Draft version February 28, 2020
Typeset using L
A
T
E
X
twocolumn
style in AASTeX62
Evidence for Returning Disk Radiation in the Black Hole X-ray Binary XTE J1550
−
564
Riley M. T. Connors,
1
Javier A. Garc
́
ıa,
1, 2
Thomas Dauser,
2
Victoria Grinberg,
3
James F. Steiner,
4, 5
Navin Sridhar,
6
J
̈
orn Wilms,
2
John Tomsick,
7
and Fiona Harrison
1
1
Cahill Center for Astronomy and Astrophysics, California Institute of Technology,
Pasadena, CA 91125, USA
2
Dr Karl Remeis-Observatory and Erlangen Centre for Astroparticle Physics,
Sternwartstr. 7, D-96049 Bamberg, Germany
3
Institut f ̈ur Astronomie und Astrophysik, Universitt T ̈ubingen, Sand 1, 72076 T ̈ubingen, Germany
4
MIT Kavli Institute, 77 Massachusetts Avenue, 37-241,
Cambridge, MA 02139, USA
5
CfA, 60 Garden St. Cambridge, MA 02138, USA
6
Department of Astronomy, Columbia University,
550 W 120th St, New York, NY 10027, USA
7
Space Sciences Laboratory, University of California Berkeley,
7 Gauss Way, Berkeley, CA 94720-7450
(Accepted February 26, 2020)
Submitted to ApJ
ABSTRACT
We explore the accretion properties of the black hole X-ray binary XTE J1550
−
564 during its
outbursts in 1998/99 and 2000. We model the disk, corona, and reflection components of X-ray
spectra taken with the
Rossi X-ray Timing Explorer
(
RXTE
), using the
relxill
suite of reflection
models. The key result of our modeling is that the reflection spectrum in the very soft state is best
explained by disk self-irradiation, i.e., photons from the inner disk are bent by the strong gravity of the
black hole, and reflected off the disk surface. This is the first known detection of thermal disk radiation
reflecting off the inner disk. There is also an apparent absorption line at
∼
6
.
9 keV which may be
evidence of an ionized disk wind. The coronal electron temperature (
kT
e
) is, as expected, lower in the
brighter outburst of 1998/99, explained qualitatively by more efficient coronal cooling due to irradiating
disk photons. The disk inner radius is consistent with being within a few times the innermost stable
circular orbit (ISCO) throughout the bright-hard-to-soft states (10s of
r
g
in gravitational units). The
disk inclination is low during the hard state, disagreeing with the binary inclination value, and very
close to 90
◦
in the soft state, recovering to a lower value when adopting a blackbody spectrum as the
irradiating continuum.
Keywords:
accretion, accretion disks – atomic processes – black hole physics
1.
INTRODUCTION
The study of accretion as a physical process has pro-
vided us with a myriad of interesting conclusions regard-
ing the nature of black holes (BHs) and strong gravi-
tational fields. This is largely due to the capabilities
we have to approach the topic across vast variability
Corresponding author: Riley M. T. Connors
rconnors@caltech.edu
timescales, distances, and scale sizes. Active galactic
nuclei (AGN), due to the linear relation between black
hole mass and dynamical timescale, are not observed to
evolve significantly on human timescales—with the ex-
ception of a few newly discovered changing-look quasars,
e.g., McElroy et al. 2016; Yang et al. 2018. However,
their smaller cousins, black hole X-ray binaries (BHBs),
exhibit high variations in flux and spectral shape over
just days to weeks (see, e.g., Nowak 1995; Homan & Bel-
loni 2005; Remillard & McClintock 2006). As such, in
arXiv:2002.11873v1 [astro-ph.HE] 27 Feb 2020
2
Connors et al.
depth modeling of BHBs as they evolve during outbursts
allows us to understand the driving physical conditions
for observable changes, and to attempt to relate this
understanding to the behavior of AGN.
Many such studies of BHB spectral evolution have
been conducted.
The broadly classified “hard” and
“soft” states are now mostly understood to be the result
of combinations of several principal components: ther-
mal blackbody emission from a multitemperature accre-
tion disk (Shakura & Sunyaev 1973; Done et al. 2007);
a hard, power-law component, originating from an opti-
cally thin gas which inverse-Compton (IC) scatters the
thermal disk photons, and is either a hot compact corona
(Haardt & Maraschi 1993; Dove et al. 1997), or sits in
the base of a relativistic jet (Markoff et al. 2005); and
a reflected component of emission, which we expect is
generated by the power-law emission illuminating the
accretion disk (Fabian et al. 1989; Garc ́ıa et al. 2014).
