QUASARS PROBING QUASARS. II.

THE ANISOTROPIC CLUSTERING

OF OPTICALLY THICK ABSORBERS AROUND QUASARS
Joseph F. Hennawi
1, 2
and Jason X. Prochaska
3
Received 2006 June 5; accepted 2006 September 20
ABSTRACT
With close pairs of quasars at different redshifts, a background quasar sight line can be used to study a foreground
quasars environment in absorption. We used a sample of 17 Lyman limit systems with column density N
H i
10
19 cm
2
selected from 149 projected quasar pair sight lines to investigate the clustering pattern of optically thick ab-
sorbers around luminous quasars at z $ 2.5. Specically, we measured the quasar-absorber correlation function in
the transverse direction and found a comoving correlation length of r
0
9.2
1.5
1.7
h
1
Mpc (comoving), assuming a
power-law correlation function / r

, with 1.6. Applying this transverse clustering strength to the line of

sight, would predict that $15%Y50%of all quasars should showa N
H i
10
19
cm
2
absorber within a velocity win-
dow of v < 3000 km s
1
. This overpredicts the number of absorbers along the line of sight by a large factor,
providing compelling evidence that the clustering pattern of optically thick absorbers around quasars is highly aniso-
tropic. The most plausible explanation for the anisotropy is that the transverse direction is less likely to be illuminated
by ionizing photons than the line of sight, and that absorbers along the line of sight are being photoevaporated. A
simple model for the photoevaporation of absorbers subject to the ionizing ux of a quasar is presented, and it is shown
that absorbers with volume densities n
H
P0.1 will be photoevaporated if they lie within $1 Mpc (proper) of a luminous
quasar. Using this simple model, we illustrate how comparisons of the transverse and line-of-sight clustering around
quasars can ultimately be used to constrain the distribution of gas in optically thick absorption-line systems.
Subject headinqg s: intergalactic medium quasars: absorption lines quasars: general surveys
Online material: color gures
1. INTRODUCTION
Although optically thick absorption-line systems, i.e., Lyman
limit systems ( LLSs) and damped Lyc systems ( DLAs), are de-
tected as the strongest absorption lines in quasar spectra, the two
types of objects, quasars and absorbers, play rather different roles
in the evolution of structure in the universe. The hard ultraviolet
radiation emitted by luminous quasars gives rise to the ambient
extragalactic ultraviolet ( UV) background (see e.g., Haardt &
Madau 1996; Miralda-Escude 2003; Meiksin 2005) responsible
for maintaining the low neutral fraction of hydrogen ($10
6
) in
the intergalactic medium (IGM), established during reionization.
However, high column density absorbers represent the rare loca-
tions where the neutral fractions are much larger. Gas clouds with
column densities log N
H i
17.2 are optically thick to Lyman
continuum(t
LL
k1) photons, giving rise to a neutral interior self-
shielded fromthe extragalactic ionizing background. In particular,
the damped Lyc systems dominate the neutral gas content of the
universe (Prochaska et al. 2005), which is the primary reservoir
for the formation of stars observed in local galaxies.
One might expect optically thick absorbers to be absent at small
separations from luminous quasars. For example, a quasar at z
2.5 with magnitude r 19 emits a ux of ionizing photons that is
130 times higher than that of the extragalactic UV background at
an angular separation of 60
00
, corresponding to a proper distance of
340 h
1
kpc. Indeed, the decrease in the number of optically
thin absorption lines (log N
H i
< 17.2, hence t
LL
P1) in the vi-
cinity of quasars, known as the proximity effect ( Bajtlik et al.
1988), provides a measurement of the UV background (Scott
et al. 2000). If nature provides a nearby background quasar sight
line, one can also study the transverse proximity effect, which
is the expected decrease in absorption in a background quasars
Lycforest caused by the transverse ionizing ux of a foreground
quasar. The transverse effect has yet to be detected, in spite
of many attempts (Crotts 1989; Dobrzycki & Bechtold 1991;
Fernandez-Soto 1995; Liske &Williger 2001; Schirber et al. 2004;
Croft 2004; but see Jakobsen et al. 2003).
On the other hand, it has long been known that quasars are as-
sociated with enhancements in the distribution of galaxies (Bahcall
et al. 1969; Yee & Green 1984, 1987; Bahcall & Chokshi 1991;
Smith et al. 2000; Brown et al. 2001; Serber et al. 2006; Coil
et al. 2006b), although these measurements of quasar galaxy
clustering are mostly limited to low redshifts P1.0. Recently,
Adelberger & Steidel (2005, hereafter AS05) measured the clus-
tering of Lyman break galaxies (LBGs) around luminous quasars
in the redshift range 2 Pz P3.5 and found a best-t correlation
length of r
0
4.7 h
1
Mpc ( 1.6), very similar to the auto-
correlation length of z $ 2Y3 LBGs (Adelberger et al. 2003).
Cooke et al. (2006) recently measured the clustering of LBGs
around DLAs and measured a best-t r
0
2.9 h
1
Mpc with
1.6, but with large uncertainties (see also Gawiser et al. 2001;
Bouche & Lowenthal 2004). If LBGs are clustered around qua-
sars and DLAs, might we expect optically thick absorbers to be
clustered around quasars? This is especially plausible in light of
recent evidence that DLAs arise from a high-redshift galaxy
population that are not unlike LBGs (Schaye 2001; Mller et al.
2002).
Clues to the clustering of optically thick absorbers around
quasars come fromproximate DLAs, which have absorber red-
shifts within 3000 km s
1
of the emission redshift of the quasars
A
1
Department of Astronomy, University of California, Berkeley, CA; joeh@
berkeley.edu.
2
Hubble Fellow.
3
Department of Astronomy and Astrophysics, University of California
Observatories/Lick Observatory; University of California, Santa Cruz, CA; xavier@
ucolick.org.
735
The Astrophysical Journal, 655:735Y748, 2007 February 1
# 2007. The American Astronomical Society. All rights reserved. Printed in U.S.A.
(see e.g., Mller et al. 1998). Recently, J. X. Prochaska et al.
(2007, in preparation) compared the line density of proximate
DLAs in the Sloan Digital Sky Survey (SDSS) Data Release 5
quasar catalog to the average number density of DLAs in the uni-
verse ( Prochaska et al. 2005) and found that the abundance of
DLAs is enhanced by a factor of $1.4 near quasars. A similar
study by Russell et al. (2006; see also Ellison et al. 2002) mea-
sured a comparable enhancement, which they attributed to the
clustering of DLA galaxies around quasars.
Projected pairs of quasars with small angular separations (0 P
5
0
) but different redshifts can also be used to study the clustering
of absorbers around quasars. In this context, clustering is mani-
fest as an excess probability, above the cosmic average, of detect-
ing an absorber in the neighboring background quasar spectrum
near the redshift of the foreground quasar. In the rst of this series
of four papers on optically thick absorbers near quasars (Hennawi
et al. 2006b, hereafter Paper I), we used background quasar sight
lines to search for optically thick absorption in the vicinity of
foreground quasars: 149 projected quasar pairs were system-
atically surveyed for Lyman limit Systems ( LLSs) and Damped
Lyc systems ( DLAs) in the vicinity of 1.8 < z < 4.0 luminous
foreground quasars. Asample of 27 newabsorbers were uncov-
ered with transverse separations R < 5 h
1
Mpc (comoving)
from the foreground quasars, of which 17 were super-LLSs with
N
H i
10
19
cm
2
.
The distribution of foreground quasar redshifts, transverse
separations, and ionizing ux ratios probed by the projected pair
sight lines studied in Paper I are illustrated in Figure 1. The lled
symbols outlined in black indicate the sight lines that have an
absorption-line system with N
H i
10
19
cm
2
within a velocity
interval of v j j 1500 km s
1
of the foreground quasar redshift ,
and the open symbols represent sight lines with no such absorber.
The line density of absorbers per unit redshift at z $ 2.5 with
column densities N
H i
10
19
cm
2
is dN/dz 0.8 (OMeara
et al. 2007), whichimplies that the expected number of random
quasar-absorber coincidences withinjvj 1500 km s
1
is $3%.
The expected number of random quasar-absorber associations
in the sample of 149 sight lines is hNi 4.4, whereas 17 sys-
tems were discovered in Paper I compelling evidence for
clustering.
This clustering is particularly conspicuous on small scales: out
of eight sight lines with R < 500 h
1
kpc, four quasar-absorber
pairs were discovered, which implies a small-scale covering factor
of $50%, whereas hNi 0.18 would have been expected at ran-
dom. This excess of $25 over randommay not come as a surprise
when one considers that galaxies are strongly clustered around
quasars on small scales. However, the light from every isolated
quasar in the universe traverses these same small scales on the
way to Earth. Naively, we would then expect a comparable cov-
ering factor of N
H i
10
19
cm
2
super-LLSs along the line of
sight, which is denitely not observed to be the case.
How can we quantify the clustering of absorbers around qua-
sars? Howdo we compare the transverse clustering pattern to the
number of proximate absorbers observed along the line of sight?
What can the quasar-absorber clustering pattern teach us about
quasars and absorbers? These questions are addressed in this
second paper of the series. We briey review the Paper I survey
of projected quasar pairs and discuss the quasar-absorber sample
in x 2. A formalism for quantifying the line of sight and trans-
verse clustering of absorbers around quasars is presented in x 3.
In x 4 we introduce a maximum likelihood technique for estimat-
ing the quasar-absorber correlation function, which we use to
measure the clustering of our quasar-absorber sample in x 5. Our
measurement of the transverse clustering is used to predict the
expected number of proximate absorption-line systems that should
be observed along the line of sight in x 6. In x 7, we introduce a
simple analytical model for the photoevaporation of optically thick
absorbers by quasars, which is used to illustrate how comparisons
of the line-of-sight and transverse clustering can constrain the dis-
tribution of gas in optically thick absorbers. We summarize and
discuss our results in x 8.
Paper III of this series (J. F. Hennawi et al. 2007, in prepa-
ration) investigates uorescent Lyc emission from our quasar-
absorber pairs, and echelle spectra of several of the quasar-LLS
systems published here are analyzed in Paper IV (J. X. Prochaska
& J. F. Hennawi 2007, in preparation). Throughout this paper
we use the best-t WMAP (only) cosmological model of Spergel
et al. (2003) with
m
0.270,

