Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/~pgreen/Papers/metals02.ps Äàòà èçìåíåíèÿ: Fri May 23 01:31:08 2003 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 00:47:18 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: cluster gas

L77
The Astrophysical Journal, 581:L77--L80, 2002 December 20
# 2002. The American Astronomical Society. All rights reserved. Printed in U.S.A.
EVIDENCE OF A STRONG N v/C iv CORRELATION BETWEEN EMISSION AND ABSORPTION LINES
IN ACTIVE GALACTIC NUCLEI
Joanna K. Kuraszkiewicz and Paul J. Green
Harvard­Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138;
jkuraszkiewicz@cfa.harvard.edu, pgreen@cfa.harvard.edu
Received 2002 September 20; accepted 2002 November 6; published 2002 November 18
ABSTRACT
The narrow absorption lines (NALs) that are seen in the rest­frame ultraviolet near the systemic redshift of
active galactic nuclei (AGNs) are not always intrinsic to the near­nuclear region but may originate in the host
galaxy or in neighboring galaxies intervening along the line of sight. A variety of criteria have been sought---and
several identified---as evidence of an intrinsic origin. We have measured both emission and absorption lines in
a Hubble Space Telescope Faint Object Spectrograph sample of objects with both C iv and N v NALs within
#5000 km s of the systemic redshift. We find a strong (199.5% confidence) linear correlation between the
#1
N v/C iv ratio in broad emission lines and that in NALs. A control sample of AGNs with NALs separated by
larger velocities shows no such correlation. Our finding thus identifies an additional test for the intrinsic nature
of NALs in any given object. The correlation shows that the chemical­enrichment histories and/or ionization
parameters of the NAL clouds are closely related to those of clouds that produce the broad emission lines.
Subject headings: galaxies: abundances --- galaxies: active --- quasars: absorption lines ---
quasars: emission lines
1. INTRODUCTION
Quasar spectra are characterized by strong, broad emission
lines (BELs) that are thought to form in a large number of small
gas clouds photoionized by the central continuum source (pre­
sumably a black hole with an accretion disk). These broad line­
emitting clouds lie at distances of 10 13 --10 18 cm from the central
engine, spanning a range of densities cm and
8 12 #3
n # 10 --10
e
covering 10% of the ionizing source. Modeling of the BEL region
suggests that although the clouds populating the BEL region
span a wide range of distances and densities, only those clouds
within an optimum range of density and ionizing flux (different
for different lines) dominate the observed line flux. This is the
so­called locally optimally emitting clouds model (Baldwin et
al. 1995).
Hamann & Ferland (1992) proposed that the N v/C iv ratio
can be used as an abundance indicator, affording a probe of
galactic nucleosynthesis measurable to high redshifts. This is
because studies of high (10.2 solar) abundance galactic H ii
regions found that nitrogen goes up roughly as the square of
the metallicity as a result of secondary nucleosynthetic pro­
cessing. At lower metallicities, primary nitrogen is more im­
portant (see the work by Izotov & Thuan 1999 on blue compact
dwarf galaxies). Many high­redshift/high­luminosity quasars
(QSOs) have N v/C iv 1 0.1, indicating supersolar abundances
(Hamann & Ferland 1993).
QSO spectra often show absorption lines that are adjacent
to the emission lines (mostly in the UV; Crenshaw et al. 1999).
More than 10% of the optically selected QSOs have highly
ionized, broad absorption lines (BALs), which are blueshifted
relative to the emission lines by tens of thousands of kilometers
per second, with velocity widths of at a least a few thousand
kilometers per second. These BALs are undoubtedly intrinsic
in origin based on the inferred partial coverage of the contin­
uum source and on the supersolar abundances (e.g., Weymann
et al. 1991). However, for the narrow absorption line (NAL)
systems, with velocity widths of a few hundred kilometers per
second, there is less consensus on the location of absorbing
gas. NAL clouds may originate from gas ejected from the nu­
cleus (similar to the BALQSOs), in the host galaxy of the QSO,
or from galaxies nearby (e.g., within a galaxy cluster that hosts
the QSO). Intervening gas clouds or galaxies at cosmological
distances lying along our line of sight may also be responsible.
