Hubble Space Telescope Spectroscopy of the Unexpected 2001 July Outburst of the Dwarf Nova WZ Sagittae

Abstract

We present Hubble Space Telescope (HST) Space Telescope Imaging Spectrograph E140M spectra of the dwarf nova WZ Sge, following the early superoutburst of 2001 July. Our four far-ultraviolet (FUV) spectra, obtained over a time span of 4 months, monitor changes in the hot component of the system during the decline phase. The spectra cover the wavelength interval 1150-1708 Å. They reveal Stark-broadened Lyα and He II (1640) absorption and absorption lines due to metals (Si, C, N, Al) from a range of ionization states. Single-temperature white dwarf models provide reasonable qualitative agreement with the HST spectra. We find that the white dwarf appears to dominate the spectra from October through December. However, it is not clear that the absorption lines of metals form in the white dwarf photosphere. Therefore, the derived abundances and rotational velocity must be viewed with caution. Only a modest improvement in the fits to the data results when an accretion belt component is included. If the FUV spectra arise from the white dwarf alone, then we measure a cooling in response to the outburst of at least 11,000 K (29,000-18,000 K). The absence of broad underlying absorption features due to metals at this stage suggests slow rotation (~200 km s-1). It is possible that the white dwarf envelope has expanded due to the heating by the outburst or that the relatively narrow absorption features we observe are forming in an inflated disk atmosphere or curtain associated with the outburst.

Full text

HUBBLE SPACE TELESCOPE SPECTROSCOPY OF THE UNEXPECTED 2001 JULY OUTBURST OF THE
DWARF NOVA WZ SAGITTAE1
Edward M. Sion
Department of Astronomy and Astrophysics, Villanova University, 800 Lancaster Avenue, Villanova, PA 19085; emsion@ast.villanova.edu
Boris T. Ga
¨nsicke
Department of Physics and Astronomy, University of Southampton, Southampton SO17 1BJ, UK; btg@astro.soton.ac.uk
Knox S. Long
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218; long@stsci.edu
Paula Szkody
Department of Astronomy, University of Washington, Seattle, WA 98195; szkody@astro.washington.edu
Fu-Hua Cheng
Center for Astrophysics, University of Science and Technology of China, Hefei, Anhui 230026, People’s Republic of China
Steve B. Howell
Institute of Geophysics and Planetary Physics, University of California, Riverside, CA 92521; howell@psi.edu
Patrick Godon2
Department of Astronomy and Astrophysics, Villanova University, 800 Lancaster Avenue, Villanova, PA 19085; patrick.godon@villanova.edu
William F. Welsh
Department of Astronomy, San Diego State University, San Diego, CA 92182-1221; wfw@sciences.sdsu.edu
Sumner Starrfield
Department of Physics and Astronomy, Arizona State University, Tempe, AZ 85287; sumner.starrfield@asu.edu
Christian Knigge
Department of Physics and Astronomy, University of Southampton, Southampton SO17 1BJ, UK; knigge@astro.soton.ac.uk
and
Warren M. Sparks
XNH, MS F664, Los Alamos National Laboratory, Los Alamos, NM 87545; wms@lanl.gov
Received 2002 November 25; accepted 2003 April 17
ABSTRACT
We present Hubble Space Telescope (HST ) Space Telescope Imaging Spectrograph E140M spectra of the
dwarf nova WZ Sge, following the early superoutburst of 2001 July. Our four far-ultraviolet (FUV) spectra,
obtained over a time span of 4 months, monitor changes in the hot component of the system during the
decline phase. The spectra cover the wavelength interval 1150–1708 A
˚. They reveal Stark-broadened Ly
and He ii (1640) absorption and absorption lines due to metals (Si, C, N, Al) from a range of ionization states.
Single-temperature white dwarf models provide reasonable qualitative agreement with the HST spectra. We
find that the white dwarf appears to dominate the spectra from October through December. However, it is
not clear that the absorption lines of metals form in the white dwarf photosphere. Therefore, the derived
abundances and rotational velocity must be viewed with caution. Only a modest improvement in the fits to
the data results when an accretion belt component is included. If the FUV spectra arise from the white dwarf
alone, then we measure a cooling in response to the outburst of at least 11,000 K (29,000–18,000 K). The
absence of broad underlying absorption features due to metals at this stage suggests slow rotation (200 km
s1). It is possible that the white dwarf envelope has expanded due to the heating by the outburst or that the
relatively narrow absorption features we observe are forming in an inflated disk atmosphere or curtain
associated with the outburst.
Subject headings: novae, cataclysmic variables — stars: individual (WZ Sagittae) — white dwarfs
1. INTRODUCTION
WZ Sge is the widely known, extensively studied proto-
type of a group of H-rich cataclysmic variables that have
extreme properties: the largest outburst amplitudes, shortest
orbital periods, longest outburst recurrence times, lowest
mass Roche lobe–filling secondaries, lowest accretion rates,
and coolest white dwarf primaries of any class of dwarf
novae (Howell et al. 1999, 2002; see Kato et al. 2001 for a
recent review). It is also the brightest dwarf nova during
outburst and arguably the closest cataclysmic variable, with
a distance of only 43 8 pc (J. Thorstensen 2001, private
communication). The inclination of the binary is high
enough (78) that the secondary star eclipses the disk rim
but not the white dwarf. Steeghs et al. (2001) found the
radial velocity semiamplitude, K2, of the secondary star to
1On the basis of observations with the NASA/ESA Hubble Space
Telescope, obtained at the Space Telescope Science Institute, which is
operated by the Association of Universities for Research in Astronomy,
Inc., under NASA contract NAS5-26555.
2Visiting at the Space Telescope Science Institute, 3700 San Martin
Drive, Baltimore, MD 21218.
The Astrophysical Journal, 592:1137–1150, 2003 August 1
#2003. The American Astronomical Society. All rights reserved. Printed in U.S.A.
1137

