Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/~pgreen/Papers/pg0946_00.ps Äàòà èçìåíåíèÿ: Mon Sep 19 19:15:54 2005 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 00:56:58 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: m 5

L79
The Astrophysical Journal, 533:L79--L82, 2000 April 20
# 2000. The American Astronomical Society. All rights reserved. Printed in U.S.A.
THOMSON THICK X­RAY ABSORPTION IN A BROAD ABSORPTION LINE QUASAR, PG 0946#301
S. Mathur, 1,2 P. J. Green, 2 N. Arav, 3 M. Brotherton, 4 M. Crenshaw, 5 M. deKool, 6 M. Elvis, 2
R. W. Goodrich, 7 F. Hamann, 8 D. C. Hines, 9 V. Kashyap, 2 K. Korista, 10 B. M. Peterson, 1
J. C. Shields, 11 I. Shlosman, 12 W. van Breugel, 13 and M. Voit 14
Received 2000 February 1; accepted 2000 March 1; published 2000 March 24
ABSTRACT
We present a deep ASCA observation of a broad absorption line quasar (BALQSO) PG 0946#301. The source
was clearly detected in one of the gas imaging spectrometers, but not in any other detector. If BALQSOs have
intrinsic X­ray spectra similar to normal radio­quiet quasars, our observations imply that there is Thomson thick
X­ray absorption ( cm #2 ) toward PG 0946#301. This is the largest column density estimated so far
24
N # 10
H
toward a BALQSO. The absorber must be at least partially ionized and may be responsible for attenuation in
the optical and UV. If the Thomson optical depth toward BALQSOs is close to 1, as inferred here, then spectroscopy
in hard X­rays with large telescopes like XMM would be feasible.
Subject headings: galaxies: active --- quasars: absorption lines --- quasars: individual (PG 0946#301) ---
X­rays: galaxies
1. INTRODUCTION
About 10%--15% of optically selected QSOs have optical/
UV spectra showing deep absorption troughs displaced blue­
ward from the corresponding emission lines. These broad ab­
sorption lines (BALs) are commonly attributed to material
flowing toward the observer with velocities of up to #50,000
km s #1 . Broad absorption line quasars (BALQSOs) are prob­
ably normal QSOs viewed at a fortuitous orientation passing
through a BAL outflow, thus implying a BAL ``covering factor''
at least 10%--15% in all QSOs. BALQSOs thus provide a
unique probe of conditions near the nucleus of most QSOs.
The absorbing columns typically inferred from the UV spectra
for the BAL clouds themselves are cm #2 (Ko­
20 21
N # 10 --10
H
rista et al. 1993). It has been noted, however, that UV studies
underestimate the BAL column densities because of saturation
(Korista et al. 1993; Arav 1997; Hamann 1998). BALQSOs,
as a class, show higher optical/UV polarization than other radio­
quiet QSOs (Schmidt & Hines 1999; Ogle et al. 1999). Polar­
ization studies reveal multiple lines of sight through high col­
umn density gas (Goodrich & Miller 1995; Cohen et al. 1995).
With the absorbing column densities as estimated from the
1 Ohio State University, Columbus, OH 43220; smita@astronomy.ohio­
state.edu, peterson@astronomy.ohio­state.edu.
2 Harvard Smithsonian Center for Astrophysics, Cambridge, MA 02138;
pgreen@cfa.harvard.edu, elvis@cfa.harvard.edu, kashyap@cfa.harvard.edu.
3 University of California, Berkeley, Berkeley, CA 94720; arav@mars
.berkeley.edu.
4 National Optical Astronomy Observatories, Tucson, AZ 85726; mbrother@
ohmah.tuc.noao.edu.
5 Catholic University of America and NASA Goddard Space Flight Center,
Greenbelt, MD 20771; hrsmike@hrs.gsfc.nasa.gov.
6 Research School of Astronomy and Astrophysics, Australian National Uni­
versity, Weston Creek, ACT 2611, Australia; dekool@mso.anu.edu.au.
7 W. M. Keck Observatory, Kamuela, Hawaii, HI 96743; goodrich@keck
.hawaii.edu.
8 University of Florida, Gainesville, FL 32611; hamann@astro.ufl.edu.
9 Steward Observatory, University of Arizona, Tucson, AZ 85721; dhines@
as.arizona.edu.
10 Western Michigan University, Kalamazoo, MI 49008; korista@cloud9
.pa.uky.edu.