0.73, and h 0.72. Unless

otherwise specied, all distances are comoving.
4
It is help-
ful to remember that in the chosen cosmology, at a redshift of
z 2.5, an angular separation of 0 1
00
corresponds to a co-
moving transverse separation of R 20 h
1
kpc, and a velocity
Fig. 1.Distribution of foreground quasar redshifts, transverse separations,
and ionizing ux ratios probed by the background quasar sight lines. The upper
plot shows the ratio of the ionizing ux to the UVbackground q
UV
; note the gen-
eral R
2
trend. The lower plot shows foreground quasar redshift vs. transverse
comoving separations. The squares have a Keck spectrum of the background
quasar, triangles have Gemini background spectra, upside-down triangles have
MMT background spectra, and circles have SDSS background spectra. Filled
symbols outlined in black have a super-LLS within v j j < 1500 km s
1
at the
foreground quasar redshift (see Table 1) and open symbols have no absorber.
The region to the left of the dotted line is excluded by the SDSS ber collision
limit of 0 55
00
, which explains the paucity of SDSS background spectra there.
The Keck/Gemini/ MMT spectra probe angular separations 1 order of magnitude
smaller than the ber collision limit, allowing us to study the foreground quasar
environment down to 70 h
1
kpc, where the ionizing ux is $10,000 times the UV
background. [See the electronic edition of the Journal for a color version of this
qure.]
4
Note that in Paper I proper distances were quoted.
HENNAWI & PROCHASKA 736 Vol. 655
difference of 1500 km s
1
corresponds to a radial redshift space
distance of s 15 h
1
Mpc. For a quasar at z 2.5, with an
SDSS magnitude of r 19, the ux of ionizing photons is
130 times higher than the ambient extragalactic UVbackground at
an angular separation of 60
00
(comoving R 1.2 h
1
Mpc). Fi-
nally, we use the terms optically thick absorbers and LLSs
interchangeably, both referring to quasar absorption-line systems
with log N
H i
17.2, making them optically thick at the Lyman
limit (t
LL
k1).
2. QUASAR-ABSORBER SAMPLE
In this section we briey summarize the quasar-absorber sam-
ple and the parent sample of projected pairs from which it was
selected; for full details see Paper I. In x 5 we quantify the clus-
tering of absorbers around quasars in the transverse direction by
comparing the number of quasar-absorber pairs discovered to
the number from random expectation. Although lower column
density systems were published in Paper I, we focus here on ab-
sorbers with N
H i
10
19
cm
2
, or so called super-LLSs. This
choice reects a variety of concerns. First, we wanted to maximize
the number of quasar-absorber pairs; had we restricted consider-
ation to only DLAs we would have been left with ve systems.
Second, this is the lowest column density at which we believe we
can identify a statistical sample of absorbers with the moderate
resolution spectra used in Paper I. Finally, the column density dis-
tribution of optically thick absorption-line systems has been de-
termined for log N
H i
19 ( Peroux et al. 2005; OMeara et al.
2007), whereas it is unknown at the lower column densities
17.2 Plog N
H i
P19.
Modern spectroscopic surveys select aqainst close pairs of
quasars because of ber collisions. For the SDSS, the nite size
of optical bers implies only one quasar in a pair with separation
<55
00
can be observed spectroscopically on a given plate,
5
and a
slightly smaller limit (<30
00
) applies for the Two Degree Field
Quasar Survey (2QZ; Croom et al. 2004). Thus, for subarc-
minute separations, additional spectroscopy is required both to
discover companions around quasars and to obtain spectra of
sufcient quality to search for absorption-line systems. For wider
separations, projected quasar pairs can be found directly in the
SDSS spectroscopic quasar catalog. Hennawi et al. (2006a) used
the 3.5 m telescope at Apache Point Observatory (APO) to spec-
troscopically conrm a large sample of photometrically selected
close quasar pair candidates and published the largest sample of
projected quasar pairs in existence.
In Paper I we combined high signal-to-noise ratio (S/ N) mod-
erate resolution spectra of the closest Hennawi et al. (2006a) pro-
jected pairs, obtained from Gemini, Keck, and the MMT, with a
large sample of wider separation pairs from the SDSS spectro-
scopic survey. The Keck, Gemini, and MMT spectra were all of
sufcient S/ N to detect absorbers with column densities N
H i

10
19
cm
2
, and a S/ Ncriterion was used to isolate the SDSS pro-
jected pair sight lines for which such an absorber could be detected
in the background quasar spectrum. We considered all projected
quasar pair sight lines that had a comoving transverse separation
of R < 5 h
1
Mpc, a redshift separation v 2500 km s
1
be-
tween the foreground and background quasar (to exclude physical
binaries), and which satised our S/ Ncriteria, and we arrived at a
total of 149 projected pair sight lines in the redshift range 1.8 <
z < 4.0.
A systematic search for optically thick absorbers with redshifts
within jvj < 1500 km s
1
of the foreground quasars was con-
ducted by visually inspecting the 149 projected pair sight lines,
where the velocity window was chosen to bracket the uncer-
tainties in the foreground quasars systemic redshift (see e.g.,
Richards et al. 2002 and x 4 of Paper I ). Voigt proles were t
to systems with signicant Lycabsorption to determine the H i
column densities. We uncovered 27 new quasar absorber pairs
with column densities 17.2 < log N
H i
< 20.9 and transverse
comoving distances 71 h
1
kpc < R < 5 h
1
Mpc fromthe fore-
ground quasars, of which 17 were super-LLSs with log N
H i
19.
The mean redshift of the foreground quasar in the parent sample
was hzi 2.47 and that of the 17 quasar Ysuper-LLS pairs was
hz
abs
i 2.55.
The completeness and false-positive rate of the Paper I survey
are a signicant source of concern. Prochaska et al. (2005) dem-
onstrated that the spectral resolution ( FWHM 150 km s
1
)
and S/ N of the SDSS spectra are well suited to constructing a
complete (k95%) sample of DLAs ( log N
H i
20.3) at z 2.2;
however, the completeness at the lower column densities log N
H i

19 considered here has not been systematically quantied. Fur-
thermore, aggressive S/ N criteria were employed in Paper I in
order to gather a sufcient number of projected quasar pairs. Line
blending can signicantly depress the continuumnear the Lycpro-
le and mimic a damping wing, biasing column density measure-
ments high or giving rise to false positives. Any statistical study
will thus suffer froma Malmquist-type bias because line blend-
ing biases lower column densities upward, and the line density
of absorbers dN/dz is a steep function of column density limit.
Based on visually inspecting the 149 background quasar spec-
tra and a comparison with echelle data for three systems, we es-
timated that the Paper I survey was $90%complete for log N
H i

19.3 for all the Keck/Gemini / MMTspectra and$3/4 of the SDSS
spectra, which accounts for about 125 of the 149 spectra searched.
To address the false-positive rate, we compared with echelle data
for three systems and found that our column density was over-
estimated by $2.5 o for one system, raising it above the super-
LLS ( log N
H i
19) threshold. However, this absorber was lo-
cated blueward of the quasars Lyu emission line, in a part of the
spectrum crowded by the presence of both the Lyc and Lyu
forests. Amore careful examination of the completeness and false-
positive rate of super-LLSs identied in spectra of the resolution
and S/ Nused in Paper I is denitely warranted. In x 5 we explore
how our clustering measurement changes if we discard systems
within %1 o of the log N
H i
19 threshold.
Relevant quantities for the super-LLSYquasar pairs that are
used in our clustering analysis are given in Table 1. The distribu-
tion of foreground quasar redshifts, transverse separations, and
ionizing ux ratios probed by all of our projected pair sight lines
is illustrated in Figure 1. The lled symbols outlined in black
indicate the sight lines that have an absorption-line system near
the foreground quasar with log N
H i
19 (see Table 1), and open
symbols are sight lines with no such absorber.
3. QUANTIFYING QUASAR-ABSORBER CLUSTERING
In the absence of clustering, the line density of absorption-line
systems per unit redshift above the column density threshold N
H i
is given by the cosmic average
dN
dz
( )
(N
H i
. z) nAf
cov
c
H(z)
. 1
where n is comoving number density of the galaxies or objects
that give rise to absorption-line systems, A is their absorption
5
An exception to this rule exists for a fraction ($30%) of the area of the SDSS
spectroscopic survey covered by overlapping plates. Because the same area of sky
was observed spectroscopically on more than one occasion, there is no ber col-
lision limitation.
QUASARS PROBING QUASARS. II. 737 No. 2, 2007
cross section (in comoving units), f
cov
is the covering factor, and
H(z) is the Hubble constant. Note that the line density is degener-
ate with respect to the combination nAf
cov
, and only their product
can be determined by measuring the abundance of absorption-
line systems.
At an average location in the universe, the probability of nd-
ing an absorber in a background quasar spectrum within the
redshift interval z 2(1 z)v/c, corresponding to a veloc-
ity interval 2v, is simply P hdN/dziz. For a projected pair
of quasars, clustering around the foreground quasar will increase
the probability of nding an absorber in the background quasar
spectrum in the vicinity of the foreground quasar, whereas the
foreground quasars radiation eld could reduce this enhancement
by photoevaporating absorption systems. If the quasar sight lines
have a comoving transverse separation R, and assuming that one
searches a velocity interval v about the foreground quasar
redshift ( because of redshift uncertainties), we can express the
increase in line density near the foreground grounds in terms of
a transverse correlation function
?
(R) as
dN
dz

dN
dz
( )
1
?
R. v . 2
where
?
(R. v) is given by an average of the three-dimensional
quasar-absorber correlation function,
QA
(r), over a cylinder
with volume V A2v/aH(z), where, a is the scale factor and
2v(1 z)/H(z) is the length of the cylinder in the line-of-sight
direction. We can thus write

?
(R)
1
V
Z
V
dV
QA
(r)
%
aH(z)
2v
Z
v, aH(z)
v, aH(z)
dZ
QA
(

R
2
Z
2
p
). 3
where Z is a comoving distance in redshift space. The last ap-
proximation in equation (4) assumes that the volume average
over the cylinder can be replaced by the line average along the
line-of-sight direction, which is valid provided that we are in
the far-eld limit, i.e., the transverse separation is much larger
than the diameter of the cylinder R 3A
1/2
. Thus, provided we
consider distances R much larger than the dimension of the ab-
sorber, the transverse clustering is independent of the absorption
cross section.
For proximate absorbers along the line of sight, we can sim-
ilarly dene a line-of-sight correlation function
dN
dz

dN
dz
( )
1
k
(v)