Adopted from Hamann & Ferland (1999), the following
properties are considered strong evidence of intrinsic absorp­
tion: the time variability of the absorption lines, the partial
coverage of the background light source (implied by multiplet
ratios), the high gas densities (inferred from fine­structure
lines), and the well­resolved profiles that are smooth and broad
(compared with both thermal line widths and the velocity dis­
persions expected in intervening clouds). Barlow & Hamann
(1997) suggested a few other properties that are weaker indi­
cators for intrinsic systems: high metallicity, high ionization,
and . It is often assumed that NAL systems found
z # z
abs em
within 5000 km s of the emission redshift are intrinsic to the
#1
QSO (however, see Richards et al. 1999 and references therein
for examples of intrinsic NALs with velocities 15000 km s ).
#1
This assumption is supported by a statistical excess of such
systems over the number expected from cosmologically dis­
tributed absorbers and by a correlation between the number of
such systems and the QSO's luminosity and radio properties
(Foltz et al. 1986, 1988; Aldcroft, Bechtold, & Elvis 1994;
Wills et al. 1995; Mo˜ ller, Jakobsen, & Perryman 1994).
We show here that a comparison of the absorption­ and
emission­line properties of QSOs---their N v/C iv ratios---
yields another useful indicator of intrinsic absorption and adds
new insight into the nature and composition of clouds near the
nucleus. Early NAL studies using low­quality data noted larger
N v/C iv ratios in systems compared with
z # z z K
abs em abs
(Weymann, Carswell, & Smith 1981) and were later con­
z em
firmed with higher quality data, for which the column densities
and abundances could be estimated (e.g., Petitjean, Rauch, &
Carswell 1994; Savaglio, D'Odorico, & Mo˜ller 1994). It was
found that systems usually have abundances
z # z Z 1
abs em
, at least an order of magnitude higher than in
Z z K z
, abs em
systems, where .
Z ! 0.1 Z ,
The N v/C iv ratio can also be used as an abundance indicator

L78 N v/C iv CORRELATION Vol. 581
TABLE 1
List of Objects
Designation
Systemic
Redshift
N v/C iv
Emission
N v/C iv
Absorption a Reference b
Program Sample: km s #1
Dv ! 5000
0050#124 . . . . . . 0.061 0.30 #0.06
#0.06 1.03 # 0.09 1
0350#073 . . . . . . 0.962 0.13 #0.05
#0.04 0.56 # 0.04 2
0955#326 . . . . . . 0.533 0.31 #0.05
#0.05 1.41 # 0.33 2
1114#444 . . . . . . 0.144 0.06 #0.02
#0.02 0.71 # 0.02 1
1130#111 . . . . . . 0.510 0.11 #0.02
#0.02 0.72 # 0.08 1
1309#355 . . . . . . 0.184 0.20 #0.09
#0.03 1.38# 0.11 1
1340#606 . . . . . . 0.961 !0.01 #0.01
#0.01 0.80 # 0.07 2
1351#640 . . . . . . 0.088 0.26 #0.04
#0.04 1.39 # 0.21 1
0.26 #0.04
#0.04 0.72 # 0.04 1
1404#226 . . . . . . 0.098 0.03 #0.12
#0.03 0.23 # 0.06 1
1425#267 . . . . . . 0.366 0.08 #0.02
#0.02 1.19 # 0.39 1
0.08 #0.02
#0.02 0.96 # 0.15 1
1538#478 . . . . . . 0.770 0.22 #0.11
#0.01 0.66 # 0.03 1
