4. ROSAT observations
4.1. Observational details
AG Dra was scanned during the All-Sky-Survey over a time span of 10 days. The total observation time resulting from 95 individual scans adds up to 2.0 ksec. All the ROSAT data analysis described in the following has been performed using the dedicated EXSAS package (Zimmermann et al. 1994).
Due to the scanning mode the source has been observed at all possible off-axis angles with its different widths of the point spread function. For the temporal and spectral analysis we have used an , extraction radius to ensure that no source photons are missed. No other source down to the 1 level is within this area. Each photon event has been corrected for its corresponding effective area. The background was determined from an equivalent area of sky located circle , south in ecliptic latitude from AG Dra.
4.1.2. Pointed observations
Several dedicated pointings on AG Dra have been performed in 1992 and 1993 with the ROSAT PSPC (Table 5 gives a complete log of the observations). All pointings were performed with the target on-axis. During the last ROSAT observation of AG Dra with the PSPC in the focal plane (already as a TOO) the Boron filter was erroneously left in front of the PSPC after a scheduled calibration observation.
Table 5. Summary of ROSAT observations on AG Dra. Given are for each pointing the observation ID (column 1), the date of the observation (2), the detector (P=PSPC, H=HRI) without or with Boron (B) filter (3), the nominal exposure time (4), and the total number of counts (5).
For each pointed observation with the PSPC in the focal plane, X-ray photons have been extracted within , of the centroid position. This relatively large size of the extraction circle was chosen because the very soft photons (below channel 20) have a much larger spread in their measured detector coordinates. As usual, the background was determined from a ring well outside the source (there are no other X-ray sources within , of AG Dra). Before subtraction, the background photons were normalized to the same area as the source extraction area.
When AG Dra was reported to go into outburst (Granslo et al. 1994) we immediately proposed for a target of opportunity observation (TOO) with ROSAT. AG Dra was scheduled to be observed during the last week of regular PSPC observations on July 7, 1994, but due to star tracker problems no photons were collected. For all the later ROSAT observations only the HRI could be used after the PSPC gas has been almost completely exhausted. Consequently, no spectral information is available for these observations. The first HRI observation took place on August 28, 1994, about 4 weeks after the optical maximum. All the following HRI observations and the single PSPC observation with the Boron filter (described above) were performed as TOO to determine the evolution of the X-ray emission after the first outburst. With the knowledge of the results of the first outburst the frequency of observations was increased for the second optical outburst.
Source photons of the HRI observations have been extracted within , and were background and vignetting corrected in the standard manner using EXSAS tasks. In order to compare the HRI intensities of AG Dra with those measured with the PSPC we use a PSPC/HRI countrate ratio for AG Dra with its supersoft X-ray spectrum of 7.8 as described in Greiner et al. (1996).
4.2. The X-ray position of AG Dra
We derive a best-fit X-ray position from the on-axis HRI pointing 180073 of R.A. (2000.0) = 40 8, Decl. (2000.0) = , with an error of . This position is only , off the optical position of AG Dra (R.A. (2000.0) = 40 94, Decl. (2000.0) = 09 7).
4.3. The X-ray lightcurve of AG Dra
The mean ROSAT PSPC countrate of AG Dra during the all-sky survey was determined (as described in paragraph 4.1.1) to (0.99 0.15) cts/sec. Similar countrates were detected in several PSPC pointings during the quiescent time interval 1991-1993.
The X-ray light curve of AG Dra as deduced from the All-Sky-Survey data taken in 1990, and 11 ROSAT PSPC pointings (mean countrate over each pointing) as well as 7 HRI pointings taken between 1991 and 1996 is shown in Fig. 3. The countrates of the HRI pointings have been converted with a factor of 7.8 (see paragraph 4.1.2) and are also included in Fig. 3.
This 5 yrs X-ray light curve displays several features:
4.4. The X-ray spectrum of AG Dra in quiescence
For spectral fitting of the all-sky-survey data the photons in the amplitude channels 11-240 (though there are almost no photons above channel 50) were binned with a constant signal/noise ratio of 9 . The fit of a blackbody model with all parameters left free results in an effective temperature of = 11 eV (see Table 6).
Table 6. Summary of blackbody model fits to the ROSAT PSPC observations of AG Dra during quiescence. Fluxes are in photons/cm2 /s and temperatures kT in eV. The absorbing column is in units of 1020 cm-2 for the three parameter fit and was fixed at the galactic value of 3.15 1020 cm-2 for the two parameter fit.