11 Ohio University, Athens, OH 45701, shields@helios.phy.ohiou.edu.
12 University of Kentucky, Lexington, KY 40506, shlosman@is.pa.uky.edu.
13 University of California, Lawrence Livermore National Laboratory, Liv­
ermore, CA 94550; wil@igpp.llnl.gov.
14 Space Telescope Science Institute, Baltimore, MD 21218; voit@stsci.edu.
earlier UV studies, we would have expected very little soft X­
ray absorption in the BALQSOs. However, BALQSOs are
found to be markedly underluminous in X­rays compared to
their non­BALQSO counterparts (Bregman 1984; Singh, Wes­
tergaard, & Schnopper 1987; Green et al. 1995). Green &
Mathur (1996, hereafter GM96) studied 11 BALQSOs ob­
served with ROSAT and found that just one was detected with
a ox
15 about 2. BALQSOs thus have unusually weak soft X­ray
emission, as evidenced by large a ox (#1.9 compared to
[from Laor et al. 1997] for radio­quiet qua­
a = 1.51 # 0.01
ox
sars). If BALQSOs are indeed normal radio­quiet QSOs, then
their weak X­ray flux is most likely due to strong absorption.
Unfortunately, due to the low observed flux, there are no ob­
served X­ray spectra of BALQSOs to confirm the absorption
scenario, with one exception: the archetype BALQSO PHL
5200 (Mathur, Elvis, & Singh 1995a). The ASCA spectrum of
PHL 5200 is best fit by a power law typical for non­BALQSOs
in the 2--10 keV range, with intrinsic absorption 2--3 orders of
magnitude higher than inferred from UV spectra alone (Mathur
et al. 1995a). However, the PHL 5200 spectrum suffers from
a low signal­to­noise ratio, and while the above was a preferred
fit, a model with no intrinsic absorption also fits the data. Re­
cently Gallagher et al. (1999, hereafter G99) studied a sample
of six new BALQSOs with ASCA, of which two were detected.
G99 derived column densities of # cm #2 to explain
23
5 # 10
the nondetections, even higher than the ROSAT estimates (as­
suming a neutral absorber with solar abundances unless stated
otherwise).
How are the X­ray and UV absorbers related to each other?
Are they both part of the same outflow? If so, then the kinetic
energy carried out is a significant fraction of the bolometric
luminosity of the quasar (see Mathur, Elvis, & Wilkes 1995b
for a discussion). With all QSOs likely to contain a BAL out­
flow, it becomes very important to measure the absorbing col­
umn density accurately to understand the energetics and dy­
namics of quasars. We attempt this with a deep ASCA
observation of a typical BALQSO, PG 0946#301.
15 The slope of a hypothetical power law connecting 2500 A š and 2 keV is
defined as , so that a ox is larger for objects with weaker
a = 0.384 log L /L
ox o x
X­ray emission relative to optical.

L80 THOMSON THICK ABSORPTION Vol. 533
TABLE 1
ASCA Count Rates for PG 0946#301
Background SIS0 SIS0 Hard SIS1 GIS2 GIS3 GIS3 Hard
Background 1 a . . . . . . 0.855 (2 j), !1.2 !0.74 !1.2 !0.75 1.84 (7.9 j) 1.4 (8.7 j)
Background 2 b . . . . . . 1.29 (3 j) 0.51 (2 j), !0.74 !1.2 !0.75 0.52 (2 j), !0.83 0.48 (2 j), !0.63
Note.---Values are given in units of 10 #3 photons s #1 . The significance of detection is given in parentheses. For non­
detections, the 3 j upper limit is given. For 2 j detections, the 3 j upper limit is listed as well. For SIS0 and GIS3, hard
band count rates are listed as well.
a With background from a source­free region on the detector.
b With background from blank sky observations.
2. OBSERVATIONS AND DATA ANALYSIS
2.1. Observations
We observed PG 0946#301 with ASCA (Tanaka, Holt, &
Inoue 1994) on 1998 November 12. ASCA contains two sets
of two detectors: the solid­state imaging spectrometer (SIS)
and the gas imaging spectrometer (GIS). The effective exposure
times in SIS0, SIS1, GIS2, and GIS3 were 72,024, 69,668,
80,910, and 80,896 s, respectively. SIS was operated in 1 CCD
mode with the target in the standard 1 CCD mode position.
GIS was operated in pulse height mode. The data were reduced
and analyzed using FTOOLS and XSELECT in a standard
manner (see ASCA Data Reduction Guide or Mathur et al.