. 4
where

k
(v)
aH(z)
Av
Z
v, aH(z)
Z
cut
dZ
Z
dA
QA
(r). 5
The lower limit of the radial integration is set to Z
cut
, a cutoff
that is introduced to parameterize our ignorance of the geom-
etry of the absorbers. For instance, if the absorbers were pan-
cake shapes that were always oriented perpendicular to the line
of sight ( like face-on spiral galaxies), then we would not cut off
the line-of-sight integration at all (Z
cut
0), since face-on pan-
cakes at zero separation can still obscure the quasar. For a hard
sphere, we would set Z
cut
2 A/
1/2
, presuming that the ob-
scuring cross section of the absorber drops to zero for points
interior to it. As smaller values of Z
cut
will correspond to larger
line-of-sight clustering, we henceforth conservatively assume
the hard sphere case and use Z
cut
2 A/
1/2
.
Note that one is no longer in the far-eld limit for the in-
tegral in equation (5), and the clustering amplitude
k
explicitly
TABLE 1
Super-LLSs Near Quasars from Paper I
Name z
bg
z
fg
0
(arcsec)
R
(h
1
kpc) z
abs
v j j
( km s
1
)
v
fg
( km s
1
)
log N
H i
(cm
2
) q
UV
Telescope
SDSS J02250739...................... 2.99 2.440 214.0 4310 2.4476 690 500 19.55 0.2 5 SDSS
SDSS J02390106...................... 3.14 2.308 3.7 72 2.3025 540 1500 20.45 0.2 6369 Keck
SDSS J0256+0039....................... 3.55 3.387 179.0 4195 3.387 20 1000 19.25 0.25 20 SDSS
SDSS J03380005...................... 3.05 2.239 73.5 1415 2.2290 960 1500 20.9 0.2 13 SDSS
SDSS J0800+3542....................... 2.07 1.983 23.1 415 1.9828 40 300 19.0 0.15 488 Keck
SDSS J0833+0813....................... 3.33 2.516 103.4 2112 2.505 980 1000 19.45 0.3 18 SDSS
SDSS J0852+2637....................... 3.32 3.203 170.9 3917 3.211 550 1500 19.25 0.4 13 SDSS
SDSS J1134+3409....................... 3.14 2.291 209.2 4073 2.2879 320 500 19.5 0.3 11 SDSS
SDSS J1152+4517....................... 2.38 2.312 113.4 2216 2.3158 370 500 19.1 0.3 30 SDSS
SDSS J1204+0221....................... 2.53 2.436 13.3 267 2.4402 370 1500 19.7 0.15 625 Gemini
SDSS J1213+1207....................... 3.48 3.411 137.8 3246 3.4105 30 1500 19.25 0.3 39 SDSS
SDSS J1306+6158....................... 2.17 2.111 16.3 302 2.1084 200 300 20.3 0.15 420 Keck
SDSS J1312+0002....................... 2.84 2.671 148.5 3129 2.6688 200 500 20.3 0.3 23 SDSS
SDSS J1426+5002....................... 2.32 2.239 235.6 4529 2.2247 1330 500 20.0 0.15 19 SDSS
SDSS J14300120...................... 3.25 3.102 200.0 4517 3.115 960 1500 20.5 0.2 26 SDSS
SDSS J1545+5112....................... 2.45 2.240 97.6 1873 2.243 320 500 19.45 0.3 30 SDSS
SDSS J1635+3013....................... 2.94 2.493 91.4 1861 2.5025 820 500 >19 111 SDSS
Notes.Optically thick absorption-line systems near foreground quasars. The background and foreground quasar redshifts are denoted by z
bg
and z
fg
, respectively.
The angular separation of the quasar pair sight lines is denoted by 0, which corresponds to a transverse comoving separation of R at the foreground quasar redshift.
Absorber redshift is indicated by z
abs
, and v j j is the velocity difference between the absorber redshift and our best estimate of the redshift of the foreground quasar. Our
estimated error on the foreground quasar redshift is denoted by v
fg
. Foreground quasar redshifts and redshift errors were estimated according to the detailed procedure
described in x 4 of Paper I. The logarithm of the column density of the absorber from a t to the H i prole is denoted by log N
H i
. The column labeled Telescope
indicates the instrument used to observe the background quasar. The quantity q
UV
1 F
QSO
/F
UVB
is the maximum enhancement of the quasars ionizing photon ux
over that of the extragalactic ionizing background at the location of the background quasar sight line, assuming that the quasar emission is isotropic (see Appendix A of
Paper I ). We compare to the UV background computed by F. Haardt & P. Madau (2007, in preparation).
HENNAWI & PROCHASKA 738 Vol. 655
depends on the cross section of the absorber. It is easy to under-
stand the nature of this dependence if one considers that the line
density /nA is xed by measurements of hdN/dzi. Thus, smaller
cross sections correspond to larger volume number densities, but
smaller scales near the quasar are being sampled by the integral
in equation (5), where the correlation function
QA
/ r

is large.
Conversely, larger cross sections sample regions farther from the
quasar, and thus
QA
is averaged over a larger volume and hence
diluted. Figure 2 illustrates the dependence of
k
(v) on the
proper radius of the absorber cross section, r
abs
(aA/)
1/2
, for a
range of sizes.
In equations (4) and (5)
QA
(r) represents a real space correla-
tion function, and it may appear that we have neglected redshift
space distortions. Strictly speaking, the redshift space correla-
tion function
QA
(R. Z) should be appear in equations (4) and (5).
This
QA
(R. Z) is the convolution of the real space correlation
function (r) with the velocity distribution in the radial direction,
which can have contributions from both peculiar velocities and
uncertainties in the systemic redshifts of the quasar. However,
provided that the distance in redshift space over which we pro-
ject v contains most of the probability under this distribution, it
is a good approximation to replace the redshift space correlation
function, under the integrals in equations (4) and (5), with the real
space correlation function, because radial velocities will simply
move pairs of points within the volume.
On small spatial scales one should be cognizant of the distinc-
tion between absorption intrinsic to the host halo of the quasar
itself and correlated absorption due to nearby galaxies. The cor-
relation function description presented above is valid in either
case; however, intrinsic absorbers may result in excess small-scale
clustering relative to the expectation for galaxies. For the trans-
verse clustering measurement, only the closest projected pairs can
probe separations small enough to resolve the host galaxy of the
quasar, whereas proximate absorbers along the line of sight can
probe scales arbitrarily close to the quasar, although peculiar ve-
locities and quasar redshift uncertainties imply that their distance
to the quasar cannot be measured. For our projected pair sample,
only the closest sight line with R 22 h
1
kpc (proper) is suf-
ciently small to probe the host of the quasar. A comparison of
transverse and line-of-sight clustering is thus complicated by the
fact that the line of sight can in principle probe very small scales
near the quasar where intrinsic absorption due to the quasar host
can play a role.
4. ESTIMATING THE CORRELATION FUNCTION
4.1. Maximum Likelihood Estimator
Given a quasar-absorber correlation function
QA
, equa-
tions (2) and (4) describe howto compute the probability of nd-
ing an absorber in the redshift interval z 2(1 z)v/c at
a transverse distance R from a foreground quasar: P(R. z)
(dN/dz)z. Considering that we only have 17 quasar-absorber
pairs selected from 149 sight lines, it will be difcult to measure
more than a single parameter with reasonable errors. Hence, we
assume the quasar-absorber correlation function to have a power-
law form

QA
C r,r
0

. 6
where C is a clustering amplitude and r
0
is the correlation
length. The amplitude C is degenerate with r
0
, but we choose to
estimate C because it allows for the possibility of anticorre-
lation (C < 0), which could result fromthe QSO ionizing radia-
tion eld. Motivated by the slope of the LBG autocorrelation
function(Adelberger et al. 2005), we choose to x 1.6. Asim-
ilar procedure was employed by AS05, who measured cluster-
ing of LBGs around luminous quasars (2 Pz P3.5) and found
a best-t correlation length of r
0
4.7 h
1
Mpc. We set r
0

4.7 h
1
Mpc as a ducial value; thus. Ccan be interpreted as the
quasar-absorber clustering amplitude relative to the AS05 quasar-
LBG result.
Consider an ensemble of N projected pair sight lines with
background quasars at transverse separations R
i
from foreground
quasars at redshifts z
i
. Given
QA
, we can compute the associated
probabilities P
i
P(R
i
. z
i
). Suppose that N
SLLS
of these N sight
lines show absorption from super-LLSs with log N
H i
19. The
likelihood of the data, given the model parameter C, is then
L(C)
Y
N
SLLS
i
P
i
Y
NN
SLLS
j
(1 P
j
). 7
where the probabilities P
i
1 are capped at 1. By maximizing
the likelihood with respect to the parameter C, we estimate the
clustering amplitude from the data.
4.2. Monte Carlo Simulations
We must verify that the maximumlikelihood estimator in equa-
tion (7) is unbiased, and we would also like to knowhowto assign
error bars to an estimate of C. Both of these points can be ad-
dressed with Monte Carlo methods. The distribution of redshifts
and transverse separations in Figure 1 is not uniform, and it is thus
important that we preserve this distribution when constructing
mock data sets to assign errors.
For a true value of the clustering amplitude C, we can com-
pute the probabilities P
i
P(R
i
. z
i
) of observing an absorber for
each of the N projected pair sight lines. Mock data sets can then
be constructed by generating an N-dimensional vector of devi-
ates from the uniform distribution x
i
and assigning sight lines
with x
i
< P
i
, an absorption-line system. The same maximum
Fig. 2.Dependence of line-of-sight correlation function on the ( proper)
size of the absorber cross section r
abs
. Avelocity interval of v 3000 km s
1
was assumed. Small cross sections sample smaller scales near the quasar where
the correlation
QA
/ r

function is large, giving a larger value for the average

in eq. (5) for
k
, whereas large cross sections dilute
QA
on the average, giving
smaller values of
k
. The solid line is for a correlation length of r 4.7 h
1
Mpc
and slope of 1.6. The dotted line is for a correlation function with slope
2.0, normalized to have the same clustering amplitude as the shallower 1.6
model at r 1 h
1
Mpc. [See the electronic edition of the Journal for a color ver-
sion of this qure.]
QUASARS PROBING QUASARS. II. 739 No. 2, 2007
likelihood estimator is applied to these mock data sets to deter-
mine an estimate

C for each one, thus allowing us to measure the
probability distribution P(

CjC) of the estimate about the true

model.
4.3. Binned Correlation Function
An alternative to the maximum likelihood technique described
above would be to estimate the correlation length by tting a
power-lawmodel to estimates of the correlation function in bins of
transverse separation. However, because we have only 17 quasar-
absorber pairs, the best-t correlation length from this procedure
would be very sensitive to the binning chosen. We compute the
correlation functionin bins because it provides a useful way to vis-
ualize the data, although we do not t the binned data for C.
A simple estimator for the transverse correlation function in
equation (2) is
6

