A&A 452, 285-294 (2006)
DOI: 10.1051/0004-6361:20054077
K. Pottschmidt1 - M. Chernyakova2,
- A. A. Zdziarski3 - P. Lubinski3,2 - D. M. Smith4 - N. Bezayiff4
1 - Center for Astrophysics and Space Sciences, University of
California, San Diego, La Jolla, CA 92093-0424, USA
2 - INTEGRAL Science Data Centre, Chemin d'Écogia
16, 1290 Versoix, Switzerland
3 - Nicolaus Copernicus Astronomical Center, Bartycka 18, 00-716
Warszawa, Poland
4 - Department of Physics, University of California, Santa Cruz,
Santa Cruz, CA 95064, USA
Received 20 August 2005 / Accepted 4 February 2006
Abstract
The Galactic Center black hole candidate (BHC)
GRS 1758-258 has been observed extensively within INTEGRAL's
Galactic Center Deep Exposure (GCDE) program in 2003 and 2004, while
also being monitored with RXTE. We present quasi-simultaneous PCA,
ISGRI, and SPI spectra from four GCDE observation epochs as well as
the evolution of energy-resolved PCA and ISGRI light curves on time
scales of days to months. We find that during the first epoch GRS 1758
258
displayed another of its peculiar dim soft states like the one
observed in 2001, increasing the number of observed occurrences of
this state to three. During the other epochs the source was in the
hard state. The hard X-ray emission component in the epoch-summed
spectra can be well described either by phenomenological models,
namely a cutoff power law in the hard state and a pure power law in
the dim soft state, or by thermal Comptonization models. A soft
thermal component is clearly present in the dim soft state and might
also contribute to the softer hard state spectra. We argue that in the
recently emerging picture of the hardness-intensity evolution of black
hole transient outbursts in which hard and soft states are observed to
occur in a large overlapping range of luminosities (hysteresis), the
dim soft state is not peculiar. As noted before for the 2001 dim soft
state, these episodes seem to be triggered by a sudden decrease
(within days) of the hard emission, with the soft spectral component
decaying on a longer time scale (weeks). We discuss this behavior as
well as additional flux changes observed in the light curves in terms
of the existence of two accretion flows characterized by different
accretion time scales, the model previously suggested for the 2001
episode.
Key words: black hole physics - stars: individual: GRS 1758-258 - gamma rays: observations - X-rays: binaries - X-rays: general
The hard X-ray source GRS 1758
258was discovered in 1990
(Mandrou 1990; Sunyaev et al. 1991) during observations of the Galactic
Center region performed with the Granat satellite. Most of
the time the source displays X-ray properties similar to the canonical
hard state of Galactic black hole binaries, i.e., a comparatively hard
spectrum with power law indices of
and an
exponential cutoff above 100 keV
(Kuznetsov et al. 1999; Lin et al. 2000; Main et al. 1999) as well as strong short term
variability on frequencies up to 10 Hz which can be characterized by
a flat-topped power spectrum (Smith et al. 1997; Lin et al. 2000). Based on these
X-ray properties and on the detection of a radio counterpart (a
point source plus a double-sided jet structure, Rodriguez et al. 1992)
GRS 1758
258is considered to be a microquasar.
With the exception of rare dim soft states that can last up to several
months, the X-ray emission of GRS 1758
258is persistent. In contrast to the
canonical picture for persistent black hole binaries, however, GRS 1758
258
most likely has a low mass binary companion and is accreting via Roche
lobe overflow. Three faint IR counterpart candidates have been
identified recently, multi-band photometry and near-infrared
spectroscopy characterizing the brightest of them as a probable K0 III
giant and the other two as main sequence A stars
(Heindl & Smith 2002a; Marti et al. 1998; Rothstein et al. 2002; Goldwurm et al. 2001; Eikenberry et al. 2001).
Taking into account the
day binary orbit
(Smith et al. 2002a), the radius of a K giant is consistent with Roche
lobe overflow, while the other counterpart candidates are too small
(Rothstein et al. 2002).
From 1990 to 1998 the source was monitored in the hard X-ray regime
with Granat/SIGMA, establishing it as persistent hard state
black hole binary but also revealing a factor of eight variability in
the 40-150 keV flux (Kuznetsov et al. 1999). At softer X-rays
monitoring of GRS 1758
258(and its "sister source''
1E 1740.7-2942) with RXTE started with monthly
observations in 1996 and is still on-going in 2005, with two
observations each week (Smith et al. 2002a,1997,2002b; Main et al. 1999; Smith et al. 2001a, this
work). This campaign
led to the discovery that in GRS 1758
258(and 1E 1740.7-2942) the
observed spectral hardness is not anti-correlated with the 2-25 keV
flux but rather correlated with the flux derivative - in the sense
that the spectrum is softest when the 2-25 keV count rate is
dropping. This behavior is different from the prototype hard state
black hole binary Cyg X-1 which is showing the canonical
anti-correlation of spectral hardness and soft X-ray flux. It has been
interpreted as an indication for the presence of two
independent accretion flows supplied with proportional
amounts of matter at large radii which are then accreted on different
time scales (Smith et al. 2002b; Main et al. 1999): a hot, e.g., ADAF-type,
accretion flow, reacting quickly to changes in the accretion rate, and
a large accretion disk with a long viscous time scale (consistent with
accretion via Roche lobe overflow). In addition, Lin et al. (2000)
performed a radio to
-ray multi-wavelength study of the hard
state as observed in 1997 August, including spectral modeling and high
time resolution analyses. Applying the thermal Comptonization model
compTT (Titarchuk 1994) they obtained a temperature of 52 keV
and an optical depth of 3.4 for the hot plasma. Sidoli & Mereghetti (2002)
obtained very similar values using compTT to model a broad
band BeppoSAX spectrum of the source obtained in 1997 April.
