A&A 407, L17-L20 (2003)
DOI: 10.1051/0004-6361:20030983
D. Porquet1 - P. Predehl1 - B. Aschenbach1 - N. Grosso2 - A. Goldwurm3 - P. Goldoni3 - R. S. Warwick4 - A. Decourchelle3
1 - Max-Planck-Institut für extraterrestrische Physik,
PO Box 1312, Garching bei München 85741, Germany
2 - Laboratoire d'Astrophysique de Grenoble, Université Joseph-Fourier, BP 53, 38041 Grenoble Cedex 9, France
3 - CEA, DSM, DAPNIA, Service d'Astrophysique, C.E. Saclay, 91191 Gif-Sur-Yvette Cedex, France
4 - Department of Physics and Astronomy, University of Leicester, Leicester LE1 7RH, UK
Received 26 May 2003 / Accepted 26 June 2003
Abstract
We report the high S/N observation on October 3, 2002 with XMM-Newton
of the brightest X-ray flare detected so far from Sgr A*
with a duration shorter than one hour (2.7 ks).
The light curve is almost symmetrical with respect to the peak flare,
and no significant difference between the soft and hard X-ray range is detected.
The overall flare spectrum is well represented by an absorbed power-law
with a soft photon spectral index of
,
and a peak 2-10 keV luminosity of 3.6
erg s-1,
i.e. a factor 160 higher than the Sgr A* quiescent value.
No significant spectral change during the flare is observed.
This X-ray flare is very different from other bright flares
reported so far: it is much brighter and softer.
The present accurate determination of the flare characteristics
challenge the current interpretation of
the physical processes occuring inside the very
close environment of Sgr A* by
bringing very strong constraints
for the theoretical flare models.
Key words: Galaxy: center - X-rays: individuals: Sgr A* - X-rays: general - radiation mechanisms: general
We report here the detection and the analysis of the up-to-now brightest X-ray flare from Sgr A*, observed in October 2002 with XMM-Newton. The present data provide the highest signal to noise measurements of a Sgr A* X-ray flare, reported so far.
The Galactic Center was observed on October 3, 2002
with Sgr A* at the center of the XMM-Newton/EPIC field of view
(see Turner et al. 2001; Strüder et al. 2001
for the description of the EPIC cameras MOS and PN, respectively).
The MOS (exposure time 16.8 ks) and PN (
13.9 ks)
cameras were both operated
in the prime full window mode, with the medium
and the thick filter selected, respectively.
The data were processed with the
XMM S CIENCE A NALYSIS S OFTWARE (version 5.4.1).
X-ray events with pattern 0-12 and 0-4 are used
for the MOS and PN, respectively.
We find a strong increase of the count rate during about 3 ks.
To identify the counterpart of this X-ray flare,
we improve the astrometry by cross-correlating the 8 PN sources detected
in the 0.5-1.5 keV energy band during the whole observation
with the 2MASS All-Sky Point Source Catalog using the eposcorr SAS task.
We find a reliable 2MASS counterpart for each of these
serendipitous PN sources, from which three are Tycho sources,
by introducing boresight correction and rotational offset of (
), and 2.7
,
respectively.
The refined position of the X-ray flaring source is then
,
,
with a 90% confidence level error box of 2.3
.
The radio counterpart of Sgr A* at
,
(Yusef-Zadeh et al. 1999), is located
only 0.8
away
from the flaring source. Since no other X-ray source was found by Chandra
in our error box, we conclude that the X-ray flare is very likely to be associated with Sgr A*.
For the light curves and spectral analysis, we use an extraction region
of 10
radius around Sgr A*, where the solar flare
background is negligible with a maximum count rate of
cts s-1.
![]() |
Figure 1:
EPIC light curves (MOS 1+MOS 2+PN) and hardness ratio (HR)
within a radius of 10
![]() ![]() ![]() |
Open with DEXTER |
The EPIC light curve in the 2-10 keV energy range shows
for the first 10 ks a quiescent period and
then a very sharp and intense
flare with a duration shorter than one hour
(2.7 ks, see Fig. 1 top panel).
The flare shape is almost symmetric relative to the peak.
Both the rise and decay phases can be fitted by
an exponential with characteristic times of 610 s and 770 s, respectively.
We also observe a "micro-structure'' at about 11 200 s
with a significance of at least 3
on a time-scale as short as 200 s, which
can further constrain the emitting region to a size of
about 7 Schwarzschild radii for the black hole mass
of about
.
The 2-10 keV count rate
at the flare peak is
cts s-1,
i.e. 9.2
+3.3-2.2 times higher than the mean value
measured during the quiescent period (
cts s-1).
B03 using Chandra found that the inner 10
radius region
around Sgr A* contains six other X-ray point sources as well as a diffuse component,
and estimated that the flux from Sgr A* during its quiescent state represents about 10%
of the total 2-10 keV flux.
