A&A 426, 787-796 (2004)
DOI: 10.1051/0004-6361:20047110
A. Wolter - G. Trinchieri
Osservatorio Astronomico di Brera, via Brera 28, 20121 Milano, Italy
Received 20 January 2004 / Accepted 30 June 2004
Abstract
We present the results
from the high resolution Chandra observation of
the Cartwheel galaxy. Many individual sources are resolved in the
image, mostly associated with the outer ring. All detected sources
have a very high X-ray luminosity (
erg s-1)
that classifies them as
Ultra Luminous X-ray sources (ULX). The brightest of them is possibly the
most luminous individual non-nuclear source observed so far,
with
erg s-1 (at D=122 Mpc).
The spatial extent of this source is consistent
with a point source at the Chandra resolution.
The luminosity function of individual X-ray sources extends about
an order of magnitude higher than previously reported in other galaxies.
We discuss this in the context of the "universal'' luminosity function
for High Mass X-ray Binaries and we derive a Star Formation Rate
higher than in other starburst galaxies
studied so far.
A diffuse component, associated with hot gas, is present.
However, deeper
observations that we will obtain with XMM-Newton
are needed to constrain its properties.
Key words: galaxies: general - X-rays: galaxies - galaxies: luminosity function, mass function - galaxies: individual: Cartwheel - galaxies: interactions
The Cartwheel galaxy is a spectacular object, with the peculiar appearance
reminiscent of a wheel (hence the name), most probably the result of an
impact with one of the companion galaxies.
It is located in a tight, compact group (SCG 0035-3357;
Iovino et al. 2003) of 4 members,
very close in space (0.3 Mpc
and velocity
400 km s-1from one another, see Taylor & Atherton 1984).
Whether the impact was due to G3 (at
NE, Higdon 1996) or G2
(at
to the North; Athanassoula et al. 1997),
two rings are now visible as a result: the outer one has the
largest linear diameter measured in ring galaxies: 80'' (
100 kpc)
along the major axis; the inner one, close to the core, is elliptical in
shape with obvious dust lanes crossing it
(Struck et al. 1996).
![]() |
Figure 1: Chandra image of the Cartwheel in the ( left) soft (0.3-2 keV) and ( right) hard (2.0-7.0 keV) energy bands. An adaptive smoothing algorithm has been applied to the data (see text). Numbers identify sources as given in Table 1, with regions used to derive the source counts (from wavdetect). |
Open with DEXTER |
Many detailed observations of the Cartwheel are available,
ranging from radio line (Higdon 1996) and continuum (Higdon 1996),
to near- (Marcum et al. 1992) and far-infrared (Appleton & Struck-Marcell
1987), optical (Theys & Spiegel 1976; Fosbury & Hawarden 1977) and
H
images (Higdon 1995) and line spectroscopy (Fosbury &
Hawarden 1977). All have confirmed
the presence of a recent starburst in the outer ring, without a
corresponding activity in the inner ring, nucleus or spokes, believed
to be relatively devoid of gas. Most
of the activity is confined in fact in the S-SW portion of the ring,
where massive and luminous HII regions characterized by large
H
luminosities and equivalent widths are found (Higdon 1995).
Both dynamical considerations and stellar evolution models suggest
an age of
yr for the star burst. The estimated supernova
rate, as high as 1 SN/yr (i.e. almost two orders of magnitude higher than in
normal galaxies), coupled with the evidence of a very low metallicity
environment (as measured from O, N and Ne)
also supports the view that star formation in the ring is a
recent phenomenon and that the gas currently forming stars
was nearly primordial at the time
of the impact (Fosbury & Hawarden 1977; Higdon 1995; Marcum et al. 1992).
We have imaged the Cartwheel for the first time in the X-ray band, using the HRI on board ROSAT (Wolter et al. 1999), and were able to attribute most of the emission to the outer ring, stronger in the Southern quadrant and very clumpy in nature. We therefore asked and obtained Chandra data with the ACIS-S in imaging mode, to better study the spatial distribution of the emission. We present here the Chandra data, and a discussion of the detection of both a diffuse component and a number of Ultra-Luminous X-ray sources well in excess of expectations.
The paper is organized as follows: in Sect. 2 we present the Chandra data; in Sect. 3 the results of the analysis for the individual sources and for the extended component; in Sect. 4 we discuss the results and present the X-ray Luminosity Function of the Cartwheel sources. Section 5 summarizes our findings.
