Astron. Astrophys. 361, 629-640 (2000)

3. Analysis with homogeneous atmospheres

3.1. Hydrogen and helium lines

The simplest approach is the assumption of a homogeneous distribution of hydrogen and helium. We use line-blanketed LTE model atmospheres; the absorption of helium and metals in the extreme ultraviolet is completely taken into account. A recent description of the procedures for the calculation of theoretical atmospheres and synthetic spectra is given by Finley et al. (1997).

We started by assuming  K and and determined an approximate helium abundance from the lines at 1640 Å and 4471 Å by a visual comparison of synthetic spectra and observations. Subsequently, and were calculated from the hydrogen Balmer line profiles using a method. The result is  K and .

The He I line at 4471 Å can be reproduced best with (Fig. 1). This abundance is somewhat too low for the He II line at 1640 Å which can be fitted better with (Fig. 2). The discrepancy between these lines can be removed if is increased to about 40 000 K. Then, both lines can be fitted with , but the temperature required is out of the error range from the Balmer line analysis. The different helium abundances do not influence the determination of and from the Balmer lines.

Fig. 1. He I line compared to homogeneous hydrogen-helium models with  K, , and  % (from top to bottom)

Fig. 2. He II line compared to homogeneous hydrogen-helium models with  K, , and  % (dotted line) and 3.0 % (full line). The observed spectrum is corrected for the radial velocity of HS 0209+0832 (see Sect. 3.2)

The continuum slope and the L line can be used as additional temperature constraints. However, both the E140M and G230LB spectra suffer from uncertainties in the flux calibration. The flux in the L core reaches negative values because the scattered light in the echelle modes of STIS cannot be removed correctly with the currently available calibration software. For the G230LB grating there are problems with the throughput correction for the aperture used (). This can be seen in the overlap region of the E140M and G230LB spectra where the G230LB flux is lower than the E140M flux. Nevertheless, the model for  K can reproduce the continuum slope in the optical and near ultraviolet ( K) if we fit the model to the G430L spectrum at 5500 Å (see Fig. 3). From the shape of the L line (with the exception of the core and regions with metal lines)  K and can be derived (see Fig. 4).

Fig. 3. Optical and near ultraviolet spectrum compared to models with  K, 35 500 K, and 40 000 K. All models have and  %

Fig. 4. L line compared to a model with  K and . The observed spectrum is rebinned to a resolution of 0.1 Å for clarity.

Our temperature and gravity determinations are consistent with the results of Jordan et al. (1993,  K) and Heber et al. (1997,  K, ). The latter analysis is the mean of five individual measurements varying by about 500 K so that our single measurement is compatible although the formal errors of the mean results do not overlap.

The new helium abundance is at the high end of the observed range: The analyses of both the 1990 and 1996 observations by Heber et al. give a mean value of whereas only the 1995 observation gives (Table 2). We have re-analyzed the 1990 observations with new model atmospheres resulting in . The new observations are clearly at variance with an abundance as low as .

Table 2. Helium abundance determinations for HS 0209+0832

Heber et al. analyzed all optical helium lines together and did not derive a seperate helium abundance for He II 4686 Å. From their fits it is not excluded that He II in the 1996 observations requires also a higher abundance but the line may be too shallow for a clear determination. In the next section the problem of the helium abundance is discussed together with the abundances of heavier elements.

3.2. Metal lines

The FUV spectrum of HS 0209+0832 exhibits about 250 lines from metals. We have determined their central wavelengths using a gaussian fit routine and have tried to identify all lines with the help of the lists from the Kurucz CD-ROM No. 23 (Kurucz & Bell 1995). Two systems with different apparent Doppler velocities can be distinguished. There are several resonance lines from C II, N I, O I and Si II with a mean Doppler velocity of (). These lines are most probably interstellar because these low ionization stages cannot exist in an atmosphere as hot as  K. The other system has an apparent velocity of . Since it contains several higher ionized lines with non-zero excitation energy (e.g. C III at 1175 Å) we believe that all lines with this velocity are of photospheric origin. This system comprises lines of C III, C IV, Si III, Si IV, Al III, Ca III, Ti III, Ti IV, Ni III, Ni IV, Zn III, and Zn IV. The identified interstellar and the most prominent photospheric lines ( mÅ) are listed in Table 3 and Table 4.