Therefore we infer for the present flare a count rate increase of a factor of about 90
between the quiescent state of Sgr A* and the flare peak.
The shapes of the light curves in the soft (2-5 keV)
and hard (5-10 keV)
energy bands are very similar (Fig. 1, middle panel).
We report in Fig. 1 (lower panel),
the hardness ratio (HR) during the flare period:
.
We removed the quiescent level contribution,
assuming that it is constant during the flare:
mean count rates of
cts s-1,
and of
cts s-1, respectively for
the soft and hard energy bands.
This corresponds to a HR value for the quiescent period
of
.
We find a flare HR of
,
indicating a rather soft spectrum.
No significant spectral evolution is observed
between the [2-5] keV [5-10] keV ranges
(a similar result is found for a different energy range, e.g. [2-3.5] keV).
We find an indication that the flare presents a higher
HR value, compared to the quiescent,
in a restricted time period of the rise phase
(i.e.
s).
However, this does not imply a change
in spectrum of the flaring source since
the emission within 10
during the quiescent period is not dominated by Sgr A*.
![]() |
Figure 2:
EPIC spectra of the flare, within a radius of 10
![]() |
Open with DEXTER |
Figure 2 shows the extracted spectra for the overall flare period.
The overall flare spectra and the rise and decay phase spectra
were binned to a minimum of 20 counts and 10 count per bin, respectively.
XSPEC V11.2.0 is used for the spectral analysis.
We use the response matrices from the RMFGEN and ARFGEN
SAS tasks. The errors and upper limits
quoted correspond to 90% confidence level for one interesting parameter.
The quiescent spectrum including the contribution of Sgr A* in
its quiescent phase and the contribution of the other sources
in a radius of 10
around Sgr A*
is subtracted from the flare spectrum.
We fit the data with an absorbed power-law model taking into account
the dust scattering effects along the line-of-sight,
using the SCATTER model soon available in XSPEC
(Predehl & Schmitt 1995).
We fix for the SCATTER model,
a visual extinction value
of 30 mag,
as determined from IR observations of stars
close to Sgr A* (e.g., Rieke et al. 1989).
The interstellar medium (ISM) column density
is fitted using the model TBABS/XSPEC (Wilms et al. 2000),
which includes updated cross-sections for X-ray absorption by the ISM.
We also use their revised ISM abundances, which produce
changes in ISM cross-sections of up to 30% with respect to
solar abundances generally used as reference abundances for the ISM.
This means that the absorption column values
found here are about 30% higher than the ones
obtained assuming solar abundances.
The power-law continuum model with a soft photon index
gives a good representation of the data (see Fig. 2 and Table 1),
and we find at the flare peak a 2-10 keV luminosity
of 3.6
erg s-1, assuming d=8 kpc.
Fixing
,
as found in B01, we obtain
a non satisfactory data fit (Table 1).
The presence of narrow Fe K
Gaussian emission lines
(
eV) are not statistically required
(
for one additional parameter).
We find at 6.4 keV, 6.6 keV, 6.7 keV, and 7.0 keV
upper limits on the equivalent widths of 33 eV, 122 eV, 123 eV, and 65 eV, respectively.
Likewise the presence of a broad emission line near 6.7 keV
(
eV) and an absorption edge at 7.5 keV
are not statistically required and we obtain EW<173 eV and
,
respectively.
The data are also well fitted by a bremsstrahlung continuum model
with a temperature of about 6.1 keV (Table 1).
We also test an optically thin plasma using the MEKAL model
with ISM abundances, letting the metal abundance vary as a free parameter,
we find a good fit for a relative metal underabundance of less than 0.17.
While fixing the relative metal abundance to be equal or two times higher
than the ISM abundances, the data fit is not satisfactory, with very high temperatures
and therefore a large over-prediction of a H-like iron line at 7 keV (Table 1).
Surprisingly, the spectrum can also be well fitted by a black-body with
kT=1.47 keV with no significant change of temperature all over the flare.
The apparent size of the emitting area is about 10 km2
at the peak for a distance of 8 kpc.
The spectral fits of the rise and decay phases of this flare
are statistically consistent with each other with no change of the photon index
(
fixed to
cm-2),
with
(
),
and
(
),
respectively. This confirms the lack of significant hardening
according to the HR found.
Table 1:
Fit of the EPIC flare spectra, in the 2-10 keV energy range,
with absorbed power-law (pow), bremsstrahlung (brems),
black-body (bb) or mekal models, taking into account dust scattering.
Frozen parameters are given between brackets.
a
is units of 1023 cm-2.
b Z is the metal abundance relative to the ISM one.
c The fluxes are unabsorbed and expressed in units
of 10-11 erg cm-2 s-1.