A weak soft excess is sometimes seen in the hard state spectra of
GRS 1758
258(Lin et al. 2000; Heindl & Smith 1998) or cannot be excluded to be present
(Mereghetti et al. 1997). It has also been observed in conjunction with
a slightly reduced hard X-ray flux during an intermediate state in
1993 March/April (Mereghetti et al. 1994). A similar recent episode in
2000 September has been characterized as an intermediate state based
on the RXTE monitoring observations (Heindl & Smith 2002a, and references
therein). Modeling an XMM-Newton/EPIC-MOS
spectrum obtained during this time, Goldwurm et al. (2001) found that in
addition to a comparatively soft power law (
)
a blackbody
component (
eV) is required.
On two occasions a much softer and dimmer state than the persistent
hard state with occasional softening has been observed: (i) A sudden
drop of the 2-25 keV rate between one RXTE monitoring observation
and the next occurred in 2001 February, with an estimated
decrease of 35% of the 1.3-200 keV luminosity within
20 d after the transition (Smith et al. 2001a). An extreme case of
the unusual flux derivative/hardness relation described above, this
behavior is different from the canonical black hole state behavior
where the soft state corresponds to a higher accretion rate and a
higher bolometric luminosity. For the 1996 soft state of Cyg X-1,
e.g., Zdziarski et al. (2002) find a 3-4 times higher bolometric luminosity
than for the typical hard state, an increase even higher than the
50-70% estimated previously (Zhang et al. 1997). In
Sect. 4 we suggest that the dim soft state can be
better understood when compared to outbursts of BHC transients than to
(focused) wind accretors like Cyg X-1. Smith et al. (2001a) also found
that the transition to the 2001 dim soft state of GRS 1758
258was mainly
due to a decreasing and softening power law component
(
in 2001 March), revealing a soft component
(
eV in 2001 March). As predicted by
Smith et al. (2001b) based on the two-flow accretion model, the soft
component decayed more slowly than the hard one, on a time scale of
28 days. Displaying a complex structure, including a partial
recovery of the count rates in 2001 July, the dim soft state lasted
until 2001 December (Heindl & Smith 2002a, see also this
work). Chandra/ACIS-HETGS
(Heindl & Smith 2002a,b) and XMM-Newton/RGS
(Miller et al. 2002) observations in 2001 March support the picture that
the decaying soft flux is emitted by an accretion disk. (ii) It can be
assumed that Granat/SIGMA exposures in fall 1991 and spring
1992 when the 40-150 keV flux was below the detection limit
(Gilfanov et al. 1993) found the source in a similar dim soft state
(Miller et al. 2002; Smith et al. 2001a). This is supported by the analysis of a
1992 March ROSAT/PSPC spectrum by Grebenev et al. (1997)
resulting in a power law index of
.
In the following we report results of monitoring observations of
GRS 1758
258with INTEGRAL and RXTE in 2003 and 2004. While the source was
in its usual variable hard state during most of the time, the data
obtained in spring 2003 clearly correspond to another dim soft state,
although it did not progress as far as the 2001 dim soft state before
the hard X-ray emission recovered again. In
Sect. 2 of this paper we describe the observing
strategy of the two monitoring programs and explain the data
extractions performed to obtain broad band PCA-ISGRI-SPI spectra and
multi-band PCA and ISGRI light curves. In Sect. 3.1 the
long term light curves and flux changes are described and in
Sect. 3.2 we present results of modeling the broad band
spectra with phenomenological and thermal Comptonization models. The
results, especially the observation of another weak soft state, are
discussed in the light of current black hole outburst and accretion
models in Sect. 4. Our conclusions are summarized
in Sect. 5.
During 2003 and 2004 the guaranteed time program amounted to 30-35%
of INTEGRAL's observing time and was mainly dedicated to the
Galactic Plane Scan (GPS) and Galactic Center Deep Exposure (GCDE)
projects within the INTEGRAL team's Core Programme. Especially the
GCDE provided a wealth of data on GRS 1758
258. All Core Programme data up to
January 2005 as well as all data of the source public at that time
have been included into the analysis presented here. Our INTEGRAL data
are grouped into four data sets observed in spring and fall of 2003
and 2004 which in the following shall be called epoch 1-4. See
Table 1 for more details.
