Документ взят из кэша поисковой машины. Адрес оригинального документа : http://hea-www.harvard.edu/~pgreen/post_FOS_WWW/paper.ps Дата изменения: Mon Oct 6 22:18:23 2003 Дата индексирования: Tue Oct 2 01:40:07 2012 Кодировка: Поисковые слова: cygnus

To appear in ApJS
Preprint typeset using L A T E X style emulateapj
EMISSION LINE PROPERTIES OF AGN FROM A POST-COSTAR FOS HST SPECTRAL ATLAS
Joanna K. Kuraszkiewicz, Paul J. Green
Harvard-Smithsonian Center for Astrophysics, 60 Garden St., Cambridge, MA 02138
email: jkuraszkiewicz@cfa.harvard.edu, pgreen@cfa.harvard.edu
D. Michael Crenshaw, Jay Dunn
Department of Physics and Astronomy, Georgia State University, Astronomy Oфces, One Park Place South SE, Suite 700, Atlanta, GA 30303
email: crenshaw@chara.gsu.edu,dunn@chara.gsu.edu
Karl Forster
California Institute of Technology, 1200 E. California Blvd., MC 405-47, Pasadena, CA 91125
email: krl@srl.caltech.edu
Marianne Vestergaard
The Ohio State University, Columbus, 140 West 18th Avenue, Columbus, OH 43210
email: vester@astronomy.ohio-state.edu
Tom L. Aldcroft
Harvard-Smithsonian Center for Astrophysics, 60 Garden St., Cambridge, MA 02138
email: taldcroft@cfa.harvard.edu
To appear in ApJS
ABSTRACT
This paper joins a series compiling consistent emission line measurements of large AGN spectral
databases, useful for reliable statistical studies of emission line properties. It is preceded by emission
line measurements of 993 spectra from the Large Bright Quasar Survey (Forster et al. 2001) and 174
spectra of AGN obtained from the Faint Object Spectrograph (FOS) on HST prior to the installation
of COSTAR (Kuraszkiewicz et al. 2002). This time we concentrate on 220 spectra obtained with the
FOS after the installation of COSTAR, completing the emission line analysis of all FOS archival spectra.
We use the same automated technique as in previous papers, which accounts for Galactic extinction,
models blended optical and UV iron emission, includes Galactic and intrinsic absorption lines and models
emission lines using multiple Gaussians. We present UV and optical emission line parameters (equivalent
widths, uxes, FWHM, line positions) for a large number (28) of emission lines including upper limits
for undetected lines. Further scientic analyses will be presented in subsequent papers.
Subject headings: galaxies: active|quasars: emission lines| quasars: general|ultraviolet: galaxies
1. INTRODUCTION
It is broadly acknowledged that the quasar central en-
gine (presumably a massive black hole with an accre-
tion disk) photoionizes gas lying farther out. This gas
emits broad permitted emission lines that are distinctive
of quasar spectra. At rst glance, quasar spectra look
quite similar; this may be the result of simple averaging.
Baldwin et al. (1995) showed that although the broad line
region (BLR) consists of clouds with a wide range of prop-
erties (gas density, ionization ux and column density),
the bulk of emission line ux is most likely produced in
the gas clouds with the optimum parameters for eфcient
emission in that line.
A closer look at the quasar spectra, however, reveals
that the spectra dier in detail and intriguingly, behave in
a correlated manner. For example it was found that AGN
that show strong optical iron emission (Fe II 4570) have
weaker [O III] 5007, and narrower, blue-asymmetric H
lines. This set of correlations was found to be the primary
eigenvector of the emission line correlation matrix of PG
quasars studied by Boroson & Green (1992). This eigen-
vector 1 was later found to correlate with UV properties
such as: CIV shift/asymmetry (Marziani et al. 1996) and
Si III]/C III] ratio, C IV and N V strength (Wills et al.
1999, Shang et al. 2003). Since eigenvector 1 was found
to correlate signicantly with X-ray properties (Laor et al.