XTE J1550
−
564 is a Galactic, transient BHB, first
detected by the All-Sky Monitor on board the
Rossi
X-ray Timing Explorer
(
RXTE
) on September 6 1998
(Smith 1998). Subsequent daily monitoring with
RXTE
for the following eight months (Sobczak et al. 2000) re-
vealed a significant 7-Crab X-ray flare just two weeks
into the outburst.
The dynamical characteristics of
XTE J1550
−
564 are well-determined. Optical/Infrared
observations made with the 6.5-meter Magellan tele-
scopes have led to strong constraints on the BH mass,
source distance, orbital period, and binary inclination:
M
BH
= 9
.
1
±
0
.
6
M
,
D
= 4
.
4
+0
.
6
−
0
.
4
kpc,
P
orb
= 1
.
54 days
and
i
= 75
◦
±
4
◦
(Orosz et al. 2002, 2011). Addi-
tionally, X-ray timing studies of the initial outburst
in 1998/99 with
RXTE
revealed quasi-periodic oscilla-
tions (QPOs) throughout the outburst (Remillard et al.
2002a). XTE J1550
−
564 has since gone into outburst
on four additional occasions, comprising one full spec-
tral evolution in 2000 (Rodriguez et al. 2003), and three
“failed” outbursts in 2001, 2002, and 2003 (a “failed”
outburst is one in which the source does not transition
from the hard to the soft state; Remillard & McClintock
2006). As such, the X-ray spectral and time variability
characteristics of XTE J1550
−
564 have been extensively
studied (Sobczak et al. 2000; Homan et al. 2001; Remil-
lard et al. 2002b; Rodriguez et al. 2003; Kubota & Done
2004; Dunn et al. 2010).
Several estimates have been made of the dimensionless
spin (
a
?
=
cJ/GM
2
, where
J
is the spin angular mo-
mentum) of the BH in XTE J1550
−
564 (0
.
1–0
.
9, Davis
et al. 2006; 0
.
76–0
.
8, Miller et al. 2009; 0
.
49
+0
.
13
−
0
.
20
, Steiner
et al. 2011; 0
.
34
±
0
.
01, Motta et al. 2014) using either
the thermal disk continuum fitting method (Li et al.
2005; McClintock et al. 2006), modeling of relativistic
reflection of X-rays off the accretion disk (Ross & Fabian
2005, 2007; Brenneman & Reynolds 2006), or modeling
of QPOs (Motta et al. 2014). All such modeling, whilst
not in perfect agreement quantitatively, reveals the BH
spin to be less than maximal, with a rough average value
of
a
?
= 0
.
5.
There has not yet been a detailed study of relativis-
tic reflection in XTE J1550
−
564 as the source evolves
through its outbursts. We do, however, have a gen-
eral phenomenological understanding of its hard-to-soft
spectral evolution, particularly from the first two out-
bursts in 1998/99 and 2000 (Sobczak et al. 2000; Ro-
driguez et al. 2003). The hard-to-soft spectral transi-
tion during both outbursts is well characterized by a
thermal disk component, peaking at
∼
1 keV in the
soft state, and a power-law component which persists
through the hard and hard-intermediate states. The
power law steepens significantly (Γ
∼
2
.
5–3) during
the intermediate states, typical of the long-known steep
power-law states of BHBs (e.g., Miyamoto & Kitamoto
1991; Miyamoto et al. 1993). In addition, curious be-
havior was found during the 7-Crab flare in the 1998/99
outburst. Sobczak et al. (2000) modeled the thermal
disk spectrum of XTE J1550
−
564, and found that the
inner radius of the accretion disk decreases sharply fol-
lowing the flare. However, they do note that this drop
in radius could be artificial, i.e., a color correction to the
disk spectrum, which is degenerate with disk tempera-
ture and radius through the overall flux.
We previously modeled (Connors et al. 2019; from
now on C19) the hard-intermediate state broadband (1–
200 keV) X-ray spectrum of XTE J1550
−
564 with the
most up-to-date relativistic reflection model,
relxill
(Garc ́ıa et al. 2014; Dauser et al. 2014). C19 found
that XTE J1550
−
564 appears to have an inner disk
inclination of 39
+0
.
6
−
0
.