We used version 4.2 of the Offline Scientific Analysis package for
INTEGRAL to extract spectra and light curves of GRS 1758
258obtained by the
INTEGRAL Soft Gamma Ray Imager (ISGRI; Lebrun et al. 2003) detector as
well as spectra from the SPectrometer on INTEGRAL
(SPI; Vedrenne et al. 2003)
. See
http://integral.esac.esa.int/workshops/Jan2005/session1/lubinski_cross.pdf
for information on instrument cross-calibration issues for OSA 4.2. Due to the grid nature of the observations and the usually hard
source spectrum, the smaller field of view Joint European X-ray
Monitor (JEM-X; Lund et al. 2003) did not yield any useful data
covering the epoch time scales. In order to extract the ISGRI data
products, all pointings ("science windows'', "ScWs'') in which the
offset of the source from the spacecraft pointing direction was
smaller than 10
have been taken into account. For offsets in
this range systematic effects in the Crab calibration spectra show the
same general trends as for pointings in the fully coded field of
view. This selection amounts to
1920 ScWs with an exposure of
approximately 1800 s each. Average spectra for the four epochs were
built by producing images in 12 energy bands for each ScW in a given
epoch, extracting the GRS 1758
258source flux from each image, and
averaging the source fluxes of all ScWs in a given energy band using
standard weighting techniques. This method is described in the ISGRI
user manual
(http://isdc.unige.ch/doc/tec/um/user_support/ibis/ibis_4.2.pdf)
and is the recommended procedure for all but the brightest
sources. For the coded aperture instrument ISGRI the diffuse Galactic
background is part of the background removed when reconstructing the
sky images out of detector shadowgrams
(Goldwurm et al. 2003; Terrier et al. 2003). For the spectral modeling we use the
ancillary response file "isgr_arf_rsp_0006.fits'' and a rebinned
version of the response matrix "isgr_rmf_grp_0012.fits''
distributed with OSA 4.2. In addition, light curves with a time
resolution of 1000 s were produced in three energy bands: 20-60,
60-100, and 100-200 keV.
Table 1:
INTEGRAL observing epochs for GRS 1758
258, giving the exposure times
of the summed spectra analyzed for each epoch and instrument,
including the RXTE/PCA.
During the first epoch the source was too soft to be detected by the
SPI instrument. For the remaining three epochs the same ScWs as for
ISGRI were used to produce epoch-summed SPI spectra, with the
exception of epoch 2, where successful OSA runs could only be obtained
by splitting the SPI data into two subsets. The difference in the
exposure times given for ISGRI and SPI in Table 1 are
mainly due to ISGRI's dead-time. The SPI spectra were extracted over
an energy range of 20-500 keV (25 bins) using the SPIROS package
within OSA, applying maximum likelihood optimization statistics
(Skinner & Connell 2003). We set the number of pseudo detectors to 84
(i.e., including events located near borders between the physical SPI
detectors and registered in more than one of them) and selected
background correction method 5 (detector averaged count rate
modulation model). The input catalog of sources consisted of the 18
sources seen in the ISGRI 20-60 keV mosaic images. Alternative
parameter settings were tested, like changing the number of pseudo
detectors to 18 (i.e., including only single events), using background
model 2 (each detector scaled separately), or allowing sources to be
variable (SEL_FLAG = 2). None of these alternatives lead to a
significant change in the obtained count rates. Applying an
alternative extraction method optimized for recovering spatially
extended emission, Strong et al. (2003) find that the diffuse Galactic
background spectrum is of roughly power law shape, falling with a
slope of 2.5-3. We verified that adding such a component to only the
SPI data does not change the best fit parameters of the
multi-instrument fits discussed in Sect. 3.2 and that
the normalization of the new power law is consistent with zero. Note
that according to a study by Dubath et al. (2005) based on observations
and simulations, SPI fluxes of sources with known positions
can be well recovered down to a source separation of at least
0
5. With an angular distance of 0
66 from GRS 1758
258the
nearest source, the bright LMXB GX 5-1, should therefore
not contaminate our GRS 1758
258SPI spectra. For the data sets presented
here a careful inspection of the spectra of both sources shows no
indication of contamination with the possible exception of one bin
affected by an overcorrection (undercorrection) of the 66.7 keV
background line for GRS 1758
258(GX 5-1). We find, however, that
excluding this bin from the analysis does not change the results.
In order to characterize the source behavior at softer X-ray energies
we use data from RXTE's Proportional Counter Array
(PCA; Jahoda et al. 1996) obtained during the on-going RXTE
monitoring campaign. Under this program 1.5 ks snapshots of GRS 1758
258
have been taken monthly in 1996, weekly from 1997 through 2000, and
approximately twice each week since then. For a description of the
offset observing strategy applied to avoid nearby sources and of the
extra background measurements taken to minimize the influence of the
diffuse Galactic emission see Smith et al. (1997) and
Main et al. (1999). This procedure has been successfully evaluated using
data from the pronounced dim soft state in 2001. Reduction of the PCA
data was performed using the HEASOFT package version 5.3.1. The
responses were generated using pcarsp version 10.1 (see
http://lheawww.gsfc.nasa.gov/docs/xray/xte/pca/ for more
information on the RXTE/PCA energy calibration under this HEASOFT
version). Average spectra for the four INTEGRAL observing epochs were
produced. In addition, long term light curves consisting of the
average count rates of each PCA monitoring pointing were generated in
three energy bands (2.5-4, 4-10, and 10-25 keV) and for the
overall 2.5-25 keV light curve. We only use data from PCA's top
layer to optimize the signal to noise ratio. For the average spectra
we additionally decided to select data from one of PCA's five
Proportional Counter Units (PCUs), namely from PCU 2,
only
. The loss
of additional PCA exposure is acceptable since our aim is to study the
broad band spectral continuum (and not, e.g., to perform deeper iron
line studies) with emphasis on characterizing the hard spectral
component, i.e., on the INTEGRAL data. This strategy also allows us to
further minimize systematic effects due to PCU cross-calibration. See
Table 1 for the total exposure times of the
epoch-summed PCA spectra. Note that the All Sky Monitor (ASM) on
RXTE is not well suited to observe GRS 1758
258: the source's daily
1.3-12.2 keV ASM rate averaged over the time of the INTEGRAL mission
up 2005 March, e.g., is 2.0
2.5 cps. Also, since the absorbed
soft flux does not change much in the dim soft state (see next
section), no change is seen in the ASM light curve around epoch 1.