1997, Brandt & Boller 1998) which are determined in the
vicinity of the central black hole, it was suggested that
dierences in emission line properties revealed by eigen-
vector 1 are caused by diering central engine parame-
ters (e.g. L=LEdd , accretion rate, orientation and/or black
hole spin). It was found that eigenvector 1 together with
eigenvector 2 provide a parameter-space in which all major
classes of broad-line sources can be discriminated, consti-
tuting a possible \H-R diagram" for quasars (Sulentic et
al. 2000; Boroson 2002).
Another famous correlation involving quasar spectra is
the anticorrelation between the equivalent width of the
broad emission lines and the UV luminosity called the
Baldwin eect (Baldwin 1977). The appeal of this cor-
relation was soon realized, since the luminosity of a dis-
tant quasar could potentially be estimated from the emis-
sion line equivalent widths, providing a standard candle
in measuring cosmological distances. In reality the scat-
ter of the Baldwin eect is too large to give meaningful
results, and studies have concentrated on understanding
and reducing this scatter (Shang et al. 2003, Dietrich et
al. (2002). Con icting results have also emerged, where
radio-loud samples and samples with a wide range of lumi-
nosities show a stronger eect (e.g. Baldwin et al. 1978,
Wampler et al. 1984, Kinney, Rivolo, & Koratkar 1990,
Wang et al. 1998), while radio-quiet samples and samples
1

2 Emission Line Properties of AGN from a post-COSTAR FOS HST Spectral Atlas
with a small luminosity range show weaker or no eect (e.g.
Steidel & Sargent 1991, Wilkes et al. 1999). A number of
explanations have been introduced to explain the Bald-
win eect. It can be either due to geometry as in Netzer,
Laor & Gondhalekar (1992), where the inclination of the
disk changes the apparent luminosity, or due to changes in
spectral energy distribution with luminosity, where more
luminous objects have softer ionizing continuum (Zheng &
Malkan 1993, Green 1998) or due to a decrease of covering
factor of the broad emission line clouds with luminosity
(Wu, Boggess, & Gull 1983). There have also been claims
that the Baldwin eect is aected by evolution (Green,
Forster, & Kuraszkewicz 2001) or may be due to selection
eects (continuum beaming, biases in selection techniques
- see Sulentic et al. 2000, Yuan, Siebert, & Brinkmann
1998).
Despite a vigorous study of emission line properties of
AGN in the last 30 years, which resulted in few thou-
sand published articles, questions about the structure and
kinematics of the BLR and their relationship to the cen-
tral engine (accretion mechanism, origin of the fuel, etc.)
have not been answered. Nor is it clear how the BLR re-
lates to the other components seen in AGN spectra: broad
and narrow absorption lines, X-ray warm absorbers, high
ionization emission lines and scattering regions. Despite
attempts to unite these components (Elvis 2000, Laor &
Brandt 2002, Ganguly et al. 2001, Murray & Chiang 1995)
denitive tests have been elusive. Progress has been ham-
pered by lack of large datasets with uniform and reliable
measurements of emission lines that would consistently
measure the continuum, and account for blended iron
emission which heavily contaminates emission lines such
as: H, Mg II, and C III] and forms a pseudo-continuum
complicating the measurements of the broadband contin-
uum, the weaker lines and the wings of strong emission
lines (Wills et al. 1985, Boroson & Green 1992, Vester-
gaard & Wilkes 2001). Most studies have concentrated
either on large non-uniform samples where emission line
measurements have been compiled from literature (Zheng
& Malkan 1993, Zamorani et al. 1992, Corbin & Boroson
1996, Dietrich et al. 2002) or small samples with uniform
measurements (Boroson & Green 1992, Wills et al. 1999,
Wilkes et al. 1999).
