARA&A, 43, 861 [NASA ADS] [CrossRef] [MathSciNet] [Google Scholar]
  30. York, D. G., Adelman, J., Anderson, Jr., J. E., et al. 2000, AJ, 120, 1579 [Google Scholar]
  31. Zwaan, M. A., & Prochaska, J. X. 2006, ApJ, 643, 675 [NASA ADS] [CrossRef] [Google Scholar]
  32. Zwaan, M. A., van der Hulst, J. M., Briggs, F. H., Verheijen, M. A. W., & Ryan-Weber, E. V. 2005, MNRAS, 364, 1467 [NASA ADS] [CrossRef] [Google Scholar]

All Tables

Table 1

N(H   i) distribution function at ⟨z⟩ = 2.5.

Table 2

and DLA incidence (dN/dz) in different redshift bins.

All Figures

thumbnail Fig. 1

Redshift sensitivity function g(z) of our full DR9 sample (dotted) and statistical sample (black) compared to that of DR7 (N09, grey).

In the text
thumbnail Fig. 2

Column density distribution functions from synthetic (left) and real data (centre) at ⟨z⟩ = 2.5. Horizontal bars represent the bin over which f(NHI) is calculated and vertical error bars represent Poissonian uncertainty. The difference between output and input mock distributions is shown at the bottom of panel a). The double power-law and Γ-function fits to the DR7 distribution (N09, ⟨z⟩ = 2.9) are shown as red dashed lines. f(NHI)(z = 0) are taken from Braun (2012, purple) and Zwaan et al. (2005, green). Right: the contribution of DLAs in a given N(H   i) range to the total mass census of neutral gas. DR9 values are corrected for systematics.

In the text
thumbnail Fig. 3

Cosmological mass density of neutral gas in DLAs as a function of redshift (Z05: Zwaan et al. 2005; B12: Braun 2012; R06: Rao et al. 2006; PW09: Prochaska & Wolfe 2009; DR9: this work).

In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.