3.3. Dependence of Zero-Point with Time

  One may use the large SWP camera datasets just discussed to determine if trends are present in science observations with respect to time. Table 2 exhibits the linear regression slopes with time for the large-aperture data for the three cameras. The data represented are in most cases the same as in Table 1. The second row of each data group lists the ratio of the regression slope to the r.m.s. error $\sigma$ in the slope. For the SWP data the regressions for three stars are significant to at least three sigma. For one of these three stars, $\zeta$ Oph, the non-zero slope was caused mainly from data by an intensive campaign conducted by one observer. Investigation showed that these differences could be traced to use of a different reference position in the FES. Thus, the derived slope for this star is questionable.

$$\tau \zeta \eta \lambda \zeta$$ Star/Cam.: 10 Lac HD 60753 +75^o325 SWP

Slope: -0.01 -0.05 -0.29 -0.49 -0.13 -0.03 -0.53 -0.34

$$\sigma$$ (<1) (<1) 2.1 (1.5) (1.0) (<1) >4 4 LWP

$$\tau \eta \zeta$$ 10 Lac +28^o4211 +75^o325 RR Tel HD93521

Slope: -0.44 -0.10 0.01 0.04 0.28 -0.03 0.05 .03

$$\sigma$$ (<1) (<1) (<1) (<1) <1 <1 <1 <1 LWR

$$\sigma$$ (<1) (<1) 3.1 (1.5) (1.0) (<1)

$$\tau \zeta \zeta$$ Star: +28^o4211

Slope: -.57 -0.55 0.28 -.28 -0.34 -1.7

$$\sigma$$ (<1) (<1) (<1) (<1)

Unlike the case of $\zeta$ Oph, the SWP data for two well-observed stars, HD 60753 and BD+75^o325 in Table 2, show clear negative trends with time (see Figure 5) that cannot be readily attributed to observational practices or instrumental conditions. A similar trend may also be present in the data of the A5 V star HD 11636 (not shown), for which the far-UV fluxes are faint. We might speculate that this apparent trend is actually the norm for UV-faint sources. For example, it is possible that these trends, if real, are by-products of an unexplained source of scattered light, i.e., ``the streak," which appeared in 1991. In fact, the scattered light from this artifact introduced a gradient in the light admitted across both the small and large apertures required that IUE telescope operators to adopt a new acquisition reference point in the FES to avoid shifts in the spectra of several km s^-1 (Pitts 2000). This displacement is similar to the change in zero-points documented for HD 60753 and BD+75^o325.

Figure 5
Apparent velocity trends with time: (a) HD 60753 and (b) BD+75325. All data were obtained through the large aperture.

Overall, we find that all but one of the 11 stars in the sample represented in Table 2 shows a negative slope. When subjected to a statistical ``Sign Test" (Dixon and Massey 1951)," the signs of these slopes are weakly significant at the 2$\sigma$ level. We conclude that over a 15 year interval a trend of several km s^-1 might exist for faint objects. We stress that this trend is not apparent in our bright objects results.

In contrast to the SWP camera results, we have found no clear velocity trends for spectra observed with the LWP and LWR camera. The LWP sample consisted of observations of eight stars: $\tau$ Sco, 10 Lac, $\eta$ UMa, $\zeta$ Cas, HD 93521, RR Tel, BD+28^o4211, and BD+75^o325. The distributions of shifts of all these datasets showed equal numbers of positive and negative slopes with time, none of which were significant above the level of 1.5$\sigma$. The occurrence in October 1983 of a large ``flare," a developing hole in the high voltage assembly, thereby producing a high, localized background, complicated the analysis of late-epoch data for this camera. The flare caused the IUE Project to reduce the camera voltage, resulting in a loss of sensitivity. This sequence of events led to a curtailment in the camera's use and an abbreviated time interval used to monitor instrumental trends of well observed objects. Among our sample of program bright stars and white dwarfs, we were able to find only two stars, $\zeta$ Cas and $\tau$ Sco, with an appreciable number of same-aperture observations before and after the flare. Of these two targets, $\zeta$ Cas is the better sampled in time. Its cross-correlation shifts exhibited no trend with time, but they did exhibit a pronounced increase in scatter (about a factor of two) after 1984.0. The data for $\tau$ Sco showed an apparent slope of -1 km s^-1 yr^-1, but this could be an artifact of a greatly increased scatter after 1984.0. An examination of the shifts as a function of wavelength for this star revealed no wavelength-dependence. This would be surprising if the flare were the cause of wavelength shifts. Our conclusion from this investigation is that the wavelength zero-points of LWR observations taken after the 1983 flare event are less reliable than before. However, this may not be a direct consequence of the flare itself. Rather, we suspect that the increased scatter arises from the decreased usage of the camera and the resulting lower accuracy of the necessary redetermination of positions in the FES used to acquire an object.