4.1. Wavelength-Systematics?

  In our earlier discussion we focused only the temporal results because we did not find any systematic dependences on wavelength. In Figure 9 we demonstrate this further by exhibiting the averages and confidence belts (± 1 ${sigma}$) of shifts for 16 stars at each wavelength interval in our SWP camera study. The mean r.m.s. over wavelength for all stars is ± 2.4 km s^-1 while the adjacent order-to-order shift is only ± 1.0 km s^-1. These errors are generally wavelength-independent except below $\approx\$ $\lambda\$ 1230. Below this wavelength the errors increase because the detector loses its sensitivity to flux. Because there is no clear trend in wavelength, we will refer to the zero-point error in terms of apparent velocity rather than wavelength.

Fig. 9
Mean shifts of wavelength zero-points as a function of wavelength extracted from a large number of IUE echellograms for 16 stars (10 Lac, $\tau$ Sco, $\zeta$ Cas, $\zeta$ Oph, $\eta$ UMa, $\lambda$ Lep and the white dwarfs BD+75^o325, BD+28^o4211, Sirius B, WD0005+511, NGC246, REJ0623-3, EG102, Wolf 1346, GD 394, Gl191-B2B). These shifts were obtained by cross-correlating wavelengths in each high-dipersion echelle order for a number of test images of a star against a reference echellogram. Belts of ± 1 $sigma}$ around the mean shifts are drawn from the distribution of the shifts for the individual stars.

A comparison of the velocity fluctuations resulting from the cross-correlations among different stars suggests that the precision of a zero-point determination varies as a weak power of the reciprocal of the net flux (detector counts). In addition, we find that order-to-order shifts are measurably larger for spectra of white dwarfs and rapid rotators (e.g., $\eta$ UMa, $\zeta$ Cas). This fact suggests that the random cross-correlation errors increase slightly with line broadening, namely from about ± 3 km s^-1 (e.g., $\tau$ Sco) to ± 4 km s^-1 (e.g., $\eta$ UMa).