A&A 383, 182-187 (2002)
DOI: 10.1051/0004-6361:20011718
A. Tiengo1,2 - E. Göhler3 - R. Staubert3 - S. Mereghetti1
1 -
Istituto di Fisica Cosmica "G. Occhialini'', via Bassini 15, 20133 Milano, Italy
2 -
XMM-Newton Science Operation Center, Vilspa ESA, Apartado 50727, 28080 Madrid, Spain
3 -
Univ. Tübingen, Inst. fur Astronomie und Astrophysik, Abt. Astronomie, Waldhauserstr., 72076 Tübingen, Germany
Received 18 September 2001 / Accepted 21 November 2001
Abstract
We report on an observation of the anomalous X-ray pulsar 1E 1048.1-5937 performed
with the XMM-Newton satellite.
The phase averaged spectrum of 1E 1048.1-5937 is well described by the sum of a power law with photon index 2.9 and a
blackbody with temperature
0.6 keV, without evidence for significant absorption or emission lines.
The above spectral parameters do not vary during the phases corresponding to
the broad pulse, while the off pulse emission shows a different spectrum
characterized by a soft excess at energies below
1.5 keV.
The XMM-Newton observation and a re-analysis of archival BeppoSAX
data show that the spectral parameters and flux of 1E 1048.1-5937 did not change significantly
during observations spanning the last four years.
All the data are consistent with a 2-10 keV luminosity varying
in the range
(5-7)
1033 erg s-1 (for a
distance of 3 kpc).
Key words: stars: individual: 1E 1048.1-5937 - X-rays: stars
The pulsar 1E 1048.1-5937 is a persistent source of X-rays with flux
of 10-11erg cm-2 s-1 and a significant
modulation at a period of
6.45 s (Seward et al. 1986; Mereghetti 1995;
Corbet & Mihara 1997; Oosterbroek et al. 1998).
It belongs to the small group of Anomalous X-ray Pulsars
(AXPs, Mereghetti & Stella 1995; van Paradijs et al. 1995).
The nature of these pulsars, that show properties clearly different
from those of the
90
X-ray pulsars powered by accretion from high mass companion stars,
is currently unclear (see Mereghetti 2001, for a review).
The lack of evidence for bright optical counterparts led to models for AXPs
involving isolated neutron stars,
but their typical values of P and
yield rotational energy
losses inadequate to power the observed luminosity.
Although low mass companions cannot be completely ruled out (Mereghetti et al. 1998), it has been proposed that the AXPs consist of isolated
neutron stars accreting from residual disks (van Paradijs et al. 1995;
Ghosh et al. 1997;
Chatterjee et al. 2000).
Another class of models is based on isolated, strongly
magnetized (B>1014 G) neutron stars, or "Magnetars''
(Thompson & Duncan 1993, 1996; Heyl & Hernquist 1997),
powered by the decay of their magnetic energy.
Here we report on a new observation of 1E 1048.1-5937 performed with the XMM-Newton satellite. The new data provide an accurate measurement of the flux, phase resolved spectroscopy and a more precise localization of this pulsar. We have also reanalyzed the BeppoSAX data originally reported by Oosterbroek et al. (1998) in order to allow a more accurate comparison with the new observation.
Obs. | Pow Law | Absorption | kT | ![]() |
2-10 keV Flux | 2-10 keV Flux |
photon index | 1022 cm-2 | (keV) | absorbedd | unabsorbed (erg cm-2 s-1 ) | ||
PN |
![]() |
![]() |
![]() |
132.1/129 |
![]() |
4.3 ![]() |
MOS1 |
![]() |
![]() |
![]() |
94.7/100 |
![]() |
5.1 ![]() |
MOS2 |
![]() |
![]() |
![]() |
100.2/96 |
![]() |
4.8 ![]() |
EPICb |
![]() |
![]() |
![]() |
340.5/332 |
![]() |
4.9 ![]() |
BeppoSAXc |
![]() |
![]() |
![]() |
185.8/180 |
![]() |
7.1 ![]() |
a Errors are at the 90% c.l. for a single interesting parameter.
b EPIC = PN+MOS1+MOS2. c Reanalysis of the observation reported by Oosterbroek et al. (1988). d The errors on the flux values are dominated by the systematic uncertainties (statistical errors are of the order a few percent). |
![]() |
Figure 1: EPIC spectrum of 1E 1048.1-5937 fitted with the sum of a power law and a blackbody (see the EPIC parameters in Table 1). The upper spectrum is from the PN camera. |
Open with DEXTER |
EPIC consists of two MOS CCD detectors and a PN CCD instrument. In total, the 3 EPICs provide >2500 cm2 of collecting area at 1.5 keV. The mirror system offers an on-axis FWHM angular resolution of 4-5'' and a field of view (FOV) of 30' diameter.