The Cartwheel was observed by Chandra on 26-27th May 2001, with ACIS-S in imaging mode, operated in the standard full-frame mode, with an integration time of 3.2 s. The back-illuminated ACIS-S S3 chip was chosen for its soft response to detect the low temperature component. Telemetry was in faint mode. For a description of the Chandra mission see Weisskopf et al. (2000).
The data were reduced with the standard Chandra pipeline with the CIAO software (version 2.3) and the most recent available calibration products, as described in http://asc.harvard.edu/. The corrections applied are those appropriate for the ACIS-S instrument.
No evidence of in-orbit high background was found in the data, so the net exposure time is 76.1 ks.
Figure 1 shows the adaptively smoothed images of
the Cartwheel
in two energy bands (0.3-2.0 keV and 2.0-7.0 keV respectively;
the energy limits are chosen as a good compromise that maximizes
the signal and minimizes
the particle background contribution and calibration uncertainties).
Both images show a very clumpy emission, mostly confined to the outer ring
(see also Wolter et al. 1999; Wolter & Trinchieri 2003; Gao et al. 2003).
Individual sources that appear point-like at the Chandra resolution
(1.5 kpc at the Cartwheel distance, see Sect. 3.1.1) account
for most of it, in particular in the harder energy band (right panel).
A fraction of the counts (
20% of the total) is found in a
more extended component, that appears in the softer energy band.
An additional component, of very low
surface brightness, appears to connect the southern portion of the
ring to the two nearby companions G1 and G2 in the soft image
(left panel of Fig. 1).
No excess emission is detected at the position of the optical nucleus.
The X-ray contours from the smoothed 0.3-7 keV image are shown in
Fig. 2 superposed on to the
HST image
and show the correspondence of the X-ray emission with the outer ring.
We will explore all the evidence more quantitatively in the following sections.
![]() |
Figure 2:
The X-ray contours in the 0.3-7 keV Chandra image
over-plotted on the optical image in the F450W filter from the HST archives.
X-ray contours are:
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
Table 1: List of sources that are detected in the area covered by the Cartwheel system and displayed in Fig. 1. Unabsorbed fluxes are computed in the 0.5-2 keV and 2-10 keV bands assuming the best fit model of the sum of point sources (see text for details). Unabsorbed, k-corrected luminosities in erg s-1 are in the 2-10 keV band. The sources are also reported in the Appendix in RA order for completeness; however, in the Appendix the fluxes are all computed under a different spectral hypothesis (see text). These sources are identified by their number in the last column of Table A.1 for easy reference.
A wavdetect detection algorithm applied to the 0.3-7.0 keV image
provides 72 sources in the whole field. We have applied the algorithm
using scales ranging from 1
to 1
and
a significance threshold of 10-6 that corresponds to 0.25
false sources in the image considered at any given scale.
All detected sources are listed and discussed in the Appendix and in
Table A.1.
In Table 1 we list the positions and count rates (in the 0.3-7.0 keV band) of the 25 sources located in the Cartwheel region, most of which should be associated with it (see also Fig. 1 and Sect. 2).
Table 2: Spectral fit results for source N.10.
By assuming the spectrum found for the sum of point sources
(see Sect. 3.1.2), we compute the unabsorbed fluxes in the 0.5-2 keV
and in the 2-10 keV bands.
Assuming the distance of 122 Mpc for the Cartwheel we list in Table 1
the logarithm of the k-corrected
luminosity
in the 2-10 keV band.
The brightest point source in the outer ring has a luminosity of at least
erg s-1 in the 0.5-2 keV
band, and of
erg s-1 in the 2-10 keV band
, one
of the brightest individual sources ever found in
galaxies.
Even though at the Cartwheel distance the Chandra resolution defines
a region of a few kpc that could contain more than one source,
the high X-ray luminosity suggests either a single extremely
bright source, or a very dense collection of several, high
sources,
which is probably even more peculiar. A possible source variability suggested
for this source (see Sect. 4.1) further suggests that this is indeed
a single high
source
(or equivalently that a single source dominates the emission from this
region).
No source is detected at the position of the optical nucleus of the
Cartwheel, indicating that it is fainter than our detection limit
of a few
erg s-1. An active nucleus is thus either
not present or it is so heavily absorbed that even at
7 keV no emission
emerges.