1631#395 . . . . . . 1.023 0.41 #0.08
#0.07 1.36 # 0.19 2
1704#608 . . . . . . 0.371 0.15 #0.04
#0.03 1.03 # 0.14 2
2041#109 . . . . . . 0.035 0.11 #0.01
#0.01 1.11 # 0.11 2
2135#147 . . . . . . 0.200 0.04 #0.02
#0.01 0.49 # 0.05 2
2251#178 . . . . . . 0.068 0.07 #0.00
#0.04 1.04 # 0.05 1
2251#113 . . . . . . 0.323 0.03 #0.15
#0.02 0.28 # 0.05 2
Control Sample: km s #1
Dv 1 5000
0414#060 . . . . . . 0.781 0.34 #0.17
#0.02 0.74 # 0.09 2
0454#220 . . . . . . 0.534 0.02 #0.01
#0.01 0.05 # 0.02 2
0.02 #0.01
#0.01 0.24 # 0.04 2
0710#118 . . . . . . 0.768 0.20 #0.12
#0.03 0.19 # 0.05 1
0916#513 . . . . . . 0.553 0.40 #0.06
#0.05 0.40 # 0.10 2
1229#021 . . . . . . 1.045 0.03 #0.01
#0.01 0.19 # 0.03 1
0.03 #0.01
#0.01 0.14 # 0.03 1
1248#401 . . . . . . 1.030 0.23 #0.04
#0.04 0.14 # 0.03 1
0.23 #0.04
#0.04 0.36 # 0.03 1
1544#489 . . . . . . 0.400 0.08 #0.02
#0.01 0.27 # 0.07 1
1611#343 . . . . . . 1.401 0.32 #0.05
#0.04 0.98 # 0.31 2
2128#126 . . . . . . 0.501 0.13 #0.07
#0.01 0.74 # 0.28 2
2145#067 . . . . . . 0.999 0.14 #0.02
#0.02 0.11 # 0.02 2
a Ratio calculated using Bechtold et al. 2002 data, except for 1340#606,
1631#395, and 2135#147, which we refitted (details in text).
b Emission­line ratio from (1) this Letter and (2) Kuraszkiewicz et al.
2002.
in absorption­line systems as long as the effects of saturation
and ionization are understood (Hamann et al. 1997b). Since
the BALs are mostly heavily saturated and hence only weakly
sensitive to the abundances, we exclude them from further
analysis. On the other hand, high­resolution studies show that,
in general, NALs are not severely saturated. In this Letter, we
show that N v/C iv correlates strongly between the BELs and
the NALs in active galactic nuclei (AGNs) with ,
z # z
abs em
indicating that either the metallicities or ionization parameters
in the NAL and BEL clouds are intimately related.
2. SAMPLE AND SPECTRAL ANALYSIS
Bechtold et al. (2002) present a large database of narrow
ultraviolet absorption lines measured from spectra observed
with the Faint Object Spectrograph (FOS; Keyes et al. 1995
and references therein) on the Hubble Space Telescope (HST)
and gathered from the Space Telescope Science Institute ar­
chives. For each of the 271 QSO spectra, Bechtold et al. present
a list of absorption lines (with significance 13.5 j) together
with line identifications and equivalent widths. We searched
these absorption­line lists for objects that have both N v and
C iv absorption within 5000 km s of the emission redshift
#1
and for which the velocity difference between the C iv and
N v absorbing systems is less than 500 km s . All absorption
#1
lines with A š are included. Lines with A š
W # 0.2 W ! 0.2
l l
were included only if an associated line from the doublet was
present with A š . We also require that the QSO spectra
W 1 0.2
l
show complete Lya and C iv emission lines, to accurately
model the N v and C iv emission lines.