4
degrees, based on the reflection
spectrum, which is
∼
35
◦
lower than the confirmed bi-
nary inclination found by Orosz et al. (2011). However,
this constraint was based on just one simultaneous ob-
servation of XTE J1550
−
564 made with the
Advanced
Satellite for Cosmology and Astrophysics
(
ASCA
) and
RXTE
during the intermediate state. Here, we seek
to model the evolution of the reflection spectrum of
XTE J1550
−
564 in order to better characterize the ge-
ometry and thermal properties of its inner accretion disk
and corona.
In this paper we explore the disk, coronal, and reflec-
tion properties of XTE J1550
−
564 during its first two
complete outbursts in 1998/99 and 2000, by physically
modeling a sample of archival
RXTE
observations. In
Section 2 we describe the
RXTE
data reduction process.
In Section 3 we outline our spectral modeling strategy
Reflection Spectroscopy of XTE J1550
−
564
3
and procedure, and detail the results. In Section 4 we
detail the implications of our reflection modeling results,
and in Section 5 we give a concluding summary. The
most striking result of our modeling, as discussed in Sec-
tion 4.2, is that we find the best-fit spectral reflection
model during the very soft state of XTE J1550
−
564 is
produced by an irradiating blackbody continuum; we
have found evidence for emission returning from the in-
ner disk onto itself due to the strong gravity of the BH.
2.
RXTE
DATA REDUCTION
RXTE
observed XTE J1550
−
564 over 400 times, with
more than half of these observations taken during the
first outburst in 1998/99. All the data from these obser-
vations is publicly available on the
RXTE
archive via the
HEASARC (High Energy Astrophysics Science Archive
Research Center). We extracted data from the Pro-
portional Counter Array (PCA) lying within 10 min of
the South Atlantic Anomaly (SAA). Since proportional
counter unit (PCU) 2 has the best calibration of all the
PCUs, and the best coverage (all PCA exposures), we
use only the data from this PCU, using all three PCU 2
layers. We then corrected all the PCU 2 spectra using
the tool
pcacorr
(Garc ́ıa et al. 2014), and subsequently
added 0.1% systematics to all the PCU 2 channels—
these comparatively low systematics are made possible
by the reduction in systematic residuals provided by
the
pcacorr
tool. The corrections provided by
pca-
corr
result from utilizing observations of the Crab with
the PCA, and iteratively reducing systematic residuals
present in averaged powerlaw fits to a summed Crab
spectrum. We refer the reader to Garc ́ıa et al. (2014)
for the details of this correction, and just note here that
for PCA spectra composed of
&
10
7
counts, there is up
to an order of magnitude increase in sensitivity to faint
spectral features. We then group the PCU 2 spectra at
a signal-to-noise of 4 based upon visual inspection of the
faintest spectra and their backgrounds, such that there
are sufficient counts per bin up to high energies (
>
20).
We restrict our spectral fitting to 3–45 keV.
3.
MODELING
We model the changing disk, corona, and reflection
components of XTE J1550
−
564 as it evolves from the
hard to soft states during its first two outbursts. Our
modeling strategy stems from several key motivating
factors:
1) In C19 we modeled simultaneous
ASCA
and
RXTE
observations taken in the hard-intermediate state during
the first XTE J1550
−
564 outburst, using the reflection
model
relxillCp
. We found that the disk inclination is
at
∼
40
◦
, significantly lower than the binary inclination
Table 1.
Properties of the selected
RXTE
-PCA (PCU 2) X-ray
spectra from the first and second outbursts of XTE J1550
−
564.
ObsID
MJD
HR
a
N
counts
b
cts s
−
1 c
Outburst 1
(10
6
)
30188-06-03-00 51064.0
0.91
6
986
30188-06-01-01 51065.3
0.82
4
1767
30188-06-04-00 51067.3
0.67
9
2807
30188-06-09-00 51071.2
0.52
13
3873
30191-01-33-00 51108.1
0.38
33
3571
40401-01-50-00 51241.8
0.25
13
4133
40401-01-27-00 51211.7
0.09
12
4525
Outburst 2
50137-02-06-00 51654.7
0.82
2
680
50134-02-01-00 51658.6
0.72
0
.
7
816
50134-02-01-01 51660.1
0.54
4
1008
50134-02-02-00 51662.2
0.38
2
1978
50134-02-02-01 51664.4
0.30
3
1429
a
Hardness
ratio
given
by
source
counts
in
[5
.
8–18 keV]/[3–5
.
8 keV] bands.
b
Number of counts in the 3–45 keV band of the PCU 2 spectra.
c
Total 3–45 keV count rate.