A number of individual sources coincide with G1, for a total
of 220 net counts, that correspond to a flux
erg cm-2 s-1 and
erg cm-2 s-1; for a luminosity
erg s-1 (0.5-10 keV band)
(assuming the same spectrum of the sum of point sources, see Fig. 6
and Sect. 3.1.2).
The disturbed optical morphology of G1 and the large
number of discrete high
sources detected in the galaxy area
are both consistent with an interaction scenario (G1 with G2, as suggested
by Higdon 1996).
A source that might be positionally coincident with the nucleus
is also detected at the location of
G3 at 3.9
with a total of 9.6
3.3 counts (see Appendix A).
The corresponding flux obtained with the same spectrum as for G1
is
erg cm-2 s-1;
erg cm-2 s-1; for
a total luminosity
erg s-1 (0.5-10 keV band).
While we were preparing this paper, a list of sources in the Cartwheel
region was presented by Gao et al. (2003). We compare our list
with that presented by them. Positions are generally consistent,
however there is a small systematic shift of
in RA and
in declination between our positions and those
of Gao et al. A few sources are only present in either list (see
notes to Table 1): it is likely that
differences in the detection process and thresholds assumed cause the slightly
different source lists; the effect is however limited to faint or
confused sources.
The count rates of the common sources are generally higher in the Gao et al.
list. This might be the result of different bands used (the detection band
is not indicated in the Gao et al. paper) and/or background
subtraction considered and/or a different detection algorithm
(under the IDL software).
![]() |
Figure 3: Surface brightness profile of source N.10 in the soft range (0.3-2 keV) ( top left) and in the total range (0.3-6 keV) ( top right) compared with the PSF from chart in the same energy ranges. Surface brightness profile of source N.14 in the energy range 0.3-2 keV ( bottom left) and 0.3-6 keV ( bottom right) compared with the PSF from chart in the same energy ranges. Source N.13 has been masked out before computing the profiles. |
Open with DEXTER |
In order to investigate the nature of the sources we have studied the radial profile of the brightest ones (i.e. with sufficient counts to perform a significant comparison with the Point Spread Function, PSF, obtained by ray tracing the photons distributed with the same spectral shape as the source under test with the CIAO task chart). The profiles of two of these sources are plotted in Fig. 3 in two different energy bands, after subtracting the detected nearby sources, together with the PSF binned in the same energy range.
The sources are unresolved at the resolution of the instrument, i.e. they
have a core of
.
An extended source of lower surface
brightness is present and it is stronger at soft energies for
,
most evident for source N.10.
We infer that
this is due to the underlying ring of star formation in the Cartwheel.
The thickness of this component,
,
is
comparable to that in the optical band.
We have also performed a number of projections, orthogonal to the ring, in a few spots of the southern ring. We plot two of these in Fig. 4. The location is given relative to the sources identified in Fig. 1. The X-axis is approximately along the radial distance from the center of the ring, from the inside towards the outside of the Cartwheel. The peak in the distribution corresponds to the ring, which appears as an annular plateau over which point sources stand out. These plots also suggest that there is a higher level of emission "inside'' the ring (i.e. low "X-dim'' values) than "outside'' (i.e. high "X-dim'' values) and that a rather sharp drop confines the outer limb of the ring. With the limited number of counts detected "inside'' the ring, detailed studies of this component are not possible and will have to wait for more sensitive observations.
![]() |
Figure 4:
Cuts in regions orthogonal to the star-forming ring. See text for comments.
The regions used are slits of 5.6
![]() ![]() |
Open with DEXTER |
The spectrum of the brightest source in the Cartwheel ring has
already been
described by Gao et al. (2003). However, with our analysis we find
slightly different results; unfortunately not enough details
are given in the Gao et al. paper for us to ascertain if the
differences are based on different hypotheses or different treatment
of the data. Therefore we present here in detail the spectral analysis
of the brightest source (the only one for which a spectral analysis can
be attempted, with 380 net counts) and for the sum of the other
individual sources detected around the ring (for which we collect in total
more than 600 net counts).
The background has been computed in large circles
devoid of bright sources at about the same off-axis position as the Cartwheel.
![]() |
Figure 5:
ACIS-S spectrum of source N.10.
The solid line corresponds with a power law model
with
![]() ![]() |
Open with DEXTER |
For the brightest source (N.10) we bin the data to have at least 30 total
counts in each bin, and we apply a simple power-law model with low
energy absorption (see Table 2).