The sample of QSOs chosen this way includes 17 objects
(two of which have double absorption systems with Dv #
km s : 1351#640 and 1425#267) with redshifts #1 and
#1
5000
is presented in Table 1. Six of the objects are radio­loud. In
1340#606 and 1631#395, we have refitted the N v absorption
lines (using Sherpa) since the partially resolved doublet was
fitted in Bechtold et al. (2002) with a single Gaussian, resulting
in an underestimate of the N v equivalent width. For the same
reasons, we refitted the C iv absorption in 2135#147. In
0050#124, the N v l1238 absorption line at z p 0.05386
abs
is contaminated by N v l1242 from an absorption system with
. Since it was difficult to resolve the two com­
z p 0.05113
abs
ponents in this object, we left the fits unchanged, but the N v
equivalent width should be treated with caution since it is
slightly overestimated.
The emission­line measurements for the spectra obtained
before the installation of the Corrective Optics Telescope Axial
Replacement (COSTAR; 1993 December) were taken from Ku­
raszkiewicz et al. (2002). The calibrated and dereddened spectra
obtained after the COSTAR installation were retrieved from
Bechtold et al. (2002). To obtain emission­line measurements
consistent with the pre­COSTAR data, we modeled (and in
some cases merged) post­COSTAR spectra just as described
in Kuraszkiewicz et al. (2002). 1
3. CORRELATION BETWEEN N v/C iv EMISSION AND ABSORPTION
The N v/C iv emission­ and absorption­line equivalent
width ratios (from both lines of the doublet) are presented
1 We fitted each spectrum with a power­law continuum, modeled blended
iron emission, included Galactic and intrinsic absorption lines, and performed
multicomponent fits to the emission­line profiles. For detailed spectral mod­
eling and emission­ and absorption­line measurements, we refer the reader to
our Web site http://hea­www.harvard.edu/#pgreen/HRCULES.html.
in Table 1 with 1 j errors (which do not include errors from
the uncertainty of the continuum placement that we estimate
to be #10%). In Figure 1a, we show the dependence between
the N v/C iv absorption­line ratio (N v/C iv) abs and the
N v/C iv emission­line ratio (N v/C iv) em . QSOs that had
more than one associated absorption system have their data
points either circled (1351#640) or surrounded by an open
square (1425#267). 2 Figure 1a shows a strong correlation
between (N v/C iv) abs and (N v/C iv) em in QSOs, with the
probability of a correlation occurring by chance %
P p 0.5
in the generalized Kendall rank test and in the
P p 0.8%
Spearman rank test (using the survival analysis package of
Lavalley, Isobe, & Feigelson 1992 to allow for the presence
of a lower limit in 1340#606). The best­fit linear regression
yields a slope consistent with unity. Lines
a p 1.06 # 0.33
with #1 j are shown as dotted lines in Figures 1a and 1b.
We also studied a control sample selected with identical criteria
to the program sample except that we now require Dv 1 5000
km s for both N v and C iv absorption. Based on this change,
#1
these NAL systems are more likely to be intervening. The control
sample includes objects with a range of redshifts 0.4 ! z ! 2.4
and is presented in Table 1 and Figure 1b. As is immediately
2 In these objects, the absorption systems with higher abundances have a
larger velocity difference relative to the systemic redshift ( ).
Dv

No. 2, 2002 KURASZKIEWICZ & GREEN L79
Fig. 1.---(a) N v/C iv emission­line ratio vs. N v/C iv absorption­line ratio with 1 j errors for QSOs with NALs within #5000 km s of the emission­line
#1
redshift. QSOs 1351#640 and 1425#267, which have two intrinsic absorption­line systems, are indicated by an open circle and an open square, respectively.
The sizes of the data points are proportional to the C iv NAL equivalent width. The dotted lines show the #1 j best­fit regressions, which include all measurements
plotted. (b) Control sample ( km s ).
#1
Dv 1 5000
evident, the control sample shows no correlation ( ) be­
P # 16%
tween (N v/C iv) em and (N v/C iv) abs .
We also tested the above correlations using N v/C iv flux
ratios. While flux ratios may be more strongly affected by
continuum placement, they also show a strong correlation
( % in the Kendall rank test and % in the
P # 0.01 P # 0.04
Spearman rank test) and an identical regression slope (1.07 #
0.15) in the program sample and no correlation in the control
sample ( %).