We have therefore undertaken a major study of AGN
emission lines, where our largely automated procedure ac-
counts for Galactic extinction, models blended optical and
UV iron emission, includes Galactic and intrinsic absorp-
tion lines, and models emission lines using multiple Gaus-
sians. Using the same modeling procedure we have pre-
viously analyzed and published measurements of emission
lines of two large datasets. The rst, 993 spectra from the
Large Bright Quasar Survey has been presented by Forster
et al. (2001; hereafter Paper I) together with detailed de-
scription of our analysis methods. The second includes
174 FOS/HST spectra obtained before the installation of
COSTAR and was presented in Kuraszkiewicz et al. (2002;
hereafter Paper II). In the current paper we present the
measurements of emission lines and plots of spectral ts of
the remaining 220 FOS/HST spectra that were observed
after the installation of COSTAR, completing the analysis
of all archival FOS/HST spectra. Statistical comparison
of the emission-line parameters and continuum parame-
ters of these large samples will hopefully bring us closer
to building an accurate model of emission line regions and
their dependence on the central engine.
2. THE POST-COSTAR FOS AGN SAMPLE
The sample was assembled by cross-correlating the
Veron-Cetty and Veron (1996) catalog of AGN with the
MAST (Multimission Archive at Space Telescope) hold-
ings. BL Lac objects were ignored, as their spectra show
no emission lines. Starburst galaxies and broad absorp-
tion line (BAL) quasars (where emission lines are heavily
disrupted by absorption features) were not included. We
chose all available (UV and optical) spectrophotometric
archival data that have been observed with the Faint Ob-
ject Spectrograph (FOS, Keyes et al. 1995 and references
therein) on HST after the installation of COSTAR (i.e.,
after December 1993). FOS spectra obtained prior to De-
cember 1993 have been analyzed by us in Paper II. We
include all spectra taken with the high resolution gratings
(G130H, G190H, G270H, G400H, G570H, G780H; spec-
tral resolution =
  1300 ). Low resolution (G160L,
G650L; spectral resolution =
  250) gratings were also
included when high resolution gratings were not available
in the matching wavelength range. Spectra obtained with
the prism were excluded as their extremely low resolution
precludes any reasonable emission line measurements. We
analyzed only spectra with a mean signal-to-noise (S/N)
per resolution element  5.
The FOS spectra were uniformly calibrated to account
for temporal, wavelength{ and aperture{dependent varia-
tions that are seen in the instrumental response. We use
the most recent version of the FOS calibration pipeline
with the ST-ECF POA version of calfos. This pipeline
provides an improved correction to the zero point o-
sets in the BLUE high resolution spectra, removes hot
pixel/hot diode regions from individual exposures and cal-
ibrates spectra to the 4.3" aperture.
We interpolated all of the spectra to a linear wavelength
scale, retaining the original approximate wavelength inter-
vals (in Angstroms per bin), and, for each object, we aver-
aged all of the spectra obtained at a particular wavelength
setting if the ux did not dier by more than 20%. If the
dierence in ux was larger, the spectra were analyzed
separately. To obtain a reliable continuum t for each
object, we combined spectra obtained at dierent wave-
length settings and observed at dierent times if the ux
levels did not dier by more than 20% in the overlap re-
gion. High resolution gratings (G130H, G190H, G270H,
G400H, G570H, G780H) were merged separately from the
low resolution gratings (G160L, G650L). In both cases
the longer wavelength spectrum was scaled to match the
shorter wavelength spectrum and the spectra were then
spliced at wavelengths in continuum regions away from
emission or absorption lines by retaining as much of the
higher S/N spectrum as possible.
At this point the sample consisted of 327 spectra. Spec-
tra which showed no emission lines, due to a too low S/N
(<5) in the line regions, or a redshift that placed strong
emission lines outside the spectrum's wavelength range
(mostly chosen for studies of the Ly forest) were then re-
moved. The nal sample consists of 220 spectra of the 180
AGN listed in Table 1. In the rst column the coordinate

Kuraszkiewicz, Green, Crenshaw, Dunn, Forster, Vestergaard, Aldcroft 3
designation based on the equinox J2000 position is given,
followed by the AGN name (column [2]), AGN type and
redshift (from the NASA/IPAC Extragalactic Database)
and Galactic NH in units of 10 20 cm 2 (columns [3]-[5]).