![]() |
Figure 1:
INTEGRAL/ISGRI and RXTE/PCA light curves of GRS 1758
![]() |
Open with DEXTER |
![]() |
Figure 2:
Evolution of the 2.5-25 keV PCA count rates of GRS 1758
![]() |
Open with DEXTER |
Figure 1 shows the evolution of the INTEGRAL and RXTE
light curves of GRS 1758
258during the monitoring in 2003 and 2004 in
several energy bands, spanning a total energy range of
2.5-200 keV. The INTEGRAL light curves have been rebinned to a
resolution of one day, for RXTE the average count rate of each PCA
data set is plotted, normalized to one PCU. The count rates during
epoch 1 are significantly lower in all energy bands above
4 keV. Above 100 keV the source is not detected in epoch 1. The PCA
measurements during and between epochs 1 and 2 suggest that the former
almost exactly covers the last two months of a
3 months long
period during which the source was in a dim soft state and that a
transition to the more common hard started with the end of
epoch 1.
This picture is supported by the average flux values determined from our spectral fits (Table 2). The 4-100 keV flux is considerably reduced in epoch 1. Consistent with the light curves the absorbed 2.5-4 keV flux does not change much between the soft and hard state epochs. The unabsorbed flux extrapolated to slightly lower energies, namely 2-4 keV, reveals an overall brightening at very low energies in epoch 1, however. A similar behavior was also observed during the onset of the 2001 dim soft state (Smith et al. 2001a). Different from 2001, though, the 2003 dim soft state starts with a short peak in the soft 2.5-4 keV light curve, coinciding with the decay above 10 keV. The 4-10 keV light curve shows a superposition of both trends, with the flare dominating first and then the decay.
Table 2: Average model flux for each epoch based on the best fit values of Table 3.
To put the INTEGRAL observing epochs into the broader context of the
source history, Fig. 2 shows the 2.5-25 keV light
curve from the PCA monitoring since 2000 as well as the 20-60 keV
ISGRI light curve again. The dim soft state in 2001 is readily
apparent, including two instances within the off phase (2001 May and
2001 July/August) where the source partly turns on again. The
2.5-4 keV and 10-25 keV overview light curves shown by
Heindl & Smith (2002a) include these turn-ons. In their Fig. 1 it can be
seen that the soft emission only reaches its minimum after the second
turn-on. The soft emission decays slower than the hard emission after
each turn-on, consistent with the two-flow scenario. The decrease of
the 2.5-25 keV rate in 2003 February is slower than the rapid
initial drop in 2001 February, but the 10-25 keV light curve in
Fig. 1 reveals that the hard emission decreases on a
similar time scale as in 2001, i.e., within about a week. The 2003 dim
soft state is shorter and not declining below 4 cps/PCU in the
2.5-25 keV band, though. In addition to the dim soft states there is
considerable long term variability present in the light curves, the
INTEGRAL 20-60 keV ISGRI rate, e.g., varies by a factor of 30-40%
within each hard state epoch. Furthermore, the 2.5-25 keV PCA light
curve shows several sudden count rate drops, less severe or shorter
than the dim soft states, e.g., in 2002 March or between epochs 3 and 4 or during epoch 4.
Table 3:
Best fit parameters for the power law models. The full model
in XSPEC notation for epoch 1 is constphabs
[diskbb+gauss+power], for epoch 2 it is
const
phabs[gauss+ cutoffpl] and for epochs 3 and 4
const
phabs[diskbb+gauss+ cutoffpl]. Parameters shown
are the hydrogen column density
,
the inner accretion
disk temperature
and its normalization
,
the power law index
and the power law cutoff energy
,
the energy
and equivalent width
of the Gaussian Fe K
line, and the flux
normalization constants of the individual instruments with respect to
the PCA,
and
.
Below 100 keV GRS 1758
258has often been modeled by an absorbed power law
(Mereghetti et al. 1997; Main et al. 1999; Sunyaev et al. 1991). Kuznetsov et al. (1999)
found that an exponential cutoff power law fits the 1990-1997
GRANAT data above 100 keV better than a simple power law and
Lin et al. (2000) obtained good fits to their joint RXTE/PCA,
RXTE/HEXTE, and CGRO/OSSE spectrum of 1997 with a cutoff
power law. Thermal Comptonization has also been shown to provide a
good description of the hard state spectra
(Kuznetsov et al. 1999; Keck et al. 2001; Lin et al. 2000). As reported in
Sect. 1, an additional weak thermal component can be
present in the hard state (Lin et al. 2000; Heindl & Smith 1998) which is more
clearly revealed in the intermediate
(Mereghetti et al. 1994; Goldwurm et al. 2001) and especially the dim soft state
(Heindl & Smith 2002a; Smith et al. 2001a; Miller et al. 2002).
From initial power law fits to our INTEGRAL data alone, we found that
for epochs 2 to 4 a cutoff is required. This is imposed by the ISGRI
data sets. Due to SPI's comparatively small effective area, the SPI
data do not carry enough weight to further constrain the cutoff
energy. Our basic phenomenological model for the simultaneous fits to
the summed INTEGRAL/RXTE spectra of each epoch thus consists of an
absorbed cutoff power law plus a Gaussian Fe K
line (see
Sect. 3.2.2 for a discussion of the need to include the
line), with an additional multicolor disk blackbody component if
required. We also applied a thermal Comptonization model
(compTT; Titarchuk 1994) to all four epochs, again including
absorption, Fe emission, and the optional disk blackbody as well as
allowing for a reflected Comptonized component
(Magdziarz & Zdziarski 1995). Normalization differences between the
instruments are taken into account in all fits by a multiplicative
constant, set to 1 for the PCA. The exact model compositions of both,
the phenomenological and the physical model, can be found in the
captions of Tables 3 and 4 for all
epochs.
