Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/~garcia/apj.555.477.ps Äàòà èçìåíåíèÿ: Wed Feb 9 00:13:34 2011 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 21:18:54 2012 Êîäèðîâêà: koi8-r Ïîèñêîâûå ñëîâà: crab nebula

THE ASTROPHYSICAL JOURNAL, 555 477õ482, 2001 July 1
( 2001. The American Astronomical Society. All rights reserved. Printed in U.S.A.
COMPLETE AND SIMULTANEOUS SPECTRAL OBSERVATIONS OF THE BLACK HOLE
X­RAY NOVA XTE J1118]480
J. E. MCCLINTOCK,1 C. A. HASWELL,2 M. R. GARCIA,1 J. J. DRAKE,1 R. I. HYNES,3 H. L. MARSHALL,4 M. P. MUNO,4
S. CHATY,2 P. M. GARNAVICH,5 P. J. GROOT,1 W. H. G. LEWIN,4 C. W. MAUCHE,6 J. M. MILLER,4 G. G. POOLEY,7
C. R. SHRADER,8 AND S. D. VRTILEK1
Received 2000 October 28 ; accepted 2001 March 5
ABSTRACT
The X­ray nova XTE J1118]480 su+ers minimal extinction (b \ 62¡) and therefore represents an out­
standing opportunity for multiwavelength studies. Hynes et al. conducted the ïrst such study, which was
centered on 2000 April 8 using UKIRT, EUV E, HST , and RXT E. On 2000 April 18, the Chandra
X­Ray Observatory obtained data coincident with a second set of observations using all of these same
observatories. A 30 ks grating observation using Chandra yielded a spectrum with high resolution and
sensitivity covering the range 0.24õ7 keV. Our near­simultaneous observations cover B80% of the elec­
tromagnetic spectrum from the infrared to hard X­rays. The UV/X­ray spectrum of XTE J1118]480
consists of two principal components. The ïrst of these is an B24 eV thermal component that is caused
by an accretion disk with a large inner disk radius : The second is a quasi power­law com­
Z35R Schw .
ponent that was recorded with complete spectral coverage from 0.4 to 160 keV. A model for this two­
component spectrum is presented in a companion paper by Esin et al.
Subject headings : accretion, accretion disks õ binaries : close õ stars : individual (XTE J1118]480) õ
ultraviolet : stars õ X­rays : stars
1. INTRODUCTION
X­ray novae (a.k.a. soft X­ray transients) are a type of
X­ray binary that typically remains quiescent for decades
before brightening by as much as 107 in X­rays in a week. In
outburst, the 2õ10 keV X­ray spectrum of most X­ray novae
is dominated by thermal emission from the inner accretion
disk. However, we now know of ïve X­ray novae that
have failed to show this soft thermal component during out­
burst (Brocksopp et al. 2001). One of these ïve is
XTE J1118]480. This source is further distinguished by its
strikingly low X­rayõtoõoptical ÿux ratio (Hynes et al.
2000).
XTE J1118]480 was discovered on 2000 March 29
(Remillard et al. 2000). During the month of March, the
Rossi X­Ray T iming Explorer (RXT E) All­Sky Monitor
data show that the X­ray intensity of the source increased
steadily to B35 mcrab (2õ12 keV). Thereafter, the intensity
remained near that level for about three months before
declining abruptly. The optical counterpart brightened
from quiescence by about 6 mag to V B 13 (Uemura et al.
2000). Optical observations in outburst and quiescence
1 Harvard­Smithsonian Center for Astrophysics, 60 Garden Street,
Cambridge, MA 02138 ; jem=cfa.harvard.edu, mgarcia=cfa.harvard.edu,
jdrake=cfa.harvard.edu, pgroot=cfa.harvard.edu, svrtilek=
cfa.harvard.edu.
2 Department of Physics and Astronomy, The Open University, Walton
Hall, Milton Keynes MK7 6AA, UK ; c.a.haswell=open.ac.uk, s.chaty=
open.ac.uk.
3 Department of Physics and Astronomy, University of Southampton,
Southampton SO17 1BJ, UK rih=astro.soton.ac.uk.
4 Center for Space Research, MIT, Cambridge, MA 02139
hermanm=space.mit.edu, muno=space.mit.edu, lewin=space.mit.edu,
jmm=space.mit.edu.
5 Physics Department, University of Notre Dame, Notre Dame, IN
46556 ; pgarnavi=nd.edu.
6 Lawrence Livermore National Laboratory, L­43, 7000 East Avenue,
Livermore, CA 94550 ; mauche=cygnus.llnl.gov.
7 Mullard Radio Astronomy Observatory, Cavendish Laboratory,
Madingley Road, Cambridge CB3 0HE, UK ; ggp1=cam.ac.uk.
8 Laboratory for High­Energy Astrophysics, NASA Goddard Space
Flight Center, Greenbelt, MD 20771 ; shrader=grossc.gsfc.nasa.gov.
conïrm that the orbital period is 4.08 hr (Patterson 2000 ;
Uemura et al. 2000 ; McClintock et al. 2001 ; Wagner et al.