P 1 30
4. DISCUSSION
The HST/FOS spectra that we use here have only moderate
resolution, which can lead to an underestimate of the optical
depths, and hence column densities, of the absorption­line sys­
tems. Many objects in the program sample have absorption­
line doublet ratios N v l1238/N v l1242 and C iv l1548/
C iv l1550 close to 2, implying that the lines are not saturated;
however, higher resolution spectra are needed to confirm this.
The conversion of measured columns to metal abundances de­
pends directly on the ionization fractions of those species mea­
sured, whereas without further (e.g., X­ray absorption) mea­
surements, we have little information available on the ionization
state of the absorbers. Finally, our measurements are likely to
compound the absorption from an ensemble of clouds along the
line of sight with a range of columns, thermal and bulk velocities,
and ionization levels. However, the correlation in Figure 1a
shows that, whatever its quality as an abundance indicator in
these data, N v/C iv correlates strongly between the BEL gas
and the NAL gas. If ionization or other effects differed between
the NAL and BEL clouds, this would merely add scatter to the
correlation. The fact that a strong correlation persists suggests
that the metallicity and/or ionization parameters in BEL and NAL
clouds are related. Otherwise the differences in these properties
must conspire to cancel each other so that the observed corre­
lation might persist. However, we now also consider possible
selection effects in our measurements.
The emission lines in both the control and program samples
are easy to detect and are measured consistently using the same,
semiautomatic measuring technique. However, the presence of
underlying emission lines could affect the measurement of the
absorption lines in the program sample differently. In general,
the N v NAL is the more difficult to detect and measure. If
stronger emission lines yielded preferentially better detection and
measurement of N v absorption, the observed correlation in
Figure 1a could be spuriously enhanced. If so, we might expect
that (1) the minimum absorption line of the program sample
W l
would be larger than for the control sample and (2) some de­
pendence of the absorption­line measurements on the emission­
line strengths. We find that the minimum absorption line for
W l
the program sample is , similar to the
0.14 # 0.02 0.16 #
found for the control sample. We also find that objects with
0.03
different C iv (and, as we found, N v) equivalent widths (W l
proportional to the data point size in Fig. 1a) are evenly dis­
tributed along the (N v/C iv) em versus (N v/C iv) abs correlation.
Hence, the underlying continuum#emission­line model does not
significantly affect the absorption­line measurement or detection.
It could also be argued that the lack of points in the upper
left­hand corner of Figure 1a could be caused by objects that
have measurable C iv absorption but undetectable N v absorp­
tion. However, in the control sample, we find many objects in
that region, which additionally provides evidence for the gen­
erally lower column of intervening absorbers. Similarly, the lack
of objects in the lower right­hand corner of Figure 1a might be
suggested as a selection effect caused by objects that have mea­
surable N v absorption but undetectable C iv absorption. We
found two such objects in the Bechtold et al. (2002) sample:
1118#1252 (with km s ), for which C iv absorp­
#1
Dv p #790
tion by pure accident coincides with strong Galactic Fe ii ab­
sorption and hence has not been identified, and 1111#4053 (with
km s ), for which C iv absorption is visible in
#1
Dv p #1027
the spectrum but is too weak to be detected. We have modeled
the emission lines in the latter object and found that log (N v/

L80 N v/C iv CORRELATION Vol. 581
C iv) em 1 #0.6, which, together with the possibly large (N v/
C iv) abs ratio, would place 1111#4053 in the upper right­hand
corner of Figure 1a without effecting the overall correlation.
A few objects from the control sample extend into the region
occupied by the program sample in Figure 1b. Since some ab­
sorbers are known to extend beyond the km s #1
Dv p 5000
criterion (e.g., BALQSOs), it is possible that NALs among the
control sample objects that lie near the program sample corre­
lation may be intrinsic. There is mounting evidence that some
QSO absorption systems with velocities 5000 km s ! !