Table 4:
Best fit parameters for the compTT model. The full
model in XSPEC notation is
constphabs[diskbb+ gauss+compTT+reflect(compTT)],
where for epoch 2 the diskbb component and for epochs 1 and 4
the reflect(compTT) component turned out not to be
required. The parameters shown are mostly the same as in
Table 3 but instead of the power law related
parameters, the physical parameters associated with Comptonization and
reflection are shown, namely the electron temperature of the
Comptonizing plasma
and its optical depth
,
and
the covering factor of the cold reflecting medium
.
in this table is either the
temperature of the diskbb component and/or the seed photon
input for the hot plasma.
We used XSPEC version 11.3.1t to perform the fits. Consistent with the
recommendations of the calibration teams, systematic errors of 0.5%
and 2% had to be added to all PCA and ISGRI spectra,
respectively. PCA data from 3-20 keV, ISGRI data from 20-150 keV,
and SPI data from 40-200 keV were taken into account in all fits,
with the exception of epoch 1 where PCA data up to only 18 keV and
ISGRI data up to only 100 keV were included. Both modeling approaches
resulted in good descriptions of the data and produced similar
values for given
epochs. Tables 3 and 4 list the best
fit parameters and
values for the power law and the
Comptonization models, respectively. Single parameter uncertainties
are given on a 90% confidence level. The results quoted for epoch 2
contain both SPI data sets. Without the 2.5 Ms of SPI data, the
values of the epoch 2 fits are in better
agreement with the quality of the other fits, e.g.,
for the epoch 2 cutoffpl fit (with no significant
changes of the best fit parameters). Figures 3 and 4 show the counts spectra, best fit models, and
residuals for the compTT fits.
![]() |
Figure 3:
Summed counts spectra for the GRS 1758
![]() |
Open with DEXTER |
![]() |
Figure 4: The same as Fig. 3 but for the observations of 2004 spring (epoch 3) and 2004 fall (epoch 4). |
Open with DEXTER |
Since the calibration of the INTEGRAL instruments, especially ISGRI, is work in progress, we expect that the best fit parameters characterizing the hard spectrum will be refined and updated in future iterations of this work. In this iteration we interpret them as indicators for general trends (e.g., the state change, qualitative consistency with canonical values, etc.). Modeling the spectra with Comptonization models also taking non-thermal electron distributions into account like compPS (Poutanen & Svensson 1996) or eqpair (Coppi 1999) will also be part of future work. However, consistency checks have been performed, applying the compPS model in a form comparable to our compTT fits (thermal electrons, slab geometry, optional multicolor disk blackbody). We obtain fits of similar quality, with seed photon temperatures, plasma temperatures and optical depths consistent with the compTT results. The reflection fraction obtained with compPS is systematically higher, though, e.g., 24% for epoch 2 compared to 10% with compTT. A similar trend of lower reflection fractions obtained with compTT was also observed between eqpair and compTT fits to Cyg X-1 INTEGRAL/RXTE spectra (Pottschmidt et al. 2004,2003). While the compTT/eqpair discrepancy is likely due to the omission of relativistic smearing of the reflection spectrum in compTT, as recently suggested by Wilms et al. (2006) on the basis of compTT and eqpair fits to several hundred RXTE monitoring observations of Cyg X-1, this is not the case here since our compPS fits do not include relativistic smearing.
As a Galactic Center source, GRS 1758
258is known to be strongly absorbed
and the
value adopted in most studies is
cm-2, as derived by
Mereghetti et al. (1997) from ASCA observations. However,
Keck et al. (2001) report
cm-2 from
ROSAT observations, Lin et al. (2000) find
cm-2 from RXTE observations, and
Goldwurm et al. (2001) determine
cm-2 with XMM. Modeling
PCA data starting at 3 keV,
and the blackbody parameters
are known to be strongly correlated and not well constrained. Here, we
obtain best fits with
values generally well consistent
with the canonical value of 1.5
cm-2 for epochs 1
and 2 (for the epoch 2 cutoffpl fit, though,
is closer to the lower value of Keck et al. 2001). In the case
of epoch 1 this includes a diskbb component which is
obviously required, while in the case of epoch 2 no thermal component
is needed. For epochs 3 and 4 the best fits with free
result in too low values of
cm-2 and
freezing
to the canonical value does not produce
acceptable fits. Adding a disk blackbody component, however, allows
for good fits with the canonical
(frozen with exception
of the epoch 4 cutoffpl fit). For the Comptonization fit of
epoch 4 this procedure results in a somewhat higher plasma temperature
and lower optical depth compared to the other hard state observations
than without including the disk component
(Table 4). The same tendency in presence of a disk
blackbody is seen when holding
at the lower value of
Keck et al. (2001). For all Comptonization fits the blackbody
temperature has been tied to the seed photon temperature of the
compTT component.