2001). A radio counterpart has also been observed (Pooley
& Waldram 2000). The most uncommon property of XTE
J1118]480 is its exceptionally high galactic latitude,
b \]62¡ and its correspondingly low reddening :
E(B[V ) B 0.013 mag cm~2 ; Hynes et al.
(N H B 7.5 ] 1019
2000). XTE J1118]480 is the least reddened of all known
X­ray binaries.
Recently, dynamical measurements by two groups have
established that XTE J1118]480 has a very large mass
function, which sets a hard lower limit of 6 on the mass
M _
of the compact X­ray source (McClintock et al. 2001 ;
Wagner et al. 2001). Since this greatly exceeds the maximum
allowed stable mass of a neutron star in general relativity
(Rhoades & Ruffini 1974), we refer to the compact primary
as a black hole.
The ïrst epoch of an intensive multiepoch, multi­
wavelength observing campaign was reported by Hynes et
al. (2000). Here we report on the second­epoch obser­
vations, centered on 2000 April 18, which are unique in
including a 0.24õ7 keV grating spectrum obtained using the
Chandra X­Ray Observatory (Chandra). This spectrum is
very important and forces modiïcation of the conclusions
reached by Hynes et al. (2000). Our near­simultaneous
observations cover B80% of the electromagnetic spectrum
from the infrared to hard X­rays ; these multiwavelength
data are modeled in a companion paper by Esin et al.
(2001).
2. NEAR­SIMULTANEOUS OBSERVATIONS
A journal of the near­simultaneous observations made on
or near April 18 is given in Table 1. The April 18 Extreme
Ultraviolet Explorere (EUV E) data are discussed exten­
sively by Hynes et al. (2000), but they were not used in their
spectral energy distribution (SED). The only data in
common between the SED presented by Hynes et al. and
the SED presented here is the April 18 UKIRT data. In the
following, we brieÿy discuss each data set in turn.
477

478 MCCLINTOCK ET AL. Vol. 555
TABLE 1
OBSERVATIONS ON OR NEAR 2000 APRIL 18 UT
Observation Interval Net Observing Time
Observatory Instrument Bandpass log (l) (UT) (ks)
RXT E . . . . . . . . . . . . . . . HEXTE 15õ200 keV 18.56õ19.68 18 Apr 19 28õ18 Apr 23 : 00 1.1
PCA 2.5õ25 keV 17.78õ18.78 18 Apr 19 28õ18 Apr 23 : 00 2.8
Chandra . . . . . . . . . . . . . LETG/ACIS­S 0.24õ7 keV 16.76õ18.23 18 Apr 18 16õ19 Apr 02 : 16 27.2
EUV E . . . . . . . . . . . . . . . SW 0.10õ0.17 keV 16.38õ16.61 16 Apr 21 34õ19 Apr 14 : 28 81.1
HST . . . . . . . . . . . . . . . . . STIS 1155õ10250 A# 14.47õ15.41 18 Apr 13 40õ18 Apr 17 : 44 6.2
UKIRT . . . . . . . . . . . . . . IRCAM/TUFTI 1õ5 k 13.78õ14.48 18 Apr 12 00 0.01õ0.06
Ryle Telescope . . . . . . 15.2 GHz receiver 2.0 cm 10.18 18 Apr 17 11õ18 Apr 18 : 36 5.1
RXT E spectra were produced using 128 channel Pro­
portional Counter Array (PCA) data collected with PCUs
0, 2, and 3, and 64 channel data collected with the HEXTE
(High­Energy X­Ray Timing Experiment) clusters A and B.
Following the suggestions of Jahoda (2000), we have selec­
ted only the top layers of each PCU and have combined the
data for all three PCUs. A response matrix was generated
using version 2.43 of ```` pcarsp,îî and a background esti­
mate was generated using the 2000 January 31 blank­sky
model for gain epoch 4 combined with version 2.1e of
```` pcabackest.îî A 1% systematic error was added to the sta­
tistical error estimate and each model was ït between 2.5
and 25 keV. This systematic error was required because the
statistical uncertainty for the PCA is signiïcantly less than
the uncertainty in the PCA response matrix, which is esti­
mated to be about 1% (Jahoda 2000). For the HEXTE we
used the standard response matrices and modeled the data
between 15 and 200 keV, allowing for a constant normal­
ization between the PCA data and the data from the
two HEXTE clusters. No systematic errors were added to
these data. The background was estimated using the
script ```` hxtback îî from FTOOLS version 5. The combined
PCA and HEXTE data are compatible with a power­law
model for the emission. The best­ït photon index was
!\ 1.782 ^ 0.005 with a 2õ200 keV ÿux of 4.2 ] 10~9 ergs
cm~2 s~1 and a s2 of 164 for 139 degrees of freedom (dof).