Note
—Observation 40401-01-50-00 actually follows observation
40401-01-27-00 temporally, but we selected our data in this way
to maximize the coverage of spectral hardness. Our sample of
observations from outburst 1 covers the period from 8 Septem-
ber 1998 to 4 March 1999. Outburst 2 data covers the period
from 20–30 April 2000.
of
∼
75
◦
(Orosz et al. 2011). Therefore, in this paper
we set out to test whether this is true across all spectral
states, and whether there is any evolution in the disk
inclination.
2) Garc ́ıa et al. (2015) paved the way for global BHB
reflection studies using the
RXTE
archive, with the goal
of characterizing the disk and coronal parameters of
GX 339
−
4, such as disk inner radius,
R
in
, and coronal
electron temperature,
kT
e
, and optical depth,
τ
. Garc ́ıa
et al. (2015) found that the inner disk remains within
∼
10
r
g
(
r
g
=
GM/c
2
, where
G
is the gravitational con-
stant,
M
is the mass of the BH, and
c
is the speed of
light) during the rise of the hard state, the corona cools,
and optical depth increases. However, the focus was on
the rise of the hard state, and did not follow the tran-
sition from hard to soft toward the outburst peak. We
want to model the disk and coronal physics as BHBs
transition from the hard to the soft state (similarly to,
e.g., Sridhar et al. 2020).
4
Connors et al.
2000
2001
PCU 2 Intensity (counts/s)
1998/99
0
10000
1000
100
10
1
2000
1998/99
Time [MJD]
2002
51600
51400
Hard Colour
51200
2003
1
0.5
Figure 1. Left
: Hardness-intensity diagram including all
RXTE
observations of XTE J1550
−
564. The hard color is defined
as the ratio of source counts in the hard and soft bands, [8
.
6–18]
/
[5–8
.
6] keV.
Right
: Light curve showing just the first two
outbursts of XTE J1550
−
564. Large circles indicate the selected data for this study, seven observations from the first outburst,
and five from the second outburst.
3) XTE J1550
−
564, as shown in Figure 1, shows wide
variability in the nature of its outbursts. The initial out-
burst in 1998/99 was bright, approaching the Eddington
limit, double peaked, and reached a very soft spectral
state , with a hardness ratio (HR)
∼
0
.
05. The second
outburst in the year 2000 peaked at lower luminosities,
and decayed after reaching HR
∼
0
.
3, so did not become
as soft. The following three outbursts were all ‘failed’,
remaining spectrally hard and peaking at luminosities a
factor of 10 lower. Thus, within the same source we can
look for key differences in the accretion physics between
outbursts.
Given these motivators, we selected observations cov-
ering the transition from the hard to soft states in out-
bursts 1 and 2. This selection is shown in Figure 1,
highlighted by the large red and blue points. We chose
seven
observations from outburst 1, and
five
from out-
burst 2, based on having enough photon statistics to
constrain reflection model parameters, and in the case
of outburst 2, the availability of data—the source transi-
tion rapidly during outburst 2, and thus there are only
a few
RXTE
exposures during the hard-to-soft transi-
tion. The seven observations taken from outburst 1 span
HR = 0
.
09–0
.
91, and the five taken from outburst 2 span
0.30–0.82. Table 1 shows the details of all selected data.
Due to the complexity of the data modeling, we only
include these 12 observations in the remainder of this pa-
per. We fit all 12 observations with a model including
a Comptonized multi-temperature disk blackbody com-
ponent, relativistically broadened and distant, unbroad-
ened reflection components, and interstellar absorption:
crabcorr * TBabs * (simplcut
⊗
diskbb + relxillCp
+ xillverCp)
.
Crabcorr
(Steiner et al. 2010) corrects the detector
response of a given instrument to retrieve the normaliza-
tion and power-law slope obtained from fits to the Crab
spectrum, provided by Toor & Seward (1974). The val-
ues adopted by the PCA instrument are
N
= 1
.
097 and
∆Γ = 0
.
01.
TBabs
is a model for interstellar absorption
using the elemental abundance tables of Wilms et al.
(2000). We use the atomic cross sections of Verner et al.
(1996).