be in the range 493–595 km s1, giving an upper limit to the
mass of the secondary, M2<0:11 M, while they found an
upper limit to the radial velocity semiamplitude of the pri-
mary, K1<37 km s1, implying a mass range 0:7M<
Mwd <0:9Mfrom the mass function. A 28 s periodicity
was detected in both the optical (Patterson 2002) and the
far-ultraviolet (FUV; Welsh et al. 1997). If due to rotation,
this oscillation period would correspond to Vsin i¼1200
km s1for a 1 Mwhite dwarf. The outburst is widely held
to arise from an extended period of enhanced mass accre-
tion onto the white dwarf, during which the accretion disk is
expected to be luminous. This disk radiates the FUV spec-
trum during the outburst phase, but as the accretion rate
declines, the disk luminosity also falls, and the white dwarf
emerges as the dominant source of FUV radiation. Further
flux declines are attributed to the cooling of the white dwarf
photosphere, which was heated by the outburst.
Following the 1978 December outburst, its relatively
clocklike outburst regularity of every 33 yr was unexpect-
edly disrupted by a 10 yr outburst on 2001 July 23, first
reported by T. Ohshima (Ishioka et al. 2001). What
followed was the most thoroughly observed dwarf nova out-
burst in history (Patterson et al. 2002; Kuulkers et al. 2002;
Knigge et al. 2000).
Shortly following the discovery of the superoutburst, two
director’s discretionary proposals for HST observation
were approved, one (GO-9287) to provide dense coverage of
the outburst phase and the other (GO-9304) to cover, with a
preliminary analysis, the emergence of the accretion-heated
white dwarf and monitor its evolution (cooling and other
properties) in response to the superoutburst. The results of
the latter program, awarded four orbits over a 4 month time
span, are reported in this paper.
2. THE HST OBSERVATIONS
The observations took place on 2001 September 11
(03:03:28 UT; 50 days postoutburst), October 10 (12:57:13
UT; 79 days postoutburst), November 10 (11:45:32 UT; 110
days postoutburst), and December 11 (00:34:59 UT; 141
days postoutburst) with HST Space Telescope Imaging
Spectrograph (STIS) using the FUV MAMA detector con-
figuration in time-tag mode with the E140M grating and
0>20>2 aperture, giving a wavelength coverage of 1140–
1735 A
˚centered on 1425 A
˚. We binned the echelle spectra in
0.1 A
˚steps and analyzed the average STIS spectra from
each individual HST visit. The data reduction was carried
out with the standard STScI pipeline reduction system. The
spectral line measurements were carried out with the soft-
ware SPLOT in IRAF by fitting Gaussians. The results of
the line identifications and line strength measurements are
presented in Tables 1–4, where, in each, column (1) lists the
centroid wavelength, column (2) lists the central depth line
flux in ergs cm2s1, column (3) lists the equivalent width
TABLE 1
Line Identifications and Absorption Line Measurements, 2001 September Spectrum
Line Center
(A
˚)
(1)
Flux
(ergs cm2s1A
˚1)
(2)
EW
(A
˚)
(3)
FWHM
(A
˚)
(4)
Line Identification
(5)
1175.04 .................... 3.8 1012 6.0 6.7 C iii 1175.26, 1175.59, 1175.71, 1175.99
1191.27 .................... 1.7 1013 0.5 1.0 Si ii 1190.41
1193.71 .................... 5.3 1013 1.6 2.0 Si ii 1193.29, 1194.50
1197.12 .................... 1.4 1013 0.5 1.1 Si ii 1197.39
1200.38 .................... 3.6 1013 1.3 2.1 N i1200.22, 1200.71
1206.21 .................... 5.0 1013 2.1 2.7 Si iii 1206.50
1228.29 .................... 2.4 1013 0.7 3.5 Si ii 1228.43, 1228.61; Si ii 1228.74
1240.52 .................... 1.7 1012 4.4 6.5 N v1240
1250.55 .................... 1.6 1013 0.3 1.4 S ii 1250.50
1253.82 .................... 1.3 1013 0.2 1.0 S ii 1253.79
1260.33 .................... 9.3 1013 1.6 2.3 Si ii 1260.42
1264.71 .................... 1.3 1012 2.3 3.6 Si ii 1264.73, 1265.00
1300.50 .................... 7.0 1012 11.6 13.8 Si iii 1298.89, 1298.96, 1301.15; Si iii 1303.32
1309.26 .................... 2.7 1013 0.6 1.3 Si ii 1309.27
1317.21 .................... 1.5 1013 0.2 1.5 Ni ii 1317.22
1329.24 .................... 1.1 1013 0.2 1.6 C i1329.09, 1329.10, 1329.2
1335.09 .................... 1.7 1012 3.5 3.7 C ii 1334.53, 1335.66, 1335.71
1369.86 .................... 2.2 1013 0.4 3.0 Ni ii 1370.13
1393.31 .................... 1.7 1012 4.0 5.9 Si iv 1393.76; Ni ii 1393.32
1401.70 .................... 1.6 1012 3.9 5.9 Si iv 1402.77
1434.60 .................... 4.7 1013 1.1 5.6 ?
1493.06 .................... 2.9 1013 0.7 5.4 N i1492.03, 1492.82, 1484.68
1526.25 .................... 3.5 1013 1.0 2.1 Si ii 1526.70
1532.78 .................... 3.2 1013 0.9 1.9 Si ii 1533.43
1548.58 .................... 2.0 1012 5.2 6.6 C iv
1640.49 .................... 1.3 1012 4.6 14.3 He ii 1640.33, 1640.35, 1640.39, 1640.47
1658.12 .................... 2.6 1013 1.0 4.7 C i1657.91, 1658.12
1671.31 .................... 5.7 1013 2.1 3.5 Al ii 1670.85 or Fe ii 1673.46
1701.84 .................... 2.3 1013 0.8 4.0 Fe ii 1701.94, 1702.04
1719.60 .................... 2.1 1013 0.9 3.3 Al ii 1719.49 + Fe ii
1724.25 .................... 2.3 1013 1.0 3.4 Al ii 1725.11
1138 SION ET AL. Vol. 592