The photon index is consistent with the value measured 10
days earlier : !\ 1.8 ^ 0.1 (Hynes et al. 2000). A Gaussian
feature at 6.4 keV was evident in the residuals to the ït
(s2 \ 152 for 136 dof) with an equivalent width (EW) of 40
eV. This feature is very probably not caused by the source
for two reasons. First, its strength is consistent with a
feature seen in ïts to the spectrum of the Crab Nebula,
where no Fe­K line is expected ; thus it is probably a system­
atic feature in the response matrices that we used (Jahoda
2000). Second, the feature does not appear in the Chandra
grating data described directly below. For an assumed
narrow 6.4 keV line, *E/E\ 0.04, the Chandra data imply a
3 p upper limit of EW\ 24 eV, which strongly rules against
the RXT E candidate Fe line. For a broader assumed line
width, *E/E\ 0.10, the Chandra 3 p upper limit is
EW\ 38 eV, which marginally rules against the RXT E
spectral feature. On the other hand, a very broad line, with
a width comparable to the energy resolution of the PCA
detector (*E/E B 0.2) and an EW of 40 eV would have
escaped detection by Chandra.
Chandra observations were performed using the Low
Energy Transmission Grating (LETG) and the ACIS­S
detector, which yielded a spectrum from 0.24 to 7 keV
(1.8õ52 with a resolution of about 0.04 The ```` Level 1 îî
A# ) A# .
data were reduced using custom IDL procedures in several
steps : (1) sky coordinates were transformed to grating coor­
dinates by correcting for spacecraft roll ; (2) the location of
zeroth order was determined by ïtting a pair of one­
dimensional Gaussian proïles ; (3) pulse height was con­
verted to energy using preÿight gains and then
(E PH )
corrected (node by node) to match the energies inferred
from the dispersion of the LETG ; and (4) events were selec­
ted using where was [0.70, 1.15]
r l \E PH /E LETG \ r h , r l , r h
for ]1 events, [0.80, 1.20] for [1 events with j \ 30 and
A# ,
[0.65, 1.25] for [1 events with j [ 30 These selections
A# .
net greater than 99% of the observed counts. All events
within of the dispersion line were included, except for
2A. 5
the events near detector gaps, which were ignored. Any
aperture size between and 3A would have yielded the
1A. 5
same results. No background was subtracted because the
background rates were negligible compared to the source
rates (\1%), except near the C­K edge, where the back­
ground rate was up to 5% of the source rate. The events
were binned and the e+ective areas were integrated over the
bins. After correcting for the fraction of counts dispersed by
the LETG ïne support structure, we estimate that the ÿuxes
are accurate to D5% in the 2õ5 keV range. XTE
J1118]480 was the brightest source observed with this
grating and detector combination, so these data are being
used extensively to verify and update the instrumental area,
which could still have systematic errors of order 10%õ20%
in the 0.2õ1.0 keV band (Marshall 2001).9 The largest uncer­
tainties are in the 0.3õ0.5 keV band. The spectrum was
searched at high resolution (0.01 for lines and edges with
A# )
assumed widths comparable to the instrumental resolution.
None was found apart from two weak, unidentiïed lines,
13.24 (0.936 keV) and 9.36 (1.325 keV), which were
A# A#
detected at signiïcance levels of 5.6 and 4.6 p, respectively ;
the latter feature was detected only in the third­order spec­
trum. The probability of ïnding a feature in one bin at º4.6
p among the 2500 bins searched is only 0.5%. The complete
high­resolution spectrum is shown in Figure 1. Given the
near absence of spectral features, the data were binned
heavily on a logarithmic grid (*E/E \ 0.02) to provide an
accurate measure of the continuum. On timescales ks,
Z1
the total X­ray intensity was constant to D3% during the
27 ks observation. For the energy range 2õ7 keV, the
power­law photon index was !\ 1.77 ^ 0.04, which is con­
sistent with the RXT E values quoted above.
EUV E observations were performed 3 arcmin o+­axis to
extend the wavelength coverage of the EUV E short­
wavelength spectrometer (SW) shortward of its normal
cuto+ (Marshall et al. 1996). Data from the SW spectro­
9 Marshall, H. L. 2001, unpublished, (LETG/ACIS­S calibration web
site http ://space.mit.edu/CXC/calib/letg -- acis/ck -- cal.html)

No. 1, 2001 BLACK HOLE X­RAY NOVA XTE J1118]480 479
FIG. 1.õFirst­order ACIS/LETG count spectrum binned at 0.025 as
A#
obtained in a 27 ks observation with Chandra. The dashed line corre­
sponds to a power­law model spectrum folded through the instrument
response with !\ 1.78 (° 3) and cm~2. Apart from a
N H \ 1.3 ] 1020
weak, unidentiïed feature centered at 13.24 (° 2), this ïrst­order spec­
A#
trum is devoid of lines.
meter were binned in 2 intervals and corrected for inter­
A#
stellar (IS) absorption using the compilation of H and He
photoionization cross sections of Rumph, Bowyer, &
Vennes (1994), with a mixture of neutral hydrogen to
neutral and ionized helium in the ratio 1 : 0.1 : 0.01. Fortu­
nately, the shape of the extinction curve is largely indepen­
dent of these details of the gas mix. Flux calibration was
achieved using the ïlter­corrected exposure time and the
o+­axis e+ective area curve of H. Marshall et al. (2001, in
preparation). An independent analysis of these data is dis­
cussed by Hynes et al. (2000) ; there are no signiïcant di+er­
ences between their extracted spectrum and the one
presented herein.
