Copyright © 2005 by Annual Reviews. All rights reserved
|
| Annu. Rev. Astron. Astrophys. 2005. 43:
861-918 Copyright © 2005 by Annual Reviews. All rights reserved |
SAMPLES BIASED BY DUST?
Surveys for damped Ly
systems have the greatest impact if they represent a fair sample of the
neutral gas in the Universe, allowing a clear probe of the evolution
with redshift of the neutral hydrogen content and the metallicity of
neutral gas. However, it has long been a major concern that the sample
of damped
Ly systems suffers from
"dust bias," i.e., the absence from a magnitude-limited QSO sample of
those QSOs that suffer obscuration from dusty foreground damped
Ly systems, leading to
underrepresentation of dusty damped
Ly systems in the
overall sample. The easiest way to probe the existence and abundance of
dust in damped
Ly systems would be to
find the 2175 Å bump feature superimposed in absorption on
background QSO spectra. While
Junkkarinen et al. (2004)
found at least one strong example, this does not appear to be the rule
(Pei, Fall & Bechtold
1991).
Without such sharp features to look for and given the wide range of
intrinsic QSO spectral slopes, reddening from dust in damped
Ly systems must be
searched for statistically by checking if the sample of QSOs with
foreground damped Ly
systems is redder on average than a "control sample" of QSOs without
foreground damped
Ly systems.
10.1. Observational Estimates of Reddening
Ostriker & Heisler
(1984)
pointed out that optically selected QSO samples are biased toward those
QSOs with little foreground dust extinction.
Fall & Pei (1989)
showed that dust in damped
Ly systems did not
appear to cause the famous drop in the QSO number abundance at z
3.
Pei, Fall & Bechtold
(1991)
detected reddening from damped
Ly systems at the 4
confidence level and
inferred dust-to-gas ratios between 1/20 and 1/5 that of the Galaxy,
enough to explain the lack of observed Lyman
emission from damped
Ly systems. This led to
the prediction that 10-70% of QSOs are missing from optically selected
samples, leading to an order of magnitude uncertainty in
g, <
Z>, and other quantities estimated from damped
Ly systems
(Fall & Pei 1993).
However, the dust-to-gas ratios estimated from high-resolution echelle
spectroscopy of QSOs with foreground damped
Ly systems are lower
than the dust-to-gas ratios predicted by
Fall & Pei (1993),
reducing the uncertainty in quantities such as
g to
factors of 2-3.
Pettini et al. (1997a)
combined a metallicity of 1/15 solar with a dust-to-metals ratio of 1/2
that in the Milky Way to find a typical damped
Ly system dust-to-gas
ratio of 1/30 Galactic. Using an SMC reddening curve, they predicted a
dust extinction of only 0.1 magnitudes at 1500 Å in the spectrum of
background QSOs due to damped
Ly system dust. If a
nucleosynthetic floor exists in damped
Ly systems at [Si/Fe]
0.3, then the
dust-to-gas ratios are even lower than this, closer to 1/200 Galactic in
most systems. Indeed, the detection of reddening due to damped
Ly systems by
Pei, Fall & Bechtold
(1991)
conflicts with the recent finding by
Murphy & Liske (2004)
that E(B - V) < 0.01 magnitudes using 81 damped
Ly systems found in a
homogeneous set of SDSS Data Release 2 QSOs. The resolution of the
conflict is not clear at present.
10.2. Surveys of Radio-Selected QSOs
An insidious (but not physically motivated) possibility would be the
existence of gray dust associated with the damped
Ly systems that could
cause obscuration without the telltale effect of reddening. Even this
effect could be overcome by using a radio-selected sample of QSOs. The
reason this has not typically been done is twofold: (1) The ability to
select QSOs within a preferred redshift range makes optical color
selection more efficient. (2) Conducting optical spectroscopic follow-up
on a radio-selected sample of QSOs is far more time-consuming precisely
because they do not have a strict optical magnitude limit; half of the
total exposure time can be required by the dimmest one or two
objects. Obviously, dropping those from the survey would defeat the
entire purpose of radio selection. One radio-selected sample has been
published: the CORALS survey of
Ellison et al. (2001)
found 19 intervening damped
Ly systems toward 66
zem
2.2 radio-selected QSOs
from the Parkes quarter-Jansky sample
(Shaver et al. 1996),
yielding a marginal increase in
d
/ dX and
g at <
z> = 2.37 versus optically selected QSO samples. These results
imply that at most half of damped
Ly systems are missing
from optically selected samples. For
g, the
radio sample yields 1.4 × 10-3 as opposed to the value
of 6.7 × 10-4 found for optically selected samples at
this redshift
(Prochaska &
Herbert-Fort 2004),
but this is only a 1.5
difference given the small sample size.
10.3. Empirical Estimates of Damped
Ly System
Obscuration
A third way to estimate the effects of dust obscuration by damped
Ly systems is to infer
this from the observed chemical abundances. Taking the observed H I
column densities and the dust-to-gas ratios implied by the depletion
patterns of the damped
Ly systems (see Equation
7), it is possible to estimate the extinction in the rest-frame UV of
the QSO for an assumed extinction curve. Given the lack of the observed
2175 Å bump feature, it appears more reasonable to assume an SMC
(Prevot et al. 1984)
rather than Galactic (Cardelli, Clayton & Mathis 1989)
dust extinction law.
Prochaska & Wolfe
(2002)
used this technique (see their figure 24; see
Prochaska 2004
for an update) to correct the observed QSO magnitudes by this inferred
extinction and then to compare the implied true magnitude with the
magnitude limit of the survey used to search for damped
Ly systems (which is
typically shallower than the limit of the survey used to discover the
QSOs). These quantities were then compared to a bootstrap prediction of
how many QSOs are expected to be missing from observed samples due to
extinction by foreground damped
Ly systems. The typical
range of extinction corrections runs from 0 to 0.3 magnitudes, even
though half of the QSOs are so much brighter than the survey limit that
they could have been seen with up to 1 magnitude of extinction. This
shows that damped
Ly systems that cause
between 0.3 magnitudes and 1 magnitude of extinction are rare and
predicts that at most 10% of QSOs are missing from the samples probed
for damped Ly systems
due to a damped Ly dust
bias.