TABLE 2
Line Identifications and Absorption Line Measurements, 2001 October Spectrum
Line Center
(A
˚)
(1)
Flux
(ergs cm2s1A
˚1)
(2)
EW
(A
˚)
(3)
FWHM
(A
˚)
(4)
Line Identification
(5)
1152.03 .................... 4.9 1013 1.298 2.419 O i1152.15
1156.58 .................... 3.3 1013 0.8 2.7 P ii 1156.96
1164.48 .................... 1.5 1013 0.4 1.7 N ii 1164.28, 1164.58
1168.17 .................... 2.6 1013 0.7 3.3 N i1167.45, 1168.54
1175.58 .................... 1.0 1012 3.2 3.3 C iii 1175.26, 1175.59, 1175.71, 1175.99
1190.04 .................... 1.5 1013 1.0 1.3 Si ii 1190.41
1193.54 .................... 2.4 1013 2.1 2.3 Si ii 1193.29, 1194.50
1197.01 .................... 7.9 1014 0.9 1.1 Si ii 1197.39
1200.11 .................... 1.7 1013 1.9 2.6 N i1200.22, 1200.71
1206.04 .................... 1.8 1013 2.3 3.1 Si iii 1206.50
1228.23 .................... 1.4 1013 1.9 4.3 Si ii 1228.74, 1228.43, 1228.61
1238.64 .................... 2.6 1013 1.6 1.9 N v1240
1242.79 .................... 2.8e 1013 1.4 1.7 N v1240
1250.45 .................... 1.4 1013 0.6 1.2 S ii 1250.50
1253.67 .................... 1.5 1013 0.5 1.1 S ii 1253.79
1260.30 .................... 8.2 1013 2.5 2.7 Si ii 1260.42
1264.81 .................... 9.4 1013 3.1 3.4 Si ii 1264.74
1272.14 .................... 1.2 1013 0.3 1.8 Fe ii ?
1302.62 .................... 4.1 1012 10.6 12.8 Si iii 1298.89, 1298.96, 1301.15; Si iii 1303.32, O i1302.17
1309.30 .................... 3.3 1013 1.1 1.4 Si ii 1309.27
1317.09 .................... 1.1 1013 0.3 1.4 Ni ii 1317.22
1329.27 .................... 8.4 1014 0.2 1.8 C i1329.09, 1329.10,1329.2
1335.34 .................... 1.3 1012 3.8 3.7 C ii 1334.53, 1335.66, 1335.71
1370.28 .................... 1.5 1013 0.4 2.3 Ni ii 1370.13
1393.73 .................... 8.7 1013 2.5 2.8 Si iv 1393.76; Ni ii 1393.32
1402.77 .................... 7.1 1013 2.2 2.5 Si iv 1402.77
1435.12 .................... 3.8 1013 1.2 9.2 ?
1466.74 .................... 2.4 1013 0.8 7.5 Ni ii 1467.26
1492.26 .................... 1.0 1013 0.4 1.2 N i1492.03, 1492.82
1494.16 .................... 8.9 1014 0.3 1.3 N i1494.68
1526.33 .................... 3.6 1013 1.5 2.0 Si ii 1526.70
1533.05 .................... 4.5 1013 1.8 2.2 Si ii 1533.43
1549.30 .................... 1.5 1012 5.0 5.7 C iv
1658.10 .................... 2.3 1013 1.3 4.5 C i1657.91, 1658.12
1671.05 .................... 3.3 1013 1.7 1.9 Al ii 1670.85
1673.67 .................... 2.5 1013 1.3 5.0 Fe ii 1673.46
1701.60 .................... 2.0 1013 1.1 3.6 Fe ii 1701.94, 1702.04
1719.78 .................... 1.9 1013 1.1 2.6 Al ii 1719.49 + Fe ii
1724.11 .................... 1.8 1013 1.1 3.1 Al ii 1725.05, 1725.07, 1725.11
TABLE 3
Line Identifications and Absorption Line Measurements, 2001 November Spectrum
Line Center
(A
˚)
(1)
Flux
(ergs cm2s1A
˚1)
(2)
EW
(A
˚)
(3)
FWHM
(A
˚)
(4)
Line Identification
(5)
1175.45 .................... 5.0 1013 2.6 2.6 C iii 1175.26, 1175.59m, 1175.71, 1175.99
1189.66 .................... 1.1 1013 1.1 1.7 Si ii 1190.41
1193.27 .................... 2.1 1013 2.7 2.7 Si ii 1193.29, 1194.50
1196.68 .................... 7.0 1014 1.3 1.8 Si ii 1197.39
1199.51 .................... 1.5 1013 2.2 3.3 N i1200.22, 1200.71
1206.14 .................... 1.8 1013 3.1 4.0 Si iii 1206.50
1238.44 .................... 1.6 1013 1.7 2.1 N v1240
1242.48 .................... 5.9 1014 0.4 1.2 N v1240
1250.34 .................... 6.0 1014 0.3 1.2 Si ii 1250.50
1253.52 .................... 6.3 1014 0.3 1.0 S ii 1253.79
1260.19 .................... 4.4 1013 2.2 2.4 Si ii 1260.42
1264.58 .................... 7.0 1013 3.0 3.6 Si ii 1264.74
1296.53 .................... 6.0 1014 0.3 1.1 Si iii 1296.73
1298.74 .................... 1.3 1013 0.7 1.1 Ti iii 1298.70

