Comparison to Kurucz 92 models

In Figs. 1-3 we compare the PHOENIX synthetic spectra with ATLAS9 spectra for a range of effective temperatures with fixed solar abundances. Note that while Kurucz 92 uses the [Anders & Grevesse(1989)Anders and Grevesse] solar abundances, we use the more recent solar abundances given in Table 5 of [Jaschek & Jaschek(1995)Jaschek and Jaschek]. The changes in the abundances will introduce small but systematic differences between the Kurucz 92 models and our model structures and between the spectra because the abundances of the key elements C, N, O, and Fe are different. The PHOENIX spectra have been degraded by convolution with a Gaussian kernel to more closely resemble the resolution of the ATLAS9 spectra. The spectra for the Kurucz models were obtained using the IDL routine KURGET1 and the corresponding database of models available in the IUE reduction and data analysis package IUERDAF.

In general, our models agree well with the ATLAS9 spectra for effective temperatures higher than about $5000\,{\rm K}$. We expect better agreement between PHOENIX and ATLAS9 spectra for higher effective temperatures since we have included the latest version of Kurucz atomic line list into PHOENIX and atomic lines dominate the overall opacity in this temperature range. The remaining differences between the spectra are probably caused by the slightly different abundances, different methods for treating the line blanketing (ODF in ATLAS9 versus direct opacity sampling in PHOENIX), the fact that the ODF's in the ATLAS9 models were constructed using a previous version of the Kurucz line lists, and by differences in the treatment of, e.g., scattering and b-f and f-f opacities. These different treatments alter the model structures, resulting in systematic variations between the spectra. In the ultraviolet (UV), the two model grid spectra differ in several regions (cf. Fig. 1), which might in part be due to the fact that the ATLAS9 spectra are based on ODF's. The general features of the UV spectra are similar. Some of the features in the PHOENIX spectrum are deeper than the ATLAS9 features. This could partly be due to the different abundances and to changes in the atomic line list itself since the calculation of the ATLAS9 models.

We compare the models in the region of the H I Balmer lines in Fig. 2. The lines widths agree well, however, the H I lines in the PHOENIX spectrum seem to be a little deeper than the ATLAS9 lines. This is probably simply an artifact of the ODF versus opacity sampling treatment of the lines. In addition, we have inserted wavelength points in the cores on the H I lines. For comparison, we also show the PHOENIX spectrum at its original resolution with the fluxes scaled by a factor of 0.5. The opacity sampling method results in many more line features appearing in the spectrum, similar to ATLAS12 results shown in [Castelli & Kurucz(1994)Castelli and Kurucz]. Figure 1 compares our models to Kurucz 92 for $\log(g)=4.0$, $\hbox{$\,T_{\rm eff}$}=10000\,{\rm K}$ and for solar abundances. The differences between the spectra are essentially the same over the whole range of gravities and are therefore not shown. The PHOENIX spectra show less flux in some UV regions (e.g., around $1500\hbox{\AA}$) and the UV lines are generally deeper than in the ATLAS9 spectra. The flux that is intercepted in the UV is redistributed into the Paschen continuum so that the total energy flux through the atmosphere is given by $\sigma \hbox{$\,T_{\rm eff}$}^4$. Thus, the differences in the spectra are probably caused mainly by a larger UV opacity in the opacity sampling PHOENIX models.

For lower effective temperatures, the differences between the PHOENIX and the ATLAS9 spectra become larger. In Fig. 3 we show a more detailed comparison between models with $\hbox{$\,T_{\rm eff}$}=4000\,{\rm K}$, $\log(g)=4.0$and solar abundances. The figure clearly shows the spectral regions where the two spectra do not agree well. In particular, the ``pseudo-continuum'' between about $0.9\,\mu$ and $1.6\mu$ lies at about a 10% lower flux level in the ATLAS9 spectrum than in the PHOENIX spectrum. In the optical spectral region between $6000\hbox{\AA}$ and about $8000\hbox{\AA}$ the PHOENIX spectrum shows a number of molecular bands (due to TiO and VO?) that are not seen in the ATLAS9 spectrum. These bands are more important in the VLMS and Brown Dwarf range of the NextGen grid. The Kurucz TiO line list results in much weaker TiO features and different band shapes than the Jørgensen TiO list that we use [Schweitzer(1995)Schweitzer]. This is more important for lower effective temperatures and will cause more significant deviations between the PHOENIX and the ATLAS9 models.

There are also discrepancies between the spectra in the near infrared (NIR) spectral region. The CO bands are somewhat different, which is probably caused by the ODF versus opacity sampling treatment of the line blanketing of the calculations. The CO line data that we use are very similar to the Kurucz CO line list [Schweitzer(1995)Schweitzer]. We include water lines in the model calculations and the spectral synthesis, which is not included in the ATLAS9 models. This causes the differences in the molecular absorption band around $2.5\,\mu$, this feature is due to a strong water vapor band. Therefore, it is likely that the differences for low effective temperatures can be explained mainly by the different opacity data used in PHOENIX and ATLAS9. At effective temperatures as high as $4000\,{\rm K}$, differences between the number of molecules and the molecular data used in the EOS's of PHOENIX and ATLAS9 are probably unimportant. At lower temperatures, however, the larger EOS of PHOENIX will likely increase the differences between the two sets of models.