A&A 365, L128-L133 (2001)
- W. Brinkmann1
- P. T. O'Brien2 - J. N. Reeves2
- K. A. Pounds2
- M. Trifoglio3 - F. Gianotti3
Send offprint request: M. Gliozzi
Max-Planck-Institut für extraterrestrische Physik, Postfach 1312, 85741 Garching, Germany - X-ray Astronomy Group, Department of Physics and Astronomy, University of Leicester, LE1 7RH, UK - Istituto TESRE, CNR, Via Gobetti 101, 40129 Bologna, Italy
Received 20 October 2000 / Accepted 6 November 2000
We present the temporal analysis of X-ray observations of the radio-loud Narrow-Line Seyfert 1 galaxy (NLS1) PKS 0558-504 obtained during the XMM-Newton Calibration and Performance Verification (Cal/PV) phase. The long term light curve is characterized by persistent variability with a clear tendency for the X-ray continuum to harden when the count rate increases. Another strong correlation on long time scales has been found between the variability in the hard band and the total flux. On shorter time scales the most relevant result is the presence of smooth modulations, with characteristic time of 2 hours observed in each individual observation. The short term spectral variability turns out to be rather complex but can be described by a well defined pattern in the hardness ratio-count rate plane.
Key words: galaxies: active - galaxies: fundamental parameters - galaxies: nuclei - X-rays: galaxies
Author for correspondance: firstname.lastname@example.org
|00/2/7 UT11:15-13:19||7.2 ks||FF||M|
|00/2/7 UT14:28-16:02||5.4 ks||LW||M|
|00/2/10-11 UT23:26-3:23||14.0 ks||FF||T2|
|00/3/2 UT18.15-21:42||12.4 ks||FF||M|
|00/5/24 UT17:21-21:07||13.6 ks||SW||M|
|a Full Frame (FF), Large Window (LW), Small Window (SW).|
|b Medium (M), Thin 2 (T2).|
In this paper we report the results of four XMM-Newton observations of PKS 0558-504 taken with the European Photon Imaging Camera (EPIC) PN, which is the best suited instrument on board the ESA satellite for timing analysis purposes, due to its high time resolution. In Sect. 2 we describe the observations and data reduction. The long and short term X-ray variability are discussed in Sects. 3 and 4, respectively. Section 5 contains the main conclusions and a summary.
|Figure 1: EPIC PN 0.2-10 keV light curve of PKS 0558-504 for orbit 30, 32, 42 and 84|
|Open with DEXTER|
|Figure 2: Hardness ratio (hard-soft)/(hard+soft) versus total count rate. Every data point corresponds to 2000 s integration time. Soft band (0.2-1 keV), hard band (1-10 keV)|
|Open with DEXTER|
We sought evidence for spectral variability by computing the hardness ratios
as a function of the total 0.2-10 keV flux. As hardness ratio we define
the difference between the count rates in the
band and those in the
band, divided by the 0.2-10 keV count rate. These ranges where chosen
to provide a good signal to noise in both the soft and the hard light curves.
The hardness ratios versus the total count rate with data from each observation
binned in 2000 s intervals is shown in Fig. 2.
The presence of a positive
correlation was quantitatively tested by performing a linear least square fit,
which confirmed the result at 5
confidence level. On long time scales
the spectral variability is correlated with flux variations such that the
spectrum becomes harder when the count rate increases.
|Figure 3: Ratio of excess variances versus total mean count rate. Each filled circle is the average over the individual observation. The triangles represent the average values over the two exposures during orbit 30|
|Open with DEXTER|
Another interesting correlation was found between the variability in the hard band and the total count rate. The variability was studied using the excess variance (Nandra et al. 1997), which is obtained by computing the variance of the overall light curve, subtracting the variance due to measurement error and dividing by the mean squared. Care must be taken in interpreting the excess variance of different observations as it is related to the length and the sampling of the time series (e.g. Leighly 1999a). However this drawback is circumvented in our analysis by considering the ratio between the excess variance in the hard band and that in the soft range during each individual observation. Figure 3 shows a clear correlation between the hard to soft excess variance ratio and the mean count rate in the 0.2-10 band: the contribution of the hard band to the variability increases and becomes dominant as the mean count rate increases. From a direct inspection of the individual values of the excess variance in the soft and hard band, this behaviour turns out to be due more to the actual increase of the hard variability than to a depletion of the soft variability level.
Based on ROSAT and ASCA data Fiore et al. (1998) and Leighly (1999b) reported
the presence of correlations between variability and X-ray luminosity
and the steepness of the X-ray spectrum. A first
analysis seems to indicate that the XMM data of PKS 0558-504
are consistent with the previous results, however a detailed study of such
correlations is beyond the scope of this paper. A spectral study
of the XMM observations of PKS 0558-504 can be found in O'Brien et al. (2001).
|Figure 4: Individual EPIC PN 0.2-10 keV light curves of PKS 0558-504 for orbit 30, 32, 42 and 84 with time binning of 200 s|
|Open with DEXTER|
A visual examination of the individual light curves indicates that
all of them show a similar long term variability pattern with
a common rise time of about 2 hours
and, in some cases, small amplitude flares superimposed.