?
(R. v)
hQAi
hQRi
1. 8
where hQAi is the number of quasar-absorber pairs in a trans-
verse radial bin centered on R, and hQRi
P
i
hdN/dziz is the
number of quasar-randompairs expected. Poisson error bars are
used for the binned correlation function. This is reasonable be-
cause pairs are spread out over the entire sky; hence, there will
be no covariance between the bins, and cosmic variance is also
negligible.
4.4. The Column Density Distribution
Before we can estimate the correlation amplitude with equa-
tion (7), we require the cosmic average line density of super-LLSs
hdN/dzi. The line density is the zeroth moment of the column
density distribution
dN
dz
( )
( N
H i
. z)
Z
1
N
H i
f
H i
N
H i
. z dN
H i
. 9
OMeara et al. (2007) measured the column density distribu-
tion for super-LLSs in the range 19 < log N
H i
< 20.3 at z $
2.7 and found good agreement with a power law f (N )
A(N
H i
/10
19
cm
2
)
b
, where A 1.10 ; 10
20
and b 1.43.
For log N
H i
20.3, Prochaska et al. (2005) measured the col-
umn density distribution from the SDSS quasar sample in red-
shift bins.
To evaluate the integral in equation (9), we use the OMeara
et al. (2007) power-law t in the range 19 < log N
H i
< 20.3 and
a spline t to the Prochaska et al. (2005) results in the redshift bin
centered at z $ 2.7. For the redshift evolution, we simply scale
the line density of all absorbers by the evolution of DLAs with
log N
H i
20.3 measured by Prochaska et al. (2005; see their
Figure 8). This procedure assumes that the abundance of super-
LLSs evolves similarly to that of DLAs, an assumption that is
consistent with but not conrmed by OMeara et al. (2007).
5. CLUSTERING RESULTS
We applied the maximumlikelihood estimator to the 17 quasar-
absorber pairs (see Table 1) selected from the 149 projected pair
sight lines shown in Figure 1. The maximum likelihood value of
the clustering amplitude is C 2.9, which corresponds to a cor-
relation length of r
0
9.2 h
1
Mpc.
In Figure 3 we showthe probability distribution of the cluster-
ing amplitude maximum likelihood estimates P(

C) for a model
with true value equal to our measurement of C 2.9. This dis-
tribution was created by applying the maximum likelihood es-
timator in equation (7) to 100,000 mock realizations of our 149
projected pair sight lines, as described in x 4.2. The mean of this
distribution is h

Ci 2.94, the dispersion is o 0.86, and the dot-

ted curve shows a Gaussian with the same mean and dispersion.
The Monte Carlo simulation indicates that our estimator is in-
deed unbiased and that the error distribution is reasonably well
approximated by a Gaussian. Using the full distribution from
the Monte Carlo simulation, we compute the 68% condence
interval about the true input value, which we use to quote
errors on our clustering measurement, C 2.9 0.8 or r
0

9.2
1.5
1.7
h
1
Mpc. If we assume a steeper slope for the correla-
tion function 2, we obtain r
0
5.8
1.0
0.6
h
1
Mpc.
We detect strong clustering of super-LLSs around quasars. Fig-
ure 4 shows the binned transverse correlation function
?
com-
pared to the two maximum likelihood ts ( 1.6 and 2)
and to the prediction from the quasar-LBG correlation function
measured by AS05. For 1.6, the quasar-absorber clustering
amplitude is 3 times larger than the quasar-LBG measurement of
AS05 and inconsistent with it at the k1 o level.
The autocorrelation functions of high-redshift galaxies tend
to become progressively steeper on (comoving) scales R P
1 h
1
Mpc characteristic of the sizes of dark matter halos (Coil
et al. 2006a; Ouchi et al. 2005; Lee et al. 2006; Conroy et al.
2005); thus, similar behavior might be in cross-correlation around
quasars. Indeed, Hennawi et al. (2006a) detected an order of mag-
nitude excess quasar autoclustering on scales P100 h
1
kpc. For
this reason, we also quote our results for 2, although a steeper
slope is not clearly favored by the binned data in Figure 4.
At rst glance, it may seem odd that the relative error on our
measurement of r
0
is $20%using only 17 quasar-absorber pairs,
whereas AS05 quote a $30%error on r
0
from$200 quasar-LBG
pairs around a sample of $50 quasars. The estimator used here is
qualitatively similar to that used by AS05 (see also Adelberger
6
This is actually an estimator of
?
(R. v) averaged over the volume of the
bin. We ignore this distinction, because the differences are substantially smaller
than our errors and we are not tting parameters from this estimate.
Fig. 3.Distribution of clustering amplitude estimates from Monte Carlo
simulations. The black histogram shows the probability distribution of the clus-
tering amplitude maximumlikelihood estimates P(

C) for a model with true value

equal to our measurement of C 2.9. This distribution was created by applying
the maximum likelihood estimator in eq. (7) to 100,000 mock realizations of our
149 projected pair sight lines, as described in x 4.2. The mean of this distribution is
h

Ci 2.94, the dispersion is o 0.86, and the dotted curve shows a Gaussian
with the same mean and dispersion. The full distribution (histoqram) is used to as-
sign errors to our measurement. [See the electronic edition of the Journal for a
color version of this qure.]
HENNAWI & PROCHASKA 740 Vol. 655
et al. 2005), in that both techniques rely on the radial clustering
of pairs of objects at xed angular positions, because the angular
selection functions are unknown (see Adelberger 2005). It is thus
worth explaining that our smaller errors arise from a few factors.
First, the relative error decreases with clustering amplitude
and our best t has a factor of 3 larger clustering strength. Sec-
ond, AS05 exclude scales R P1.2 h
1
Mpc from their analy-
sis, whereas four out of our 17 quasar-absorber pairs have R <
1.2 h
1
Mpc. Small-scale pairs are effectively worth many
large-scale pairs because the S/ N per pair is much higher. Finally,
the AS05 errors are determined from the eld-to-eld dispersion
in the data, which includes a contribution from cosmic variance.
Because the projected pairs of quasars are distributed over the en-
tire sky, our measurement does not suffer from cosmic variance
errors.
To illustrate that the errors from our Monte Carlo technique
are sensible and comparable to the AS05 result, we randomly
resampled our pair sight lines with R 1.2 h
1
Mpc to create a
mock projected pair sample 30 times larger, but with the same
distribution of redshifts and transverse distances. We increased
the search windowv 3000 km s
1
, in closer agreement with
the l 30 h
1
Mpc radial window averaged over by AS05. The
average number of quasar-absorber pairs expected from this hy-
pothetical enlarged sample is hNi
P
i
P
i
206, which is of or-
der the number of LBG-AGNpairs used by AS05. Assuming
1.6 and r
0
4.7 h
1
Mpc, our Monte Carlo simulation gives
r
0
4.7
0.9
1.0
h
1
Mpc, or a relative error of $20%, comparable
to the AS05 errors but slightly smaller.
5.1. Systematic Errors
5.1.1. Malmquist Bias
As discussed in x 2, the identication of log N
H i
19 in spec-
tra at the S/ N and resolution used in Paper I, can result in a sig-
nicant Malmquist-type bias because line blending scatters lower
column density absorbers upward, and the line density of absorb-
ers dN/dz is a steep function of column density limit. If absorbers
with column densities log N
H i
< 19 scatter up into our sample,
this error would bias our clustering measurement high. To inves-
tigate the impact of this bias, we redo the clustering analysis but
ignore the six quasar-absorber pairs in Table 1 that have column
densities within 1 o of the threshold log N
H i
19: SDSS J0256+
0039, SDSSJ0800+3542, SDSSJ0852+2637, SDSSJ1152+4517,
SDSS J1213+1207, and SDSS J1635+3013. The eleven remain-
ing quasar-absorber pairs give a maximum likelihood clustering
amplitude of C 1.7 0.7, or r
0
6.4
1.7
1.8
h
1
Mpc, compared
to our measurement of C 2.9 0.8, or r
0
9.2
1.5
1.7
h
1
Mpc,
for the full sample. This illustrates that even under very con-
servative assumptions about the possible effect of Malmquist
bias, the clustering amplitude would be only be reduced by
$1.5 o.
5.1.2. Redshift Errors
Another possible source of bias in our sample could arise
from the determination of the foreground quasar redshifts. Note
that the clustering analysis properly takes the redshift uncer-
tainties into account by averaging the correlation function over a
window of jvj 1500 km s

1
. However, assigning redshifts
to the foreground quasar (see x 4 of Paper I ) is not a completely
objective process. The quasar emission lines sometimes exhibit
mild broad absorption line ( BAL) or metal line absorption, and
the line centering can be sensitive to howthese features are masked.
It is possible that we were biased toward including quasar-absorber
pairs in our sample, and thus tended to assign redshifts result-
ing in velocity differences jvj < 1500 km s
1
. To address this
issue, we redo the analysis discarding the three quasar-absorber
pairs in Table 1 that have velocity differences larger than the
quotederror, v
fg
jvj: SDSSJ02250739, SDSSJ1426+5002,
and SDSS J1635+3013. The 14 remaining quasar-absorber pairs
give a maximumlikelihood clustering amplitude of C 2.4
0.8
0.7
,
or r
0
8.1
1.6
1.7
h
1
Mpc. Furthermore, if we discardthe ve quasar-
absorber pairs with the largest velocity differences, we measure
C 1.9 0.7, or r
0
7.1
1.6
1.7
h
1
Mpc.
Another possible source of concern is that for quasars with
large redshift errors
fg
k1000 km s
1
, the effective volume
that we searched for LLSs would be larger than the jvj <
1500 km s
1
that we quote. The actual volume searched would
be the result of the convolution of a top-hat distribution of width
jvj < 1500 km s
1
with a Gaussian with dispersion set by the
redshift error v
f g
. For the largest errors quoted in Table 1, v
fg

1500 km s
1
, this would correspond to a volume about 33%
larger, or an effective search window of jvj < 2000 km s
1
.
To determine the degree to which this biases our results, we re-
estimated the correlation function assuming a search window
of jvj < 2000 km s
1
and measured C 2.5 0.8, or r
0