The values of NH are in general taken from the Bell Lab-
oratory H I survey (Stark et al. 1992). In a few cases
for which NH had been specically measured, we quote
the values from the literature (Lockman & Savage 1995,
Elvis et al. 1989); for objects with declination > 40 o , NH
is from Heiles & Cleary (1979). The last column of Ta-
ble 1 gives the list of spectra that were analyzed for each
object. The name of the spectrum consists of the coordi-
nate designation from column (1), followed by a two letter
designation: \o" indicates a post-COSTAR spectrum (in
Paper II pre-COSTAR spectra were designated with \r");
a second letter (a to z) indicates whether the AGN in ques-
tion has more than one spectrum available. A capital let-
ter indicates a spectrum of a lensed component as e.g. in
1001+5553oA and 1001+5553oB. In Table 2 we show a
detailed list of FOS gratings, and datasets with exposure
times that were used to compile spectra listed in Table 1
(see the ApJ Web site 1 for full version of Table 2).
3. ANALYSIS OF SPECTRA
3.1. Continuum and blended iron tting
Since our goal was to assemble a uniform database of
emission line measurements, we have analyzed our post-
COSTAR spectra following the same tting procedures as
those used in the LBQS and pre-COSTAR/FOS spectral
analysis (for details see Papers I and II). We used the mod-
eling software Sherpa 2 (Freeman, Doe & Siemiginowska
2001) developed for the Chandra mission, where the model
parameters were determined from a minimization of the
 2 statistic with modied calculation of uncertainties in
each bin (Gehrels 1986) and using the Powell optimiza-
tion method for continuum, iron emission and rst emis-
sion line ts and the Levenberg-Marquardt optimization
method in the nal emission line ts (see below). First
we t a reddened power-law continuum 3 to regions of the
spectrum redwards of Ly and away from strong emission
lines and blended iron emission. We use the same contin-
uum windows as in the analysis of pre-COSTAR contin-
uum spectra (see Table 2 in Paper II), with the addition of
a new window redwards of H at 6990{7020  A rest frame.
Most of the post-COSTAR spectra were tted by a single
power law. However in 21 spectra that covered a large
wavelength range, two power laws were introduced: one
(UV) extending at  rest < 4200  A and another (optical) at
 rest > 4200  A, both normalized at  = 4200  A. In Table 3
we present the slopes of the dereddened UV and optical
continua (column [2] and [5] respectively) with the normal-
ization of the continuum in units of 10 14 erg cm 2 s 1  A 1
(column [3]) at the observed wavelength  norm (column
[4]). The slopes and normalizations are quoted with 2 er-
rors. For spectra with only one continuum window present,
a constant slope of = 1 is quoted without errors. This
value was adopted since the mean slope of the pre and
post-COSTAR FOS sample is 0.970.09 (see the ApJ Web
site 4 for full version of Table 3).
The next step in our tting procedure was to model
the blended iron emission lines. In the UV we used the
Vestergaard & Wilkes (2001) iron template covering rest
frame wavelengths between 1250{3100  A, while in the opti-
cal we used the Boroson & Green (1992) template covering
4250{7000  A. First a crude estimate of the template's ux
normalization was obtained by tting the 2000 km s 1
FWHM template to regions where iron emission is known
to be strongest (see column [2] in Table 2 of Paper II).
Then the FWHM of iron emission was estimated by com-
paring the spectrum with a grid of templates with FWHM
between 900 and 10,000 km s 1 in steps of 250 km s 1 .
This was followed by a t of both the FWHM and ux
normalization at the iron tting windows, followed by two
iterations of the continuum and iron ts (refer to Paper I
for more details). At this point the continuum and iron
ts results were inspected and adjustments were made to
spectra not tted successfully (5% of spectra needed ad-
justments of the continuum t and 3 spectra needed ad-
justment of iron ts).
3.2. Emission and absorption line tting
The emission lines were generally tted with one Gaus-
sian. However since most (95%) FOS spectra have high
S/N, the strong emission lines (Ly, C IV, C III], Mg II,
H, H) were tted using two components: the very broad
line region (VBLR) component and the intermediate line
region (ILR; see Brotherton et al. 1994) here referred to
as the broad and narrow components respectively. We use
exactly the same emission line inventory as in the pre-
COSTAR spectra (see Table 3 in Paper II).