#1 v
75,000 km s are, if not intrinsic, at least affected by the illu­
#1
minating QSO. For example, Richards et al. (1999) show that
the distribution in velocity space of such systems is dependent
on the QSOs' radio properties (luminosity, radio spectral index,
and radio morphology; see also Borgeest & Mehlert 1993, Pe­
titjean et al. 1994, and Hamann et al. 1997a). Based on the
evidence here, we consider ratios lying in the region of the
program sample correlation to provide corroborative evidence
for an intrinsic nature, which may extend to absorbing systems
at these higher velocities.
5. SUMMARY
One reasonable interpretation of the (N v/C iv) em versus
(N v/C iv) abs correlation we report here is that the NAL and
BEL clouds share related chemical­enrichment histories and/
or ionization parameters. In Elvis (2000), the NALs and BELs
are cospatial, situated in gas outflowing from the disk, where
NAL gas is the warm highly ionized medium that confines the
BEL clouds. In Sabra, Hamann, & Shields (2002), NALs orig­
inate from the narrow line­emitting region when the observer
is looking down through the line­emitting gas toward the con­
tinuum source. It may also be possible that the NAL gas is
enriched in the vicinity of the QSO nucleus and then ejected
into the host galaxy of the QSO or farther out into the inter­
galactic medium. But high abundance does not guarantee an
intrinsic origin. 3 Conversely, lower abundance in NAL gas does
not exclude an intrinsic origin. In the lower left region of Fig­
ure 1a, there are NALs in the program sample that lie directly
along the correlation, but with ratios indicating lower abun­
dances, similar to those in the control sample.
The wide velocity range of clouds contributing to BELs
makes it likely that BELs are produced by clouds spanning a
wide range of physical conditions, with an ionization level and
a density of principal interest to line production (Baldwin et
al. 1995). BEL or BAL measurements are thus likely to inte­
grate over thousands of clouds or perhaps over a wind structure
(e.g., de Kool & Begelman 1995; Murray & Chiang 1995;
Proga, Stone, &Kallman 2000) covering wide ranges of density
and ionizing flux (Korista et al. 1997; Baldwin et al. 1995).
By contrast, the velocity width of NALs is lower. The NAL
clouds may probe fewer lines of sight, which facilitates a sim­
pler and perhaps partially independent physical interpretation
of absorption­line measurements. We thus suggest that even
using spectra of moderate resolution and signal­to­noise ratio
(S/N), the strong correlation that we report is supportive evi­
dence for the validity of BEL ratios as either ionization or
abundance indicators. Investigations at higher resolution and
S/N are clearly warranted.
The authors gratefully acknowledge the support provided for
this project by NASA through grant NAG5­6410 (LTSA).
P. J. G. acknowledges support through NASA contract NAS8­
39073 (CXC). We thank the anonymous referee, Fred Hamann,
Martin Elvis, and Smita Mathur for valuable comments. The
data in this Letter are based on observations made with the
NASA/ESA Hubble Space Telescope, obtained from the data
archive at the Space Telescope Science Institute, which is op­
erated by the Association of Universities for Research in As­
tronomy, Inc., under the NASA contract NAS 5­26555.
3 Tripp, Lu, & Savage (1996) studied an associated absorption system in a
QSO with a high metal abundance (11 Z , ) NAL, which at first suggested
z p 3
that the absorbing gas could originate near the nucleus of the QSO. However,
a lack of the excited­state absorption in C ii* l1336 (compared with detected
C ii l1335) in this system indicates a very low (!7 cm ) density absorber
#3
suggesting 1300 kpc distances from this QSO.