The model
simplcut
(Steiner et al. 2017) is a variant
of the model
simpl
(Steiner et al. 2009), and functions
as a coronal plasma, inverse-Compton (IC) scattering
the disk photons in a convolution kernel. The
simpl-
cut
model includes a coronal electron temperature (
kT
e
)
and thus contains a high-energy cutoff in the power-law
continuum. It is important to be aware of the effects of
selecting particular coronal IC continuum components
in our modeling. We prefer to use
simplcut
over more
physically motivated models, such as
nthComp
(Zdziarski
et al. 1996;
̇
Zycki et al. 1999), because
simplcut
con-
serves the disk photon flux when calculating the portion
of scattered photons, which is set by the parameter
F
sc
,
with a maximal value of unity resulting in all disk pho-
tons being upscattered. As we show explicitly in Sec-
tion 3.3, since
simplcut
adopts the same spectral shape
for the scattered photons as given by
nthComp
, the phys-
ical constraints of the corona are identical between the
two models. However, since
nthComp
is normalized in-
dependently of the disk flux, one can arrive at spurious
Reflection Spectroscopy of XTE J1550
−
564
5
estimations of the disk flux when modeling hard-state
spectra in the 3–45 keV
RXTE
band. Thus, adopting
simplcut
allows us to constrain the co-evolving disk and
corona properly, tracking the inner disk temperature
and flux, along with the coronal properties. Similarly,
we decide against using the more self-consistent
eqpair
model (Coppi 2000). The
eqpair
model calculates the
plasma thermodynamics based upon parameterization
of the coronal and disk compactness and coronal optical
depth. However, given both that we only model data in
the PCA energy band, and need a simple way to relate
the coronal properties to the irradiating continuum for
reflection, we prefer
simplcut
. In Section 4 we show ex-
plicit comparisons of the PCA residuals when applying
these different continuum components. The disk pho-
tons in our model are provided as a multi-temperature
blackbody component
diskbb
(Mitsuda et al. 1984).
The models
relxillCp
and
xillverCp
are flavours
of the
relxill
suite of relativistic reflection models
(Dauser et al. 2014; Garc ́ıa et al. 2014), they are used
to calculate the reflection spectrum resulting from the
illumination of an IC spectrum atop the accretion disk.
XillverCp
provides the reflection spectrum resulting
from this illumination, which produces fluorescent line
emission, the most prominent being Fe K emission, as
well as Compton down-scattering of higher energy pho-
tons, giving the characteristic ‘
Compton hump
’.
Relx-
illCp
includes the full ray tracing calculations from the
irradiating source to the disk and onward to the ob-
server, allowing for a full calculation of the relativis-
tic effects which distort the spectrum, including light-
bending effects, Doppler shifts, and gravitational red-
shifts.
In all our fits we treat the model parameters as fol-
lows. The
crabcorr
parameters for offset normalization
and photon index are fixed at
N
= 1
.
097 and ∆Γ = 0
.
01,
respectively. We fix the interstellar absorption hydrogen
column density at
N
H
= 10
22
cm
−
2
in accordance with
Galactic H
I
surveys (Kalberla et al. 2005). Though we
found a value of 9
.
228
+0
.
007
−
0
.
009
×
10
21
cm
−
2
in C19, in the
3–45 keV band occupied by the PCA data this difference
is not impactful on our modeling results, and keeping its
value fixed reduces degeneracies. The disk temperature
(
T
in
) and normalization (
N
disk
) in the model compo-
nent
diskbb
are both kept free. The
simplcut
ReflFrac
parameter is fixed to 1, positing only up-scattering in
the coronal IC calculation. The photon index of the IC
spectrum (Γ) and electron temperature (
kT
e
) are both
kept free. We fix the black hole spin to
a
?
= 0
.
5 in
rough accordance with the previous spectral continuum
fitting, reflection fitting, and time-variability modeling
results for XTE J1550
−
564 (Davis et al. 2006; Miller
et al. 2009; Steiner et al. 2011; Motta et al. 2014). We
fix the emissivity index for the illumination of the disk to
q
= 3 throughout the disk, since the emissivity profile is
typically shallow for non-maximal BH spin (Dauser et al.
2013). The reflection fraction is fixed to -1 such that the
reflection components of
relxillCp
and
xillverCp
ex-
clude the illuminating continuum, already provided by
simplcut
⊗
diskbb
. The photon index (Γ) and electron
temperatures (
kT
e
) are tied to the corresponding values
in
simplcut
. The disk inclination (
i
) and iron abun-
dance (
A
Fe
) are all left as free parameters, and tied
between the
relxillCp
and
xillverCp
models. The
disk ionization (log
ξ
) is left to vary freely in the
relx-
illCp
component, and fixed at log
ξ
= 0 in the
xil-
lverCp
component, representing distant, near-neutral
reflection. The inner-disk radius in the
relxillCp
com-
ponent,
R
in
, is left free, and influences the relativistic
effects as calculated in the model. The
xillverCp
and
relxillCp
components are normalized independently.