The fitted
value is consistent with other absorption measures
in the Carthweel: the best fit is higher than the line-of-sight
Galactic value but consistent with the intrinsic absorption
measured e.g. in the optical band:
the value of
measured (Fosbury & Hawarden 1977)
corresponds to
cm-2(using
,
for
,
see e.g. Bohlin et al. 1978) well within the range of the fitted
values
in Table 2.
The power law slope would indicate a High Mass X-ray Binary (HMXB), as also
proposed by Wolter
et al. (1999) based on luminosity arguments.
Although a more complex model is not required by the data, we also used
the absorbed
Multicolor Disk model (diskbb model in XSPEC; Mitsuda et al. 1984)
used for other ULX (see e.g. Bauer et al. 2003; Zezas et al. 2002),
which gives an equally good fit
(see Table 2).
The formal errors in the temperature/
parameters (90% confidence region for one
interesting parameter) are given in Table 2.
Unfortunately, given the statistics, a
thermal plasma model, like a Raymond-Smith or a Mekal with fixed
solar abundance, also gives formally
acceptable fits (
of the order of 1.1), with a temperature
(
2.5 keV) not strongly constrained.
Therefore we feel that more detailed spectral models will not convey
meaningful additional information for this source.
However, we notice the same small excess at
1.5 keV that was pointed out by Gao et al. (2003).
This line could correspond to a feature from either Mg, or Al, but if these
are due to a low temperature plasma, there should be more prominent lines
around 1 keV.
Lines produced by low ionization states of Mg, Si, and S in an optically thin gas
(see e.g. Iwasawa et al. 2003) could also be an explanation:
e.g. in the spectrum of source #11 in the Antennae
a MgXIII line is fitted at
1.50 keV (Zezas et al. 2002).
While adding a Gaussian component would clearly reduce
the minimum
,
we feel that this is not required, either
statistically (the
is about 1 even without
the line) or from the distribution of the residuals.
Moreover it is also clear from the Gao et al. results that their
are small (<1 always), reflecting the very low
significance of each bin.
Fluxes and luminosities are given in Table 2 for different bands,
including a luminosity in the 0.05-100 keV range,
often assumed to be a measure
of the bolometric luminosity. All luminosities are k-corrected.
As the values in Table 2 show, the "bolometric'' luminosity
strongly depends on the model and on the
value assumed,
given the large extrapolation to both low and high energies,
therefore caution should be used when comparing "bolometric''
luminosities derived
from data extracted in significantly smaller energy ranges and
extrapolated using different models.
To derive an average spectrum for the individual sources in the
Cartwheel we also accumulated the counts from
circles around the positions associated
to the other individual sources detected.
Source N.10 is not included since it is by far the brightest and its
inclusion would heavily bias the results. We also
do not include sources N.11 and N.22 that appear to have a
lower soft/hard count ratio than the average source,
suggesting either an intrinsically different
spectrum or a larger absorbing column.
Since their location is also not coincident with the star-forming ring,
they could be background sources seen through
the absorbing material in the Cartwheel. Optical observations (imaging
and spectral) are needed to confirm the identification.
The spectral data are binned so that each bin has a
significance
after background subtraction.
The resulting spectrum is shown in
Fig. 6. The data are fitted by a power law with best fit values
cm-2
and
[2.04-2.34] (reduced
for 50 d.o.f.).
Even in this case, the derived low energy absorption is higher than galactic,
but consistent with the reddening observed in the HII regions.
The contour plot of the uncertainties in spectral index and
low energy absorption is shown in Fig. 7.
The photon index we obtain is steeper than what is observed in other bright
binaries in nearby galaxies when fitted with a simple power law model,
and in particular steeper than the spectrum of source N.10, but
a few examples of steep sources are present e.g. in the Antennae
(Zezas et al. 2002) or
in other galaxies (e.g. Humphrey et al. 2003).
Since it is likely that this slope results from a combination of different
slopes, and perhaps even of different intrinsic absorption, depending on the
location of the source with respect to the star forming region, this
result should be taken with caution and might not be indicative of
a population of steeper spectra sources.
The measured unabsorbed flux is 3.0/
erg cm-2 s-1
(in the 0.5-2.0/2.0-10 keV band),
which corresponds to a total luminosity of
/
(0.5-2.0/2.0-10.0 keV) erg s-1.
The sum of the luminosity of the detected point sources is then roughly
twice that of N.10 in the soft band and comparable to that of N.10 in the hard band.
