3 Stellar parameters and spectral synthesis

3.1 Atmospheric parameters

The atmospheric parameters derived from the uvby$\beta$ colour indices by using the Moon & Dworetsky (1985) code and adopting the previously-derived reddening, E(b-y)=0.050, are $T\rm _{eff }$=7910 K and $\log g\,=\,3.79$.

We have also used the Geneva photometry indices to compute $T\rm _{eff }$  and log g, by adopting the same reddening value and by using the Künzli et al. (1997) calibration; in this case the parameters are: $T\rm _{eff }$= 7673 $\pm$ 65 K, log g = 3.98 $\pm$ 0.06 and $\rm [M/H] = -0.77 \pm 0.24$.

3.2 HR diagram

These data allow us to put the star in the HR diagram. For that purpose we used the evolutionary tracks computed by Schaller et al. (1992) and the isochrones by Meynet et al. (1993) for the solar chemical composition.

The absolute magnitude corrected for the reddening is: M_V=0.27; by applying the bolometric correction, taken from the Bessell et al. (1996) tables, the bolometric magnitude results $M_{\rm bol}=0.22$. By adopting $M_{\rm bol\odot}=4.75$ the luminosity is derived: log  $(L/L_{\odot})=1.81$. According to the position of the star in the HR diagram the value of log g is about 3.6, slightly lower than the two photometrically-derived values 3.79 and 3.98.

3.3 Computed spectra

We have computed a synthetic spectrum by using the Kurucz model with $T\rm _{eff }$=8000 K, $\log g=4.0$ and solar abundances to investigate the correlation with the observations. The normalized cross correlation indices between the two observed spectra and the theoretical one are displayed in Fig. 1. The two peaks obtained for the 1992 spectrum are evident, while only a broad and asymmetric cross correlation curve results from the 1993 data; however we note the low value of the normalized correlation peak even at this date.

A further correlation (Fig. 2) of only the metallic lines with, as a template, the spectrum computed by using the model with the Geneva atmospheric parameters, shows even better the two peaks for the 1992 spectrum and allows us to measure the differential radial velocity of the two components of this spectroscopic binary system at this phase: $\Delta RV= 52$ kms^-1.

Figure 2: The cross correlation curve for only the metal lines of the 1992 spectrum obtained by using as a template the synthetic spectrum computed with the model with the Geneva atmospheric parameters.

Figure 3: The comparison of the two observed spectra in the region 4200-4300 Å; HD174005a and HD 174005b refer to spectra taken on June 9, 1992 and on Sep. 7, 1993 respectively.

Figure 4: The H$_{\gamma }$ profile of HD 174005 spectrum observed in 1992 (solid line) compared with the spectra from the models: $T\rm_{eff }$=7673 K, log g=3.98, $v\sin i =100$ kms-1 and solar abundances (dotted line), the abundances reduced by 10 (dash-dot-dot line) and $T\rm_{eff }$=7910 K, log g=3.79 and the same broadening (dashed line).

Figure 5: The poor fit of the metal lines with the computed spectra computed assuming the model with $T\rm_{eff }$=7673 K, log g=3.98 and solar abundances (dotted line) and metal abundances reduced by a factor of 10 (dashed line); we note in particular that the Mg II 4481 doublet in the spectrum taken in 1992 does not have a rotationally broadened profile and the feature at 4415 Å is not reproduced by the computations. The straight lines in the observed spectra correspond to bad CCD columns.

The large variations in the line profiles of the metal lines are shown in Fig. 3, where a sample of the 2 spectra, corrected for the radial velocity values of the brighter component, are plotted. The star appears as a clear SB2 on the 1992 spectrum; we note in particular the well-resolved Sr II 4215 line, the double peaks of the Fe I 4271 and of the Cr I-Ti II 4290 blends and the general broader profiles in this spectrum labelled as HD 174005a.

As a further sign of duplicity we recall that the observed core of the Balmer lines should be deeper than that in the spectrum computed by using Kurucz SYNTHE code, which does not include NLTE effects. On the contrary, the spectrum taken in 1992 has the H$_{\gamma }$ core shallower than or as deep as the computed ones. In Fig. 4 this observed profile is compared with those computed by using the models with the parameters derived in Sect. 3.1. The best fit of H$_{\gamma }$ is obtained by using the parameters $T\rm _{eff }$=7673 K, log g=3.98 and solar abundances; however this choice does not reproduce the metallic lines as may be seen in Fig. 5.

The broad and apparently weak metal lines in the 1992 spectrum are given in Fig. 5 and are compared with two computed spectra ( $T\rm _{eff }$=7673 K, log g=3.98), one with solar abundances and the other one with these reduced by 10.; we note in particular the non-rotationally broadened profile of Mg II 4481 which is characterized by a very flat and square core in the 1992 spectrum.

Having only two spectra at our disposal and taken more than one year apart, we cannot make any guess regarding the possible period of this system.

Figure 6: The observed and dereddened UV colour indices ( $V-m_{\rm UV}$)$_{\rm o}$compared with those expected for a star with the parameters $T\rm_{eff }$=7750 K, log g =4.0 and [Fe/H] = 0. (bottom), -0.5 (middle) and -1. (top) labelled with filled triangles; the symbols for the observed indices are +: undereddened values; square: E(B-V)=0.069; star: E(B-V)=0.152.

3.4 UV photometry

We have compared the S2/68 TD1 observations, dereddened by using E(B-V)=0.069 or 0.152, with theoretical indices computed by integrating the Kurucz fluxes for several metal-abundances, over the band-width of the four S2/68 TD1 channels.

The observed and theoretical colour indices ( $V-m_{\rm UV})_0$ are shown in Fig. 6. The flux energy distribution of HD 174005 in the UV does not indicate a significant metal underabundance unless the unrealistically high reddening E(B-V)=0.152 is assumed. The observations fit the computations with a moderate underabundance between -1.0 and -0.5 dex. This is coherent with the value of -0.77 derived from the Geneva photometry.

The coherence of the flux distribution over a large wavelength range from the optical to the UV reinforces the result obtained from spectroscopy in the optical: the two stars have a very similar luminosity and effective temperature.