3.1. The X-ray light curves
In Fig. 1 we plot the LECS and MECS light curves binned over 360 sec. Two major flares were detected, but the first one on Nov 4 was not observed by the LECS because this instrument was not operative at that time. In the 1.7-10 keV band of the MECS, the Nov 04 flare was characterized by a rather flat maximum reached at 06:04 UT and maintained for about 1.3 h. The total energy released in this band was 6.51034 erg. The second major flare was detected on Nov 11 with the flare maximum at 18:38 UT, a 1/e decay time of 6.4 h, and a total energy of 8.91034 erg in the MECS and 1.61035 erg in the LECS bandpasses, respectively. Apart from these two flares, the light curves in the range 0.1-10 keV show several minor flare-like episodes. For instance, the MECS light curve obtained during the first segment of observations, binned over 3600 sec and folded with the orbital period (see Fig. 2) shows at least five low-amplitude flares which are superimposed on the sinusoidal curve which is the best fit to the outside of eclipse and outside of flare data points drawn with filled symbols. The five low amplitude flares show increases at light maximum between 20 and 30%, values much larger than the background fluctuations as seen in Fig. 2. The sinusoidal curve can be interpreted as representative of the system X-ray quiescent emission. It indicates the presence of a well defined rotational modulation with semiamplitude 30%. The maximum emission occured at phase 0.2, indicating the presence of a bright X-ray emitting structure on the leading hemisphere of the G star or on the trailing hemisphere of the K star. As show in Fig. 2, no return to the quiescent level has followed the major flare on Nov 4, before the onset of the successive uncompletely observed flare at 15:46 UT. The data acquired during quiescence in the second observing run are not enough to see if a rotational modulated variability is still present.
Our Beppo-SAX observations covered one primary eclipse, and three secondary eclipses. During the primary eclipse on Nov 3, the LECS was not operative. The MECS count rate (see Fig. 2) shows a 25% reduction centered at the primary eclipse (orbital phase 1.0) with the dip confined to totality. We have not observed any count rate dips at secondary eclipses, but this could be due to the flaring state of the source during all the annular eclipses we observed (Nov 2, Nov 4, and Nov 12).
In Fig. 1 we also plot the PDS light curve binned over 3600 sec. About a half of the PDS data are upper limits. The PDS count rates oscillate around 0 with a standard deviation of 0.15 and 0.13 cts s-1 in the two observing runs, respectively. Several sporadic increases of the hard X-ray flux were observed exceeding , but not apparently correlated with the soft X-ray flux behaviour. However, the two following cases could be exceptions: i) on Nov 4, at the end of the 1st observing run, the increase in the PDS count rate exceedes and appears to be coincident with the soft X-ray flare observed by the MECS at 15:40 UT (see also Fig. 2 at phase 1.6); ii) on Nov 11, during the major soft X-ray flare oberved by the MECS and LECS, the hard X-ray flux appears to be decaying exponentially from a greater than initial value.
3.2. The X-ray spectra
The spectra from the LECS and from the MECS detectors were analysed using an optically-thin plasma emission model with two discrete temperature components and variable global metal abundances (the abundances of individual elements have the same ratios as in the solar case). In fact, the limited spectral resolution of LECS and MECS does not permit a reliable determination of coronal abundances of individual elements. The MEKAL code under XSPEC 10.0, based on the model of Mewe and Kaastra with Fe L calculations by Liedahl (Mewe et al. 1995), was used for this analysis. The assumed solar abundances are those of Anders & Grevesse (1989). The effect of interstellar absorption was taken into account using the model of Morrison & McCammon (1983) as in the XSPEC-WABS code. Taking into account a neutral hydrogen volume density of 0.07 cm-3 (Paresce 1984) and the distance d=42 pc measured by H IPPARCOS (Perryman et al. 1997), the neutral hydrogen column density would amount to NH1019 cm-2. However, this is an approximate value because the H I extintion is inhomogeneous. Walter (1996) and Griffiths & Jordan (1998) estimated a much lower value of NH21018 cm-2, using the ratio between the Fe XVI lines at 335 and 361 Å as observed by EUVE. Linsky et al. (1998) estimated in the direction of AR Lac NH=(5.92.5)1018 cm-2 by interpolating measurements of the H I extintion in directions closeby to AR Lac in the local interstellar cloud.