![]() |
Figure 6:
ACIS-S spectrum of the combined individual
sources. The solid line corresponds with a power law fit with
![]() ![]() |
Open with DEXTER |
![]() |
Figure 7:
Confidence contours of the two parameters ![]() ![]() |
Open with DEXTER |
To investigate the spectral properties of the "extended'' emission
we collect counts from a region that includes virtually all of the
Cartwheel extent, but excludes all detected point sources analyzed
previously.
The spectrum has been binned to have a significance
after
background subtraction in each bin.
In spite of the limited statistics,
we found that the spectrum requires at
least two components.
A single component fit (e.g. power law, plasma model with fixed abundance)
is formally acceptable since the errorbars are large, but it has badly
distributed residuals.
We tried different combinations of two components, which are all
equally acceptable, given the statistics.
A good representation (Fig. 8) is
given by a combination of a power law
(
,
cm-2)
and a plasma model (Raymond-Smith component with kT=0.2 keV and
an abundance fixed at 0.5
solar to reflect the low metallicity
of the gas measured in optical-IR).
The power law component, which could represent fainter unresolved individual
binaries, has a slope consistent with that found for the combined point
sources (see Sect. 4.1.3). The unabsorbed flux of this component is
/
erg cm-2 s-1 (0.5-2.0/2.0-10.0 keV),
about 25/10% of the resolved point source flux in the soft and hard band
respectively.
The flux of the diffuse hot gas component is
erg cm-2 s-1 (0.5-2.0 keV band).
It contributes mostly at 0.6-0.9 keV as expected from the temperature
found. The total luminosity of the gas is of the order of
erg s-1, i.e. only about a factor 4 less than the
soft gaseous component in the most X-ray luminous starburst galaxy known,
NGC 3256 (Moran et al. 1999).
![]() |
Figure 8: ACIS-S unfolded spectrum of the diffuse component: even if the statistics are low, the spectrum cannot be fitted by a single component. We show here the fit with a hot plasma component (dashed line) plus power law (dashed gray line) plus the total fit (solid gray line). See text for details. |
Open with DEXTER |
![]() |
Figure 9: The ROSAT-HRI contours from the adaptively smoothed image are over-plotted onto the Chandra soft adaptively smoothed image in gray scale. |
Open with DEXTER |
In Fig. 9
we show the HRI contours overlaid on the adaptively
smoothed Chandra image in the soft band.
The only striking difference in the image is in source labeled "A'' in
the figure,
which is not present in the new Chandra image. However,
the source is probably not related to the Cartwheel itself, so it might be
any kind of variable source from an AGN to a transient Galactic source.
The total soft luminosity
seen by Chandra (adding up all the different components, from point sources
to diffuse gas)
of
erg s-1 is consistent with that of
the HRI of
cgs, once the same H0 is assumed,
and taking into account the different
spectral parameters used to derive the luminosities.
A strict comparison of the contribution from each individual source is
not easy, given the very different size of the PSF of the two instruments.
However,
by comparing the main knots of emission in the HRI with the corresponding
regions in Chandra,
we notice a variation in the flux level in at least one region.
If we make the hypothesis that the region around G1 has not varied,
then the region around sources N.10-13-14 shows a factor of 3 increase.
This is consistent with variability observed in other ULX,
if one source only is responsible for the flux increase.
Since N.10 is the dominant source we can probably attribute the
variation to this one source; otherwise, as already suggested in
Sect. 3.1, we should assume that there is
a large compact cluster of sources that vary together in a region
of
1.5-2 kpc. For comparison, in a similar region in
the Antennae, closer and therfore more resolved,
there are up to 7 sources brighter than 1038 erg s-1, of which
only 2 are above 1039 erg s-1 (see
Fig. 1 and Table 1 of Zezas et al. 2002), which would contribute
only about 1/10 of the observed luminosity of source N.10.
The flux variation implies that an origin of the X-ray emission from SNR is probably less likely than the association with an accreting compact object.
Even if N.10 is by far the brightest source in the ring, it is clear
from Fig. 9 that the HRI was also detecting the entire
ring emission.
We estimate in fact that in the 0.2-2.0 keV band,
the brightest
source N.10 contributes only 1/4 of the total luminosity.
Gao et al. (2003) instead propose that most
of the emission detected by the ROSAT HRI is from this source only;
however they use a much broader band than the 0.2-2.0 keV that
matches the ROSAT HRI energy band.