TABLE 3—Continued
Line Center
(A
˚)
(1)
Flux
(ergs cm2s1A
˚1)
(2)
EW
(A
˚)
(3)
FWHM
(A
˚)
(4)
Line Identification
(5)
1301.88 .................... 1.0 1013 0.7 1.5 Si iii 1298.89, 1298.96, 1301.15; Si iii 1303.32; O i1302.17
1304.25 .................... 8.5 1014 0.7 1.2 Si ii 1304.37
1309.10 .................... 2.0 1013 0.9 1.3 Si ii 1309.27
1317.01 .................... 1.3 1013 0.2 1.4 Ni ii 1317.22
1329.27 .................... 8.4 1014 0.2 1.7 C i1329.09, 1329.10, 1329.2
1335.06 .................... 1.0 1012 4.2 4.1 C ii 1334.53, 1335.66, 1335.71
1393.40 .................... 6.1 1013 2.3 2.5 Si iv 1393.76; Ni ii 1393.32
1402.48 .................... 5.1 1013 2.0 2.3 Si iv 1402.77
1433.62 .................... 2.8 1013 1.1 7.9 S i1433.30; Ca ii 1433.75
1438.71 .................... 2.2 1013 0.9 4.1 ?
1465.15 .................... 4.7 1013 2.0 12.9 Fe ii 1465.04
1492.99 .................... 1.5 1013 0.7 5.3 N i1492.03, 1492.82
1494.16 .................... 8.9 1014 0.3 1.3 N i1494.68
1526.01 .................... 2.8 1013 1.4 1.9 Si ii 1526.70
1532.78 .................... 4.0 1013 1.9 2.5 Si ii 1533.43
1548.79 .................... 1.2 1012 5.0 5.4 C iv
1657.88 .................... 2.6 1013 1.6 5.2 C i1657.91, 1658.12
1670.91 .................... 3.8 1013 1.8 2.0 Al ii 1670.85
1719.98 .................... 9.8 1014 0.7 2.1 Al ii 1719.49 + Fe ii
1723.79 .................... 1.5 1013 1.2 3.4 Al ii 1725.05, 1725.07, 1725.11
TABLE 4
Line Identifications and Absorption Line Measurements, 2001 December Spectrum
Line Center
(A
˚)
(1)
Flux
(ergs cm2s1A
˚1)
(2)
EW
(A
˚)
(3)
FWHM
(A
˚)
(4)
Line Identification
(5)
1175.73 .................... 4.8 1013 3.1 3.0 C iii 1175.26, 1175.59, 1175.71, 1175.99
1190.51 .................... 6.5 1014 1.1 1.4 Si ii 1190.41
1193.76 .................... 7.9 1014 1.8 2.1 Si ii 1193.29, 1194.50
1197.22 .................... 3.3 1014 0.9 1.4 Si ii 1197.39
1200.10 .................... 4.7 1014 1.2 1.7 N i1200.22, 1200.71
1206.29 .................... 7.1 1014 1.8 2.1 Si iii 1206.50
1238.78 .................... 1.2 1013 1.7 2.1 N v1240
1242.05 .................... 1.2 1013 1.4 1.8 N v1240
1250.58 .................... 4.8 1014 0.4 1.5 S ii 1250.50
1253.82 .................... 5.2 1014 0.4 0.8 S ii 1253.79
1260.44 .................... 2.6 1013 1.9 2.1 Si ii 1260.42
1264.85 .................... 3.8 1013 2.6 3.2 Si ii 1264.74
1294.64 .................... 5.0 1014 0.3 0.7 Si iii 1294.55; Ti iii 1294.67; Ti iii 1294.72
1296.90 .................... 4.4 1014 0.3 1.0 Si iii 1296.73
1299.05 .................... 1.0 1013 0.7 1.2 Si iii 1298.89, 1298.95
1302.15 .................... 5.5 1014 0.5 1.1 O i1302.17
1304.60 .................... 6.0 1014 0.6 1.1 Si ii 1304.37
1306.06 .................... 1.6 1014 0.9 1.3 O i1306.03
1309.46 .................... 1.6 1013 0.9 1.3 Si ii 1309.27
1317.15 .................... 7.3 1014 0.3 1.7 Ni ii 1317.22
1335.43 .................... 8.0 1013 3.8 3.8 C ii 1334.53, 1335.66, 1335.71
1393.86 .................... 4.7 1013 2.3 2.4 Si iv 1393.76; Ni ii 1393.32
1402.89 .................... 4.0 1013 1.9 2.2 Si iv 1402.77
1418.86 .................... 9.9 1014 0.5 3.1 ?
1492.91 .................... 1.6 1013 0.9 5.7 N i1492.03, 1492.82
1526.46 .................... 1.8 1013 1.2 1.5 Si ii 1526.70
1533.21 .................... 3.1 1013 1.8 2.5 Si ii 1533.43
1549.28 .................... 8.7 1013 4.4 5.4 C iv
1658.08 .................... 1.4 1013 1.2 3.8 C i1657.91, 1658.12
1671.29 .................... 1.7 1013 1.4 1.5 Al ii 1670.85
1673.67 .................... 2.5 1013 1.3 5.0 Fe ii 1673.46

(EW in A
˚), column (4) lists the FWHM (in A
˚), and column
(5) gives the line identification(s).
The analysis of the FUV light curves obtained in the
time-tag mode will be reported elsewhere. The September
observation took place during a rebrightening, as shown in
Figure 1 of Long et al. (2003), where the placement of the
Far Ultraviolet Spectrographic Explorer (FUSE ) and HST
STIS observations is compared with the optical light curve
of the outburst and decline. This rebrightening was one of
several closely spaced so-called echo outbursts. These echo
outbursts, smaller in amplitude and shorter in duration than
the main outburst, are widely thought to be smaller accre-
tion events. The exact physical mechanism responsible for
the echo outbursts is not yet understood. The spectroscopic
results are discussed in the next section.
3. THE SPECTRAL FEATURES
The four spectra corresponding to the September, Octo-
ber, November and December HST observations are dis-
played as flux versus wavelength plots in Figures 1–4,
together with the line identifications labeled in the figures
with vertical tick marks. First, there is a steady drop in the
continuum flux level (measured at 1450 A
˚) from 4:21013
ergs cm2s1A
˚1in September to 1:91013 ergs cm2s1
A
˚1in December, suggesting that the white dwarf is cooling
and/or shrinking after being bloated by the outburst heat-
ing. Although it would be useful to compare this continuum
flux level decline, there is an unfortunate gap in the IUE cov-
erage of the postoutburst interval following the 1978
December 1 superoutburst. Therefore, there are no IUE
spectra for direct comparison with our spectra at a compa-
rable epoch in time. The earliest postoutburst spectrum fol-
lowing the 1978 outburst was taken a full 7 months later,
whereas our 2001 December spectrum was taken 5 months
postoutburst. The flux level of SWP5761 at 1450 A
˚,7
months postoutburst onset, is 1:61013 ergs cm2s1A
˚1
compared with the flux level of our 2001 December HST
spectrum of 1:91013 ergs cm2s1A
˚1, 5 months
postoutburst onset.
The most prominent line feature in all four spectra is
the very broad Lyabsorption, which we attribute to the
high-gravity white dwarf photosphere. One does not see the
H2quasi-molecular absorption feature (centered around
Fig. 1.—Observed flux, F(ergs cm2s1A
˚1) vs. wavelength (A
˚) for the 2001 September 11 observation of the early superoutburst of WZ Sge. Vertical tick
marks: Strongest line features.
OUTBURST OF WZ SGE 1141