Hubble Space Telescope (HST ) observations were made
using the Space Telescope Imaging Spectrograph (STIS)
and the E140M, E230M, G430L, and G750L gratings. An
average calibrated spectrum for April 18 was constructed
from standard HST pipeline data products. The spectrum
exhibits emission lines of Ha (weak), He II (1640, 4686 A# ),
Si IV (1394, 1403 and N V (1239, 1243 as well as Lya,
A# ) A# ),
higher order Balmer lines, and the Balmer jump in absorp­
tion. To identify the continuum spectrum more clearly, and
remove near­infrared (NIR) fringing, the spectra were
rebinned with the regions dominated by Lya and N V
masked out.
The 3.8 m United Kingdom Infrared Telescope (UKIRT)
was used to make NIR photometric observations at
JHKL M with IRCAM/TUFTI. The measured magnitudes
and comments on the data analysis can be found in Hynes
et al. (2000).
The Ryle Telescope was used to monitor the ÿux density
at 15.2 GHz using techniques similar to those described in
Pooley & Fender (1997). The phase calibrator used was
J1110]440, and the ÿux­density scale was established by
observations of 3C 48 and 3C 286.
3. SUPPORTING OBSERVATIONS
The intensity of XTE J1118]480 was relatively constant
during the 81 ks EUV E observation (Hynes et al. 2000) and
during the 27 ks Chandra observation (° 2). In addition, the
source was very stable on timescales hours at all wave­
Z
lengths for weeks before and after our April 18 observing
campaign. To illustrate this constancy, Figure 2 shows
selected data for a D7 week period around the time of April
18, which is indicated by a dashed line in the ïgure. Also,
indicated by an arrow in Figure 2a is the time of the April 8
observing campaign, which occurred near the end of the
rising phase of the outburst (Hynes et al. 2000).
The 2õ12 keV X­ray data shown in Figure 2a are consis­
tent with a mean intensity of counts s~1
I X \ 2.89 ^ 0.41
(rms), which corresponds to a relative intensity of 38 mcrab.
Optical photometric data are shown in Figure 2b. For these
data, we ïnd mean magnitudes of B\ 13.02 ^ 0.04 (rms)
and I \ 12.68 ^ 0.05 (rms). Figure 2c shows 15.2 GHz radio
data with a mean ÿux density of mJy (rms).
S l \ 8.68 ^ 0.87
Thus the rms variability of the source during this 7 week
interval was 14% in the X­ray, 5% in the optical, and 10%
in the radio. Moreover, much of this apparent variability is
caused by measurement error. The typical measurement
error for the ASM X­ray detectors was ^10% (Fig. 2a), and
for the Ryle Telescope it was ^5% (Fig. 2c). Thus we con­
clude that for several weeks around April 18, XTE
J1118]480 was stable in intensity to better than 10% on
timescales of D1 day in the radio, optical, and X­ray bands.
Because of the long­term stability of the source, we have
summed 16 consecutive RXT E spectra obtained between
April 13.39 and May 15.38 in order to improve the counting
statistics above 100 keV. We used the analysis techniques
described in ° 2. These 16 observations can all be ït individ­
ually by a power­law model with photon indices that range
between 1.77 and 1.81. The summed spectrum has a total
integration time of 45.9 ks in the PCA and 16.8 ks in the
HEXTE (only cluster A was used). A power­law model ït
the summed spectrum with !\ 1.779 ^ 0.003 for a s2 of
FIG. 2.õSeven­week record of the intensity of XTE J1118]480 mea­
sured at X­ray, optical, and radio frequencies. The dashed line corresponds
to 2000 April 18.5 UT, the nominal time of the intensive observing cam­
paign reported on herein. The arrow is drawn at 2000 April 8.5 UT, the
nominal time of the observations conducted by Hynes et al. (2000). (a) 2õ12
keV RXT E ASM light curve. The data shown are daily­average intensities,
which were obtained from the MIT ASM web page. (b) UBV RI optical
photometric data were obtained on 17 occasions using the 1.2 m telescope
at the F. L. Whipple Observatory. Only the B and I light curves are shown
here. (c) 15.2 GHz radio light curve obtained using the Ryle Telescope.

HST
EUVE
CXO
RXTE
480 MCCLINTOCK ET AL. Vol. 555
116 for 95 dof. The best­ït power law that includes a simple
exponential cuto+ at high energies, i.e., N(E) PE~!e~E@Ec,
yielded a cuto+ energy of 940 keV for a s2 of 103 for 94 dof.
Arbitrarily ïxing the cuto+ energy at 300 keV produced an
unacceptable ït with a s2 of 150 for 95 dof. Of course, these
results do not rule against an abrupt, breaking cuto+ at
energies of D150 keV since these energies are near the limit
of HEXTEîs response. Compared to the simple power­law
model, a somewhat poorer ït to the data was achieved
using a Comptonization model (```` compTT îî in XSPEC ;
Arnaud & Dorman 2000 ; Titarchuk 1994) with an electron
temperature of 207 keV and an optical depth of 1.0 for a s2
of 125 for 94 dof.