1995a and G99 for details of data reduction).
2.2. Image Analysis
2.2.1. XSELECT Analysis
We used XSELECT to create full and hard (2--9.5 keV) band
images for each of the four detectors. We also created combined
SIS and GIS images. We looked for the target in these images
displayed with SAOIMAGE. While there were sources seen
within the GIS field of view, there was no obvious source seen
at the target position in any of the four detectors. We then
smoothed the images with a Gaussian function of pix­
j = 1--2
els. A faint source at the position of the target was then evident
in the GIS3 hard­band image and a trace of a source was seen
in the full GIS image, but not in any other image. Note that
for a standard pointing position, the target lies closest to the
optical axis in SIS0 and GIS3. GIS3 is more sensitive in hard
X­rays than SIS0. The fact that the source is seen by eye in
the GIS3 detector only suggests that the source is faint with
flux mainly in the hard band.
We extracted the total counts in a circular region with a 3#
radius centered on the source position. Because our source is
observed to be so faint, background subtraction is crucial in
determining the net source count rate, so we have done careful
background subtraction using different background estimates.
Background counts were extracted in two different ways:
(1) from a source­free region on the detector and (2) from
exactly the same region as the source in the blank sky back­
ground files provided by the ASCA guest observer facility. The
significance of the source detection was therefore different for
different background estimates. For SIS, the blank sky back­
ground is underestimated because it is available in the BRIGHT
mode only, while the source counts were extracted in the
BRIGHT2 mode. So the SIS detections are less reliable with
background 2. We found that the source was detected in GIS3
and the GIS3 hard band and is marginally detected in SIS0
(2 j). It was not detected in any other detector in either band­
pass. The significance of detection for the source in different
detectors and the resulting net count rate is given in Table 1.
For nondetections, we give a 3 j upper limit of the count rate
(see G99 for exact formulation of the detection and corre­
sponding count rate estimate).
2.2.2. XIMAGE Analysis
Determination of whether or not the source is detected is
extremely important to our results. As an independent check,
we performed image analysis with XIMAGE (Giommi, An­
gellini, & White 1997), which is designed for detailed image
analysis. The DETECT algorithm in XIMAGE locates point
sources in an image by means of a sliding­cell method. We
used DETECT on all of our images and looked for a source
at the position of the target. Again, we found the source to be
detected in the GIS3 hard band. To minimize the number of
spurious sources detected, the threshold used by DETECT is
somewhat conservative. As a result, sources with intensity just
above the image background can be missed. We found that the
source was detected in the full band GIS image if we lowered
the detection threshold. The source was not detected in other
detectors. These results are consistent with those from the
XSELECT analysis discussed above.
2.2.3. CIAO Analysis
We applied more sophisticated wavelet­based techniques
(Freeman et al. 2000) to provide independent support to the
above detections. Software developed for Chandra Interactive
Analysis of Observations (CIAO) allows us to decompose the
image such that structures at different scales are enhanced. We
analyzed the central 20# region of GIS3 images in both the full
spectral range and in the harder range. Wavelet analysis of the
GIS image at scales approximating the size of the point­spread
function shows that detection of PG 0946#301 is complicated
by the presence of a strong nearby source #5# away. In the
GIS hard band image, this source is significantly weaker, and
we detect PG 0946#301 at a probability of spurious detection
of 10 #4 , with a net count rate of counts
#3
(1.26 # 0.25) # 10
s #1 (90% confidence). This is consistent with the results dis­
cussed above.
2.3. Column Density Constraints
Consistency among the methods discussed above gives us
confidence in our measurements and in our resulting detections
in GIS3 and nondetections in other detectors. If the low ob­
served X­ray count rate is due to intrinsic absorption, we can
estimate the absorbing column density in PG 0946#301. Since
the source did not yield enough net counts in any detector to
perform spectral analysis, we use the method discussed in
GM96 to determine the column density. We first calculate the
flux expected from the source if there was no intrinsic ab­
sorption. This was done using the observed B­magnitude of
the source ( mag) and assuming . The redshift
B = 16.0 a = 1.6
ox
of the source ( ) and the Galactic column density
z = 1.216

No. 2, 2000 MATHUR ET AL. L81
TABLE 2
Column Density Constraints
Detector G Detection a 3 j Lower Limit b 2 j Detection
GIS3 . . . . . . . . . . . . 1.7 0.95 2.1 3.3
2.0 0.52 1.2 1.95
GIS3 hard . . . . . . 1.7 1.2 2.55 3.2
2.0 0.67 1.55 1.95
SIS0 . . . . . . . . . . . . 1.7 1.4 1.42 1.95
2.0 0.9 0.92 1.3
SIS0 hard . . . . . . 1.7 ) 2.12 2.73
2.0 ) 1.42 1.86
Note.---Values are given in units of 10 24 atoms cm #2 .