1400 A
˚) in any of the four spectra, although this feature is
prominent during quiescence when the temperature of the
white dwarf is below about 20,000 K (Sion et al. 1995c).
The other features in Figures 1–4 are predominantly in
absorption, except for apparent broad emission wings at
Civ (1548, 1550) flanking deep absorption. There is, how-
ever, a weak, double-peaked Lyfeature in emission that
could be associated, at least in part, with the system. A close
examination reveals that the feature appears double-
peaked, especially in the November and December data,
and that the profile resembles that of C iv in that the blue
emission peak is higher than the red peak for October–
December. In the September data, it is more difficult to see,
but it looks plausible that the red peak is higher, just as it is
in C iv. A rough measurement of the peak-to-peak separa-
tion of the peaks is approximately 6 A
˚.AtLy,1A
˚corre-
sponds to about 245 km s1, so that separation corresponds
to roughly 1500 km s1. This leads to a value of the disk
velocity, Vdisk sin i’750 km s1. For comparison, Skid-
more et al. (2000) measure Vdisk sin i¼630 km s1in qui-
escence from the Balmer H. Therefore, these values for
Lysuggest a disk-formed line. This is extremely important
because it bears directly upon the basic question of the rees-
tablishment of the disk following the outburst. On the other
hand, a chromosphere (temperature inversion) on a rapidly
rotating white dwarf might produce lines with these widths.
The other line features cover a broad range of ionization.
Of immediate interest is the mix of ions, the excitation/
ionization states of the transitions, and the differences one
sees between these spectra and the spectra of WZ Sge
obtained during quiescence. First, the presence of N vin
absorption in the four spectra is extremely important since it
was never seen in spectra obtained during quiescence. For
example, there is no evidence of the N v(nor C iii)absorption
in the lower quality, lower resolution IUE spectra from early
1979 July. It is noteworthy here that N vis seen in absorption
in the dwarf nova U Gem during quiescence but does not
share the same velocity as the gravitationally redshifted white
dwarf photosphere (Long et al. 1994). This feature is thought
to arise in an extended hot region of gas near the white dwarf
in U Gem. Moreover, the N vfeature in WZ Sge is too sharp
to be consistent with formation in a velocity-broadened disk
region at the high inclination of WZ Sge unless it arises from
absorption in the outer disk of continuum photons from the
inner disk and/or white dwarf.
There is a feature at 1608 A
˚that has an interesting behav-
ior: the absorption becomes stronger with time. While the
feature is likely identified as Fe ii, other identifications are
Fig. 2.—Same as Fig. 1, except for the 2001 October 10 observation
1142 SION ET AL. Vol. 592

possible but less likely (e.g., N v,Ciii). We note the presence
of artifacts of STIS in the form of pseudoabsorptions at
approximately 1652, 1671, 1690, and 1710 A
˚.
There are noteworthy temporal variations seen in the line
strengths. A number of line features became increasingly
narrow with time, and emission wings present in September
are not seen in December. The most striking example of this
change in profile shape is the Si iv doublet. It is quite possi-
ble that the Si iv feature in the September spectrum may be
associated with an accretion belt on the white dwarf spun
up by the high accretion that took place during the outburst.
In that spectrum, the red wing of the C iv doublet compo-
nent at 1550 A
˚exhibits P Cygni emission. However, by the
time of the October spectrum and continuing through the
December spectrum, it is the blue wing of C iv 1548 that
appears in emission. While double-peaked profile structure
is typically associated with formation in the accretion disk
during dwarf nova quiescence, the feature may possibly sig-
nal a major change in the outflow characteristics and there-
fore the emergence of the underlying accreting white dwarf
accompanied, however, by signs of disk accretion. It is also
interesting that He ii (1640 A
˚) is seen in broad, shallow
absorption in the September spectrum but only marginally
present in the October, November, and December spectra.
This feature would not be expected to form in the photo-
sphere of a 20,000–30,000K white dwarf. The weakening
of He ii with time is presumably indicative of declining
temperature in its formation region.
Finally, numerous absorption lines of neutral, singly and
doubly ionized sulfur, silicon, and carbon are present.
Neutral nitrogen lines also appear. Given the distance to
WZ Sge of only 43 pc, it is unlikely these features are, in any
significant measure, interstellar in origin because they
appear to be too broad for interstellar features. Some of the
low-ionization features, e.g., Si ii 1260, 1265, 1527, and 1533
A
˚, have similar shapes and widths to N v, suggesting the
possibility that these low-ionization features form in the
same nonphotospheric region. Other low-ionization lines,
e.g., C ii 1335 A
˚, do not resemble expected photospheric fea-
tures. The predominance of absorption lines from neutral
carbon one sees in deep quiescence spectra is not found in
the four spectra. Nevertheless, the white dwarf spectrum
has emerged recognizably by the time of the September
observation (see Fig. 1), on the basis of our comparative
examination of the archived 2001 August spectra of Knigge
et al. (2002).
There are also unidentified features between 1430 and
1440 A
˚present, to some degree, in all four spectra, which we
Fig. 3.—Same as Fig. 1, except for the 2001 November 11 observation
No. 2, 2003 OUTBURST OF WZ SGE 1143

tentatively identify as a blend of Si iand C i. Such features
were seen in the quiescence spectra of WZ Sge (Sion et al.
1995; Cheng et al. 1997).
It is noteworthy that the disk emission seen in quiescence
(Sion et al. 1995c; Cheng et al. 1997) when the disk is opti-
cally thin is not seen at all in the four spectra (September,
October, November, and December) following the 2001
July outburst. However, it is possible that the accretion disk
is optically thick in the lines but optically thin in the contin-
uum once it is reestablished following the outburst. As we
discuss below, we cannot conclusively rule out an optically
thick, steady state accretion disk.
4. SYNTHETIC SPECTRAL FITTING
The model atmosphere (TLUSTY; Hubeny 1988), and
spectrum synthesis (SYNSPEC; Hubeny, Lanz, & Jeffrey
1994; Hubeny & Lanz 1995) codes and details of our 2min-
imization fitting procedures are discussed in detail in Sion et
al. (1995a, 1995b) and Huang et al. (1996a, 1995b) and will
not be repeated here. To estimate physical parameters, we
took the white dwarf photospheric temperature Teff ,logg,
Si and C abundances, and rotational velocity (Vrot ) as free
parameters and computed the reduced 2. Since our white
dwarf model spectra are normalized to 1 Rat a distance
of 1 kpc, the distance of a source is computed from
d¼1000 ðpcÞðRwd =RÞ=
ffiffiffiffi
S
p, where Sis the scale factor
S¼4Rwd=R
ðÞ
2d=kpcðÞ
2. In all cases, we first found the
best-fitting model and then computed the stellar radius from
the scale factor by using the distance of 43 pc. Before the
model fitting, any emission lines, including those in the core
of Ly , were masked out in each individual STIS spectrum.
In Table 5, we provide the results for accretion disk–only
fits to the four spectra. This exercise utilized model accretion
disks for the full range of white dwarf masses (0.35–1.2
M), accretion rates (108.5 to 1010:5Myr1), and disk
inclination angles (18–81) presented in the grid of Wade &
Fig. 4.—Same as Fig. 1, except for the 2001 December 10 observation
TABLE 5
Accretion Disk Only
Observation
_
MM
ðMyr1)
i
(deg) 2Scale Factor
Mwd
ðMÞ
Sep.................... 109.0 18 13.11 1.93 1010.55
Oct.................... 109.0 18 16.92 3.17 1010.35
Nov................... 109.5 18 12.13 1.19 1000.35
Dec ................... 109.5 18 7.87 9.54 1010.35
1144 SION ET AL. Vol. 592