As a rst stage, the FWHM and peak amplitude of the
Gaussians are modeled while keeping the position of the
emission line xed at the expected wavelength (calculated
from redshift). Then the position of the line is freed and
modeled together with the FWHM and peak amplitude
using Powell optimization. In the next step all Gaussian
parameters are retted, this time using distinct high and
low sigma rejection criteria. We found that  = 3 for low
rejection omits most of the absorption lines superimposed
on the emission lines, while  = 7 for high rejection by-
passes most spikes not associated with the emission line
(e.g. geocoronal Ly, cosmic rays, etc.). At this stage we
use the Levenberg-Marquardt optimization method, which
is faster than the Powell method, but only works well if
the statistical surface is well-behaved (after two runs of the
emission line parameter tting with the Powell method,
this was certainly the case). It is nearly impossible to de-
sign a fully automated procedure that can deal with the
wide range of spectral shapes that AGN show, so at this
point the ts were inspected and adjustments were made
to spectra where necessary. About 5% of spectra needed
adjustments at least in one emission line t.
For each spectrum, the continuum, iron and emission
line model obtained in the Sherpa tting was next used as
an input \continuum" to the FINDSL routine (Aldcroft
1 http://www.journals.uchicago.edu/ApJ
2 http://cxc.harvard.edu/sherpa/index.html
3 We use the the reddening curves of Cardelli, Clayton & Mathis 1989 to account for Galactic extinction; see Paper I for details
4 http://www.journals.uchicago.edu/ApJ

4 Emission Line Properties of AGN from a post-COSTAR FOS HST Spectral Atlas
1993), which identies narrow absorption lines and ts
them with Gaussian proles. We set the routine to nd ab-
sorption lines away from the Ly forest region (blueward
of  rest =1065  A) and outside the Balmer continuum re-
gion (3360{3960  A), where the global power law continuum
may not t the spectra well. The minimum signicance
level for identication of absorption lines was set to 4.5
(see Paper I for more details). We detect and t absorp-
tion lines with W   0:3  A. The absorption line parameters
were then used in the next iterative modeling step where
the position, peak amplitude, and FWHM of the absorp-
tion line were modeled simultaneously by the Sherpa pro-
gram, followed by another iteration of the emission line
tting. After this stage the results were inspected and
spectra retted if the automated procedure did not per-
form well. An example of a full spectral tting is shown
in Fig.1. The top panel shows the reddened power-law
continuum t redwards of Ly to the observed spectrum,
followed below by panels showing blended iron and emis-
sion line modeling of Ly, C IV, C III], and Mg II. Since
the whole post-COSTAR FOS Spectral Atlas includes 220
spectra, we present similar plots of other spectral ts only
on our Web site. 5
3.3. Error analysis
The error analysis follows the procedure from Paper I
(see Section 3.5 of that paper for details), in which the 2
errors for each emission line parameter were determined
from the  2 condence interval bounds (
 2 =4.0) using
the uncertainty procedure in Sherpa. The upper limits of
equivalent widths were determined by xing the line posi-
tion at the expected wavelength, the FWHM at the value
of the median FWHM found for that line in the LBQS
sample (see column [3], Table 3 of Paper II) and by set-
ting the amplitude of the line to the 2 positive error.
4. EMISSION LINE MEASUREMENTS
In Table 4 we present the rest frame emission line mea-
surements for one example object NGC 3516 (spectrum
1106+7234oe). Due to its large size the full table for the
post-COSTAR FOS Spectral Atlas is available only in elec-
tronic from the ApJ Web site 6 and at our Web site 5 . The
format of Table 4 is exactly the same as the format of the
electronic tables of emission line measurement presented
for the LBQS and pre-COSTAR FOS samples, making it
simple to analyze the LBQS and FOS samples together.