During this observation the medium filter was used. The MOS cameras were operated in Small Window mode to limit source photon pile-up effects and to improve the time resolution. Due to its shorter frame integration time (73 ms), the PN could operate in the standard Full Frame mode. The net exposure time was 7135 s in each MOS camera and 4744 s in the PN.
The EPIC data reduction was performed using version v5.0 of the XMM-Newton Science Analysis System. An additional data cleaning was required for MOS2 data in order to remove some residual low energy electronic noise.
The target was clearly detected in the three EPIC cameras at coordinates
RA = 10
50
06.3
Dec = -59
53'17'' (J2000).
Based on the current performance of the XMM-Newton satellite attitude
reconstruction we conservatively estimate an uncertainty of
4'' on this position
that is consistent with previous, less accurate measurements (Mereghetti et al. 1992).
The spectral analysis with the PN camera was based
on counts extracted from a 100''100'' box
centered at the source position. This box was inside a
single CCD chip.
For the background extraction we tried several source-free
regions close to the target, without finding significant
differences in the results.
All the spectra were extracted in the nominal 0.2-10 keV range and
rebinned so that at least 20 counts were included in each energy bin and the
channels did not oversample the PN energy resolution.
We first analyzed spectra based on both single and double events.
This gives good statistics without significantly degrading
the spectral performances.
Single component models, i.e. power law, thermal bremsstrahlung and
blackbody, gave unacceptable results. A very good fit was obtained
with the "canonical'' AXP model based on the sum of a
blackbody and a power law. The best fit
parameters were a blackbody temperature kT
0.6 keV and a power law
photon index
2.7.
The two MOS spectra were based on all the counts of the small CCD windows
(100100'') used in this observation. The background
could only be extracted from the peripheral CCD chips, excluding
small regions around sources and correcting the intensity for the
vignetting effect that is significant in this region more than
5' offset from the telescope axis.
We only considered the 0.2-10 keV energy range and applied the same
rebinning as in the PN case.
For both MOS spectra a combination of blackbody and
power law gave the best fit result, with spectral parameters consistent
with those obtained with the PN camera.
We therefore proceeded to a joint spectral analysis of the three data sets, allowing only the relative normalization to vary. The results are reported in Table 1 for the blackbody plus power law model and the corresponding spectrum is shown in Fig. 1.
The observed flux in the 2-10 keV energy range
is 4.1
10-12 erg cm-2 s-1
in the two MOS cameras.
The value obtained with the PN is
about 10% smaller. Part of the discrepancy can be explained
by the fact that in our
analysis the PN "Out of Time'' events are excluded.
They are estimated to be the
6.2% of the total flux in Full Frame mode.
The remaining discrepancy is within the uncertainties in the relative
calibration of the instruments.
The best fit spectral parameters reported in Table 1
are consistent with those derived from previous ASCA
observations (Paul et al. 2000), but they show some differences
with respect to those observed with BeppoSAX by
Oosterbroek et al. (1998), who found a harder powerlaw
(
2.5) and a smaller absorption
(
0.45
1022 cm-2). In order to check if this discrepancy is really due to a spectral variation we present a reanalysis of the
BeppoSAX data in Sect. 2.4.
![]() |
Figure 2: Residuals of the PN single events spectrum. |
Open with DEXTER |
![]() |
Figure 3: EPIC PN light curves of 1E 1048.1-5937 folded at the best period for different energy ranges. |
Open with DEXTER |
To fully exploit the energy resolution provided by the EPIC CCD detectors, we performed a spectral analysis by selecting only single events. Spectra based on single events, i.e. those for which the collected charge is not split between adjacent CCD pixels, provide the best energy resolution, at the expenses of a slightly reduced efficiency.
To extract and rebin source and background spectra from the PN camera
we followed the same procedure described in the previous section. The residuals
of the best fit spectrum (power law plus blackbody) are shown in Fig. 2.
Possible features are seen at energies of 0.95 keV and in the
region between
4 and 5 keV.
Although the highest point in the low energy feature is at 3.5
from the best fit model, the fact that its width is smaller than the
instrumental resolution at this energy (FWHM
60-70 eV) suggests
that it is most likely an artifact.