a If 3 j or better detection (Table 1), either background. Note: the column
densities derived here are lower than the 3 j lower limits because the two
are from two different background measurements; e.g., for GIS3, detections
are with background 1, while the 2 j detections and 3 j lower limits are with
background 2.
b The upper limit on count rate gives the lower limit on the column density.
The 2 j detections and 3 j lower limits are with the same background.
( atoms cm #2 ; Murphy et al. 1996) were taken
20
N = 1.6 # 10
H
into account to predict the 2--10 keV flux in the observed band
(= ergs s #1 cm #2 ). An X­ray power­law slope with
#13
7.2 # 10
photon index was used. We then entered this model
G = 1.7
into the X­ray spectral analysis software XSPEC (Arnaud
1996), with normalization consistent with the expected flux
and simulated spectra using SIS and GIS response matrices.
The response of the telescope and detectors was taken into
account as well. The column density at the redshift of the source
was an additional parameter used in the simulation. If there
was no intrinsic absorption, then the predicted count rate was
found to be typically an order of magnitude larger than the
observed one. We then varied the value of the intrinsic ab­
sorption, keeping the normalization constant, until the predicted
and observed column densities matched. The values of intrinsic
column density estimated in this way are given in Table 2.
This estimate of N H depends upon G and a ox . Given the
observed range of a ox (§ 1), our adopted value of a = 1.6
ox
gives conservative estimates of column densities. X­ray spec­
tral slopes also vary among quasars. So we have estimated N H
for as well as . Flatter spectra result in even
G = 2.0 G = 1.7
higher derived column densities. As shown in Table 2, even
the conservative estimate results in Thomson thick X­ray ab­
sorption in PG 0946#301, i.e., cm #2 . The column
24
N # 10
H
density estimates are consistent with the detection in GIS3 and
nondetection in SIS0.
Alternatively, is it possible that PG 0946#301 (and BAL­
QSOs in general) is intrinsically X­ray weak? Earlier work
(GM96; G99) could not rule out this possibility. To test this,
we estimated the observed SIS0 hard band count rate for flux
consistent with detection in GIS3 hard band, but no intrinsic
absorption. We find that the source would have been detected
in the SIS0 hard band at greater than 8 j significance (with
; 17 j with ). So we conclude that the observed
G = 1.7 G = 2.0
X­ray weakness of BALQSOs is due to absorption and not due
to intrinsic weakness. We cannot, however, rule out the pos­
sibility that the source is intrinsically X­ray weak with an un­
usual spectral shape (turning up at around 10 keV, rest frame).
It is also possible that the observed flux is only the scattered
component, from a line of sight different from the absorbing
material. This is unlikely in PG 0946#301, which is not
strongly polarized ( %; Schmidt & Hines 1999).
0.85 # 0.14
However, if true, it again implies the existence of X­ray--thick
matter along the direct line of sight.
3. DISCUSSION
We have clearly detected the quasar PG 0946#301 in our
deep ASCA observation, and we infer that there is Thomson
thick X­ray absorption ( cm #2 ) toward this BAL­
24
N # 10
H
QSO. The use of a detection, rather than upper limits, to de­
termine the absorption is highly significant. In earlier work,
GM96 and G99 had estimated absorbing column densities of
a few times 10 22 and 10 23 cm #2 , respectively. However, these
were based on nondetections only and hence yielded only lower
limits to the column density. A detection provides a much
stronger estimate.