8.4
1.6
1.7
h
1
Mpc. Note that this test is clearly very conservative,
since only 6 of our 17 pairs have v
fg
as large as 1500 kms
1
. For
a more typical smaller error of
fg
1000 km s
1
, for the fore-
ground quasars we obtain C 2.8 and r
0
9.0, or a 0.1 o differ-
ence in the clustering strength.
Naively one might have expected the clustering amplitude
C to change proportionally with the search volume, but in fact
our results are much less sensitive to the exact value of v
used. This is because the probability of having an absorber P
dN/dz h i1
?
(R. v)z (z 2(1 z)v/c) depends on
v directly (through z), but there is an additional dependence
from
?
(R. v). The transverse correlation function
?
decreases
with increasing v, which counters the direct increase in P from
v, and thus our likelihood analysis is very insensitive to the ef-
fective search volume.
In conclusion, we nd that the larger effective volume caused by
quasar redshift errors, or the inclusion of several quasar-absorber
Fig. 4.Binned transverse quasar-absorber correlation function for 17 quasar-
absorber pairs selected from 149 projected pair sight lines. The lines indicate the
best-t model from our maximum likelihood analysis, and the shaded regions
indicate the range allowed by 1 o errors estimated from Monte Carlo simulations.
The solid line and the medium gray shaded region are for an assumed correlation
function slope of 1.6. The steeper dotted line and dark gray shaded region are
for 2.0. The dashed curve and light gray region showthe transverse correlation
function
?
if we set the quasar-absorber correlation function
QA
to the QSO-LBG
correlation function measured by AS05 (r
0
4.7 1.3 h
1
Mpc; 1.6). [See
the electronic edition of the Journal for a color version of this qure.]
QUASARS PROBING QUASARS. II. 741 No. 2, 2007
pairs with the velocity differences jvj 1500 km s
1
, would
only reduce our measured clustering amplitude by $1 o.
6. ANISOTROPIC CLUSTERING OF ABSORBERS
AROUND QUASARS
If the clustering pattern of optically thick absorbers around
quasars is isotropic, then we can use our estimate of the quasar-
absorber correlation function in x 5 to predict the number of
proximate super-LLSs that should be observed in a given ve-
locity window, v, along the line of sight. According to equa-
tion (5), the clustering enhancement
k can be computed by
averaging
QA
over a cylinder of length v/aH(z) and cross sec-
tional area A.
However, because the cross section of the absorbers is un-
known, relating the transverse clustering to the line-of-sight clus-
tering requires an assumption about the size of the cross section.
Very little is known about the sizes of DLAs and LLSs. Briggs
et al. (1989) detected H i 21 cm absorption of a DLAwith N
H i

5 ; 10
21
cm
2
against an extended radio source, allowing them to
place a lower limit of r k12 h
1
kpc on the size of the absorbing
region. Prochaska (1999) estimated an absorbing path length of
$2 h
1
kpc for a super-LLS with log N
H i
19.1 by comparing
the column density to the total volume density, which was de-
termined from the collisionally excited C ii k1335 transition.
Adelberger et al. (2006) constrained the size of a DLAwith
log N
H i
20.4 to be k8 h
1
kpc, based on their detection of
uorescent Lyc emission from the edge of a DLA galaxy sit-
uated $8 h
1
kpc from the location of DLA absorption. Lopez
et al. (2005) measured identical column densities of N
H i

10
20.5
cm
2
for both DLAs at z 0.9313 detected in the indi-
vidual images of the gravitational lens HE 05123329, allow-
ing them to constrain the size of the DLA to be k4 h
1
kpc (see
also Smette et al. 1995). It is not clear how to relate these mea-
surements to the absorption cross sections of the log N
H i
19
of interest to us here.
To determine the cross section size as a function of limiting
column, r
abs
N
H i
aA N
H i
/
1/2
, we adopt the simple
approximation that the comoving number density of absorption-
line systems, n, and the covering factor, f
cov
, are independent of col-
umn density, which gives the simple scaling r
abs
/ hdN/dzi
1/2
from equation (1). Two families of sizes, small and large
are considered, which we believe bracket the range of possi-
bilities. Fiducial physical sizes of r
abs
5 and 20 h

1
kpc are
chosen at the DLAthreshold (log N
H i
20.3) for the small and
large absorbers, respectively. The r
abs
/ hdN/dzi
1/2
scal-
ing then predicts respective sizes of r
abs
9 and 38 h
1
kpc for
super-LLSs with log N
H i
19.
In Figure 5 we show the probability that a quasar at z 2.5
will have a proximate optically thick absorption-line systems
within v < 3000 km s
1
, as a function of limiting column den-
sity. The left (right) panel shows the prediction for the small
( large) family of cross section sizes, and the dot-dashed curves
shows the prediction in the absence of clustering (
k
0), which
is simply the integral of the column density distribution in equa-
tion (9). Since we have measured the transverse clustering only for
log N
H i
19, this gure assumes that clustering is independent
of limiting column density.
The transverse clustering overpredicts the fraction of quasars
that have a proximate absorber by a large factor. For example, for
small absorption cross sections and 1.6, the quasar-absorber
correlation function measured from the transverse direction
predicts that a fraction P 0.30 0.07 of all quasars should
show a proximate super-LLS ( log N
H i
19) within v
3000 km s
1
. Even the AS05 clustering amplitude, which is a fac-
tor of $3 smaller than our best t, would predict P 0.12
0.08
0.06
.
A steeper correlation function results in even more proximate
absorbers. For the best-t transverse clustering amplitude with
2, the probability would be P 0.49
0.17
0.10
.
Making the absorbers larger changes this prediction by factors
of $2Y3. The large absorbers predict that a fraction P 0.14
0.03 of quasars should have a proximate absorber, given our
best-t clustering amplitude for 1.6. Our steeper 2 t
gives P 0.14
0.02
0.04
, and the AS05 result predicts P 0.07
0.03. Although the line density of proximate super-LLSs near
quasar has yet to be measured, it is incontrovertible that 15%Y50%
of quasars do not show a super-LLS within v < 3000 km s
1
along the line of sight.
J. X. Prochaska et al. (2007, in preparation) recently measured
the line density of proximate DLAs (log N
H i
20.3) with v <
3000 km s
1
from a sample of $100 of these systems selected
Fig. 5.Probability of a quasar at z $ 2.5 having a proximate absorber within 3000 km s
1
as a function of limiting column density. The left panel is for small
absorber cross sections and the right panel is for large absorbers. Our measurement of the quasar-absorber correlation function
QA
from the transverse direction is used
to predict the number of proximate absorbers along the line of sight. The curves and shaded regions show the predictions and 1 o errors for the quasar-absorber
correlation function
QA
estimated fromthe transverse direction in x 5, as well as the clustering strength measured by AS05. The line styles and shading correspond to the
same models as in Fig. 4. The lower dotted line indicates the prediction in the absence of clustering, which is just the integral of the column density distribution in eq. (9).
The point with the error bar represents the J. X. Prochaska et al. (2007, in preparation) measurement for proximate DLAs (log N
H i
20.3) at z $ 2.5. [See the electronic
edition of the Journal for a color version of this qure.]
HENNAWI & PROCHASKA 742 Vol. 655
from $6000 quasars from the SDSS DR5. They found that the
number of proximate DLAs was enhanced by a factor of 1.4
0.1 over the expectation fromthe statistics of intervening absorb-
ers ( Prochaska et al. 2005), over the redshift range 2.2 Pz P4.
At lower redshifts z $ 2.5, characteristic of the mean of our pro-
jected pair sample, they found $10 proximate DLAs out of
$2000 SDSS quasars with z < 2.7, corresponding to an over-
density of 1.2 0.15 over the expectation from the cosmic
average. The point with the error bar in Figure 5 indicates the
probability of having a proximate DLA from the J. X. Prochaska
et al. (2007, in preparation) measurement, P
DLA
0.0095
0.001. Our best-t clustering amplitude predicts that P 0.12
0.3 (P 0.25
0.09
0.05
) of quasars should have a nearby DLA for
small absorbers and 1.6 ( 2). Large absorbers change this
prediction to P 0.055 0.013 (P 0.069
0.023
0.013
).
The caveat should be included that our prediction for the num-
ber of proximate absorbers from the transverse clustering is very
sensitive to the small-scale behavior of the correlation function.
Specically, using equation (5) to predict the line-of-sight clus-
tering implicitly assumes that the power-law model of the corre-
lation function, / r

, is valid down the lower limit of the

integration, Z
cut
, which we have set to be the diameter of the
absorbers. For super-LLSs, we used (proper) radii of r
abs
9
and 38 h
1
kpc, for small and large absorbers, respectively, cor-
responding to (comoving) Z
cut
63 and 266 at z 2.5. The
smallest transverse separation of our projected pair sample is
R 72 h
1
kpc, and there are just ve sight lines with R P
300 h
1
kpc, three of which have an absorber (see Fig. 1 and
Table 1). The transverse measurement assumes a single power
law from R 70 h
1
kpc to 5 h
1
Mpc. One should bear in mind
that only a handful of projected pairs probe the small scales (R P
300 h

1
kpc) that the line-of-sight clustering is very sensitive to,
but it is reassuring that closest bin in Figure 4 is consistent with the
power-law t.
We have shown that the transverse clustering that we quanti-
ed in x 5 overpredicts the abundance of proximate absorbers
along the line of sight by a large factor, $4Y20, under reasonable
assumptions about the sizes of the cross sections of these absorb-
ers. The clustering pattern of absorbers around quasars is thus
highly anisotropic. The most plausible explanation for this ani-
sotropy is that the transverse direction is less likely to be illu-
minated by ionizing photons than the line of sight and that the
optically thick absorbers along the line of sight are being photo-
evaporated. We discuss the physical effects that could give rise to
this anisotropy in x 8. Next, we introduce a simple model that
provides physical insight into the problem of optically thick ab-
sorbers subject to the intense ionizing ux of a nearby quasar.
7. PHOTOEVAPORATION OF OPTICALLY
THICK CLOUDS
The problem of an optically thick absorption-line system ex-
posed to the ionizing ux of a nearby luminous quasar is analo-
gous to that of a neutral interstellar cloud being exposed to the
ionizing radiation of an OB star, a problem that was rst inves-
tigated by Oort & Spitzer (1955). Bertoldi (1989) classied the
behavior of photoevaporating clouds based on their initial column
density and the ionization parameter at the location of the cloud,
and he developed an analytical solution to follow the radiation-
driven implosion phase of the cloud. Interestingly, although
Bertoldi (1989) was primarily concerned with the fate of inter-
stellar clouds near OB stars in H ii regions, he commented briey
on the applicability of the same formalism to Lyc clouds exposed
to the ionizing ux of a quasar.
7.1. Cloud Zappinq
Following Bertoldi (1989), we model an optically thick ab-
sorber as a homogeneous spherical neutral gas cloud with total
number density of hydrogen n
H
that is embedded in photoion-
ized intergalactic medium at temperature T 20. 000 K, corre-
sponding to an isothermal sound speed c
i
16.5T
1/2
20
km s
1
,
where T
20
is the temperature in units of 20,000 K. If the cloud is
at a distance r from a luminous quasar that is emitting S ionizing
photons per second, then the ionizing ux at the cloud surface,
F
i
S/4r
2
, will drive an ionization front into the neutral gas. If
the ux is large enough to make the conditions R-type (see Spitzer
1978 for a discussion of how ionization fronts are classied), an
R-type ionization front will propagate through the cloud without
dynamically perturbing the neutral gas. As the front propagates
into the cloud, it ionizes an increasing column of neutral gas,
steadily reducing the ionizing ux at the front, until the condi-
tions become M-type, at which point the front will stall and drive
a shock into the neutral upstream gas. This shock will implode
the cloud and compress it until conditions become D-type, allow-
ing the front to continue propagating and establishing a steady
photoevaporation ow (Bertoldi 1989).
The cloud will be zapped, or completely photoevaporated,
if the ionizing ux is large enough relative to the column density
such that the entire cloud can be ionized in a recombination time.
In this case, the R-type front will completely cross the cloud with-
out stalling. Part of the cloud will remain neutral and a shock will
form provided that
c 494N
1
H. 20.3
1.1 ; 10
4
T
1,2
20
h i
< 1. 10
where N
H. 20.3
is the total hydrogen column in units of 10
20.3
cm
2
,
and the ionization parameter is dened by F
i
/n
0
c, with
2.58 ; 10
5
S
56
r
2
Mpc
n
1
H.1
. 11
where S
56
is the ionizing ux in units of 10
56
s
1
, r
Mpc
is the phys-
ical distance in units of Mpc, and n
H.1
is the total hydrogen num-
ber density in units of 1 cm
3
. The condition that the ionization
front be R-type at the surface of the cloud is 1.1 ; 10
4
T
1/2
20
.
Consider our ducial example of a foreground quasar with
r 19 at an angular separation of 0 1
0
from an absorber,
corresponding to a transverse proper distance of 485 kpc. The
ionizing ux is enhanced by q
UV
130 over the UVbackground
and S
56
5.2. For a DLAwith total hydrogen column of N
H