The best method to quantify
time variability without the problems encountered in the traditional Fourier
analysis technique in case of unevenly sampled data is a structure function
analysis (e.g. Simonetti et al. 1985; Hughes et al. 1992). The
first-order structure function is the mean deviation for data points separated
by a time lag ,
One of the
most useful features of the structure function is its ability to discern the
range of time scales that contribute to the variations in the data set: the
characteristic time scales of the variability are identified by the maxima
and slope changes in the
plane. For a stationary random process the
structure function reaches a plateau state for lags longer than the longest
correlation time scale. If a light curve contains cycles of period P, the
SF will rise to a maximum at
(Smith et al. 1993).
|a The errors on the count rate represent the dispersion around the mean.|
|b Excess variances are in 10-3 units.|
|Figure 5: Structure functions of orbits 30, 32, 42 and 84. Time lags are in seconds. The dashed lines have been drawn to mark the interesting time lags interval. Most of the error bars are smaller than the symbol size|
|Open with DEXTER|
Figure 5 shows that all the observations have either a relative or absolute maximum around 2 hours. This means that a common typical time scale (which probably reflects the similar rise time in the light curves) characterizes all observations. In addition, during orbits 30 and 84 a further characteristic time scale around 3 hours seems to be present. A first confirmation for the presence of quasi-periodicity in the temporal behaviour of PKS 0558-504 comes from a periodogram analysis, which yielded a strong signal at 2.3 hr. However, further longer observations are necessary for a firm confirmation.
In order to seek evidence for spectral variability and for its origin
on short time scales we have split each of the broad band light curves into a
hard and a soft light curve and plotted the hardness ratios versus the
time. While during orbits 32 and 84 (when the flux was at intermediate values)
no significant spectral variability
was found, orbits 30 and 42 present an interesting and somewhat puzzling
spectral and temporal behaviour shown in Figs. 6 and 7, respectively. The mean count rates and the excess
variances in the soft, hard and broad bands are summarized in Table 2.
|Figure 6: Top panel: soft (0.2-1 keV) and hard (1-10 keV) light curves during orbit 30; data are binned in 200 s intervals. The soft light curve divided by the mean (empty squares) has been multiplied by 1.5, to avoid overlapping with the hard one (filled circles). Bottom: hardness ratio versus time with 200 s binning|
|Open with DEXTER|
During the first exposure of orbit 30 the variability, quantified by the excess variance is strongly dominated by the soft photons ( , for the soft and the hard band respectively), with the soft flux steadily increasing and the hard fluctuating around a constant value. As a consequence, while the total flux increases the hardness ratio decreases in time, with a slope of hr obtained from a linear least square fit. A totally different trend is observed during the second exposure, starting just one hour later. In this case the variability is dominated by the hard flux ( , for the hard and soft band, respectively), which increases faster than the soft emission and seems to peak before it. The hardness ratio is steadily increasing with the total flux and the time (slope: hr). A similar spectral behaviour is observed during orbit 42: the variability is strongly dominated by the hard flux ( ) and the spectrum becomes harder when the total count rate increases (slope: hr).
On the basis of the five exposures available (two for orbit 30 and
one each for orbits 32, 42 and 84), the spectral variability on short
time scales of PKS 0558-504
can be summarized in the following way: when the average
broad band count rate is at a low level (the mean value is around
), the variability is dominated by the soft photons
and the hardness ratio decreases as the total flux raises. Conversely,
when the average count rate is above a certain level
for the second exposure of orbit 30 and orbit 42,
respectively), the variability is dominated by the hard band and the spectrum
becomes harder when the count rate increases. At intermediate values of
count rate (the mean is around
orbit 32 and 84), the variability in the two bands is very similar
and no significant spectral variability is detected.
|Figure 7: Top panel: soft (0.2-1 keV) and hard (1-10 keV) light curves during orbit 42; data are binned in 200 s intervals. The soft light curve divided by the mean (empty squares) has been multiplied by 1.5, to avoid overlapping with the hard one (filled circles). Bottom: hardness ratio versus time with 200 s binning|
|Open with DEXTER|
The exciting results from the temporal analysis of a few short exposures demonstrate the extraordinary capabilities of XMM, in particular in light of the long observations possible in virtue of the highly eccentric orbit.
The XMM-Newton project is supported by the Bundesministerium für Bildung und Forschung/Deutsches Zentrum für Luft- und Raumfahrt (BMBF/DLR), the Max-Planck Society and the Heidenhain-Stiftung. EPIC was developed by the EPIC Consortium led by the Principal Investigator, Dr. M. J. L. Turner. The consortium comprises the following Institutes: University of Leicester, University of Birmingham, (UK); CEA/Saclay, IAS Orsay, CESR Toulouse, (France); IAAP Tuebingen, MPE Garching, (Germany); IFC Milan, ITESRE Bologna, IAUP Palermo, (Italy). EPIC is funded by: PPARC, CEA, CNES, DLR and ASI. MG acknowledges support from the European Commission under contract number ERBFMRX-CT98-0195 (TMR network "Accretion onto black holes, compact stars and protostars").