Table 3. Interstellar lines

Table 4. Photospheric lines

Table 4. (continued)

The composition of the atmosphere is rather unusual. Whereas carbon and silicon are also detected in several hot DA white dwarfs (e.g. Bruhweiler & Kondo 1981, Dupree & Raymond 1982, and later observations), calcium, titanium, nickel, and zinc have been identified for the first time in a white dwarf with  K. Therefore, we have tested whether lines from different elements belong to systems with slightly different velocities. This is not the case: The apparent velocity is for the carbon and silicon lines together, for calcium, for titanium, for nickel, and for zinc. Since these velocities are very similar a common origin of all lines seems to be plausible.

Table 5 contains a list of abundances derived from comparisons with model atmospheres. Upper limits for several elements are also given. In Fig. 5, Fig. 6, and Fig. 7 we show example fits of several photospheric lines. For the two ionization stages of zinc oscillator strengths are only available for Zn III. The wavelengths of the Zn IV lines with the largest intensities observed in laboratory spectra have been taken from Sugar & Musgrove (1995) and Crooker & Dick (1968). The inferred velocity for these lines is , in agreement with the velociy of the Zn III lines ().

Fig. 5. Carbon lines: C III (left and middle) with and C IV (right) with

Fig. 6. Ca III (), Ti IV (), and Zn III () lines

Fig. 7. Si IV (), Ca III (), Ni III and Ni IV (), and Zn III () lines

Table 5. Photospheric abundances

Before the HST observations were obtained only the C IV lines at 1548.187 Å and 1550.772 Å could be identified with IUE. Jordan et al. (1993) derived - more than one order of magnitude higher than our determination from the STIS spectra. A direct comparison of the IUE and STIS observations shows that the C IV lines in the IUE spectrum are indeed much stronger. However, we cannot be sure that the variation is real due to the low signal-to-noise ratio of the IUE observation.

The carbon and titanium abundances need some discussion, because the fit to different ionization stages gives different results. From the C III lines we derive whereas C IV gives . The situation for titanium is similar: from Ti III and from Ti IV. These discrepancies can be removed if the effective temperature is increased to 40 000 K. Then, the abundances are and . This is the same problem as with He I and He II. An effective temperature of  40 000 K is clearly ruled out by the Balmer lines, L, and the optical slope.

The discrepant abundance determinations are typical non-LTE effects. This may be the best explanation although the effective temperature of HS 0209+0832 is rather low. We have included the uncertainties from the choice of the ionization stage for the errors given in Table 5 for the abundances.

The large amount of heavy elements could also affect the flux distribution and therefore the determination of due to the line blanketing and backwarming effects. We have included the EUV opacities of the metals in the model calculations. However, most of the EUV flux is already absorbed by helium so that metals are only of minor importance and, therefore, do not influence the temperature determination.

The abundance discrepancies could also be a hint that helium, carbon, and titanium are already diffused into deeper layers with somewhat higher temperature. In comparison to a homogeneous atmosphere the lines of these elements would be formed in larger depths where higher ionization stages are favored. This would support the assumption that helium and metals have been recently accreted onto HS 0209+0832 (see Sect. 5).

It is remarkable that iron lines could not be detected in the FUV spectrum, although nickel is identified. This is in contrast to all hotter DA white dwarfs, where more observations of iron lines exist than of nickel, and both elements have approximately the same abundances. However, the upper limit for the iron abundance is only a factor of two lower than the measured abundance for nickel so that a small amount of iron may be hidden. The reason for the lower abundance of iron is unclear.

About 100 lines remain unidentified (see Table 6). After identification of the most prominent interstellar and photospheric lines we have systematically calculated differences between the list of unidentified lines and lists extracted from the Kurucz CD-ROM. With this comparison we could identify several of the weaker lines. The search for additional metals is hampered by the lack of accurate oscillator strengths for atoms with high atmomic numbers. For instance, several lines could be identified with the strongest Cu IV lines from Meinders (1976) but other strong Cu IV lines are missing in the spectrum of HS 0209+0832. A clear decission would only be possible with reliable wavelengths, oscillator strengths, and calculations of model spectra.

Table 6. Unidentified lines

We have also tested for molecular lines which are occasionally found in ultraviolet spectra of reddened stars. However, the usual features (Jenkins et al. 1973, van Dishoek & Black 1984) are not visible in the spectrum of HS 0209+0832.

© European Southern Observatory (ESO) 2000

Online publication: October 2, 2000
helpdesk.link@springer.de