Since the number of counts detected during the individual PSPC pointings allows high signal-to-noise spectra, we investigated the possibility of X-ray spectral changes with time. First, we kept the absorbing column fixed at its galactic value and determined the temperature being the only fit parameter. We find no systematic trend of a temperature decrease (lower panel of Fig. 3). Second, we kept the temperature fixed (at 15 eV in the first run and at the best fit value of the two parameter fit in the second run) and checked for changes in , again finding no correlation. Thus, no variations of the X-ray spectrum could be found along the orbit. The rather small degree of observed variation of temperature and (Table 6) over more than three years during quiescence (including the ROSAT All-Sky-Survey data) are not regarded to be significant due to the correlation of these quantities (Fig. 4).
The independent estimate of the absorbing column towards AG Dra from the X-ray spectral fitting indicates that the detected AG Dra emission experiences the full galactic absorption. While fits with as free parameter (see Table 6) systematically give values slightly higher than the galactic value (which might led to speculations of intrinsic absorption), we assess the difference to be not significant due to the strong interrelation of the fit parameters (see lower right panel of Fig. 4) given the energy resolution of the PSPC and the softness of the X-ray spectrum. We will therefore use the galactic value (3.15 1020 cm-2 according to Dickey and Lockman 1990) in the following discussion.
With fixed at its galactic value the mean temperature during quiescence is about 14-15 eV, corresponding to 160000-175000 K. These best fit temperatures are plotted in a separate panel below the X-ray intensity (Fig. 3). The small variations in temperature during the quiescent phase are consistent with a constant temperature of the hot component of AG Dra.
4.5. The X-ray spectrum of AG Dra in outburst
As noted already earlier (e.g. Friedjung 1988), the observed fading of the X-ray emission during the optical outbursts of AG Dra can be caused either by a temperature decrease of the hot component or an increased absorbing layer between the X-ray source and the observer. In order to evaluate the effect of these possibilities, we have performed model calculations using the response of the ROSAT HRI. In a first step, we assume a 15 eV blackbody model and determine the increase of the absorbing column density necessary to reduce the ROSAT HRI countrate by a factor of hundred. The result is a factor of three increase. In a second step we start from the two parameter best fit and determine the temperature decrease which is necessary to reduce the ROSAT HRI countrate at a constant absorbing column (3.15 1020 cm-2). We find that the temperature of the hot component has to decrease from 15 to 10 eV, or correspondingly from 175000 K to 115000 K.
The only ROSAT PSPC observation (i.e. with spectral resolution) during optical outburst is the one with Boron filter. The three parameter fit as well as the two parameter fit give a consistently lower temperature. But since the Boron filter cuts away the high-end of the Wien tail of the blackbody, and we have only 19 photons to apply our model to, we do not regard this single measurement as evidence for a temperature decrease during the optical outburst.
What seems to be excluded, however, is any enhanced absorbing column during the Boron filter observation. The best fit absorbing column of the three parameter fit is 4.4 1020 cm-2, consistent with the best fit absorbing column during quiescence. Since the low energy part of the spectrum in the PSPC is not affected by the Boron filter except a general reduction in efficiency by roughly a factor of 5, any increase of the absorbing column would still be easily detectable. For instance, an increase of the absorbing column by a factor of two (to 6.3 1020 cm-2) would absorb all photons below 0.2 keV and would drop the countrate by a factor of 50 contrary to what is observed.
It is interesting to note that the decrease of the X-ray flux is similarly strong in both, the 1994 and 1995 outbursts, while the optical amplitude of the secondary outburst in 1995 was considerably smaller than the first outburst. We note in passing that the intensity of the HeII and NV lines also showed a comparable large increase in the main 1981/1982 outburst and the minor outburst in 1985/1986 while the optical and the short wavelength UV continuum amplitudes again were smaller in the latter outburst (Mikolajewska et al. 1995). Since the short wavelength UV continuum is the Rayleigh-Jeans tail of the hot (blackbody) component, this behaviour suggests that a temperature decrease is the cause of the reduced X-ray intensity during outburst rather than increased absorption with the temperature decrease being smaller in the secondary outbursts.
© European Southern Observatory (ESO) 1997
Online publication: June 5, 1998