In the full Table 4, each spectrum is represented by 43
rows, one for each possible emission line measurement. In
the rst column the name of the spectrum is given, fol-
lowed by the object's redshift (column [2]), followed by
information on the emission line parameters: name of the
emission line (column [3]), FWHM in km s 1 (columns
[4] to [6] showing the value and 2 errors), the oset
of the peak of the Gaussian emission line model (all lines
except iron) in km s 1 from the expected position based
on the tabulated redshift (columns [7] to [9] value, 2
errors), the rest frame equivalent width of the emission
line in  A (columns [10]-[12]) and the observed frame ux
in units of 10 14 erg cm 2 s 1 (columns [13]-[15]). Errors
quoted for ux and W  are based on the uncertainties in
the amplitude and FWHM of the Gaussian model and do
not include an error from an uncertainty in the underlying
continuum ux level, which we estimate to be about 10%.
For emission lines where only an upper limit on ux and
W  is available, no values for the peak oset are quoted
as the position of the line was xed at the line's expected
wavelength. Also, the FWHM value in this case was set to
the median value for the LBQS sample (see Table 3 in Pa-
per II) with no associated errors. Finally, the last column
(16) in the full table gives the number of narrow absorption
features used in the emission line modeling. Our Gaussian
decomposition is not necessarily unique and may be sen-
sitive to slight shifts in continuum placement. While the
total ux and equivalent width are easily derived by sum-
ming values provided for individual Gaussian components,
no simple combination yields a FWHM representative of
the entire emission line. We therefore list in Table 4b
(electronic version only) the total line FWHM (with 2
errors) of those lines that have been modeled using two
Gaussians. These are: Ly, C IV, C III], Mg II, H, and
H where the width of the line was measured at half peak
of the dereddened emission line model after excluding iron
emission, absorption lines, and weaker emission lines (e.g.
in the Lya region we exclude NV line, in H region [NII]
and [SII]).
5. STATISTICS AND COMPARISON WITH THE
POST-COSTAR SAMPLE
The statistical properties of the rest frame W  and
FWHM distributions of the emission line measurements of
the post-COSTAR spectra are presented in Table 5. The
numbers quoted were obtained by excluding o-nuclear
spectra (e.g. the 10 dierent NLR knots of Mrk 78). To
avoid excessive weight given to a single object, in cases
of multiple spectra we tally only measurements from the
spectrum with the highest S/N and resolution. In total
1607 emission lines have been modeled among which 97
are upper limits. In Table (5) the name of the emission
line is given in column (1), followed by the total number
of emission lines modeled (column [2]) and the number of
upper limits (column [3]). The mean, standard deviation
and median of the W  and FWHM for the detected lines
are presented in columns (4)-(6) and (10)-(12) respectively.
When upper limits in W  were present we used the non-
parametric survival analysis technique and a Kaplan-Meier
estimator to reconstruct the true W  distribution and to
calculate the means and medians in columns (7)-(8) (for
reference see Isobe, Feigelson & Nelson 1986 and Lavalley,
Isobe & Feigelson 1992).
Since the strong emission lines such as: Ly, C IV,
C III], Mg II, H, H were tted using either two (broad
and narrow) components or one (single) component, we
calculated the W  and FWHM means and medians for
these components separately. The mean and median W 
for the whole line (indicated as the \sum" in Table 5) was
calculated as either the sum of the broad and narrow com-
ponents or the single component alone.
Both the pre- and post-COSTAR FOS samples are het-
erogeneous, and represent neither complete nor uniform
selection. Nevertheless, as a check on our methods and
5 See http://hea-www.harvard.edu/~pgreen/HRCULES.html
6 http://www.journals.uchicago.edu/ApJ

Kuraszkiewicz, Green, Crenshaw, Dunn, Forster, Vestergaard, Aldcroft 5
on the consistency between these samples we compare
the statistical properties of the W  and FWHM of the
pre-COSTAR sample analyzed in Paper II and the post-
COSTAR sample presented here. Overall, the means and
medians for the UV lines agree within the errors. We
did not, however, attempt to compare the W  of post-
COSTAR single components of the strong emission lines
or optical lines redwards of Ne V with the pre-COSTAR
sample measurements, since the number of available W  in
both or either samples is too small for meaningful analysis.