Giant HII regions and complex structures, typically coincident with peaks
of H
emission, have been observed in actively star-forming objects
like the interacting system "The Antennae'' (NGC 4038/9; e.g. Fabbiano et al. 2001) as extremely bright X-ray sources, with intrinsic
luminosities reaching several
erg s-1. To check this association
in the Cartwheel, we have plotted the positions of the HII knots as
measured by Higdon (1995) on the X-ray image in Fig. 10.
The positions of the circles that mark the HII regions have been
shifted
by about 1
in RA and 0.5
in Dec for better agreement with the locus of
the X-ray peaks (well within the positional uncertainty of both
the X-ray and the HII reference frame). A general trend in the location
of the X-ray and HII emission is evident. However, there is no one-to-one
correspondence with any of the X-ray sources. If anything, the X-ray emission
seems to be at the edge of the HII knots.
The same kind of general positional agreement is evident between the
X-ray emission and the Mid-IR peaks
(Charmandaris et al. 1999); the so called "hot spot'', especially at
15
m, dominates the output in this energy range. It coincides with
two large HII complexes, but not with the brightest X-ray source.
The resolution of the ISO data is however 6-7
,
so more
than one X-ray source (e.g. at least N.22 and N.24) might be
associated with the ISO emission.
The best interpretation of this similarity is that the region is active
in general, but time scales and regions of emission
are not directly linked.
![]() |
Figure 10: X-ray smoothed map in the (0.3-7.0) keV band. Over-plotted are the position of HII regions numbered as in Higdon (1995). |
Open with DEXTER |
The excessive number of very high X-ray luminosity individual sources makes it interesting and complicated to understand their nature.
The limiting luminosity for point sources
is of the order of
erg s-1,
which is already above the Eddington limit for a neutron star binary
(
erg s-1; see e.g. King et al. 2001).
Most of the detected sources are also above the limits of canonical
Ultra-Luminous X-ray sources (ULX), i.e.
erg s-1.
The main uncertainty in the
calculation is the correct association of a
few of the individual sources to the Cartwheel itself (i.e.
they might be foreground or background sources). We assume that all sources
within the optical ring belong to the Cartwheel. However
there are also a number of sources not positionally
coincident with the ring: N.6, N.8, N.11, N.20, N.22
that we consider below one by one.
N.8 is located between the Cartwheel and the G1 galaxy, so it is probably unrelated to either source, unless the encounter affected even this area. We consider this unlikely and treat it as a background source.
From comparison of the soft and hard count rates sources N.11 and N.22 appear to have a different spectrum from the rest.
If this results from higher absorption, with
the limited statistics available, we estimate
an absorber of
cm-2, consistent with a
galaxy like the Cartwheel itself.
However it is not possible to determine whether the sources
are embedded in the absorber, and therefore belong to the Cartwheel,
or are behind it, and therefore background sources. Given the
location of N.22 in the ring we consider this to belong to the
Cartwheel. The association of N.11 is less certain, however by analogy
with N.22 we also consider it as part of the galaxy and include both in
the estimate of the Luminosity Function in the next section.
N.6 and N.20 are not exactly on the peak of star formation, but close to the inner side of the ring. We consider them "leftovers'' from past star formation, and therefore related to the Cartwheel.
We discuss below the properties of the ULX in the Cartwheel on a statistical basis, by deriving their Luminosity Function.
![]() |
Figure 11:
Luminosity function of all bright isolated sources in the ring.
The solid line represents the Grimm et al. (2003) luminosity function with
![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
We have computed the luminosity function (LF) of all 16 isolated sources detected along the outer ring, assuming the distance of the Cartwheel for computing the luminosities (sources that are not used are indicated with a note in Table 1).
Grimm et al. (2003) propose that
a "universal'' LF for HMXB can be constructed by normalizing to
the Star Formation Rate (SFR)
in the formula
We notice: 1) The slope 1.61 reproduces reasonably well the distribution,
above
erg s-1;
2) the low luminosity flattening might be due to incompleteness
(lower flux sources are harder and harder to detect
above a diffuse emission plateau, see e.g. discussion in Kim & Fabbiano
2003);
3) the cut-off luminosity should be higher than in the Grimm et al. formula:
at least source N.10 is above the assumed cut-off;
4) with a cut-off at
erg s-1, if nothing else
should change in the functional form of the above equation, the nominal
SFR derived is
yr-1;
5) discarding source N.10 a resonable fit is obtained with the Grimm et al. formula with an
yr-1; the discrepancy
at low luminosities however is larger.
