Astron. Astrophys. 355, 227-235 (2000)

5. Flare line fluxes

5.1. SWP line fluxes (1200-2000 Å)

As we have concluded in Sect. 4.1, there is evidence for at least 2 separate flares in the spectra of II Peg. In Fig. 12 we show the individual spectra of these flares compared to the mean "quiescent" spectrum derived from the non-flare spectra. Apart from the dramatic increases in the line fluxes there is also evidence for a continuum, or, perhaps, a large number of unresolved weak lines between the most conspicuous ones.

Flares are generally expected to occupy a relatively small fraction of the stellar surface area. Therefore, to a good approximation, we can treat the quiescent atmosphere as unaffected by the flare and forming an additive background to the line flux from the flares themselves. We have therefore subtracted the mean quiescent fluxes in each detected line from those fluxes measured during the flares. The resulting net fluxes will be treated as a signal purely from the flares themselves.

As we have pointed out above, we see evidence of a continuous flux distribution underlying those individual lines. It is possible that there is a contribution to this "continuum" from unresolved individually weak lines. We have unsucessfully searched for evidence of the Si I continuum breaks at 1524Å and 1682Å as suggested by Phillips et al. (1992). In the absence of firm evidence either way we will treat the distribution of radiation as a true continuum extending over the entire SWP wavelength range as has been suggested by other authors (see Byrne 1996 and references therein). We have also measured the 1595 Å continuum power as defined in Phillips et al. (1992) and confirmed the relationship between this `continuum' and the C IV line power. Our spectra yield C IV line powers of 28.06 and 27.86 for SWP45531 and SWP45532 respectively and `continuum' powers of 28.08 and 27.89, all measurements in logarithmic units.

Fig. 9. The SWP 45531 flare spectra compared to the mean "quiescent" spectrum of II Peg. Individual spectral lines are identified on the top spectrum.

Fig. 10. The SWP45532 flare spectra compared to the mean "quiescent" spectrum of II Peg. Individual spectral lines are identified on the top spectrum.

5.2. LWP line fluxes (2000-3000Å)

Only one LWP HIRES spectrum was taken during a flare namely LWP23854. This was obtained in between the two SWP flare spectra on the night of 5 September (SWP45531 and SWP45532). In this spectrum all the observed emission lines are dramatically enhanced in flux and considerably broadened (Fig. 11).

Fig. 11. A sample "quiescent" IUE LWP HIRES spectrum of II Peg (LWP23864) in the neighbourhood of the MgII h&k lines. Overplotted is the large flare of 5 September 1992 (LWP23854). The vertical lines marks the position of the interstellar lines (IS).

Assuming, as we did for the SWP spectra in Sect. 5.1, that the flare was confined to a limited area of the disk and that the global flux was largely unaffected by the flare, we have subtracted LWP23864 (the post-flare, quiescent line profile) from the flare line profile. This is shown in Fig. 12

Fig. 12. The LWP line profiles for the flare spectrum (LWP23854) after subtraction of LWP23864 ("quiescent" spectrum).

Analysing a less energetic flare, Doyle et al. (1989b) detected in the excess emission from the flare two absorption features on both sides of the Mg II k line maximum. The origin of the red one is clear (it is the intestellar absorption line also found in our data) while the blue one could be attributed, according to the authors, to an erupting filament moving at a speed of 25 km s^-1 relative to the bulk of the flare. In our data we do not find such clear evidence for overlying cold material absorbing the emission from the flare, but we still detect traces of these kind of spectral features. In particular, the ones marked in Fig. 12 show similar velocities in both lines (roughly 45 km s^-1) measured relative to the maximum emission in the flare spectrum.

5.3. Differential emission measure

Following Jordan et al. (1987) we have constructed emission measure curves for those spectra with measurable Si III ] 1892 Å line fluxes, namely, the mean quiescent spectrum and the flare spectrum SWP45532. This intersystem line was used to estimate the electron density as described in Sect. 5.4. Solar abundances were adopted given the lack of estimates of transition region abundances for this RS CVn system. Only recently, Mewe et al. (1997) using ASCA spectra of II Pegasi found an underabundance of silicon of a factor 5 relative to the cosmic abundances of Anders & Grevesse (1989) used in this paper (they found other non-solar abundances of elements not relevant to our emission measure analysis). As shown below, the emission measure curve obtained with solar abundances is relatively smooth and such a change in the silicon abundance would significantly increase the scatter in the data. In order to use the new abundances in the emission measure computation we would need carbon and nitrogen abundances obtained consistently (i.e. fitting an ASCA spectrum) something which was not possible with the dataset analysed in Mewe et al. (1997).

In the emission measure formalism we have also used the equilibrium ion populations of Arnaud & Rothenflug (1975), and the collisional strengths included in Table 4. The results are shown in Figs. 13 and 14.

Fig. 13. Logarithmic emission measure diagram for the mean quiescent spectrum of II Pegasi. Solid lines represent the emission measure loci computed (from left to right) with the Si II , C II , Si IV , C IV and N V line fluxes. Dotted lines correspond to the emission measure loci of the Si III ] intersystem line assuming four different electron densities. The dashed line is the fit to the emission measure curve.

Fig. 14. As Fig. 13 but for the flare spectrum SWP45532.

Table 3. Line fluxes at Earth for some of the strongest lines in the SW ultra-violet spectrum of II Peg in 1992 during flares after subtraction of the mean "quiecent" fluxes in the same lines. ^* marks lines with low signal to noise ratio.

