next previous
Up: Models of infrared spectra 1997


Subsections

3 Results

3.1 Principal spectral features

In Fig. 1 the principal features formed by the molecular species C2, CO, and CN are displayed as separate spectra. Atomic features are also shown, as these are also present in Sakurai's Object (see Eyres et al. 1998; Geballe et al. 2002). As in the optical spectrum (Pavlenko et al. 2000), absorption of only a few molecular species accounts for the main features in the IR spectrum. Only the most abundant isotopic species of each molecule is shown. Of the less abundant isotopic species, only bands of 13CO have been detected in the infrared (Eyres et al. 1998).

3.2 Dependences on T $\mathsfsl{_{eff}}$, log g, V $\mathsfsl{_t}$ and log N(H)

The model spectra of Sakurai's Object display a strong dependence on $T_{\rm eff}$  (Fig. 2).[*] In general, the dependence of the IR SED on $T_{\rm eff}$ is determined mainly by the variations of the molecular densities with temperature. The band strengths of CN, CO and C2 all increase as $T_{\rm eff}$ decreases. Changes in the continuum fluxes are much smaller. Similar effects are seen in model optical spectra (Pavlenko & Yakovina 2000). However, there the molecular bands are numerous, whereas in the infrared only the few strongest vibration-rotation bands of CN, C2, and CO are prominent.

  \begin{figure}
\par\includegraphics[width=8.8cm,clip]{MS2440f1.eps}
\end{figure} Figure 1: Model spectra of species that produced the strongest absoption features in the 1.0-2.5 $\mu $m spectrum of Sakurai's Object spectra during 1997, computed for $T_{\rm eff}$ /log g = 5500/0.0 model atmosphere with Asplund et al. (1999) abundances for October 1996. The model spectrum due to atomic species alone, labelled VALD (see text) is also shown. Spectra are artificially shifted on the y-axis.


  \begin{figure}
\par\includegraphics[width=8.8cm,clip]{MS2440f2.eps}
\end{figure} Figure 2: Dependence of the model IR spectrum on $T_{\rm eff}$. The model spectra use Asplund et al. (1999) abundances for October 1996. The observed spectrum of Sakurai's Object on July 13, 1997 is shifted on the y-axis.


  \begin{figure}
\par\includegraphics[width=8.8cm,clip]{MS2440f3.eps}
\end{figure} Figure 3: Dependence of the model spectrum on log g.

As can be seen in Fig. 3, the dependence of the spectrum on log g is generally rather weak. However, there are differences in the responses of different spectral regions. The strong molecular bands show rather weak dependence on log g, whereas the fluxes at 1.25-1.35, 1.60-1.75, 1.9-2.2 microns show more noticeable changes.

Previous abundance analyses of the spectra of Sakurai's Object and related R CrB stars indicate microturbulent velocities $V_{\rm t}$ in the range 5-8 km s-1 (cf. Asplund et al. 2000). The value of $V_{\rm t}$ affects the spectral distribution, as is shown in Fig. 4. The effect of $V_{\rm t}$ on the IR spectra of Sakurai's Object is larger at the heads of molecular bands than elsewhere, because the heads are formed by closely packed molecular lines whose overall absorption is sensitive to $V_{\rm t}$.

The main sources of line opacity in the model atmospheres approximating Sakurai's Object in 1997 are molecular (Pavlenko et al. 2000). Thus it is not surprising that the optical spectra which match Sakurai's Object respond weakly to changes in the hydrogen abundance. This is in contrast to the behavior of models corresponding to the star a year earlier (Asplund et al. 1997). Similarly, the model IR spectra of Sakurai's Object for $T_{\rm eff}$ = 5000-6000 K depend weakly on log N(H) (Fig. 5). The magnitude of the change in the spectrum when log N(H) is changed from -2.42 (the Asplund et al. 1999 value for October 1996) to -0.97 (i.e, a change of 1.5 dex) is comparable (in a qualitative sense) to lowering log g from 1 to 0 (Fig. 3).

3.3 Changes between 1997 April 21 and July 13

Fits to the spectra of Sakurai's Object on April 21 and July 13 are shown in Figs. 6 and 7. The long wavelength portion of the H band is of special interest for the "carbon problem'', because the strongest absorption bands of the C2 molecule, the Ballick-Ramsey bands, occur just longward of 1.768 $\mu $m. In the comparatively hot atmosphere of Sakurai's Object log N(C) > log N(O) (Asplund et al. 1997, 1999) and the abundance of C2 depends mainly on the elemental abundance of carbon. Therefore, these bands may provide the most accurate determination of log N(C). The fits imply that the carbon abundance is in the range log $N\rm (C) = -1.3 \pm 0.2$. The most likely value is 0.3 dex higher than that found by Asplund et al. (1999). The accuracy of the determination of log N(C) is limited mainly by the quality of the molecular line list.

The effective temperatures that best fit the 1.0-2.0 $\mu $m spectra in 1997 April and July are $5500\pm 200$ K and $5250 \pm 200$ K, respectively, indicating that the cooling evidenced by the dramatic spectral changes seen between 1996 and 1997 (e.g., Geballe et al. 2002) continued in 1997. Our estimated uncertainties in the above temperatures are rather large, despite the comparatively good fits to the observed spectra, because of questions concerning abundances, non-sphericity effects, and dynamical phenomena, and because of contamination of the spectra by dust emission (see below).

  \begin{figure}
\par\includegraphics[width=8.8cm,clip]{MS2440f4.eps}
\end{figure} Figure 4: Dependence of the model spectrum on $V_{\rm t}$.


  \begin{figure}
\par\includegraphics[width=8.8cm,clip]{MS2440f5.eps}
\end{figure} Figure 5: Dependence of the model spectrum of Sakurai's Object on log N(H).

3.4 Hot dust

Emission by dust is evident in the 1997 spectra by the mismatch between the synthetic and observed spectra longward of 2.0 $\mu $m in Figs. 6 and 7. The difference between the observed and synthetic spectra is greater in the July spectrum, attesting to an increase in the amount of dust. The thermal emission from the dusty envelope overlaps the region of first overtone bands of 12CO and 13CO at $\lambda~>~$2.3 $\mu $m. Usually these bands are used for determination of carbon abundances and isotopic ratios (cf. Lazarro et al. 1991). The reduced equivalent widths of the CO bands in July 1997 cannot be reasonably attributed to a large decrease in the oxygen abundance, because (1) this is unlikely to have occurred in three months and because the continuum shortward of the CO bands also shows an excess. We note that in fitting spectra, the most frequent situation is that the computed spectra have excess flux due to the deficit of known or hypothetical opacities. To fit the observed spectra, opacities in the model would need to be decreased at $\lambda >$$\mu $m, an unrealistic possibility.


next previous
Up: Models of infrared spectra 1997

Copyright ESO 2002