The absorption at high energy has a width consistent with the instrumental resolution.
It can be fit with a Gaussian line
centered at
keV and with
eV.
None of these two possible features is confirmed by the MOS spectra
(an excess is seen in the residuals at 0.9 keV only in the MOS2 at
2
).
The timing analysis was done by correcting the time of arrival of
the counts to the Solar System barycenter and then folding them
to several trial periods around the expected value.
The best periods obtained for the PN and MOS data were
respectively
and
.
The PN folded light curve is shown
in Fig. 3 for different energy ranges.
The pulse profile is characterized by a broad, nearly sinusoidal profile
at all energies, as already noted in previous observations
with other satellites.
The level of modulation was derived by fitting a constant plus a
sinusoid to the normalized pulse profiles, yielding a pulsed fraction of
72%
below 1.5 keV and
94% at higher energies.
To search for possible spectral variations as a function of the spin
phase, we divided the data in four phase intervals
corresponding to the off-pulse (0.2-0.6), rising (0.6-0.8),
top (0.8-1), and declining (0-0.2) parts of the light curve.
The spectra of the four phase intervals
were fitted with the power law plus blackbody model,
allowing only the overall normalization to vary and keeping all the other
parameters fixed at the best fit values of Table 1.
The resulting residuals, see Fig. 4, clearly illustrate that the
average spectral parameters provide a satisfactory fit to the three
phase intervals covering the rise, peak and decay of the pulse.
On the other hand, the spectrum during the phase interval 0.2-0.6
(corresponding to the minimum in the folded light curve) presents
a significant excess below 1.5 keV.
Despite the limited counting statistics, we attempted to model the spectral shape
in the pulse minimum. A single power law provides
an acceptable fit with parameters
and
cm-2.
Keeping the absorption fixed at the phase averaged value of 1022 cm-2,
we obtain
.
![]() |
Figure 4: Results of the phase resolved spectroscopy. The figures show the ratios between the source spectra in the indicated phase intervals and the average spectrum. |
Open with DEXTER |
We reanalyzed the BeppoSAX data of the May 1997 observation
already reported by Oosterbroek et al. (1998).
The main difference with respect to the work of these
authors is our estimate of the background spectrum from a region
of the field close to the target.
Since 1E 1048.1-5937 lies at low galactic latitude, where significant diffuse X-ray emission
from the Carina nebula is present,
the use of standard background files from BeppoSAX observations of
extragalactic empty fields underestimates the background
level (particularly at energies below 3 keV).
Our extraction regions for the MECS (2-10 keV) were a circle of radius 4' for
the source and a circular corona from 280'' to 560'' for the background.
The corresponding LECS (0.5-10 keV) values were 8', 520'' and 960''.
The best fit results to the joint LECS+MECS spectra for the power law plus
blackbody model are given in Table 1. Both the absorption and the power law
photon index are different from those reported by Oosterbroek et al. (1998) and
consistent with the EPIC values. Thus we conclude that there is no evidence for
a variation in the source spectrum between the two observations.
The unabsorbed flux corresponding to the best fit BeppoSAX parameters is
7.110-12 erg cm-2 s-1 (2-10 keV),
about 45% higher than that observed with XMM-Newton.
Although the spectra of all AXPs are well described by the sum of a power law and a blackbody, there are no compelling reasons to attribute this spectral shape to the presence of two physically distinct emitting components (some suggestions in this sense were made, e.g., by Ghosh et al. 1997). In fact, as noted by Özel et al. (2001), the contribution of the blackbody component to the total flux is a strong function of the energy, while the pulsed fraction at different energies does not change significantly. In the case of two physically distinct spectral components this would require an ad hoc coupling in the pulse profiles of the blackbody and power law components. The fact that the pulsed emission from 1E 1048.1-5937 can be fit by the same spectral shape at all the phases also suggests that it originates from a single physical component, with a spectrum more complex than a pure blackbody. Theoretical AXPs spectra have been recently computed by several authors in the context of models of thermal emission from strongly magnetized neutron stars. These models predict modifications to pure blackbody-like spectra, in the form of hard power-law tails (Özel 2001; Perna et al. 2001) and/or narrow spectral features (Zane et al. 2001). Further observations with better statistics are clearly required to confirm the possible features in the spectrum of 1E 1048.1-5937 reported in Sect. 2.2.
The period value found in the XMM-Newton observation
is consistent with the overall spin-down trend at
an average
s s-1that 1E 1048.1-5937 has maintained in the last few years.