Clear residuals in the 6-7 keV range are present for all epochs when
no iron line is included. The
values obtained when removing
the Gaussian iron line from the models is given for reference in
Table 3. Note that the F-test may not be used to
test for the presence of a line (Protassov et al. 2002). In epoch 4 the
improvement of the fit when including the iron line is considerably
smaller than for the other epochs and an acceptable fit can be
achieved without the line (
), residuals
remain, however. The iron line is generally narrow, with widths around
or below 0.4 keV, and consistent with zero. The line energy ranges
from 6.40 to 6.73 keV and is mostly consistent with 6.4 keV, i.e.,
neutral Fe. Interestingly, the one exception is epoch 3 where we also
measure the strongest reflection component. A 3-4 times higher line
equivalent width is measured for the soft state epoch 1, consistent
with earlier measurements
(Heindl & Smith 1998; Sidoli & Mereghetti 2002; Smith et al. 2001a). This mainly reflects the
reduced level of continuum emission during that time, since the line
normalization does not change significantly between the epochs,
including the soft state. It has to be kept in mind, though, that the
Galactic diffuse emission features a strong iron line and that the
iron line parameters obtained from the fits are most likely influenced
by a non-perfect correction for this emission.
The parameters we are mainly interested in are those characterizing
the broad band continuum. We caution again that calibration
uncertainties prohibit a statistical comparison with earlier
results. Nevertheless we list earlier results for a qualitative
comparison and to illustrate the overall picture. For epochs 2 to 4 we
find values typical for hard state BHC spectra. For the
phenomenological model the power law indices lie between
1.54
+0.01-0.02 (epoch 2) and 1.69
+0.05-0.05 (epoch 4)
and the cutoff energies range from 136
+13-16 keV (epoch 3) to
246
+26-56 keV (epoch 4). Lin et al. (2000) find
(uncertainty of the order of 0.04) and
keV (uncertainty of the order of 30 keV) for their
joint RXTE and CGRO/OSSE spectrum of 1997 and
Kuznetsov et al. (1999) find
and
keV for their combined 1990-1997
GRANAT data (no cutoff was fit to shorter data sets). With
246 keV the cutoff energy for epoch 4 is at the limit of what can be
measured with these observations. However, no good fit can be obtained
without cutoff (
).
From the Comptonization models we obtain plasma temperatures of
78
+34-15 keV and 49
+29-9 keV and optical depths of
0.71
+0.16-0.07 and 1.00
+0.21-0.21 for epochs 2 and
3. Here the values of Lin et al. (2000) are
keV
(uncertainty of the order of 7 keV) and
(uncertainty of
the order of 0.3). Similarly, Sidoli & Mereghetti (2002) find values of
keV and
for
their BeppoSAX data set. In both of these cases the higher
optical depth is probably mainly due to two effects, first the fact that
no reflection component has been included in the models and second
that the sphere+disk geometry has been used. We see a moderate trend
towards higher values of
when switching from slab to
sphere+disk geometry in our fits.
Kuznetsov et al. (1999) obtain
keV and
.
However, they were using a predecessor to
compTT, namely the model of Sunyaev & Titarchuk (1980), therefore
their results cannot be directly compared to ours. Also, these values
reflect the average over a wide range of
and
values obtained from their two observation periods each year.
In the Comptonization fits we also allow for reflection of the
Comptonized radiation of a cold accretion disk and find reflection
factors of 10.0
+5.6-5.6% and 13.8
+5.0-5.5% for
epochs 2 and 3, respectively. No reflection is detected in
epoch 4. From the cutoffpl fits it is also clear that the
epoch 4 spectrum is less curved than the other two hard state spectra.
With
keV and
the Comptonization parameters of epoch 4
correspond to the hottest and most transparent plasma among the hard
state observations. While the latter might be an artifact due to the
introduction of the disk blackbody necessary to constrain
(the Compton-y changes only slightly)
, a possible physical origin for the differences
observed in the epoch 4 spectrum is suggested by the occurrence of one
of the sudden moderate drops in the PCA rate during this time (see
Fig. 2 and Sect. 3.1).
In general, the range of different results for the hard state
parameters is not too surprising in the light of the considerable
long term variations known to be present even within the hard state
(Fig. 2), however, it is also clear that INTEGRAL
calibration caveats apply. With
keV,
,
and reflection fractions of 17-24% the
compTT fits of Pottschmidt et al. (2003) to a set of INTEGRAL and
RXTE observations of Cyg X-1 result in similar parameters as
observed for epoch 3.
As expected from the long term evolution of the light curves, the
spectrum of epoch 1 differs considerably from the others. In both
models an additional soft component of comparable strength is clearly
present. We obtain a multicolor disk blackbody temperature of
477
+11-27 eV from the power law fit and of
482
+14-16 eV from the Comptonization fit. This is consistent
with the 2001 dim soft state where Smith et al. (2001a) found a disk
blackbody temperature of 4647 eV with the PCA,
Miller et al. (2002) give values of 340
10 eV and 600
10 eV,
depending on
,
for XMM observations, and
Heindl & Smith (2002b) find 505
7 eV with Chandra. Based on
these previously measured soft state values and since the soft state
spectrum is dominated by disk emission below
5 keV, we believe
that the values quoted above give a realistic measure of the
temperature. Not surprisingly the seed photon/disk temperature is
not well constrained in the hard state observations. For epoch 3,
e.g., the disk temperatures obtained from the cutoff power law and the
Comptonization fits are formally not consistent but what is consistent
is the fact that in both cases the disk component is needed if
is assumed to lie within the range of previously
measured values. Where a disk blackbody component was included in the
hard state fits it never dominates the soft spectrum. With
the power law is significantly steeper
in epoch 1 but does not quite reach the value of
observed in 2001 March (Smith et al. 2001a). No
cutoff is detected but during this time the high energy flux was
comparatively weak and the spectrum could only be obtained out to
100 keV. The steepness of the spectrum translates into a small
optical depth of 0.29
+0.43-0.13 in the Comptonization fit,
while the temperature of the hot plasma is found to be
64
+4-15 keV, i.e., not significantly different from the hard
state epochs 2 and 3.