In the following sections (3.1, 3.2, 3.3), we begin by
showing some results of phenomenological fits to our se-
lected data, move on to a discussion of interesting fea-
tures detected in the very soft state, and then show the
full results of our relativistic reflection modeling as dis-
cussed in this section.
O1, HR = 0.09
Energy [keV]
1010
O1, HR = 0.91
O1, HR = 0.82
O1, HR = 0.67
O1, HR = 0.52
O1, HR = 0.38
O1, HR = 0.25
Ratio
Energy [keV]
5
1010
O2, HR = 0.82
O2,HR=0.72
O2, HR = 0.54
Mean error
O2, HR = 0.38
O2,HR=0.30
Mean error
1
1.05
1.1
1.15
5
Figure 2.
Fe K line ratios after fitting the spectral con-
tinuum model
TBabs*simplcut
⊗
diskbb
to all our selected
data. The left panel shows the ratio residuals for outburst
1 data, and the right for outburst 2. Error bars have been
removed from the residuals for clarity, the average total
±
errors are shown in the top right of each panel.
3.1.
Hard-to-soft transition
Figure 2 shows the evolution of data residuals when
fitting the model
TBabs*(simplcut
⊗
diskbb)
to the
PCU 2 spectra in our selected sample. The goal of fit-
ting such a model is to isolate the Fe K emission and
edge features.
It is not possible to definitively quantify a shift in
the centroid energy of the Fe K line, due to the lim-
ited energy resolution of the PCA detector (
∼
1 keV at
6
Connors et al.
6 keV). However, we see more blueward line emission
as XTE J1550
−
564 transitions to the soft state. This
is particularly pronounced in outburst 1, during which
time the source is brighter. The reasons for this evolu-
tion are not clear, but it could possibly be due either
to geometrical changes in the inner flow, i.e., the disk
inclination may be varying, or alternatively the result
of distinct changes in the irradiating spectrum. It is
also possible that we are seeing excess emission in the
7–9 keV band that need not necessarily be associated
with the Fe K reflected emission.
3.2.
The very soft state: additional features
In the very soft state of XTE J1550
−
564, represented
in our selected sample by observation 40401-01-27-00,
there are prominent features in both the 4–5 keV band,
and at
∼
6
.
8–7 keV (see Figure 3). In order to ex-
plore these, we took a more comprehensive look at
the multiple observations taken during this soft branch
(HR
<
0
.
1) by selecting 11 PCA spectra within an ob-
servation window of
∼
13 days during the secondary rise
of the 1998/99 outburst (HR = 0
.
09).
40401-01-22-00
Ratio
40401-01-23-00
40401-01-21-01
40401-01-25-00
40401-01-20-00
Energy[keV]
0.95
1
1.05
1.1
5
10
15
20
Xe
Edge
4-5
keV
Abs
6.9keV
40401-01-32-00
40401-01-29-00
40401-01-28-00
40401-01-21-00
40401-01-27-00
40401-01-26-00
Figure 3.
Ratio residuals remaining after fitting a basic
spectral model,
TBabs*simpl
⊗
diskbb
, to data within the
soft branch of the 1998/99 outburst. All 11 selected spectra
show similar features in the 4–5 keV band, and at
∼
6
.
9 keV.
Figure 3 shows curious features in the PCU 2 spectra
of 11 individual observations. A striking and unexpected
absorption signature appears at
∼
6
.
8–7 keV. This fea-
ture has not been reported in previously analysed PCA
data of XTE J1550
−
564 during the 1998/99 outburst
(Sobczak et al. 2000), nor in any other observations of
the source. The reasons for this are likely that Sobczak
et al. (2000) necessarily added 0.5% systematics to the
PCA channels in their analysis, undoubtedly masking
this feature. We refer the reader to Garc ́ıa et al. (2014)
for details of the
pcacorr
tool, showing the complex sys-
tematics the tool removes (see also Appendix A). Since
we were able to reduce many of the PCA systematics
using the
pcacorr
tool, and thus add only 0.1% sys-
tematics, this feature may now have become observable.
The explanations for the feature are unclear, but could
be evidence of either of the following: (i) an absorp-
tion line from an outflowing disk wind, or (ii) a feature
inherent to the PCA detector.
Disk winds are ubiquitous in BHB soft states (Ponti
et al. 2012), thus it is not unexpected that we may see
such signatures, though they have not previously been
detected in XTE J1550
−
564. If present in a wind, this
feature is likely to coincide with the Fe
XXVI
line, pre-
viously found in BHBs in the soft state (e.g., Lee et al.