4. THE SPECTRAL ENERGY DISTRIBUTION
The most difficult problem in constructing the SED is in
properly correcting the EUV ÿuxes for IS absorption. It is
not possible to obtain a precise measurement of the IS
N H ,
column depth. For example, 21 cm measurements imply
cm~2 (Dickey & Lockman 1990). On the
N H \ 1.34 ] 1020
other hand, the COBE maps of dust IR emission imply
cm~2 (Schlegel, Finkbeiner, & Davis
N H \ 0.67 ] 1020
1998). However, neither of these values can be considered
secure, since can vary by a factor D2 on much smaller
N H
angular scales than those probed by the surveys just men­
tioned (Faison et al. 1998). A line­of­sight estimate of N H
using the Ca II lines implies a high value : N H \ 2.8 ] 1020
cm~2 ; however, the uncertainty in is at least 0.2
log (N H )
(Dubus et al. 2001). The source was too faint to make a 21
cm absorption measurement feasible.
There is an additional complication with the IS absorp­
tion in the EUV : A signiïcant fraction of the absorption
near 0.1 keV is attributable to neutral and ionized helium ;
however, there are no existing data on the absorbing
columns for this line of sight. Rigorous inclusion of He
absorption based on only the neutral hydrogen column
would then require knowledge of both the line­of­sight
hydrogen and helium ionization fractions, as well as the
abundance of helium relative to hydrogen. Consequently,
we estimate the IS absorption, parameterized by by
N H ,
assuming the neutral and ionized He number densities rela­
tive to that of neutral hydrogen stated in ° 2. Given all of
these uncertainties, we, like Hynes et al. (2000), chose to
estimate by examining the SED itself.
N H
In Figure 3 we show four realizations of the SED for four
assumed values of Here we have omitted the infrared
N H .
and radio data. At energies above a few tenths of a keV, the
spectrum can be described in terms of a power law with
varying spectral index as follows. At the highest energies,
the photon index is B1.78 (°° 2 and 3). At energies below
D2 keV, the spectrum becomes harder and the photon
index approaches B1.5. It is possible that the systematic
errors discussed in ° 2 contribute to this hardening of the
spectrum ; however, it is unlikely that they can account for
all of it.
The only appreciable gap in the SED is centered at
log (l) B 16.0, where the ISM is opaque. We now focus on
the question of how to connect the HST data with the
EUV E data across this unobserved region. It is apparent
from Figure 3 that the dereddened EUV E spectrum is very
sensitive to the choice of and is consequently very uncer­
N H
tain. Therefore, we ïrst consider the high­quality HST spec­
trum, which depends weakly on the choice of reddening. We
can be conïdent that the HST spectrum is made up of
FIG. 3.õFour realizations of the SED for the indicated values of the
column density. Inspection of panels (a)õ(d) shows that the HST UV spec­
trum and the low­energy portion of the Chandra X­ray spectrum are only
mildly sensitive to the choice of However, the EUV E spectrum is
N H .
extremely sensitive. Solid curves represent multicolor disk blackbody
model spectra, which have been normalized to the HST ÿux at
log (l) \ 15.1. Models are speciïed by the temperature at the inner edge of
the accretion disk, The four models in each panel are equally spaced by
T in .
2 eV, and the lowest and highest values of are indicated in the ïgure.
kT in
substantial thermal emission because the Balmer jump in
absorption is apparent at log (l) B 14.9. In the following,
we assume that the UV emission is caused chieÿy by the
accretion disk (see ° 5).
We therefore consider the simple model spectrum of a
steady accretion disk as implemented in XSPEC (Arnaud &
Dorman 2000 ; Mitsuda et al. 1984 ; Makishima et al. 1986).
We do not include a power­law component, which would
increase the EUV E model ÿuxes somewhat. These multi­
color disk blackbody spectra are meant to be illustrative ;
they are not ïts to the data. Each disk spectrum shown in
Figure 3 is speciïed by two parameters : (1) the temperature
at the inner edge of the disk, and (2) the normalization,
T in ,
which we arbitrarily take to be the dereddened HST ÿux at
log (l) \ 15.1. We ïrst consider the SED for N H \ 1.6
] 1020 cm~2 (Fig. 3d). We ïnd that no simple disk spec­
trum can match the steeply rising EUV E spectrum, and the
mismatch is only worse for higher values of We there­
N H .
fore do not consider values of cm~2
N H Z 1.6 ] 1020
further.