![]() |
Figure 5:
Unfolded, unabsorbed model spectra corresponding to the
compTT fits listed in Table 4. The
![]() |
Open with DEXTER |
Is the dim soft state of GRS 1758
258really that different from the soft
states observed in other sources? In Fig. 5 the
unfolded, unabsorbed model spectra corresponding to the
compTT fits are shown. The typical pivoting between the soft
state spectrum and the three hard state spectra is seen. With
3 keV the pivot energy lies considerably lower than for Cyg
X-1, where a value of about
10 keV is observed
(Wilms et al. 2006; Zdziarski et al. 2002). However, taking the nature of GRS 1758
258as low
mass X-ray binary (LMXB) and Roche lobe accretor into account, its
behavior might be more akin to the state transitions displayed by LMXB
BHC transients than to those of the high mass X-ray binary (HMXB) and
focused wind accretor Cyg X-1, i.e., hysteresis might play an
important role. In the following the term "hysteresis'' is used to
describe the existence of an "overlap region'' in luminosity in which
both, soft and hard states can occur (see,
e.g., Zdziarski & Gierlinski 2004; Meyer-Hofmeister et al. 2005; Miyamoto et al. 1995)
. According
to the rough estimate for the bolometric luminosity that can be
derived from our fits (using a distance estimate of 8.5 kpc based on
the assumption of a near-GC location of GRS 1758
258), the 2003 dim soft state
is 0-20% less luminous than the hard state, depending on the hard
state epoch and spectral model used for comparison. For a
10
black hole the hard state luminosities that we measure
correspond to 2-3%
.
The differences between the
states in terms of fluxes in different energy bands have been
presented in Table 2.
Before comparing the range of hysteretic behavior observed in GRS 1758
258
to other sources, we note that another possible reason for observing
reduced soft state luminosities might be a geometric effect introduced
by the inclination i of the system: In the hard state the
geometrically thick hot plasma is present which can be assumed to
radiate approximately isotropically. In the soft state only the
decaying accretion disk remains which is geometrically thin with a
luminosity
(Frank et al. 1992). If the system is viewed
close to edge-on the projected area of the inner disk is comparatively
small, allowing only for a small percentage of the disk luminosity to
reach the observer. In addition, X-rays from the inner disk may be
further obscured due to flaring of the outer disk (Narayan & McClintock 2005).
GX 339-4 is the source for which hysteresis in the above sense has
been best studied so far. Depending on the energy range, its lowest
soft state flux can lie a factor of 2.5-10 below the brightest hard
state flux (Nowak et al. 2002; Belloni et al. 2005; Zdziarski et al. 2004). Nowak et al. (2002),
e.g., find that the 3-9 keV soft state flux can be less than half
the hard state flux, similar to what we find for GRS 1758
258
(Table 2). The bolometric flux of GX 339-4 in the
soft state can be up to an order of magnitude lower than in the hard
state (Zdziarski et al. 2004), an even more extreme behavior than indicated by
our bolometric estimates for GRS 1758
258. Accordingly, the schematic picture
which has recently been developed of the "q-shaped'' tracks
followed by black hole transients in the hardness-intensity diagram
over an outburst includes a large range of soft state intensities
(
,
Fender et al. 2004), not necessarily
exceeding the highest hard state ones. Although we concentrate on
average soft state parameters in this work, we want to note that the
hardness-intensity diagrams for the 2003 soft state that can be
derived from the energy-resolved PCA light curves show a
counterclockwise evolving pattern comparable to transients. Due to the
pronounced soft state it is rather slightly "p-shaped'' but
otherwise qualitatively very similar (a detailed quantitative
comparison is beyond the scope of this work). Overall it seems that
GRS 1758
258's dim soft state of 2003 - and also the even dimmer one of 2001
- are no remarkable states for a non high mass BHC (see
also Remillard 2005). This is especially true if part of the
luminosity reduction in the dim soft state is due to the inclination
effect described above. How about the overall state evolution, though?
Can the occurrence of the dim soft states be understood in the frame
of the outburst evolution scheme mentioned above? In the following two
paragraphs we discuss this question on the basis of the light curves
displayed in Figs. 1 and 2. Note,
however, that a detailed spectral analysis of the individual PCA
pointings is beyond the scope of this paper.
The initial phase of the state change consists of a sudden drop of the
>4 keV count rate around JD 2 452 680 (2003 February) and a
simultaneous moderate brightening of the thermal component, observed
as a 70
increase in the 2.5-4 keV count rate. Only two
monitoring observations find the source in this phase, i.e., it lasted
roughly a week. During the following weeks the dim soft state is
observed: the hard emission does not recover until end of 2003 April
and after the initial increase the soft emission decays slowly to a
low hard state level. The initial outburst-like phase is similar to a
canonical transition to a soft state but with the system not settling
into a state with a stable thermal component. In this sense the
episode is a "failed state transition''. The short soft flare may
reflect an actual change in the accretion disk parameters (e.g., a
temperature change and/or a change of the inner disk
radius). Alternatively, the increase in soft photon flux could at
least partly be caused by the disappearance of the Comptonizing
medium, i.e., the soft photons acting as seed photons in the hard
state are now emerging without being Comptonized. While the 2001 dim
soft state showed no initial flaring of the 2.5-4 keV count rate, a
soft excess compared to the hard state level also became visible in
the unabsorbed spectrum.