2002; Miller et al. 2006). Thus, to test the validity of
the claim that we may be seeing the same feature in
our PCU 2 data in the soft state, we performed full
phenomenological fits to the softest observation in our
sample.
We fit observation 40401-01-27-00 (HR = 0
.
09) using
the model
[crabcorr * TBabs * smedge(simplcut
⊗
diskbb + gau + gau) * edge]
. The first Gaussian com-
ponent represents the Fe K emission line due to reflec-
tion, and the second Gaussian has negative normaliza-
tion to represent the Fe
XXVI
absorption line from the
disk wind. The energies of the emission and absorption
line are fixed at 6.4 keV and 6.9 keV respectively. The
width of the absorption line is fixed at
σ
= 0
.
01 keV,
but we allow the emission line width to vary freely such
as to represent relativistic smearing at the inner disk.
The
smedge
component represents the relativistically
smeared iron edge (Ebisawa PhD thesis, implemented
by Frank Marshall). We fix the edge width to 7.1 keV,
allow the edge energy to vary between 7–9 keV, and
the optical depth
τ
to vary freely. The
edge
compo-
nent is included at
∼
4
.
8 keV, representing the xenon L
edge in the PCU 2 layers. Figure 4 shows the resultant
fit, achieving
χ
2
ν
= 64
/
39 = 1
.
6. The equivalent width
(EW) of the absorption line is
∼
33 eV, comparable
to those found for the Fe
XXVI
line in other soft-state
BHBs (e.g., Miller et al. 2006;
∼
40 eV).
However, since this component had been revealed to
us after applying the
pcacorr
tool to the PCA data,
we cannot rule out the possibility that this feature is
inherent to the PCU 2 detector. Given the softness of
the data, and thus number of X-ray counts in the low-
energy PCA channels, absorption features can manifest
where they were previously left undetected. This was
noted by Garc ́ıa et al. (2015) in their global study of
GX 339
−
4. Two apparent absorption features were de-
tected in the PCU 2 spectra of GX 339
−
4 at
∼
5
.
6 keV
and
∼
7
.
2 keV. Garc ́ıa et al. (2015) proposed that these
could have appeared due to the uncertain energy resolu-
tion of the PCA. However, in our data we only detect an
absorption feature at
∼
6
.
9keV, and it only appears dur-
Reflection Spectroscopy of XTE J1550
−
564
7
0.1
1
10
50
Flux
[keV
2
ph
cm
-2
s
-1
keV
-1
]
Absorption
line
6.9
keV
smedge*(simplcut*diskbb)*edge
Emission
line
6.4
keV
0.97
1
1.03
3
5
7
10
20
30
χ
2
/
ν
=
64/39
=
1.6
Ratio
E
[keV]
Figure 4.
Fit of model
crabcorr * TBabs * smedge
(simplcut
⊗
diskbb + gau + gau) * edge
to 40401-01-27-
00. The first Gaussian component represents the broad Fe
K emission line due to reflection. The second Gaussian com-
ponent represents absorption in a disk wind, likely Fe
XXVI
.
We fix the emission line at 6.4 keV and allow the width to
vary freely. The absorption line is fixed at 6.9 keV, with
σ
= 0
.
01 keV. The bottom panel shows the data-to-model
ratios.
ing the soft state, whereas Garc ́ıa et al. (2015) detected
both features in the bright hard state of GX 339
−
4.
The edge at
∼
4–5 keV has been previously reported
in
RXTE
observations of bright sources, and has also
been discussed in the relevant calibration papers (Ja-
hoda et al. 2006; Shaposhnikov et al. 2012). It is thus
well-known that the Xe L-edge region still requires mod-
eling, because this feature is not fully accounted for in
the calibration.
In the following Section we discuss the results of full
reflection modeling of our selected sample of data, and
include the additional features discussed here, applying
an edge component to represent xenon from the detector
(necessary in observations exceeding
∼
10
7
counts, with
significant disk emission, i.e., soft) wherever it is needed
by the data, and a Gaussian absorption feature at
∼
6
.
9 keV to model out the residual feature around the Fe
K line.
3.3.
Reflection modeling results
We have established that there are complex features
in the Fe region, particularly in the very soft branch
of outburst 1. Therefore, we now show results of our
best reflection models applied to these data, taking into
account the complex residual features as already de-
scribed. Figure 5 shows the key reflection modeling pa-
rameters and their associated uncertainties as a function
of X-ray hardness, allowing direct comparison between
outbursts 1 and 2. Tables 2 and 3 show the numerical
values corresponding to Figure 5, along with the other
model parameters. There are several interesting results
to notice.