Assuming that there is indeed Thomson thick matter cov­
ering the X­ray source, can we infer its ionization state? The
X­ray absorber will cover the optical and UV continuum
sources as well, at least partially. If the absorber is completely
neutral, it will result in significant H i opacity, which is not
observed (Arav et al. 1999). If the absorber is completely ion­
ized, then the opacity due to Thomson scattering would be the
same in the optical, UV, and X­rays (up to ). Thus, this
2
m c
e
scenario by itself cannot account for the unusually large values
of a ox . If, on the other hand, the hydrogen is mostly ionized
but there are still some hydrogen­ and helium­like heavy ele­
ments, then photoelectric absorption would still be the domi­
nant mechanism in X­rays. In the optical/UV, a Thomson opac­
ity of 1 would result in attenuation by a factor of 2.7. Based
on a statistical comparison of polarization properties of BAL­
QSOs and unabsorbed, radio­quiet QSOs, Schmidt & Hines
(1999; see also Goodrich 1997) inferred an attenuation factor
of for BALQSOs. This is consistent with the atten­
2.4 # 0.3
uation inferred here for PG 0946#301. The X­ray absorber
thus must be at least partially ionized and may be responsible
for attenuation in the optical and UV.
While PG 0946#301 suffers significant attenuation along
the line of sight, it is not highly polarized. This might be
because the scattering medium is not present or well placed in
PG 0946#301 to produce high polarization. The scattering
region may be attenuated as much or more than the direct view.
Alternatively, it might be a geometric effect in which a large
part of the UV continuum is not covered by the X­ray absorber,
so it is unattenuated. Whether the X­ray absorber has an ion­
ization state overlapping the range of UV BALs and whether
the outflow velocities are similar remain outstanding questions.
It is possible that the X­ray absorber is stationary, at the base
of winds producing BALs. The X­ray continuum source might
be preferentially covered. X­ray spectroscopy is necessary to
better probe the nuclear region in BALQSOs. For PG
0946#301, we predict about 0.015 counts s #1 with the XMM
PN. A reasonable spectrum may be obtained in about 70 ks.
We thank K. Arnaud and L. Angellini for help with XIM­
AGE. This work is supported in part by NASA grants NAG5­
8360 (P. J. G., S. M.), NAG5­3249 (S. M.), and NAG5­3841
(I. S.). The work by W. v. B. at IGPP/LLNL was performed
under the auspices of the US Department of Energy under
contract W­7405­ENG­48.

L82 THOMSON THICK ABSORPTION Vol. 533
REFERENCES
Arav, N. 1997, in ASP Conf. Ser. 128, Mass Ejection from AGN, ed. N. Arav,
I. Shlosman, & R. J. Weymann (San Francisco: ASP), 208
Arav, N., Korista, K. T., de Kool, M., Junkkarinen, V. T., & Begelman, M. C.
1999, ApJ, 516, 27
Arnaud, K. A. 1996, in ASP Conf. Ser. 101, Astronomical Data Analysis
Software and Systems V, ed. G. Jacoby & J. Barnes (San Francisco: ASP),
17
Bregman, J. M. 1984, ApJ, 276, 423
Cohen, M. H., Ogle, P. M., Tran, H. D., Vermeulen, R. C., Miller, J. S.,
Goodrich, R. W., & Martel, A. R. 1995, ApJ, 448, L77
Freeman, P. E., Kashyap, V., Rosner, R., & Lamb, D. Q. 2000, ApJ, submitted
Gallagher, S., Brandt, W. N., Sambruna, R., Mathur, S., & Yamasaki, N. 1999,
ApJ, 519, 549 (G99)
Giommi, P., Angelini, L., & White, N. 1997, The XIMAGE Users' Guide
(Greenbelt: NASA/GSFC)
Goodrich, R. 1997, ApJ, 474, 606
Goodrich, R., & Miller, J. S. 1995, ApJ, 448, L73
Green, P. J., et al. 1995, ApJ, 450, 51
Green, P. J., & Mathur, S. 1996, ApJ, 462, 637 (GM96)
Hamann, F. 1998, ApJ, 500, 798
Korista, K. T., Voit, G. M., Morris, S. L., & Weymann, R. J. 1993, ApJS, 88,
357
Laor, A., Fiore, F., Elvis, M., Wilkes, B. J., & McDowell, J. C. 1997, ApJ,
477, 93
Mathur, S., Elvis, M., & Singh, K. P. 1995a, ApJ, 455, L9
Mathur, S., Elvis, M., & Wilkes, B. 1995b, ApJ, 452, 230
Murphy, E. M., Lockman, F. J., Laor, A., & Elvis, M. 1996, ApJS, 105, 369
Ogle, P., et al. 1999, ApJS, 125, 1
Schmidt, G., & Hines, D. C. 1999, ApJ, 512, 125
Singh, K. P., Westergaard, N. J., & Schnopper, H. W. 1987, A&A, 172, L11
Tanaka, Y., Holt, S. S., & Inoue, H. 1994, PASJ, 46, L37