In contrast to the weeks long soft X-ray flares of Cyg X-1,
however, for which the term "failed state transition'' was coined
(Pottschmidt et al. 2003,2000), GRS 1758
258does not settle back
into the hard state after the flaring but the hard component stays
"off''. In the case of the 2001 dim soft state Smith et al. (2001a)
suggested a sudden shutoff of mass transfer from the companion being
responsible for the "off'' phase. Put into the context of the "q''
pattern of transient outbursts, the dim soft states of GRS 1758
258could
therefore well represent the thermally dominated outburst phase since
the main decay track proceeds through this state
(Remillard 2005). The hard state of GRS 1758
258, also covering a
considerable range of luminosities, would then correspond to phases of
rising and peak luminosities, again consistent with transient
outbursts. As mentioned in Sect. 1, additional
intermediate states - or failed state transitions - of GRS 1758
258have
been observed (Mereghetti et al. 1994; Heindl & Smith 2002a; Goldwurm et al. 2001), further
completing the outburst picture.
Another interesting property of the dim soft state is the fact that
the decay of the hard and soft spectral components is governed by two
different time scales. This has been studied in detail for the 2001
dim soft state by Smith et al. (2001a) who found that while the power law
flux decreased by an order of magnitude from one monitoring
observation to the next, the disk black body flux decayed on a time
scale of 28 d. This behavior is also visible in the 2003 light
curves (Fig. 1), especially in the fast decline of the
10-25 keV rates and the much slower trend in the 2.5-4 keV rates
after the initial drop down from the "failed state transition''
level. The source also shows several drops of the hard component for
durations of only a few days or less, e.g., around JD 2 451 932
(2001 January, see Fig. 2 and Smith et al. 2001a),
2 452 364 (2002 March, Fig. 2), 2 452 788 (2003 May,
Fig. 1), or 2 452 855 (2003 August,
Fig. 1). All these quasi-independent changes of the
hard and soft spectral component further support the interpretation of
the behavior of GRS 1758
258in terms of two different accretion flows. As
shown by Smith et al. (2001a), the model of Chakrabarti & Titarchuk (1995) can
explain many of the observations. As already mentioned, this model
assumes that proportional accretion rate changes introduced to both
flows at large radii propagate with nearly the free-fall time scale
through the Comptonizing medium and independently on the viscous time
scale through the accretion disk. Different propagation speeds are a
general feature of the model, i.e., they are not restricted to its
high accretion rate soft state associated with bulk motion
Comptonization. For lower accretion rates complicated dependencies of
spectral hardness and accretion rate are possible, covering the
correlation between the flux derivative and the spectral hardness as
well as the dim soft state (Smith et al. 2001a; Chakrabarti & Titarchuk 1995). The
strength of these time delay effects increases for larger accretion
disks and there are indications that such a picture might be generally
applicable for Roche lobe overflow transients: a state transition due
to a sudden change in the power law component during a time when the
disk black body parameters evolved smoothly has recently also been
seen in the black hole transient H1743-322, in this case
marking the transition between the thermally dominant and the
intermediate state (Kalemci et al. 2006).
![]() |
Figure 6: Ratio of the 10-25 keV and 2.5-4 keV PCA count rates. Dashed and dotted vertical lines denote the dim soft state of 2003 and times of sudden hardening, respectively. |
Open with DEXTER |
In addition to soft episodes we also observe several occurrences of a rather sudden hardening (Fig. 6) mainly due to declines of the soft component, visible, e.g., in the 2.5-4 keV rates around JD 2 452 797 (2003 June), 2 453 148 (2004 May), or 2 453 261 (2004 September). In the first case this is clearly related to a preceding drop of the hard component. In the second case the drop happens at the end of a months long decline of the count rates in all energy bands which is especially visible in the INTEGRAL range (epoch 3). The situation is less clear for the third occurrence (during epoch 4) but the 10-25 keV light curve also indicates a preceding decline. Due to this overall picture and the probably affected broad band spectrum of epoch 4, we consider it less likely that the hard episodes are caused, e.g., by absorption events, but are rather another example of the quasi-independent behavior of the hard and soft component on different time scales. Interestingly, a similar episode of sudden hardening has also been observed for the "two-flow source'' 1E 1740.7-2942 (Smith et al. 2002b).
Finally, while the hard state parameters have been discussed in Sect. 3.2.2 already, we emphasize again that apart from small peculiarities which might be caused by spectral variations within the epochs (epoch 4) the epoch-summed hard state spectra can be well described by cutoff power law and thermal Comptonization parameters which are compatible with canonical values found for BHCs in the hard state, e.g., Cyg X-1.
Acknowledgements
We thank Jörn Wilms for helpful discussions. This work has been partly funded by NASA contract NAS5-30720 (KP) as well as by NASA grants NAG5-13576 and NNG04GP41G (DMS, NB). AAZ and PL have been supported by KBN grants 1P03D01827, 1P03D01727, 1P03D01128, PBZ-KBN-054/P03/2001 and 4T12E04727. This work is based on observations with INTEGRAL, an ESA project with instruments and science data centre funded by ESA member states (especially the PI countries: Denmark, France, Germany, Italy, Switzerland, Spain), Czech Republic and Poland, and with the participation of Russia and the USA. We thank the RXTE schedulers for making the years long monitoring campaign of GRS 1758
258possible. KP thanks the Aspen Center for Physics for its hospitality during the final stages of the preparation of this paper.