REFERENCES
Aldcroft, T. L., Bechtold, J., & Elvis, M. 1994, ApJS, 93, 1
Baldwin, J. A., Ferland, G., Korista, K., & Verner, D. 1995, ApJ, 455, L119
Barlow, T. A., & Hamann, F. 1997, in ASP Conf. Ser. 128, Mass Ejection
from Active Galactic Nuclei, ed. N. Arav, I. Shlosman, & R. J. Weymann
(San Francisco: ASP), 13
Bechtold, J., Dobrzycki, A., Wilden, B., Morita, M., Scott, J., Dobrzycka, D.,
Tran, K.­V., & Aldcroft, T. L. 2002, ApJS, 140, 143
Borgeest, U., & Mehlert, D. J. 1993, A&A, 275, L21
Crenshaw, D. M., Kraemer, S. B., Boggess, A., Maran, S. P., Mushotzky, R.
F., & Wu, C.­C. 1999, ApJ, 516, 750
de Kool, M., & Begelman, M. C. 1995, ApJ, 455, 448
Elvis, M. 2000, ApJ, 545, 63
Foltz, C. B., Weymann, R. J., Peterson, B. M., Sun, L., Malkan, M. A., &
Chaffee, F. H. 1986, ApJ, 307, 504
Foltz, C. B., Chaffee, F. H., Weymann, R. J., & Anderson, S. F. 1988, in QSO
Absorption Lines: Probing the Universe, ed. J. C. Blades, D. A. Turnshek,
& C. A. Norman (Cambridge: Cambridge Univ. Press), 53
Hamann, F., Barlow, T., Cohen, R. D., Junkkarinen, V., & Burbidge, E. M.
1997a, in ASP Conf. Ser. 128, Mass Ejection from Active Galactic Nuclei,
ed. I. Arav, I. Shlosman, & R. J. Weymann (San Francisco: ASP), 19
Hamann, F., Beaver, E. A., Cohen, R. D., Junkkarinen, V., Lyons, R. W., &
Burbidge, E. M. 1997b, ApJ, 488, 155
Hamann, F., & Ferland, G. 1992, ApJ, 391, L53
---------. 1993, ApJ, 418, 11
---------. 1999, ARA&A, 37, 487
Izotov, Y. I., & Thuan, T. X. 1999, ApJ, 511, 639
Keyes, C. D., Koratkar, A. P., Dahlem, M., Hayes, J., Christiansen, J., &
Martin, S. 1995, Faint Object Spectrograph Instrument Handbook, Version
6.0 (Baltimore: STScI)
Korista K., Baldwin, J., Ferland G., & Verner D. 1997, ApJS, 108, 401
Kuraszkiewicz, J. K., Green, P. J., Forster, K., Aldcroft, T. L., Evans, I., &
Koratkar, A. 2002, ApJS, 143, 257
Lavalley, M., Isobe, T., & Feigelson, E. D. 1992, in ASP Conf. Ser. 25,
Astronomical Data Analysis Software and Systems I, ed. D. M. Woarral,
C., Biemsderfer, & J. Barnes (San Francisco: ASP), 245
Mo˜ller, P., Jakobsen, P., & Perryman, M. A. C. 1994, A&A, 287, 719
Murray, N., & Chiang, J. 1995, ApJ, 454, L105
Petitjean, P., Rauch, M., & Carswell, R.F. 1994, A&A, 291, 29
Proga, D., Stone, J. M., & Kallman, T. R. 2000, ApJ, 543, 686
Richards, G. T., York, D. G., Yanny, B. Kollgaard, R. I., Laurent­Muehleisen,
S. A., & vanden Berk, D. E. 1999, ApJ, 513, 576
Sabra, B., Hamann, F., & Shields, J. 2002, American Physical Society April
Meeting APR02, Abstract D11.007
Savaglio, S., D'Odorico, S., & Mo˜ller, P. 1994, A&A, 281, 331
Tripp, T. M., Lu, L., & Savage, B. D. 1996, ApJS, 102, 239
Weymann, R. J., Carswell, R. F., & Smith, M. G. A. 1981, ARA&A, 19, 41
Weymann, R. J., Morris, S. L., Foltz, C. B., & Hewett, P. C. 1991, ApJ, 373,
23
Wills, B. J., et al. 1995, ApJ, 447, 139