Firstly, we see relatively consistent evolution of the
thin accretion disk properties between the two out-
bursts. The inner disk temperature (
T
in
) and normal-
ization (
N
disk
) increase and decrease respectively as the
source transitions from the hard to the soft state. This
is consistent with the inner disk moving closer to the
innermost stable circular orbit (ISCO). However, con-
straints on
R
in
from the reflection component are not
very strong, but largely consistent with being, if not at
the ISCO, within a factor of a few. The ISCO, for a
prograde BH spinning at
a
?
= 0
.
5, is at 4
.
23
r
g
, thus
the range of disk inner radii from the hard to soft states
is from a maximum of
∼
18
r
g
and
∼
34
r
g
in outbursts
1 and 2 respectively, down to 4
.
23
r
g
. We did not relate
the
R
in
parameter of the reflection model
relxillCp
to
the disk normalization (
N
disk
= (
R
in
/D
10
)
2
cos
θ
, where
D
10
is the distance to the source in units of 10 kpc,
and
θ
is the disk inclination) in our modeling. However,
the estimates of
R
in
derived from the
N
disk
constraints
broadly agree with the reflection modeling results, with
the exception of those at HR
>
0
.
8, i.e., the bright hard
state. However, this calculation does not take into ac-
count the uncertainty in the color temperature of the
inner disk (which can be up to a factor of 2; see, e.g.,
Davis et al. 2005).
Secondly, the properties of the Comptonizing plasma
show very similar behavior to that observed in previ-
ous global reflection studies of GX 339
−
4 (Garc ́ıa et al.
2015). The corona remains much hotter during out-
burst 2, the fainter outburst, whereas the photon index
(Γ) of the coronal spectrum is almost identical through-
out the transition. However, closer inspection of the
coronal temperature constraints (
kT
e
) during outburst
2 reveals that we mostly only achieve lower limits, and
those lower limits are typically far beyond the maximum
energy of the PCA (
>
45 keV). Therefore, in order to
ensure we have not limited ourselves by the exclusion of
the available HEXTE data in this case, we re-modeled
those observations (outburst 2) with the HEXTE data
included. We selected the HEXTE cluster A and B data,
including data between 20–200 keV. HEXTE B spectra
were corrected using the
hexBcorr
tool (Garc ́ıa et al.
2016b), and we grouped both HEXTE A and B spec-
tra by factors of 2, 3, and 4 in the 20–30, 30–40, and
8
Connors et al.
1.5
2.0
2.5
3.0
Γ
Outburst
1:
1998/99
Outburst
2:
2000
1
10
100
kT
e
[keV]
0.0
0.1
1.0
T
in
10
4
10
6
10
8
N
disk
10
30
50
70
90
Inclination
[deg]
1
5
10
R
in
[R
ISCO
]
1
2
3
4
log
ξ
1
10
A
Fe
0.1
0.5
1.0
0.2
0.4
0.6
0.8
F
sc
X-ray
hardness
0.1
0.5
1.0
1.5
0.2
0.4
0.6
0.8
τ
corona
X-ray
hardness
Figure 5.
Parameters and their uncertainties against spectral hardness. All data were fit with the model
crabcorr * TBabs *
(simplcut
⊗
diskbb + relxillCp + xillverCp)
with the following exceptions: in cases in which a xenon edge is required in the
4–5 keV band, and in the very soft state of outburst 1, where an absorption line commensurate with a disk wind is required at
6.9 keV. Red points show the parameter trends for outburst 1, and blue for outburst 2. The coronal optical depth is calculated
as
τ
corona
=
−
ln(1
−
F
sc
).
40–250 keV ranges respectively, in order to achieve an
oversampling of
∼
3 times the instrumental resolution.
We then grouped all HEXTE spectra by a signal-to-
noise ratio of 4, just as we did with the PCA, in order
to achieve the required statistics per bin. We fit the
PCA and HEXTE A/B data simultaneously, adopting
free normalisation constants in the
crabcorr
model to
account for cross-calibration between instruments. The
results are shown in Table 4. The coronal electron tem-
perature,
kT
e
, remains very high, and can only be con-
strained in the first two observations. Here the values
(80
+70
−
30
keV and 70
+10
−
10
keV) are still significantly higher
than those found for outburst 1, confirming our result
that the corona is hotter during outburst 2. In addition,
other key reflection properties do not differ significantly,
though we do find the inclusion of HEXTE data does al-
low for slightly tighter constraints on the disk inclination