10
20.3
cm
2
at this distance, the cloud will survive (c < 1) pro-
vided n
H
0.27 cm
3
, which would give < 0.002.
The left panel of Figure 6 shows the lower limits on the volume
density of a DLAwith neutral hydrogen column log N
H i
20.3,
set by the condition for cloud survival, as a function of physical
distance from a quasar at z 2.5. The dashed, solid, and dotted
curves correspond to r 17, 19, and 21 mag, respectively. In the
region beloweach curve, the volume densities are too small and
the clouds are photoevaporated, whereas above the curves the
clouds can survive. The right panel shows the lower limit on the
neutral hydrogen column density as a function of distance, as-
suming a volume density n
H
0.1. If DLAs with log N
H i
20.3
have sizes in the range r
abs
$ 1Y5 kpc their corresponding num-
ber densities are n
H
$ N
H
/r
abs
0.01Y1. Thus according to Fig-
ure 6 optically thick absorbers with n
H
P0.1 will be photo-
evaporated if they lie within $1 Mpc of a luminous quasar.
Note that the condition in equation (6) considers the total col-
umn density log N
H
, not the neutral column, log N
H i
. Although
DLAs are expected to be predominantly neutral N
H i
% N
H
, the
QUASARS PROBING QUASARS. II. 743 No. 2, 2007
ionic absorption lines typically observed in LLSs suggest that
they are the photoionized analogs of DLAs ( Prochaska 1999)
and that their ionization state is determined by ionization equilib-
rium with the UV background. In equation (10), we have as-
sumed that optically thick absorbers are spherical top-hat density
distributions. Hence, an LLS with column density N
H i
P20 can
be thought of as a sight line that passes through the photoionized
outskirts of a top-hat cloud that has a larger total hydrogen col-
umn N
H
, and the total column determines the ability of the cloud
to survive a blast of ionizing radiation from the quasar. Thus for
N
H i
P20, we compute an ionization correction with the standard
approach, which is to assume a slab geometry and determine the
ionization balance in a uniform background, using a photoion-
ization code such as Cloudy ( Ferland et al. 1998). To construct
the curve in the right panel of Figure 6, we interpolated through
a grid of Cloudy solutions
7
for N
H i
(N
H
).
7.2. A Toy Model to Predict the Statistics of Proximate DLAs
In this section we use a toy model to illustrate how quasar-
absorber clustering can be used to constrain the physical properties
of optically thick absorbers. The criterion c < 1 in equation (10)
gives a minimumdistance fromthe quasar R
SS
(n
H
), as a function
of volume density, at which an optically thick absorption-line
system with column density N
H i
can survive. Our approach is to
simply assume that absorbers at smaller distances are photoevap-
orated and absorbers at larger distances survive. We also assume
that the transverse direction is not illuminated by the quasar, and
hence the transverse clustering measures the intrinsic quasar-
absorber clustering, in the absence of ionization effects. Because
proximate absorbers are denitely illuminated, this intrinsic clus-
tering is then reduced along the line of sight by photoevaporation.
For the column density range log N
H i
20.3, we evaluate R
SS
at the lower limit log N
H i
20.3, which is a decent approxima-
tion because the column density distribution is steep and the
statistics will be dominated by absorbers near the threshold. To
simplify the computation, we take R
SS
to be a distance only along
the line of sight, which is valid provided that R
SS
3A
1/2
. Then
we can write that the line density of absorbers within v <
3000 km s
1
is
dN
dz

dN
dz
( )
1
aR
SS
H(z)
v

k
(v)
!
. 12
where
k
is given by equation (5) but with Z
cut
R
SS
. The as-
sumption that the transverse direction gives the intrinsic clus-
tering in the absence of ionization effects amounts to using the
correlation function
QA
, measured from the transverse cluster-
ing, in the line-of-sight integral in equation (5).
In Figure 7 we showour toy model prediction for the probabil-
ity of a quasar having a proximate DLA with log N
H i
20.3
within v < 3000 km s
1
as a function of total volume density
of hydrogen n
H
. We assumed that the DLAs have a size of r
abs

5 h
1
kpc, characteristic of the small absorbers discussed in x 6.
However, identical results are obtained for large absorbers (r
abs