Large variations in the average spin-down rate were detected in the past.
In particular, between June 1992 and March 1994 (Mereghetti 1995; Corbet & Mihara 1997),
s s-1 was about a factor 2 higher
than that measured earlier and after 1996.
More recent observations have clearly shown that the level of rotational
irregularities in 1E 1048.1-5937 is higher than in other sources of this class
(Paul et al. 2000; Baykal et al. 2000, 2001; Kaspi et al. 1999, 2001; Gavriil & Kaspi 2001).
In principle, the study of the spin-down evolution in AXPs could
discriminate between accretion-based and magnetar models.
If AXPs are powered by accretion (either from a companion star or from a residual disk)
one would expect significant
fluctuations superimposed on the long
term spin-down trend, possibly correlated with luminosity variations, due to
irregularities in the accretion flow. On the other hand,
the period evolution of a magnetar should be much more
regular, with the possible exception of "glitches'' as observed in
young radio pulsars.
It is unclear whether the spin-down variations seen in 1E 1048.1-5937 are related to changes in its X-ray flux, particularly for observations carried out before 1995. Oosterbroek et al. (1998) compared the flux of 1E 1048.1-5937 measured with BeppoSAX to all the values obtained in the previous years with the Einstein, EXOSAT, ROSAT and ASCA satellites. Although the flux values span almost an order of magnitude, the systematic uncertainties are difficult to evaluate, since some of these measurements referred to different energy ranges. Furthermore, the data from non-imaging detectors might be contaminated by the presence of the strongly variable source Eta Carinae within the field of view.
By means of a systematic program aimed at phase-coherent timing of 1E 1048.1-5937 with the RossiXTE
satellite,
Kaspi et al. (1999) showed the
presence of significant
changes between 1996 and 2000.
Their ephemerids are shown by the solid curves in Fig. 5.
Unfortunately, due to difficulties in the background
subtraction, the RossiXTE non-imaging observations can
measure only the intensity of the pulsed flux, which was found
consistent with a constant value.
Our XMM-Newton and BeppoSAX results, compared with the ASCA ones (Paul et al. 2000),
show that all the measurements
obtained since 1994 with imaging detectors are consistent with
an unabsorbed 2-10 keV flux within the range
(5-7)
10-12 erg cm-2 s-1.
Our period measurement is inconsistent with an extrapolation of the phase-coherent solution reported by Kaspi et al. (2001). This is a further demonstration of the high level of timing noise present in 1E 1048.1-5937.
![]() |
Figure 5: Flux and spin period evolution of 1E 1048.1-5937 during the years 1996-2000. The top panel reports the 2-10 keV flux corrected for absorption as measured with BeppoSAX, ASCA and XMM-Newton. The different symbols in the bottom panel are as follows. Triangles: BeppoSAX and XMM-Newton (this work); squares: RossiXTE (Mereghetti et al. 1998; Baykal et al. 2000); diamonds: ASCA (Paul et al. 2000); solid lines: RossiXTE (Kaspi et al. 2001). |
Open with DEXTER |
The large collecting area of the XMM-Newton telescope has provided the first significant evidence for spectral variations as a function of the spin phase in 1E 1048.1-5937. Contrary to other AXPs, the spectral shape changes only in the off-pulse emission, while no significant variations are seen during the broad pulse.
No convincing evidence was found for narrow spectral features. Longer XMM-Newton
observations (also exploiting the RGS instrument) and improved calibrations
of the instruments are required to verify (or not) the structures
marginally detected in the EPIC PN camera.
The XMM-Newton observation, as well as our re-analysis of the archival BeppoSAX
data, show that the spectral parameters and flux of 1E 1048.1-5937 did not change significantly
during observations spanning the last four years.
The fact that all the observations of good quality obtained with imaging detectors and covering
energies up to 10 keV are consistent with a
2-10 keV luminosity varying in the range
(5-7)
1033 erg s-1 (for d=3 kpc), casts some doubts on previous reports of variability up to a factor 10.
It is unclear whether the large timing noise of 1E 1048.1-5937 can be ascribed to variations in the torque exerted on the neutron star by such relatively small fluctuations in the mass accretion rate.
Finally, we have provided a reduction of a factor 15 in the area of
the error box of 1E 1048.1-5937. This will undoubtedly help in the search for an optical counterpart of this
intriguing X-ray source.
Acknowledgements
E. Gö. and R. St. acknowledge the support through a grant by DLR (Verbundforschung).