We next examine cm~2, which is illus­
N H \ 0.75 ] 1020
trated in Figure 3a. By considering the EUV E data as part
of the nonthermal spectrum (see Hynes et al. 2000), it is
possible to accommodate a thermal disk component with
eV. However, the comparison of the model to the
kT in [ 12
data is somewhat unsatisfying for two reasons. First, Hynes
et al. made the reasonable assumption that the EUV and

No. 1, 2001 BLACK HOLE X­RAY NOVA XTE J1118]480 481
FIG. 4.õEnlargement of the SED corrected for the adopted column
depth of cm~2. This is the same SED shown in Fig. 3c,
N H \ 1.3 ] 1020
except that the infrared data are included here. The inset shows all of the
same data plus the 15.2 GHz radio data for April 18. The following com­
ments about the individual data sets apply also to Fig. 3 : (1) UKIRT ÿux
densities reported by Hynes et al. (2000) have been multiplied by 0.8 to
force them to approximately match the HST spectrum ; the discrepancy
may be because of the short UKIRT exposures and the rapid optical
variability of the source (Patterson 2000). (2) HST data were dereddened
(Cardelli, Clayton, & Mathis 1989 ; Predehl & Schmitt 1995). No ad hoc
corrections have been applied to these data error bars are plotted on every
eighth point. (3) EUV E data become quite uncertain near the lowest
energy (0.10 keV) error bars are displayed on approximately every third
data point. (4) Chandra LETG spectrum has very small statistical errors
error bars are drawn on every tenth data point. The spectrum shown here
is truncated at 6 keV ; otherwise, no ad hoc corrections have been made to
this spectrum. (5) RXT E data begin at 3 keV and thus overlap substan­
tially with the LETG spectrum. Small ïlled circles without error bars show
the spectrum for April 18. The summed spectrum (° 3) is plotted as open
circles with error bars. The statistical signiïcance of the last two data
points in the summed spectrum are 3.8 p at 154 keV and 2.8 p at 162 keV.
Since Crab observations indicate that RXT E systematically produces
ÿuxes that are high by about 15%, and since some modest variability is
expected, the ÿuxes for April 18 were multiplied by 0.84 to match the
overlapping LETG ÿuxes ; similarly, the summed­spectrum ÿuxes were
multiplied by 0.91.
hard X­ray ÿuxes were approximately ïtted by the same
power law ; however, the Chandra data complicate this
picture by revealing a wider range of power­law slopes.
Second, the structure in the EUV E spectrum is not easily
explained, although it may possibly be attributable to the
presence of a warm absorbing medium around the source
(see Esin et al. 2001). In short, we consider N H \ 0.75
] 1020 cm~2 a viable choice ; however, we do not favor it.
Even lower values of appear less likely since at long
N H
wavelengths the EUV E spectrum plunges downward
(Hynes et al. 2000) and has the wrong inÿection to join up
with the model disk spectra.
We favor the two remaining spectra shown in Figures 3b
and 3c. First consider the case cm~2 (Fig.
N H \ 1.3 ] 1020
2c). This case provides the closest match between one of the
model disk spectra eV) and the EUV E spectrum.
(kT in \ 24
cm~2 probably corresponds to an upper
N H \ 1.3 ] 1020
limit on the value of since there may be a signiïcant
N H
nondisk component of UV emission at log (l) \ 15.1 (see °
5). In this case, the disk spectra in Figure 3c would need to
be renormalized downward, thereby creating some dis­
agreement between the data and the models : i.e., the EUV E
data would then rise more steeply than the models (cf.
Fig. 3d).
Finally, consider the spectrum for cm~2
N H \ 1.0 ] 1020
(Fig. 3b). At short wavelengths, the EUV E spectrum con­
forms closely to the model for eV, whereas at
kT in \ 22
longer wavelengths it straddles the models for kT in \ 18õ22
eV. If the normalization of the disk models were to be cor­
rected downward (as mentioned above), this model would
provide a better description of the data. In short, we con­
clude that the most probable value of the column density
lies in the range cm~2. In the follow­
N H \ 1.0õ1.3 ] 1020
ing, we adopt cm~2 and the version of the
N H \ 1.3 ] 1020
SED shown in Figure 3c. This spectrum, including the infra­
red and radio data, is reproduced in more detail in Figure 4 ;
see the legend for further details about the individual data
sets. The data shown in Figure 4 are available in digital
form by request : for the HST data contact C. Haswell or
R. Hynes ; for the remaining data contact J. McClintock.
5. DISCUSSION
Most black hole X­ray novae in outburst reach the high/
soft state (van der Klis 1994) and spend considerable time
there ; consequently, most X­ray novae are observed in this
state. The high/soft state is dominated by a D1 keV
blackbody­like spectral component, which is widely attrib­
uted to an accretion disk with its inner edge at or near
the radius of the inner­
R in \ 3R
Schw (R
Schw \ 2GM/c2),
most stable orbit (Tanaka & Lewin 1995). A very simple
and successful model for this thermal disk spectrum is the
multicolor disk blackbody model (° 3 ; Mitsuda et al. 1984 ;
Makishima et al. 1986).