Hubeny (1998), except for the models at inclinations of 8,
which have numerical problems. These fits were carried out
without regard to any observed constraints on the white
dwarf mass, disk inclination, or accretion rate. For each of
the four HST observations, the best-fitting disk model was
determined, and the results are presented in Table 5. The
best-fitting disk model in each case was for either a white
dwarf mass of 0.4 or 0.35 M, an accretion rate between
_
MM ¼108:5and 109:5Myr1, and disk inclination angles
of i¼18.
It is clear that the best-fitting disk models, selected with-
out regard to observed parameters, are implausible because
the observed absorption lines are far too sharp and there-
fore must arise from another source. The white dwarf mass
used in the best-fitting disk models is also implausibly low,
especially in view of the results reported by Steeghs et al.
(2001). In addition, the best-fitting disk inclination angle is
far too low to be applicable to WZ Sge, which, in quies-
cence, suffers hot spot eclipses and has an inclination close
to 80. This conclusion is consistent with the results of
Howell et al. (1999, 2002), who demonstrated that no inner
disk or boundary layer is present during the superoutbursts
of tremendous outburst amplitude dwarf novae (TOADS).
However, it is still possible that a model with a plausible
inclination and white dwarf mass could give a reasonable fit
to the spectra even if, in a numerical sense, the reduced 2
value is not optimal. Since the Wade & Hubeny disk models
are normalized to 100 pc, then the distance to a source in
parsecs is simply the 100 pc divided by the square root of the
scale factor. Using the distance of 43 pc as a constraint on
the fits, there are three models that provide roughly the cor-
rect flux. These models, all at a disk inclination of 75=5, are
models P, T, and Y in the nomenclature of Table 2 in Wade
& Hubeny (1998). They have white dwarf masses of 0.8, 1.0,
and 1.2 M, respectively, and accretion rates, respectively,
of
_
MM ¼109:5,10
10.0, and 1010:5Myr1. Thus, by con-
straining the fits to satisfy the distance to WZ Sge and by
adopting the most reasonable estimates of the white dwarf
mass and binary inclination, viable values of the accretion
rate are derived that do not depend on the details of
continuum slope fitting.
In an additional exercise, we kept the white dwarf mass
fixed at 1.2 M, the inclination at 81, and fitted disk models
to the October spectrum, letting the accretion rate vary. The
best-fitting disk model (lowest reduced 2) corresponded to
_
MM ¼1010:5Myr1. This is considerably lower than the
accretion rate implied by the best-fitting disk model
(
_
MM ¼109:0Myr1) in Table 5, where fits using all
combinations of the disk parameters were carried out.
Our next experiment utilized single-temperature white
dwarf–only fits to the HST data. In these fits, the white
dwarf temperature was varied from 16,000 to 30,000 K in
steps of 1000 K, the surface gravity was chosen to be
log g¼8:5, rotational velocity Vsin ivaried from 200 to
800 km s1in steps of 100 km s1, and chemical abundances
of Si and C were chosen to be either 0.01, 0.1, 1, 2, 3, 5, or 10
times the solar values. The resulting best-fit parameters and
their 3 errors for each of the four spectra are recapitulated
in Table 6.
The best-fitting single-temperature, log g¼8:5, white
dwarf models to the September, October, November, and
December observations are displayed in Figures 5, 6, 7, and
8, respectively. If log g¼8:0 is used in the fitting, the tem-
peratures in the above table become 28,000 K for the
September spectrum, 20,000 K for the October spectrum,
18,000 K for the November spectrum, and 17,000 K for the
December spectrum. These temperatures are 800–1600 K
cooler than those derived at higher gravity.
As a check on our temperature results from scaling the
models to the observations, we have computed an extensive
grid of models in fine-temperature steps of 100 K and
interpolated within this grid for log g¼8:5. The resulting
temperatures for the four spectra are 28,800 K (September),
21,600 K (October), 19,200 K (November), and 18,300 K
(December).
The single-temperature white dwarf fits indicate that
following the outburst, the accretion-heated white dwarf
cooled by at least 11,000 K (29,000–18,000 K). In Figure 9,
we display a cooling curve of temperature versus time with
the days counted starting on the first day of the outburst. In
the figure, we have denoted the STIS data points with circles
and the FUSE data points (from Long et al. 2003) with
stars.
The rotational velocity, Vsin i, from the single-
temperature white dwarf models, on the basis of the fitting
of the relatively narrow metal features and Lyprofile, is
only 200 km s1. This velocity, if it corresponds to the true
underlying white dwarf, is far lower than the value of 1200
km s1reported by Cheng et al. (1997) during quiescence.
We suspect, however, that these narrow features may not be
forming in the white dwarf photosphere (see below).
We calculated the white dwarf radius from the scale fac-
tors listed in Table 6 by fixing the distance at 43 pc. The
resulting white dwarf radii corresponding to the best-
fitting single-temperature white dwarf models to the four
spectra are 4:92 108(September), 6:68 108(October),
8:00 108(November), and 8:10 108cm (December).
The December value of the white dwarf radius implies a
white dwarf mass smaller than 0.6 M. Such a low mass is
ruled out by the radial velocity study of Steeghs et al. (2001).
We also tried various combinations of white dwarf mod-
els and disk models as composite fits. All but one such fit
TABLE 6
White Dwarf Only
Observation
T
(1000 K)
Si
(solar)
C
(solar)
Vsin i
(km s1)2Scale Factor
log g
(cm s2)
Sep............................ 29.0 +0.5/0.5 5.0 +0.3/0.3 2.0 +3.0/0.4 600 +100/100 8.43 2.71 1028.5
Oct............................ 22.0 +0.8/1.0 1.0 +0.1/0.1 0.5 +0.1/0.1 200 +50/50 10.57 5.00 1028.5
Nov........................... 19.0 +0.2/0.1 0.5 +0.1/0.1 0.1 +0.2/0.1 200 +50/50 7.63 7.14 1028.5
Dec ........................... 18.0 +0.1/0.1 0.5 +0.1/0.1 0.1 +0.1/0.2 200 +20/50 5.10 7.33 1028.5
Note.—Error bars are 3 .
No. 2, 2003 OUTBURST OF WZ SGE 1145