20 h
1
kpc). Because our toy model excludes small scales r
abs
T
R
SS
fromthe clustering integral in equation (5), we are again in the
far-eld limit where the the volume average is nearly independent
of the absorber cross section. This independence breaks down for
very large densities n
H
k10
2
, where R
SS
approaches the size of
the absorbers (see Fig. 7). The curves and shaded regions showthe
predictions and 1 o errors for the transverse clusteringmeasured in
x 5, as well as the clustering strength measured by AS05. We as-
sumed a quasar with magnitude r 19.1 at z 2.5, chosen to
match the mean magnitude and redshift of the proximate DLA
sample of J. X. Prochaska et al. (2007, in preparation). The long-
dashed horizontal lines indicate the measurement and 1 o range
measured by J. X. Prochaska et al. (2007, in preparation) of
P
DLA
0.0095 0.001.
Our toy model suggests that DLAs with volume densities in
the range 10
3
cm
3
Pn
H
P10
1
cm
3
are required to agree
with the measurement of J. X. Prochaska et al. (2007, in prep-
aration). Even the weaker AS05 clustering of LBGs around
quasars would require n
H
P1 cm
3
. For larger volume densities,
DLAs illuminated by quasars would survive at smaller radii where
the clustering is strong, giving rise to a larger number of proximate
DLAs. Note that the comparison of our toy model to the J. X.
Prochaska et al. (2007, in preparation) measurement in Figure 7
assumes that the clustering of absorbers is independent of column
Fig. 6.Regions of parameter space where an optically thick absorber can survive near a luminous quasar set by eq. (10). The dashed, solid, and dotted curves correspond to
r 17, 19, and 21 mag, respectively. Optically thick absorbers above a curve can survive, whereas those below are photoevaporated. Left : Survival regions in the volume
densityYdistance plane, assuming a neutral column of log N
H i
20.3. Riqht : Survival regions in the neutral column densityYdistance plane, assuming a volume density of
n
H
0.1 cm
3
. [See the electronic edition of the Journal for a color version of this qure.]
7
The photoionization models used Cloudy version 6.0.2 and an F. Haardt &
P. Madau (2007, in preparation) QSO+galaxies spectrumat z 2.5 and assumed a
metallicity of 1.5 and J
912
21.2.
HENNAWI & PROCHASKA 744 Vol. 655
density threshold (i.e., our measurement is for log N
H i
19).
This assumptions was made because our quasar-absorber sample
did not have sufcient statistics to measure the quasar-absorber
clustering for DLAs alone (see x 5).
Although crude, this model illustrates how a comparison of
line-of-sight and transverse quasar absorber clustering can be used
to determine the density distribution in optically thick absorbers.
Detailed models of self-shielding with radiative transfer (Zheng &
Miralda-Escude 2002a; Cantalupo et al. 2005; J. A. Kollmeier
et al. 2007, in preparation) would be required for a more accurate
treatment, and such analyses could easily include cuspy density
proles or a multiphase distribution of gas. Although better sta-
tistics and more theoretical work are necessary, the clustering
of optically thick absorbers around quasars will provide im-
portant newconstraints on the physical nature of these systems.
8. SUMMARY AND DISCUSSION
8.1. Summary
In this paper we used a sample of 17 super-LLSs (log N
H i
< 19)
selected from 149 projected quasar pairs sight lines in Paper I to
investigate the clustering pattern of optically thick absorbers around
quasars. Based on these data, we nd the following results:
1. A simple formalism is presented for quantifying the clus-
tering of absorbers around quasars in both the transverse and
line-of-sight directions. The clustering of absorbers around qua-
sars in the transverse direction is independent of the size of the
absorption cross section, whereas the line-of-sight clustering was
shown to be sensitive to the cross section size.
2. Applying this formalism to the 17 super-LLSs (log N
H i
<
19) selected from149 projected quasar pair sight lines with mean
redshift z 2.5, we determine a comoving correlation length of
r
0
9.2
1.5
1.7
h
1
Mpc for a power-law correlation function with
1.6. This is 3 times stronger than the clustering of LBGs
around quasars recently measured by AS05. If we assume a
steeper slope of 2.0, we measure r
0
5.8
1.0
0.6
h
1
Mpc.
3. The clustering of optically thick absorbers around quasars
is highly anisotropic. If we apply the clustering amplitude mea-
sured in the transverse direction to the line of sight, the frac-
tion of quasars that have a proximate absorber within v <
3000 km s
1
is overpredicted by a factor as large as $4Y20, de-
pending on assumptions about cross section sizes and the slope
of the correlation function. The most plausible explanation for
the anisotropy is that the transverse direction is less likely to be
illuminated by ionizing photons than the line of sight and that
the optically thick absorbers along the line of sight are being
photoevaporated.
4. A simple model of absorbers as uniform spherical over-
densities is discussed and we give an analytic criterion that de-
termines whether an absorber illuminated by a quasar will be
able to self-shield. This criterion indicates that optically thick ab-
sorbers with n
H
P0.1 will be photoevaporated if they lie within
$1 Mpc of a luminous quasar. We combine this criterion with a
toy model of the effect of photoevaporation on line-of-sight
clustering to illustrate how comparisons of the line-of-sight and
transverse clustering around quasars can ultimately be used to
constrain the distribution of gas in optically thick absorption-line
systems. A similar experiment applied to Mg ii, C iv, or other
metal absorption-line systems near quasars ( Bowen et al. 2006;
G. E. Prochter et al. 2007, in preparation) would also yield val-
uable insights into their physical nature.
8.2. Discussion
The anisotropic clustering pattern of absorbers around quasars
suggests that the transverse direction is less likely to be illumi-
nated by ionizing photons than the line of sight. This suggestion
gains credibility in light of the null detections of the transverse
proximity effect in the Lyc forests of projected quasar pairs
(Crotts 1989; Dobrzycki &Bechtold 1991; Fernandez-Soto et al.
1995; Liske & Williger 2001; Schirber et al. 2004; Croft 2004;
but see also Jakobsen et al. 2003). Although these studies are each
based only on a handful of projected pairs, they all come to similar
conclusions: the amount of (optically thin) Lyc forest absorption
in the background quasar sight line near the redshift of the fore-
ground quasar is larqer than average rather than smaller, the op-
posite of what is expected from the transverse proximity effect.
Two physical effects can explain both the optically thin results
and our result for optically thick systems: anisotropic emission or
variability, which we discuss in turn.
If quasar emission is highly anisotropic, the line of sight would
be exposed to the ionizing ux of the quasar, whereas transverse
absorbers would be more likely to lie in shadowed regions.
Studies of Type II quasars and the X-ray background suggest that
quasars with luminosities comparable to our foreground quasar
sample (M
B
< 23) have $30% of the solid angle obscured
( Ueda et al. 2003; Barger et al. 2005; Treister & Urry 2005),
although these estimates are highly uncertain. Naively, we would
expect the covering factor of transverse absorbers to be approxi-
mately equal to the average fraction of the solid angle obscured.
But in Paper I we found a very high covering factor (6/8) for hav-
ing an optically thick absorber with log N
H i
17.2 (see Fig. 1 of
Paper I ) on the smallest (proper) scales R < 150 h
1
kpc. Al-
though the statistics are clearly very poor, this high covering
factor is suggestive of a signicantly larger obscured fraction.
If the ionizing ux of the foreground quasar varies consider-
ably on a timescale shorter than the transverse light crossing time
Fig. 7.Toy model prediction for the probability of a quasar at z $ 2.5 with
r 19.1 having a proximate DLAwithin 3000 km s
1
as a function of volume
density. The upper x-axis shows self-shielding radius, which is the minimum
distance fromthe quasar at which a DLAcan survive for a given volume density.
The curves and shaded regions show the predictions and 1 o errors for the quasar-
absorber correlation function
QA
estimated from the transverse direction in x 5, as
well as the clustering strength measured by AS05. The line styles and shading cor-
respond to the same models as in Fig. 4. The long-dashed horizontal lines indicate
the measurement and 1 o range measured by J. X. Prochaska et al. (2007, in
preparation) P
DLA
0.0095 0.001. The short dotted horizontal line is the
cosmic average (i.e., no clustering) measured by Prochaska et al. (2005). This
model suggests that DLAs with volume densities in the range 10
3
cm
3
Pn
H
P
10
1
cm
3
are required to agree with the proximate DLA measurement of J. X.
Prochaska et al. (2007, in preparation). [See the electronic edition of the Journal
for a color version of this qure.]
QUASARS PROBING QUASARS. II. 745 No. 2, 2007
between the foreground and background sight lines, a transverse
proximity effect might not be observable. This is because the ion-
ization state of the gas along the transverse sight lines is sensitive
to the foreground quasars luminosity a light crossing time before
the light that we observe was emitted. At 60
00
(1.2 h
1
Mpc) from
a z 2.5 quasar the transverse light crossing time is 1.1 ; 10
6
yr.
Currently, the lower limit on the intermittency of quasar emis-
sion comes from observations of the (optically thin) proximity
effect ( Bajtlik et al. 1988; Scott et al. 2000) in the Lyc forests
near quasars. The presence of an optically thin proximity effect
implies that the IGM has had time to reach ionization equilibrium
with the quasars increased ionizing ux, which requires that the
duration of a burst of quasar radiation is longer than the IGMequi-
libration time, t
burst
k10
4
yr ( Martini 2004). The photoevap-
oration timescale for an optically thick absorber is $N
H i
/F or the
light crossing time, whichever is longer, where F is the ionizing
ux. At a distance of 100 kpc froman i 19 quasar, it would take
15,000 yr to photoevaporate a DLA with log N
H i
20.3, i.e.,
comparable to the light crossing time if DLAs have sizes of
$5 kpc. Thus the time it takes for the line of sight to manifest
the effects of the quasars ionizing ux is $10
4
yr for both the
optically thin and optically thick regimes. Hence, if quasars
emit in bursts of duration 10
4
yr Pt
burst
P10
6
yr, this would
sufce to explain the absence of the optically thin transverse
proximity effect as well as the anisotropic clustering pattern of
optically thick absorbers discovered here.
The optically thin proximity effects are probably the most
promising way to disentangle whether obscuration or variability
can explain why the transverse direction is less likely to be illu-
minated by ionizing photons (Schirber et al. 2004; Croft 2004).
This is because for the lower column density absorbers charac-
teristic of the Lyc forest, cosmological N-body simulations can
predict the statistical properties of the absorbers ab initio without
the complications of radiative transfer.
It might seempuzzling that we measure a clustering amplitude
that is a factor of 3 larger than the clustering of LBGs around qua-
sars at similar redshift measured by AS05. One could argue that
on small scales we are probing material intrinsic to the quasar. If
interpreted as clustering, this intrinsic contribution would lead us
to overestimate the correlation function. However, only our clos-
est sight line (R 22 h
1
kpc, proper) is sufciently small for this
to be a worry, and this would imply a very large cross section for
DLA absorption. AS05 excluded small scales fromtheir analysis,
whereas the strength of our signal is in part driven by the small-
scale systems. If we exclude sight lines with R < 500 h
1
kpc,
we measure a r
0
7.6
1.9
2.0
h
1
Mpc that is within1 o of the AS05
measurement. It is also possible that we measured a larger clus-
tering signal because the galaxies that host optically thick ab-
sorbers are more strongly biased with respect to the dark matter
distribution than LBGs ( but see Cooke et al. 2006).
By comparing our transverse quasar-absorber clustering strength
to the dark matter correlation function and the autocorrelation of
quasars, we can determine the bias of super-LLSs hosts b
SLLS

QA
/(
DM

QSO
)
1/2
. By comparing this bias to the bias of dark
matter halos b(M ) ( Efstathiou et al. 1988; Cole & Kaiser 1989;
Mo & White 1996; Sheth & Tormen 1999) a host halo mass can
be estimated. However, without modeling the occupation of super-
LLSs in dark matter halos this mass estimate is only good to a
factor of $3, but is nevertheless a good check on the sensibility
of our clustering measurement. If we interpolate between the
clustering strength ( Porciani et al. 2004) of quasars at z < 2.1
and the strong and rapidly evolving high-redshift, z 2.9, qua-
sar clustering measured by Shen et al. (2007) and average over
the redshift distribution of our projected pair sample, we nd that
the effective correlation length of quasars at the mean redshift
of our sample (z $ 2.5) is r
0
11.2 h
1
Mpc, for a correlation
function
QSO
/ r

with slope 2. Evaluating the bias on

the scale r 3 h
1
Mpc using the formula above, we nd b
LLS

3.0, given our transverse measurement of
QA
(r
0
9.2 h
1
Mpc
for 1.6) or a lower value of b
LLS
1.6for the steeper slope
2 (r
0
5.8 h
1
Mpc). These bias values correspond to host
halo masses of M 2 ; 10
12
h
1
M

and M 2 ; 10
11
h
1
M

,
respectively, using the b(M ) formula from Sheth & Tormen
(1999). This bias and mass scale for super-LLSs is sensible, and
this calculation suggests that the large clustering amplitude that
we measured arises primarily because of the very strong cluster-
ing of quasars at the redshifts of our sample. In light of this, it is
puzzling that Adelberger & Steidel (2005) did not measure a
stronger cross-correlation between LBGs and quasars, since the
redshift range of their quasar-LBG pair sample is comparable to
ours, but slightly higher.
We argued that the transverse clustering predicts that galaxies
correlated with a quasar should give rise to an absorption-line
systems with log N
H i
19 within v < 3000 km s
1
, a signi-
cant fraction of the time $15%Y50%. We postulated that these
systems are not observed because of photoevaporation, but should
not a similar argument also apply to galaxies? High-redshift gal-
axy spectra should show optically thick absorption due to nearby
correlated galaxies in addition to intrinsic absorption due to H i in
the galaxy itself. The individual spectra of high-redshift galax-
ies rarely have high enough signal-to-noise ratio to detect even
DLAs. However, the break due to Lyman limit absorption can
be detected in very deep integrations, with the added advantage
that the LLSs are much more abundant. Recently, Shapley et al.
(2006) detected emission blueward of the Lyman limit in 2 out
of 14 LBGs at z $ 3. For a galaxy-absorber correlation length
in the range r
0
5Y10 h
1
Mpc, equation (2) would predict that
25%Y75%of galaxies should showa correlated absorber within a
window of v < 500 km s
1
, comparable to the redshift uncer-
tainties of LBGS. This number assumes LLSs with log N
H i

17.2 have a size of $50 h
1
kpc, and we used the abundance of
LLSs measured by Peroux et al. (2003). While it is possible that
the lack of ux blueward of the Lyman limit in LBGs is due to
absorption by gas intrinsic to the LBG, correlated absorbers along
the line of sight could also play a signicant role.
Using the ionizing ux of a quasar to study the distribution of
neutral gas in optically thick absorbers is a powerful new way to
study these absorption-line systems. We have showed how com-
parisons of the line-of-sight and transverse clustering of absorb-
ers around quasars constrains their distribution of gas. Theoretical
models of LLSs and DLAs that include radiative transfer and self-
shielding ( Zheng & Miralda-Escude 2002a; Cantalupo et al.
2005; J. A. Kollmeier et al. 2007, in preparation) should ex-
plore how morphology, cuspy density proles, or a multiphase
mediumchange the ability of absorbers to survive near quasars.
Optically thick H i clouds in ionization equilibrium with a ra-
diation eld reemit $60%of the ionizing photons they absorb as
uorescent Lyc recombination line photons (Gould &Weinberg
1996; Zheng & Miralda-Escude 2002b; Cantalupo et al. 2005;
Adelberger et al. 2006; J. A. Kollmeier et al. 2007, in prepara-
tion). Recently, Adelberger et al. (2006) reported a detection of
uorescence froma serendipitously discovered DLA(log N
H i

