The preflight prediction of the detector UV response was underestimated. Vega (V = 0.03, A0 V, distance= 8.1 pc), a star relatively bright in the UV, gives a count rate of 0.10 counts s-1 (7 times that of the Einstein HRI), and beta Car (V = 1.68, A0 III, distance=26 pc) gives a count rate of 0.006 counts s-1.
At the 1996 May HEAD meeting in San Diego CA, S. Murray presented the results of some testing done on UV filters in conjunction with AXAF work (Zombeck et al., 1997). UV filters similar to those flown on ROSAT were also tested.
Whereas the conventional wisdom held that the primary UV ``leak'' for the ROSAT HRI occurred in the band 1500 to 1900 angstroms, there were at least two problems with that explanation (Silverman, Harris, and Kearns, 1996):
a) The predicted UV-leak count rate for Vega was about a factor of 100 lower than measured; and
b) The PHA distribution of Vega was significantly different (i.e. peaked at a smaller PHA) than that for the UV calibration lamp (primarily an emission line of mercury at 1830 angstroms). This can be seen by comparing figures 25 and 27.
Murray and his colleagues found a number of transmission bands above 2300 angstroms, and conclude that having the aluminium coating on both faces of a filter creates a ``lexan sandwich'' which then exhibits the behavior of an interference filter.
It is not possible to make precise quantitative predictions because there is no laboratory filter exactly identical to the ROSAT flight models. Several ROSAT filters were manufactured and one set has been held in storage at SAO. However, there is significant variation in the filter thickness and coating thickness compared with the actual flight article. There is also a possible aging effect on the Aluminum coating - oxidation - that differs between the flight filter and the laboratory sample.
We can however conclude that there is excess transmission in the region from 2000-4000 angstroms and this may be responsible for the high count rate of Vega (0.1 counts s-1). There is great uncertainty in the CsI photocathode response in the UV (from 1000-4000A) which makes the picture even more confused, but it is reasonable to ascribe the ``softer'' PHA distribution to the fact that softer photons (than the UV calibration lamp) cause the leak. Even if we get the filter completely right, there is no direct measurement of the photocathode.