Low Mass X-ray Binaries (LMXB) are unlikely to contribute significantly at
these luminosities, since their average LF (Gilfanov 2004) is very steep
above a few 1037 erg s-1 and has a cut-off
at
erg s-1, so that
their contribution in the luminosity range probed by this observation would be
marginal relative to HMXB.
By following the different approach suggested by Grimm et al. (2003) of
relating the SFR to the total X-ray luminosity (their Eq. (21)) we derived
a slightly higher value of
yr-1.
These results compare to the values derived earlier, based on the efficiency
of star forming galaxies of producing X-rays (Wolter et al. 1999) of
yr-1.
The higher SFR might explain the difference in luminosity of the sources found
in the Cartwheel with respect e.g. to those in the Antennae, that have
an SFR of
yr-1 (Zezas et al. 2002; Fig. 11).
The question of a higher cut-off
in this system remains open.
We have presented results from a Chandra observation of the
Cartwheel. A number of isolated and very luminous
sources is present, closely related to the
region of high star formation detected at other wavelengths, accounting for
at least 75-80% of the total luminosity of
erg s-1 in the 0.5-10 keV band.
A more extended gaseous component
coincident with the ring is also detected, and also a more tentative
diffuse component that might permeate the entire system.
The extended component has a low temperature (
keV) consistent
with an origin related to starburst superwinds as in NGC 3256 (Moran et
al. 1999; Lira et al. 2002) or in NGC 253 (Pietsch et al. 2001; Strickland et al. 2000) and a luminosity
erg s-1 in the
0.5-10 keV band.
Individual sources are consistent with being pointlike, although even with
the superb Chandra resolution we only probe the kpc scale at the
Cartwheel distance.
However several considerations prompt us to suggest that we are really
detecting individual very bright sources, among the brightest
ULX seen in external galaxies.
The most luminous, N.10, has a
(0.5-2/2-10 keV)
erg s-1, to be compared with e.g. the brightest ULX
in M 82 (
erg s-1 in 0.5-10 keV at its brightest;
Matsumoto et al. 2001 and Kaaret et al. 2001),
in NGC 4559
(
erg s-1 in 0.3-10 keV band; Soria et al. 2004)
or in NGC 2276 (
erg s-1 in the 0.5-2 keV band;
Davis & Mushotzky 2004).
All these luminosities are computed assuming isotropic emission;
the true luminosity might be lower if the X-rays are beamed.
The crude spectral analysis of the brightest source
indicates that models typical of accreting binaries (like
an absorbed power law or a multicolor disk model) could describe the
data. Further, a possible flux variation strongly suggests
that the observed emission is due to a single object.
The total luminosity of the individual sources reaches at least
erg s-1 in the 0.5-10.0 keV band.
The derived Luminosity Function is consistent with a population of
HMXB in an actively star forming object and the system requires an
SFR of at least 12-20
yr-1, in agreement with
previous estimates for the Cartwheel.
We find a good match with the assumption of the "universal''
Luminosity Function for HMXB of Grimm et al. (2003); however we suggest a
cut-off luminosity
higher.
The nature of ULX is still not clear and the wide range of their characteristics points to the possibility that they are a heterogeneous class, as indicated by close scrutiny of nearby objects (e.g., Roberts et al. 2004). We are inclined to exclude SNR because of possible long term variability. Also, the spectral properties are consistent with HMXB. Since the luminosity is so high, a black hole accreting object is more likely than a neutron star. However, no fit with MD gives a low enough temperature (the best fit is kT=1.3 keV) to indicate a very high mass compact object.
Our forthcoming XMM-Newton observation will allow us to confirm the presence of the diffuse component with higher statistical significance and to better define the nature of the individual sources through spectral analysis.
Acknowledgements
We thank the referee for comments that greatly improved the presentation of this work. We have received partial financial support from the Italian Space Agency (ASI). This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. This research has made use of SAOImage DS9, developed by Smithsonian Astrophysical Observatory.
The wavelet algorithm detects 72 sources in the S3 CCD area
(8.4
8.4
), 47 of which are not positionally
related to the Cartwheel group.