In our spectral fits we first fixed NH to the value derived by Linsky et al. (1998), and then allowed it to vary to derive the best-fit NH value. Freezing NH to 5.91018 cm-2, and contemporarly fitting together the total LECS and MECS spectra (T) we derived the best-fit model spectrum and the fit residuals shown in Fig. 3. The fit has a reduced of 1.06 with 358 degree of freedom (d.o.f). The best fit temperatures are 2.050.07 keV and 0.770.03 keV, with a ratio between the two emission measures (cool/hot) of 0.58 (EMcool=1.821053 cm-3 and EMhot=3.131053 cm-3). The global metal abundance is 0.660.06. The reduced of this best-fit is acceptable, but the fit is poor below 0.25 keV and in the region of the Fe complex at 6.7 keV. The best-fit rescaling factor is 0.7 for the LECS to the MECS. The best-fit parameters are given in Table 2.
Table 2. Results of a two-components MEKAL plasma model fit to the spectral data. The neutral hydrogen interstellar column density NH was fixed to to 5.91018 cm-2 (Linsky et al. 1998).
We then performed the analysis with NH as a free parameter, and obtained the best-fit model spectrum and the fit residuals shown in Fig. 4 (see Table 3). The fit has a better reduced of 0.95 (with 357 d.o.f). The resulting best fit temperatures do not change appreciably, but the ratio between the cool/hot emission measures increases to 0.95 and the global metal abundance decreases to 0.430.02. The NH value resulted even higher than expected, (6.11.5)1019 cm-2.
Table 3. Results of a two-components MEKAL plasma model fit to the total spectral data (T). The neutral hydrogen interstellar column density NH was left free to vary in the fit procedure.
It is well known that high values of interstellar absorption can mimic low metallicity (Tagliaferri et al. 1997). Moreover, lower best-fit metallicities are known to be correlated with higher emission measure ratios between the cool and the hot components (Favata et al. 1997a). A low metal abundance (0.33) also results from analyzing MECS data using the MEKAL model without interstellar absorption, since the interstellar absorption is negligible at energies above 1.8 keV. On the other hand, since the metal abundances are measured by the relative intensity between emission lines and continuum, the abundances inferred by analysing only the MECS data could be in error because of the lack of line-free continuum flux in the energy range covered by the MECS. The presence of a strong and systematic dependence of the best-fit metallicity on the spectral region being fitted was also pointed out by Favata et al. (1997a), who applied two component MEKAL plasma models on both simulated and real spectra of Beppo-SAX LECS and ASCA SIS instruments.
The spectral analysis of the two separate segments of observations (Nov 2-4, and Nov 11-12) was done with NH constrained to 5.91018 cm-2. We found that the two-component temperatures did not significantly change between the two time intervals (see Table 2). On the contrary, the best-fit global metal abundance seems to be slightly variable. We note, however, that the higher global metal abundance resulting from the 2nd observation could be linked to the low integration time for the Fe K line complex at 6.7 keV and the resulting poorer statistics in the 2nd run MECS spectrum. We therefore conclude that there is no evidence in these data for temporal variability of the coronal metal abundance of AR Lac.
It is worth noting that we were not able to reproduce very well the Fe K line complex region around 6.7 keV. The fit of the Fe K feature is bad both when we model the total 0.1-10 keV spectrum, and when we model the MECS spectrum only. Conversely, the spectral line seems to be narrower than expected from the model. We checked for possible calibration problems or other effects linked to the detector by discussing these questions with people on the MECS team at the IFCAI-CNR (Palermo), and we had to conclude that the observed feature should be real. A slight deficiency of the model near the Fe-K complex above 6 keV was also reported by Kaastra et al. (1996) who analyzed AR Lac ASCA spectra with the SPEX plasma model.
In order to improve our fit in the 6.7 keV region, we also let the Fe abundance to vary separately from the other elements, but we immediately faced the problem that, given the available spectral resolution, the individual element abundances cannot be constrained with sufficient confidence.
Since the Fe abundance was observed to be variable with the state, quiescent or flaring, of the source (see Stern et al. 1992, White, Pallavicini, & Lim 1995, Mewe et al. 1997 and Güdel et al. 1999), we have searched for a better fit of the Fe K line complex also by allowing the global metal abundance to be different for the two plasma components. Assuming NH=5.91018 cm-2, we found a solution with a reduced of 0.97 (357 d.o.f.) that implies almost the same temperatures reported before, an abundance of 0.40 solar for the hot component, and a super-solar abundance of 1.94 solar for the cool component, with an emission measure ratio of 0.17 between the cool and the hot components. However, the cool component abundance is not very well constrained; it can be between 1.17-6.71 (90% confidence level). Moreover, the Fe K line complex is not better fit by this model. Therefore, we cast doubt on the physical significance of this solution.
We also attempted to fit the data with a three-temperature MEKAL model, but we did not find a significant improvement by adding a third component, either of intermediate or of higher temperature (the instruments are sensitive for temperatures in the range 0.1-10 keV).
© European Southern Observatory (ESO) 1999
Online publication: June 17, 1999