20.4) situated 49
00
away from a luminous quasar at z 2.84.
Fluorescent Lyc emission from absorbers illuminated by qua-
sars offers a new window to study physical properties of ab-
sorbers, as well as a second transverse sight line to observe the
HENNAWI & PROCHASKA 746 Vol. 655
ionizing ux of a quasar. In Paper I we published seven newquasar-
absorber pairs that would have uorescent surface brightnesses of
j
Lyc
19.5Y24.3 mag arcsec
2
if the foreground quasars emit
isotropically. However, the clustering anisotropy discussed here
suggests that these transverse absorbers may not be illuminated,
and the lack of a detection of uorescent emission would provide
compelling evidence in favor of this conclusion. Constraints on
uorescent emission from these systems will be discussed in the
next paper in this series (J. F. Hennawi et al. 2007, in prepara-
tion). It is particularly intriguing that proximate DLAs observed
along the line of sight seem to preferentially exhibit Lyc emis-
sion superimposed on the Lyc absorption trough ( Mller et al.
1998; Ellison et al. 2002). Is this Lyc emission uorescence? Is
the uorescent emission from proximate DLAs observable?
In Paper I we argued that a factor of $20 more transverse
quasar-absorber pairs could be compiled in a modest amount of
observing time, which would improve our measurement of the
transverse correlation function by a factor of $5. The study of
J. X. Prochaska et al. (2007, in preparation) will yield a sample of
k100, which should be sufcient to measure the column density
distribution f (N
N
H i
) of absorbers near quasars. Any differences
between it and the distribution in the eld would provide an
important new constraint on the physical nature of optically thick
absorbers. Finally, we add that similar studies using projected
quasar pairs and searching for line-of-sight proximate absorbers
can be conducted for Mg ii, C iv, and other metal absorption lines
systems near quasars (Bowen et al. 2006; G. E. Prochter et al.
2007, in preparation), yielding similar insights into their physical
nature.
Large samples of optically thick absorption-line systems near
quasars are well within reach, for transverse systems using qua-
sar pairs as well as proximate absorbers along the line of sight.
This data will provide new opportunities to characterize the en-
vironments of quasars and the physical nature of absorption-line
systems, and it will uncover new laboratories for studying uo-
rescent emission from optically thick absorbers.
We are grateful to Jordi Miralda-Escude, Juna Kollmeier,
Piero Madau, and Zheng Zheng for reading an early version of
this manuscript and providing critical comments. We thank John
OMeara, Yue Shen, and Michael Strauss for sharing results prior
to publication and for helpful discussions. J. F. H. acknowledges
enlightening discussions with Kurt Adelberger, Doron Chelouche,
Bruce Draine, Sara Ellison, Taotao Fang, Gabe Prochter, Chris
McKee, Brice Menard, David Russell, Alice Shapley, and Michael
Strauss. J. F. H. is supported by NASA through Hubble Fellow-
ship grant 01172.01-A awarded by the Space Telescope Science
Institute, which is operated by the Association of Universities
for Research in Astronomy, Inc., for NASA, under contract NAS
5-26555. J. X. P. wishes toacknowledge fundingthroughNSFgrant
AST-0307408. The conclusions of this work are based on data
collected fromobservatories at the summit of Mauna Kea. The au-
thors wish to recognize and acknowledge the very signicant cul-
tural role and reverence that the summit of Mauna Kea has always
had within the indigenous Hawaiian community. We are most
fortunate to have the opportunity to conduct observations from
this mountain.
REFERENCES
Adelberger, K. L. 2005, ApJ, 621, 574
Adelberger, K. L., & Steidel, C. C. 2005, ApJ, 630, 50 (AS05)
Adelberger, K. L., Steidel, C. C., Kollmeier, J. A., & Reddy, N. A. 2006, ApJ,
637, 74
Adelberger, K. L., Steidel, C. C., Pettini, M., Shapley, A. E., Reddy, N. A., &
Erb, D. K. 2005, ApJ, 619, 697
Adelberger, K. L., Steidel, C. C., Shapley, A. E., & Pettini, M. 2003, ApJ, 584,
45
Bahcall, J. N., Schmidt, M., & Gunn, J. E. 1969, ApJ, 157, L77
Bahcall, N. A., & Chokshi, A. 1991, ApJ, 380, L9
Bajtlik, S., Duncan, R. C., & Ostriker, J. P. 1988, ApJ, 327, 570
Barger, A. J., Cowie, L. L., Mushotzky, R. F., Yang, Y., Wang, W.-H., Steffen,
A. T., & Capak, P. 2005, AJ, 129, 578
Bertoldi, F. 1989, ApJ, 346, 735
Bouche, N., & Lowenthal, J. D. 2004, ApJ, 609, 513
Bowen, D. V., et al. 2006, ApJ, 645, L105
Briggs, F. H., Wolfe, A. M., Liszt, H. S., Davis, M. M., & Turner, K. L. 1989,
ApJ, 341, 650
Brown, M. J. I., Boyle, B. J., & Webster, R. L. 2001, AJ, 122, 26
Cantalupo, S., Porciani, C., Lilly, S. J., & Miniati, F. 2005, ApJ, 628, 61
Coil, A. L., Newman, J. A., Cooper, M. C., Davis, M., Faber, S. M., Koo,
D. C., & Willmer, C. N. A. 2006a, ApJ, 644, 671
Coil, A. L., et al. 2006b, ApJ, 644, 671
Cole, S., & Kaiser, N. 1989, MNRAS, 237, 1127
Conroy, C., Wechsler, R. H., & Kravtsov, A. V. 2006, ApJ, 647, 201
Cooke, J., Wolfe, A. M., Gawiser, E., & Prochaska, J. X. 2006, ApJ, 636, L9
Croft, R. A. C. 2004, ApJ, 610, 642
Croom, S. M., Smith, R. J., Boyle, B. J., Shanks, T., Miller, L., Outram, P. J., &
Loaring, N. S. 2004, MNRAS, 349, 1397
Crotts, A. P. S. 1989, ApJ, 336, 550
Dobrzycki, A., & Bechtold, J. 1991, ApJ, 377, L69
Efstathiou, G., Frenk, C. S., White, S. D. M., & Davis, M. 1988, MNRAS, 235,
715
Ellison, S. L., Yan, L., Hook, I. M., Pettini, M., Wall, J. V., & Shaver, P. 2002,
A&A, 383, 91
Ferland, G. J., Korista, K. T., Verner, D. A., Ferguson, J. W., Kingdon, J. B., &
Verner, E. M. 1998, PASP, 110, 761
Fernandez-Soto, A., Barcons, X., Carballo, R., & Webb, J. K. 1995, MNRAS,
277, 235
Gawiser, E., Wolfe, A. M., Prochaska, J. X., Lanzetta, K. M., Yahata, N., &
Quirrenbach, A. 2001, ApJ, 562, 628
Gould, A., & Weinberg, D. H. 1996, ApJ, 468, 462
Haardt, F., & Madau, P. 1996, ApJ, 461, 20
Hennawi, J. F., et al. 2006a, AJ, 131, 1
. 2006b, ApJ, 651, 61 ( Paper I )
Jakobsen, P., Jansen, R. A., Wagner, S., & Reimers, D. 2003, A&A, 397, 891
Lee, K.-S., Giavalisco, M., Gnedin, O. Y., Somerville, R. S., Ferguson, H. C.,
Dickinson, M., & Ouchi, M. 2006, ApJ, 642, 63
Liske, J., & Williger, G. M. 2001, MNRAS, 328, 653
Lopez, S., Reimers, D., Gregg, M. D., Wisotzki, L., Wucknitz, O., & Guzman,
A. 2005, ApJ, 626, 767
Martini, P. 2004, in Coevolution of Black Holes and Galaxies, ed. L. C. Ho
(Cambridge: Cambridge Univ. Press), 169
Meiksin, A. 2005, MNRAS, 356, 596
Miralda-Escude, J. 2003, ApJ, 597, 66
Mo, H. J., & White, S. D. M. 1996, MNRAS, 282, 347
Mller, P., Warren, S. J., Fall, S. M., Fynbo, J. U., & Jakobsen, P. 2002, ApJ,
574, 51
Mller, P., Warren, S. J., & Fynbo, J. U. 1998, A&A, 330, 19
OMeara, J. M., et al. 2007, ApJ, in press (astro-ph/0610726)
Oort, J. H., & Spitzer, L. J. 1955, ApJ, 121, 6
Ouchi, M., et al. 2005, ApJ, 635, L117
Peroux, C., Dessauges-Zavadsky, M., DOdorico, S., Sun Kim, T., & McMahon,
R. G. 2005, MNRAS, 363, 479
Peroux, C., McMahon, R. G., Storrie-Lombardi, L. J., & Irwin, M. J. 2003,
MNRAS, 346, 1103
Porciani, C., Magliocchetti, M., & Norberg, P. 2004, MNRAS, 355, 1010
Prochaska, J. X. 1999, ApJ, 511, L71
Prochaska, J. X., Herbert-Fort, S., & Wolfe, A. M. 2005, ApJ, 635, 123
Richards, G. T., Vanden Berk, D. E., Reichard, T. A., Hall, P. B., Schneider,
D. P., SubbaRao, M., Thakar, A. R., & York, D. G. 2002, AJ, 124, 1
Russell, D. M., Ellison, S. L., & Benn, C. R. 2006, MNRAS, 367, 412
Schaye, J. 2001, ApJ, 559, 507
Schirber, M., Miralda-Escude, J., & McDonald, P. 2004, ApJ, 610, 105
Scott, J., Bechtold, J., Dobrzycki, A., & Kulkarni, V. P. 2000, ApJS, 130, 67
Serber, W., Bahcall, N., Menard, B., & Richards, G. 2006, ApJ, 643, 68
Shapley, A. E., et al. 2006, ApJ, 651, 688
Shen, Y., et al. 2007, AJ, submitted
QUASARS PROBING QUASARS. II. 747 No. 2, 2007
Sheth, R. K., & Tormen, G. 1999, MNRAS, 308, 119
Smette, A., Robertson, J. G., Shaver, P. A., Reimers, D., Wisotzki, L., &
Koehler, T. 1995, A&AS, 113, 199
Smith, R. J., Boyle, B. J., & Maddox, S. J. 2000, MNRAS, 313, 252
Spergel, D. N., et al. 2003, ApJS, 148, 175
Spitzer, L. 1978, Physical Processes in the Interstellar Medium ( New York:
Wiley)
Treister, E., & Urry, C. M. 2005, ApJ, 630, 115
Ueda, Y., Akiyama, M., Ohta, K., & Miyaji, T. 2003, ApJ, 598, 886
Yee, H. K. C., & Green, R. F. 1984, ApJ, 280, 79
. 1987, ApJ, 319, 28
Zheng, Z., & Miralda-Escude, J. 2002a, ApJ, 568, L71
. 2002b, ApJ, 578, 33
HENNAWI & PROCHASKA 748