Several models have been proposed to explain the first Sgr A* X-ray flare observed with Chandra. For instance, it could be produced by a sudden enhancement of accretion through the inner hot circularized flow predicted by the model of Liu & Melia (2002). When the viscosity increases, synchrotron self-Compton (SSC) of the millimeter to sub-millimeter photons is the dominant process, and a soft spectrum is expected. Such flares would be coupled with a similar increase in the mm and sub-mm ranges, which have not been yet observed (e.g., from March 2001 to May 2002, Zhao et al. 2003). If the viscosity decreases then thermal Bremsstrahlung is the dominant process, and a hard spectrum would be generated, consistent with the slopes observed in the two other X-ray (much lower S/N) flares. Three possible flare mechanisms have been proposed by Markoff et al. (2001) on the basis of a jet model: increased jet power or accretion rate, increased heating of quasi-thermal relativistic particles, or sudden shock acceleration. The first scenario predicts a steep spectral index, and a huge radio flare for an accretion rate increase, and simultaneous flaring at all frequencies for an jet power increase, but this have never been yet observed. The second and third scenarii predict harder spectrum and were compatible with the spectral index of the first Chandra X-ray flare.
The flare luminosity excludes normal stars as the origin of this event, but compact objects, although unlikely, cannot be totally excluded (see discussion in B01). We rule out the possibility that the symmetrical shape of the flare light curve is produced by micro-lensing of a star located behind Sgr A*. Indeed from Zinnecker (2003)'s formulae, we find that hour duration micro-lensing events imply very high tangential velocities for the lensed star and unphysical distances from the black hole, less than one Schwarzschild radius. For instance a velocity as high as 10 000 km s-1 corresponds to a distance of about only 1012 cm. The duration of the trailing edge of this flare is very short. If the physical process is thermal, this could be explained by the very rapid cooling of a plasma not magnetically confined, as proposed by Merloni & Fabian (2002).
The total column density inferred
from the spectral fit is about
cm-2 for the power-law, bremsstrahlung and mekal models.
The visual extinction
mag, inferred from IR observations of
stars very close to Sgr A* (Rieke et al. 1989),
corresponds to an absorption column density
along the line-of-sight of about
cm-2 (Predehl & Schmitt 1995).
This discrepancy between
and
(also reported by Predehl & Trümper 1994)
may indicate an additional, probably related to the source,
gaseous absorption from nearly neutral gas.
A spectral analysis of other X-ray sources
in the XMM-Newton field of view
will enable to check whether this absorption excess
is peculiar to Sgr A* (Predehl et al., in preparation).
Nayakshin et al. (2003) proposed
that such X-ray flares are due to the passage of stars through the accretion
disk around Sgr A*, and that extra absorption may be related to
the neutral accretion disk photosphere
through which the star passes.
The excess absorption could also be explained by an overabundance of heavy
metal elements from a putative Supernova remnant
or He and other heavy elements from
He stars located in the neighbourhood of Sgr A* (e.g., Eckart et al. 1995).
However, for the blackbody model no extra absorption is required.
We find also a soft X-ray flare spectrum
with
,
much softer than for the two other bright flares reported up to now
(with lower S/N):
(B01), and
(Goldwurm et al. 2003).
The present spectral index is consistent with the ones
found during the quiescent periods of Sgr A* by Chandra:
in September 1999 (B03),
and
in October 2000 (B01).
Such a soft X-ray spectrum is predicted by the scenario
proposed by Liu & Melia (2002) in case of an increased viscosity
of the accretion flow, and by the one proposed by Markoff et al.
(2001) in case of an increase of accretion rate or jet power.
However the predictions of these two latter models are not
consistent with observational constraints found at other wavelengths
(e.g., Zhao et al. 2003; Hornstein et al. 2002).
Although, the Nayakshin et al. (2003) star-disk interaction model
predicts a soft X-ray spectral index (
)
and a moderate
Fe K
line EW of about 60 eV
the allowed temperature values (kT=1 and 70 keV) are very fine-tuned
(S. Nayakshin, private communication).
Emission from a hot optically thin plasma with
keV
and a low relative metal abundance (<0.17 compared with the ISM abundances)
is also consistent with the observed spectrum.
But the high metal depletion is difficult to explain
in view of the ambient, diffuse and metal rich emission and the presence
of close supernova remnants such as Sgr A East.
If the flare is related to the accretion disk, the low metallicity could indicate a
less evolved material in the disk.
In conclusion, the present observation strongly constrains the flare modeling of Sgr A*.
Acknowledgements
This work is based on observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and the USA (NASA). We would like to thank the referee, F. K. Baganoff, for constructive comments and suggestions. D.P. is supported by a MPE fellowship.