Histograms of W  and FWHM of the emission lines blue-
wards of Mg II are presented in Figure 2. The rst and
third rows represent the W  distributions, while the sec-
ond and fourth rows give the FWHM distributions. In all
panels, solid lines represent distributions for detections,
while the dotted lines show the estimated W  distribu-
tions from the Kaplan-Meier estimator if upper limits are
present. In the panels which show the sum of Ly, C IV,
C III] and Mg II distributions, the shaded histograms rep-
resent results from single Gaussian component ts.
The luminosity and redshift range of the post-COSTAR
sample is comparable to the pre-COSTAR sample ana-
lyzed in Paper II (see Fig.3). However the post-COSTAR
sample shows a larger number of low luminosity AGN
such as Seyferts, LINERs and NLS1s. Objects with
log L(2500  A)< 30 comprise of 30% of the post-COSTAR
sample and only 15% of the pre-COSTAR sample. In Fig-
ure 4 we show the distributions of log L(2500  A) for both
samples. The two-tailed Kolmogorov-Smirnow test gave a
99.9% probability that these distributions are dierent.
6. CONCLUSIONS
We have presented the emission line measurements of a
sample of AGN which has been observed by the FOS/HST
after the installation of COSTAR. Our sample includes 180
objects and 220 spectra, that have been modeled using
an automated technique which ts multiple Gaussians to
the emission lines, taking into account Galactic reddening,
blended iron emission, and Galactic and intrinsic absorp-
tion lines. In this paper we present uniform measurements
of 1607 emission lines including equivalent widths, FWHM
and shifts from the line's expected position and calculate
upper limits for weak lines. We also present the underlying
continuum parameters (slopes and normalization). This is
the third paper in a series of papers aimed at uniformly
measuring emission line properties in large AGN samples.
It has been preceded by a presentation of emission line
properties in 1000 optical spectra from the Large Bright
Quasar Survey (Paper I) and 200 UV spectra observed
by FOS/HST in the pre-COSTAR era (Paper II). All 1387
spectral ts and tabulated results are available at our Web
site. 7 Such large uniformly measured databases will hope-
fully bring us closer to a better understanding of the origin
of the line emitting regions and their relationship to the
central engine.
PJG and JK gratefully acknowledge support provided
by NASA through grant NAG5-6410 (LTSA). PJG and
TA acknowledges support through NASA contract NAS8-
39073 (CXC). MV acknowledges nancial support for Pro-
posal number AR-09549, provided by NASA through a
grant from the Space Telescope Science Institute, which
is operated by the Association of Universities for Research
in Astronomy, Incorporated, under NASA contract NAS5-
26555. We are grateful to Todd Boroson for providing
the FeII optical template. This research was made based
on observations made with the NASA/ESA Hubble Space
Telescope, obtained from the data archive at the Space
Telescope Science Institute and using the Multimission
Archive at the Space Telescope Science Institute (MAST).
STScI is operated by the Association of Universities for
Research in Astronomy, Inc., under NASA contract NAS5-
26555. Support for MAST for non-HST data is provided
by the NASA Oфce of Space Science via grant NAG5-
7584 and by other grants and contracts. This research has
also made use of the NASA/IPAC Extragalactic Database
(NED) which is operated by the Jet Propulsion Labora-
tory, California Institute of Technology, under contract
with the National Aeronautics and Space Administration.
APPENDIX
Notes on individual objects:
0039-5117oa - two power law continua were tted to this spectrum, which were joined at a non-standard wavelength of
3000  A observed frame, for a better continuum t.
0238+1636oa - this BL Lac object is included in our sample as it shows weak emission lines.
0241-0815oa - spectrum spans a large wavelength range from 2200  A to 6800  A, so the power law continuum does not t
the spectrum well, especially at H wavelengths.
0320-1926oa, 1337+2423oa, 1959+4044oa- short spectra with only one standard continuum window, for a better con-
tinuum t we added a non-standard continuum window at red side of MgII.
0742+6510oa-oj - spectra of 10 dierent NLR knots in Mrk 78 (a Seyfert 2) showing interaction of the NLR gas with
the ISM.
1048-2509oa - very weak continuum.
1223+1545oa - very weak continuum.
1252+291