Table 4. Atomic parameters used in the computation of the emission measure distribution.
Notes:
1) Mendoza (1981), 2) Hayes & Nussbaumer (1984), 3) Lennon et al. (1975), 4) Baluja et al. (1980,1981), 5) Dufton & Kingston (1985), 6) Cochrane & McWhirter (1983)

These results can be compared with previous determinations of this emission measure distribution, in particular with the latest reexamination of the IUE observations of this RS CVn system. In the work by Griffiths & Jordan (1998) based on observations of II Pegasi taken in October 1981, a slightly different definition of the emission measure (apparent volume emission measure) is used which merely introduces a new geometric scaling factor. Once this is taken into account, the main differences affect the absolute value of the emission measure and the shape of the temperature dependence. The absolute value of the logarithmic emission measure for the data obtained in 1992 is lower by a factor at due to the considerably lower value of the C IV line flux. Given that our transition region line fluxes are amongst the lowest ever measured, this means that similar conclusions could be drawn by comparison with any other dataset. The ratio of the 1981 line fluxes to the 1992 measurements increases with increasing line formation temperature and, thus, the shape of the emission measure distribution changes. Above the Si IV formation temperature, this implies a steeper decrease of the emission measure distribution for our data. Although with a different definition of the emission measure and different atomic parameters, Byrne et al. (1987) constructed another emission measure distribution based on the same line fluxes which they fitted with a power law. Another data set obtained in 1983 was analysed by Doyle et al. (1989b) in order to obtain emission measure curves for II Peg. Using the same methodology as in Byrne et al. (1987) they obtained emission measure curves with a similar behaviour in the temperature range for the quiescent state while the flaring state showed a flattening of the distribution between the C IV and N V formation temperatures.

5.4. Electron density

The density diagnostics available for the set of emission lines present in the SWP spectra were discussed in detail in Byrne et al. (1987). Following this discussion we have computed electron densities for the mean quiescent spectrum and the two flare spectra with good signal-to-noise in the lines involved in the analysis (namely, SWP 45531 and SWP45532) using the calibration of the Si IV (1394/1403) /C III ](1909) line ratio against the quiet solar limb observations by Doschek et al. (1976). The Si IV total emission in the resonance doublet was computed from the 1394 component assuming the optically thin line ratio (2:1)
due to the blending of the weaker component at 1403 with nearby O IV and S IV lines.

Under the assumptions underlying this calibration (solar abundances of carbon and silicon and formation temperature roughly equal for both lines) we have obtained a value of =15.5 for the quiescent state of II Peg and 15.7 and 16.0 for SWP 45531 and SWP 45532 respectively. These correspond to =10.7 in quiescence and =10.9 and =11.2 for the two flaring spectra.

As we mentioned in the previous section, the Si III ] 1892 Å line allows us to estimate the electron density by an indirect method. It consists in assuming different values of the electron density in the emission measure formalism and choosing the one that provides a smooth emission measure curve. Using this method we have obtained electron densities roughly equal to 11.6 for the quiescent state and 11.9 for the flare spectrum.

It is important to point out that the quiescent state electron density is in complete agreement with the density estimate obtained by Griffiths & Jordan (1998) despite the large difference in the radiative losses of II Peg in both epochs. If we gauge the radiative losses of the transition region of this system by adding up the Si IV , C IV and N V line fluxes, we find that in 1981 this amounts to erg cm^-2 s^-1, roughly twice the value for quiescence in 1992. In fact, even within the data of Rodono et al. (1987) we can find hints that the electron density estimated in this way is independent of the level of activity as measured by the radiative losses of the transition region of II Peg. In the series of papers following Rodono et al. (1987) up to the work by Griffiths & Jordan (1998) a distinction is made between a quiet hemisphere (used for comparison above) and an active hemisphere on II Peg based on transition region fluxes and spot modelling (see Byrne et al. 1987). The ratio between the radiative losses of the transition region (measured in the same way as above) of the active hemisphere relative to the quiet hemisphere is roughly 5, but the ratio of the C II 1335 Å flux to the Si III ] 1892 Å flux (that measures the relative position of the Si III ] loci relative to the mean emission measure curve) is equal for both hemispheres.

The electron density estimate based on the Si III ] loci is one order of magnitude higher than the one obtained with the Si IV ( 1394/1403)/C III ](1909) line ratio. This can be partly due to the fact, already pointed out by Byrne et al. (1987), that a significant fraction of the Si III ] is formed at temperatures below . But under the assumption of constant pressure in the transition region, this is still insufficient to account for the difference of one order of magnitude in the estimates. As a matter of fact, even if we assume a line formation temperature of for the Si III ] line and for Si IV ( 1394/1403) and C III ](1909) we find a change in density of only 0.5 dex.

This kind of discrepancy has already been pointed out for other stars (see Linsky et al. 1995) in the sense that density sensitive line ratios yield values sistematically lower than emission measure based densities. Furthermore, densities obtained via density sensitive line ratios seem to cluster in a narrow band around independent of activity level, spectral type or luminosity class (see Ayres et al. 1998; Wood et al. 1996 or Wood et al. 1997). These puzzling results suggest that the density stratification in the transition region is still a matter of debate. It has been proposed that the surface of active stars is covered by at least two types of regions with substantially different densities but similar emission measures. Although the present data do not allow further speculation, the fact that in II Pegasi in particular, emission measure based densities are equal for two hemispheres at a time where one was by large more active than the other, seems to support this hypothesis.

© European Southern Observatory (ESO) 2000

Online publication: March 17, 2000
helpdesk.link@springer.de