The density of the sources corresponds therefore to
sources/sq. deg at the faintest detected flux of
erg cm-2 s-1 in the 0.5-10 keV band. This is entirely consistent with the density
measured in the deep Chandra surveys (e.g. Rosati et al. 2002).
None of these field sources is identified in the literature although several optical/radio associations are possible (e.g. PKS or NVSS sources, faint optical counterpart visible on the POSS, or even with magnitude from APM, etc.) We discuss here briefly the X-ray properties of the brighter ones, for which we have explored the spectral properties (i.e. those sources with more than 500 net counts).
We investigate in detail the case of CXO J003728.8-334442, the brightest
source detected in the area, which is
positionally coincident with the radio source PKS 0035-340, but so far
not identified.
The statistics of the
X-ray data allows us to collect a spectrum, which we
bin in such a way that each channel has
significance after background subtraction
(Fig. A.1). A single power law model gives an acceptable
fit with a photon index
[1.63-1.78] and
cm-2, consistent with
the classical AGN spectrum, and unabsorbed flux (0.5-10 keV band)
erg cm-2 s-1.
As is evident from the figure, even if the
is statistically
acceptable, there is an excess at
4.5 keV.
If we interpret this feature as a Fe-K
line, we can use it to measure
the still unknown redshift of the source.
We can make two different hypotheses that imply two different
spectral models: that the source is an AGN (the radio
galaxy itself); or that the X-ray source is a cluster of galaxies, of
which the radio galaxy is a tracer (see e.g., Zanichelli et al. 2001,
who however did not find a cluster around PKS0035-340).
In either case, the interpretation of the feature as the Fe-K
line gives a similar
.
An even better match with the spectral feature is obtained by fitting
both the 6.4 keV ("neutral'') and the 6.7 keV ("highly-ionized'')
Fe-line at the same redshift.
The significance of each line is only about 1
;
nevertheless
we obtain a consistent fit with z=0.425, and equivalent width of
200 eV for both lines (<500 eV at the 90% confidence level).
The derived luminosity would then be
erg s-1
(for H0=75 km s-1 Mpc-1, or
erg s-1
for H0=50 km s-1 Mpc-1), consistent both with the AGN
and the cluster hypothesis.
The formal temperature fit of kT=6.8 [5.7-8.3] keV is broadly consistent
with the expectation from the luminosity - temperature relation
of Markevitch (1998).
To our knowledge this is the first redshift derived from X-ray spectroscopy without prior measures or estimates. The statistics is however scanty and does not grant that the interpretation of the feature is correct.
We further analyzed the spectral data of CXO J003747.4-334104 and CXO J003756.3-334124.
CXO J003747.4-334104 has a faint optical counterpart visible on the DSS II
red plate.
A single power law model with Galactic
gives an acceptable
fit with a photon index
[1.66-1.85], consistent with
the canonical AGN spectrum, and flux (0.5-10 keV)
erg cm-2 s-1.
If the source is indeed an AGN with redshift in the range 0.4-2.0 (typical
of X-ray selected AGN) the derived luminosity would then be
erg s-1 (for
H0=75 km s-1 Mpc-1), consistent with the AGN hypothesis.
If the small excess at
2.4 keV were the Fe K
line, than the
redshift would be
.
CXO J003756.3-334124 does not have an optical counterpart on the DSS II plate,
and therefore the most likely counterpart is a distant X-ray cluster.
The spectrum would then need to be fitted with a thermal or
plasma emission model; however, not knowing the redshift this could be tricky.
If we use a bremsstrahlung model
the best fit temperature is kT = 9.4 [6.5-15.6] keV with a flux
(0.5-10 keV)
erg cm-2 s-1.
A mekal model with abundance = 0.3 solar, at a few selected
redshifts between 0.5 and 1.2, also gives a reasonable fit
and a similar flux.
For the same redshift range 0.5 to 1.2 (typical of X-ray selected clusters
of galaxies) the derived luminosity would then be
erg s-1 (for
H0=75 km s-1 Mpc-1), in the range of distant
cluster luminosities.
A single power law could also fit the data within the errors,
and results in a slope
,
flatter than classical AGN spectra.
For these two sources no conclusion can be drawn from X-rays alone, and a detailed optical investigation would be needed to define the counterpart.
Table A.1:
Sources detected by the wavelet algorithm in the S3 field of view.
Count rates have been converted into flux by assuming a power law with
index
and Galactic low energy absorption
cm-2. An X-ray spectral analysis is presented in the
text for sources indicated with *.