Fig. 5.—September HST STIS spectrum of WZ Sge compared with the best-fitting single-temperature white dwarf model atmosphere: Teff ¼29;000 K,
log g¼8:5, Si = 5 solar, C = 2 solar, and Vsin i¼600 km s1.
Fig. 6.—October HST STIS spectrum of WZ Sge compared with the best-fitting single-temperature white dwarf model atmosphere: Teff ¼22;900 K,
log g¼8:5, Si = 1 solar, C = 0.5 solar, and Vsin i¼200 km s1.
1146

Fig. 7.—November HST STIS spectrum of WZ Sge (flux vs. wavelength) compared with the best-fitting single-temperature white dwarf model
atmosphere: Teff ¼19;000 K, log g¼8:5, Si = 0.5 solar, C = 0.1 solar, and Vsin i¼200 km s1.
Fig. 8.—December HST STIS spectrum of WZ Sge compared with the best-fitting single-temperature white dwarf model atmosphere: Teff ¼18;000 K,
log g¼8:5, Si = 0.5 solar, C = 0.1 solar, and Vsin i¼200 km s1.
1147

failed with large 2values. The September spectrum could
be fitted with a white dwarf having Teff ¼28;000 K,
log g¼8:5 with Si elevated to 5 times solar, C with 2 times
solar plus an accretion disk with Mwd ¼0:80 M, with
_
MM ¼1010:5Myr1, and i¼41. This best-fit disk plus
white dwarf combination is displayed in Figure 10. The dot-
ted line represents the white dwarf model, the dashed line is
the accretion disk contribution, and the solid line is the com-
bination of the two components. Note that the accretion
disk contributes less than 1% of the FUV flux in this com-
bined fit. However, this model lacks plausibility because we
know the inclination of the disk in WZ Sge is close to 80.
In an additional experiment, two-temperature, differen-
tially rotating composite fits were attempted, consisting of a
rapidly spinning equatorial accretion belt at higher tempera-
ture and a cooler, slowly rotating photosphere. For example,
for the November spectrum, the best-fitting two-temperature
combination corresponded to an accretion belt with
Vsin i¼3400 km s1,Tbelt ¼28;000 K, log g¼6, and solar
abundances, combined with an 18,000 K photosphere with
log g¼8:5, Vsin i¼200 km s1, Si = 0.5 solar, and C = 0.1
solar. Overall, however, these composite fits produced only a
very modest improvement in the reduced 2value. For the
November observation, the best-fitting white dwarf plus
accretion belt corresponds to 88% of the flux emitted by the
white dwarf and 12% emitted by the accretion belt.
Finally, we have attempted to assess the effect of a layer
of cool gas in which iron peak absorbers would alter the
emergent stellar spectrum. In this model, an absorbing layer
of gas is located between the star, and the observer like a
curtain (which is why this model is referred to as the iron
curtain). We adopted curtain parameters similar to those
found in the FUSE study of WZ Sge by Long et al. (2003).
We adopted Tcur ¼10;000 K, electron density in number
per cubic centimeter Ne¼1013,vturb ¼200 km s1,and
Fig. 9.—Cooling curve of temperature vs. time for the WZ Sge white
dwarf with temperatures from HST STIS and FUSE measurements. Time
in days is counted starting from the first day of the outburst. Circles: STIS
data points. Stars:FUSE data points.
Fig. 10.—September HST STIS spectrum of WZ Sge compared with the best-fitting single-temperature white dwarf model atmosphere plus accretion disk
model. The white dwarf has Teff ¼28;000 K, log g¼8:5, Si = 5 solar, C = 2 solar (dotted line) and the accretion disk has Mwd ¼0:80 M,M¼1010:5M
yr1, disk inclination angle i¼41(dashed line). Solid line: Sum of the disk plus the white dwarf spectrum.
1148 SION ET AL. Vol. 592