A few X­ray novae have been observed only in the low/
hard state : e.g., V404 Cyg, GRO J0422]32, and
GS1354[64 in 1997 (Brocksopp et al. 2001). A number of
canonical X­ray novae (e.g., Nova Mus 1991 and
GS2000]25) and Cyg X­1 have also been observed in the
low/hard state (Tanaka & Lewin 1995). For the sources
observed in this state, the nonstellar optical/UV spectrum
has provided evidence for the presence of an outer accretion
disk. However, for these sources it has not been possible to
determine directly the properties of their inner disks, which
emit in the EUV and soft X­ray bands, because of the high
IS column depths : cm~2. Observations of XTE
N H Z 1021
J1118]480 with cm~2 provide the ïrst
N H B 1.3 ] 1020
detection of radiation from the inner accretion disk for an
X­ray nova in the low/hard state.
A simple explanation for the low thermal temperature in
XTE J1118]480 is that the inner edge of the accretion disk
is quite far out compared to This is the
R in \ 3R
Schw .
expected state of a+airs in some models of the low/hard
state (e.g., Esin, McClintock, & Narayan 1997). Moreover,
there is good indirect evidence based on Compton reÿection
models that such large, cool disks do exist in the low/hard
state of several systems (e.g., see Gierlinski et al. 1997 ;
Miller et al. 2001).
Based on our adopted value of (° 4) and a distance of
N H
1.8 kpc (McClintock et al. 2001 ; Wagner et al. 2001), we can
make a rough estimate of the inner disk radius for kT in \ 24
eV (Fig. 3c) as follows. For an intensity of 38 mcrab and a
photon spectral index of 1.78 (° 3), the 1õ160 keV X­ray

482 MCCLINTOCK ET AL.
luminosity is ergs s~1. For an assumed
L X B 1.2 ] 1036
value of the accretion efficiency, v, the X­ray luminosity
determines the mass accretion rate, and hence
M\ L X /vc2,
the inner radius of a steady state disk : R in \
(e.g., Frank, King, & Raine 1992). For
(3GM X M0 /8npT in 4 )1@3
a canonical efficiency of 10%, we ïnd and
R in \ 60R Schw ,
for an advection­dominated ÿow with an assumed efficiency
of 0.1% we ïnd A comparable value for
R in \ 285R
Schw . R in
can be obtained by using the normalization constant deter­
mined for the multicolor disk blackbody model, which
depends only on the solid angle subtended by the inner disk
and the inclination of the disk (Arnaud & Dorman 2000).
Assuming that the ÿux at log (l) \ 15.1 is caused solely by
the disk (see below) and assuming a high inclination, i \ 80¡
(Wagner et al. 2001), we ïnd for the model
R in \ 34R Schw
with kT \ 24 eV shown in Figure 3c.
We note that the radio/IR data and the HST data below
the Balmer jump can all be reasonably well ït by a single
power law, which steepens at longer radio
lF l P l1.10,
wavelengths (Hynes et al. 2000). Thus, in the IR/radio band
there is an additional component of emission above that
expected for a multicolor disk blackbody spectrum (Hynes
et al. 2000 ; Esin et al. 2001). This component may be associ­
ated with an outÿow from the system (Esin et al. 2001 ;
Fender 2001).
Could this IR/radio component a+ect our earlier conclu­
sions on the temperature of the thermal accretion disk com­
ponent? In ° 4 we assumed that the UV ÿux is dominated
by the spectrum of the accretion disk. This assumption is
supported by the presence of a B0.09 mag Balmer jump in
the spectrum. Unfortunately, the strength of the Balmer
jump in absorption cannot be used to determine quantitat­
ively the fraction of the total ÿux that is caused by the
accretion disk. In a disk there are several conditions that
can reduce or even eliminate the Balmer jump : e.g., high
inclination and X­ray heating (la Dous 1989). Indeed,
Balmer jumps are often absent or appear in emission in the
spectra of dwarf novae (la Dous 1989 ; Williams 1983) and
one was not detected in the spectrum of X­ray Nova Mus
1991 during outburst (Cheng et al. 1992). In short, we
cannot rule out the hypothesis that the IR/radio component
contributes to the UV spectrum of XTE 1118]480. How­
ever, the presence of a signiïcant Balmer jump argues for
appreciable disk emission. In this work, we assume that the
disk ÿux is dominant in the UV band.
In conclusion, the global spectrum consists of three com­
ponents : (1) a B 24 eV thermal component caused by the
accretion disk with a large inner radius (2) a
(Z35R
Schw ) ;
quasi power­law component, which extends from D0.4 keV
to at least 160 keV ; and (3) a third component that domi­
nates the spectrum at wavelengths greater than several
microns. A model for the ïrst two components and a dis­
cussion of the third one are presented in a companion paper
by Esin et al. (2001).
We thank the Chandra X­Ray Observatory Director H.