hydrogen column density log NH¼20:24. However, with
the use of the curtain option in SYNSPEC (adopting Tcur
and electron density, computing the mass density ,
constructing a curtain opacity table, and using the SYN-
SPEC curtain algorithm CIRCUS with input hvturbiand NH
for an adopted fractional coverage of the stellar source), the
resulting spectrum differs very little from the synthetic spec-
trum without a curtain and makes little difference in the
quality of the fit, at least in the STIS range. If a column is
chosen with log NH>22, the theoretical spectrum begins to
deviate significantly from the observed spectrum.
However, we find that for a two-component absorbing
layer consisting of a cool iron curtain with T¼10;000 K
and a column density of NH¼1022 plus a warmer iron cur-
tain with T¼30;000 K and a column density of NH¼1022,
a better fit to the details of absorption lines results, but not
especially to the overall shape of the continuum. In particu-
lar, the fits to the lines C iii 1175, Si ii 1190, 1193, 1197, Si iii
1206, Si ii 1260, 1265, Si ii 1309, C ii 1334, Si iv 1394, 1403,
and Si ii 1527, 1533 are improved.
We also tried a slowly rotating (200 km s1) white dwarf
model and a fast-rotating (1200 km s1) white dwarf model,
both with a two-component high column density absorbing
layer (with a turbulent velocity of 200 km s1), and did not
notice any appreciable difference. In some parts of the spec-
trum, the fast-rotating star model seems to fit the observa-
tion slightly better, but in other regions, it is the slowly
rotating models with curtains that appear to fit better.
5. DISCUSSION
We have analyzed the STIS data with fits of synthetic
spectra of optically thick accretion disks, high-gravity
photospheres, combined disks and photospheres and two-
temperature, differentially rotating, composite fits combin-
ing photospheres with rapidly spinning accretion belts. An
optically thick accretion disk added to a white dwarf photo-
sphere does not satisfactorily match the observations for the
range of accretion disk models that were available to us.
Single-temperature white dwarf models do provide reason-
able agreement with the HST spectra, but the September
spectrum is problematic to this interpretation. It appears
that there is an additional radiating component. For exam-
ple, the Lyline does not approach zero flux as one expects
for a 20,000 K white dwarf. Perhaps there is filling in of the
absorption by some light from the disk. Another point is
that, unless only a fraction of the white dwarf is exposed in
the September data, then an increasing white dwarf radius
with time is indicated as we progress to the December data.
The system was observed during an optical rebrightening
phase in which an additional source of radiation would be
expected in the September data. It seems clear that our inter-
pretation of overall changes in the white dwarf (e.g., cool-
ing, radius changes) hinges critically on what radiating
components are contributing to the September spectrum.
Among the possible sources are a circumbinary disk, ejected
shell, or corona/wind structure on the scale of several white
dwarf radii, although in the latter two sources, one would
expect signs of line emission that are not seen. Since
Chandra observations revealed a forest of emission lines just
after outburst (Kuulkers et al. 2002), this may be evidence
of a large, optically thin component extending out to a few
white dwarf radii.
The problems with applying a white dwarf photospheric
interpretation alone to the September spectrum may also
extend to the other three spectra as well. We cannot as yet
be certain that the narrow absorption lines of metals in all
four spectra really form in the white dwarf photosphere.
Therefore, the abundances and rotational velocity we have
derived must be regarded as preliminary.
The main points in favor of a white dwarf interpretation
of the FUV spectra are (1) that it is difficult to produce the
broad, observed Lyabsorption feature with anything
other than a high-gravity, hydrogen-rich photosphere and
(2) that the model fits give reasonable white dwarf radii for
the known distance (43 pc) of WZ Sge. It is possible that the
underlying white dwarf remains obscured, perhaps due to a
raised disk rim during the outburst that may gradually
become less vertically extended as the outburst progresses,
thus leading to the exposure of an increasing fraction of the
white dwarf photosphere. The obscuration of the white
dwarf by the occulting effect of a disk rim in a high-inclina-
tion cataclysmic variable has been demonstrated recently in
an HST study of the nova-like variable DW UMa by
Knigge et al. (2000). The presence of material high above
the disk plane of WZ Sge is also suggested by the X-ray data
of Kuulkers et al. (2002).
If the FUV spectra arise from the white dwarf alone, then
we measure a cooling in response to the outburst from
30,000 to 19,000 K. This is a lower limit because we did
not observe the peak temperature the white dwarf reached
during the outburst. Even if the first observation has to be
excluded until a better understanding of the disk contribu-
tion to that spectrum is obtained, the white dwarf cooled by
almost 5000 K between October and December. The
absence of the broad absorption features (as seen in deep
quiescence) suggests that the white dwarf has expanded due
to the heating by the outburst or that the relatively narrow
absorption features we observe are forming in an inflated
disk atmosphere or curtain of remnant material associated
with the outburst.
In a companion paper (Long et al. 2003), we performed a
similar analysis of the FUSE spectra obtained in WZ Sge in
a similar period. On the whole, the two analyses agree,
although the temperatures derived from the FUSE data are
somewhat higher than those obtained here. It is not entirely
clear whether the higher FUSE temperatures represent a
nonuniform temperature distribution on the surface of the
white dwarf or some effect associated with the narrow lines
in the spectrum. We are hopeful that analysis of spectra
obtained far from the outburst with FUSE and HST will
resolve these inconsistencies as well as resolve the overall
question of the rotation rate of the white dwarf.
This work is supported by NASA through grants GO-
9304 from the Space Telescope Science Institute, which is
operated by the Association of Universities for Research in
Astronomy, Inc., under NASA contract NAS5-26555. Sup-
port was also provided, in part, by NSF grant 99-01195
and NASA ADP grant NAG5-8388 (E. M. S.). B. T. G.
acknowledges support from a PPARC Advanced
Fellowship.
No. 2, 2003 OUTBURST OF WZ SGE 1149

REFERENCES
Cheng, F., Sion, E. M., Szkody, P., & Huang, M. 1997, ApJ, 484,
L149
Howell, S. B., Ciardi, D., Szkody, P., van Paradijs, J., Kuulkers, E., Cash,
J., Sirk, M., & Long, K. 1999, PASP, 111, 342
Howell, S. B., Fried, R., Szkody, P., Sirk, M., & Schmidt, G. 2002, PASP,
114, 748
Huang, M., Sion, E. M., Hubeny, I., Cheng, F. H., & Szkody, P. 1996a,
ApJ, 458, 355
———. 1996b, AJ, 111, 2386
Hubeny, I. 1988, Comput. Phys. Commun., 52, 103
Hubeny, I., & Lanz, T. 1995, ApJ, 439, 875
Hubeny, I., Lanz, T., & Jeffrey, S. 1994, Newsletter on Analysis of
Astronomical Spectra (Fife: St. Andrews Univ.), 20, 30
Ishioka, R., et al. 2001, IAU Circ. 7669
Kato, T., Sekine, Y., & Hirata, R. 2001, PASJ, 53, 1191
Knigge, C., Hynes, R. I., Steeghs, D., Long, K. S., Araujo-Betancour, S., &
Marsh, T. R. 2002, ApJ, 580, L151
Knigge, C., Long, K. S., Hoard, D., Szkody, P., & Dhillon, V. 2000, ApJ,
539, L49
Kuulkers, E., Knigge, C., Steeghs, D., Wheatley, P. J., & Long, K. S. 2002,
in ASP Conf. Ser. 261, The Physics of Cataclysmic Variables and Related
Objects, ed. B. T. Ga
¨nsicke, K. Beuermann, & K. Reinsch (San
Francisco: ASP), 443
Long, K., Froning, C., Knigge, C., Sion, E. M., & Szkody, P. 2003, ApJ, in
press
Long, R., Sion, E., Huang, M., & Szkody, P. 1994, ApJ, 424, L49
Patterson, J., et al. 2002, PASP, 114, 721
Sion, E. M., Cheng, F., Long, K. S., Szkody, P., Gilliland, R., Huang, M.,
& Hubeny, I. 1995c, ApJ, 439, 957
Sion, E. M., Huang, M., Szkody, P., & Cheng, F. H. 1995a, ApJ, 445, L31
Sion, E. M., Szkody, P., Cheng, F. H., & Huang, M. 1995b, ApJ, 444, L97
Skidmore, W., et al. 2000, MNRAS, 318, 440
Steeghs, D., et al. 2001, ApJ, 562, L145
Wade, R., & Hubeny, I. 1998, ApJ, 509, 350
Welsh, W., Skidmore, W., Wood, J., Cheng, F., & Sion, E. 1997, MNRAS,
291, 57
1150 SION ET AL.