Tananbaum for granting Directorîs Discretionary Time, the
entire Chandra team for their superb e+ort and enthusiasm,
and NASA for providing support for J. E. M., M. R. G., and
J. J. D. through grant DD0­1003X and contract NAS 8­
39073. The EUV E observations were made possible by a
generous grant of Directorîs Discretionary Time by EUV E
Project Manager R. Malina, the e+orts of EUV E Science
Planner M. Eckert, the sta+ of the EUV E Science Oper­
ations Center at CEA, and the Flight Operations Team at
Goddard Space Flight Center. This work includes obser­
vations with the NASA/ESA Hubble Space Telescope,
obtained at the Space Telescope Science Institute, operated
by the Association of Universities for Research in
Astronomy, Inc., under NASA contract NAS 5­26555. We
would like to thank the HST and RXT E support sta+ for
ongoing efficient support. UKIRT is operated by the Joint
Astronomy Centre on behalf of the UK Particle Physics
and Astronomy Research Council. C. A. H., R. I. H., and
S. C. acknowledge support from grant F/00­180/A from the
Leverhulme Trust. J. E. M. acknowledges helpful dis­
cussions with A. Siemiginowska and N. Brickhouse. We
thank an anonymous referee for several constructive com­
ments.
REFERENCES
Arnaud, I., & Dorman, B. 2000, XSPEC (An X­Ray Spectral Fitting
Package), version 11 (Greenbelt, MD : HEASARC, NASA/GSFC)
Brocksopp, C., Jonker, P. G., Fender, R. P., Groot, P. J., van der Klis, M.,
& Tingay, S. J. 2001, MNRAS, 323, 517
Cardelli, J. A., Clayton, G. C., &Mathis, J. S. 1989, ApJ, 345, 245
Cheng, F. H., Horne, K., Panagia, N., Shrader, C. R., Gilmozzi, R., Paresce,
F., & Lund, N. 1992, ApJ, 397, 664
Dickey, J. M., & Lockman, F. J. 1990, ARA&A, 28, 215
Dubus, G., Kim, R. S. J., Menou, K., Szkody, P., & Bowen, D. V. 2001,
ApJ, 553, 307
Esin, A. A., McClintock, J. E., Drake, J. J., Garcia, M. R., Haswell, C. A.,
Hynes, R. I., Marshall, H. L., &Muno, M. P. 2001, ApJ, 555, 483
Esin, A. A., McClintock, J. E., &Narayan, R. 1997, ApJ, 489, 865
Faison, M. D., Goss, W. M., Diamond, P. J., & Taylor, G. B. 1998, AJ, 116,
2916
Fender, R. P. 2001, MNRAS, 322, 31
Frank, J., King, A., & Raine, D. 1992, Accretion Power in Astrophysics (2d
ed. ; Cambridge : Cambridge Univ. Press)
Gierlinski, M., Zdziarski, A. A., Done, C., Johnson, W. N., Ebisawa, K.,
Ueda, Y., Haardt, F., & Phlips, B. F. 1997, MNRAS, 288, 958
Hynes, R. I., Mauche, C. W., Haswell, C. A., Shrader, C. R., Cui, W., &
Chaty, S. 2000, ApJ, 539, L37
Jahoda, K. 2000, unpublished talk presented 2000 March 22 at the Rossi
2000 Symp. (Greenbelt, MD GSFC/NASA)
la Dous, C. 1989, MNRAS, 238, 935
Makishima, K., Maejima, Y., Mitsuda, K., Bradt, H. V., Remillard, R. A.,
Tuohy, I. R., Hoshi, R., &Nakagawa, M. 1986, ApJ, 308, 635
Marshall, H. L., Carone, T. E., Shull, J. M., Malkan, M. A., & Elvis, M.
1996, ApJ, 457, 169
McClintock, J., Garcia, M., Caldwell, N., Falco, E. E., Garnavich, P. M., &
Zhao, P. 2001, ApJ, 551, L147
Miller, J. M., Fox, D. W., Di Matteo, T., Wijnands, R., Belloni, T., Pooley,
D., Kouveliotou, C., & Lewin, W. H. G. 2001, ApJ, 546, 1055
Mitsuda, K., et al. 1984, PASJ, 36, 741
Patterson, J. 2000, IAU Circ. 7412
Pooley, G. G., & Fender, R. P. 1997, MNRAS, 292, 925
Pooley, G. G., &Waldram, E. M. 2000, IAU Circ. 7390
Predehl, P., & Schmitt, J. H. M. M. 1995, A&A, 293, 889
Remillard, R., Morgan, E., Smith, D., & Smith, E. 2000, IAU Circ. 7389
Rhoades, C. E., & Ruffini, R. 1974, Phys. Rev. Lett., 32, 324
Rumph, T., Bowyer, S., & Vennes, S. 1994, AJ, 107, 2108
Schlegel, D. J., Finkbeiner, D. P., &Davis, M. 1998, ApJ, 500, 525
Tanaka, Y., & Lewin, W. H. G. 1995, in X­Ray Binaries, ed. W. Lewin,
J. van Paradijs, & E. van den Heuvel (Cambridge : Cambridge Univ.
Press), 126
Titarchuk, L. 1994, ApJ, 434, 570
Uemura, M., et al. 2000, PASJ, 52, L15
van der Klis, M. 1994, ApJS, 92, 511
Wagner, R. M., Foltz, C. B., Shahbaz, T., Casares, J., Charles, P. A., Star­
rïeld, S. G., &Hewett, P. 2001, ApJ, in press
Williams, G. 1983, ApJS, 53, 523