The instrument sensitivity for full-aperture observations was determined by comparing observations of the hot DA white dwarf HZ 43 with the model atmosphere for this star calculated by one of us (DF), using D. Koester's model atmosphere codes ([Koester, Schulz, & Weidemann 1979]; [Finley, Koester, & Basri 1995]). The parameters for this model are , log g = 8.0, and V = 12.91. and log g were determined from fits to Balmer line profiles (Finley, Koester, & Basri 1995), while the V magnitude was measured by the FOS ([Bohlin, Colina, & Finley 1995]). The models incorporate the Hummer-Mihalas occupation probability formalism ([Hummer & Mihalas 1988]) with the critical field strength parameter set to twice the nominal value (). The model was corrected for transmission through the interstellar medium, assuming a hydrogen column density of and b = 10 km s. Extinction by dust was assumed to be negligible. The resulting model flux was then convolved with a gaussian of 3 Å FWHM to simulate the instrumental resolution.
HZ 43 was observed three times during the mission (at 35, 64, and 329 hours MET). The last pointing contained the longest periods of clean photometric data, and was therefore used to define the effective area. Following acquisition, a 470 s integration was obtained through the round slit, followed by a 696 s integration through the round slit; both integrations were entirely within orbital night. The observed count rates were the same for the two integrations, and the count rate fluctuations on a two second time scale are consistent with Poisson noise, indicating that both portions of the observation are photometric. We therefore combined the two spectra, and applied the following corrections: (1) Pulse pile-up: This is event loss caused by the arrival of overlapping events on the photodiode array within the 1 msec readout time; this effect and the correction scheme are described in [Kruk95]. The peak correction is 9% at 1085 Å, dropping to 1% at either end of the spectrum. (2) Dark count subtraction: 0.01%--0.1%; see § 3.5. (3) Scattered light subtraction: 0.005--0.1%; also see § 3.5. (4) Second order subtraction: the equivalent of 21% of the counts Å in the 912--920 Å region contribute in second order at 1824--1840 Å. This excess is subtracted from the data. The extreme-ultraviolet (EUV) flux (< 912 Å) appearing in second order is estimated to peak at about 1% of the counts at 1550 Å, drop to 0.5% at 1300 Å and 1650 Å and drop rapidly to zero shortward of Lyman . This EUV flux was ignored in the present calibration, but will be included in future revisions. (5) Flat field corrections: see § 3.4. (6) Time dependence of the instrument throughput: see below.
To produce the instrument sensitivity curve we divided the corrected count spectrum into the model flux spectrum. Dividing this into the energy per photon as a function of wavelength and dividing by the (slightly variable) pixel width gives the effective area, which is shown in Figure 2. The peak effective area is 24.1 cm at 1160 Å, where the corresponding inverse sensitivity is . For comparison, we also show the final effective area curve for the Astro-1 version of HUT ([Kruk95]). As was the case on Astro-1, the effective area measured in-flight is in fairly good agreement with the pre-flight laboratory calibration shortward of 1050 Å, but falls roughly 30% below it at longer wavelengths. This difference is believed to be due to aging of the CsI photocathode.
The stability of the instrument throughput was monitored by making multiple observations of the DA white dwarf GD 394 (at 64, 92, 138, 211, and 341 hours MET), in addition to the three observations of HZ 43 listed above. The sensitivity was found to decline gradually over the course of the mission in a wavelength-dependent fashion. The total decline was relatively small at long wavelengths (5% at 1800 Å), and rose steadily towards shorter wavelengths (27% at 912 Å). The rate of decline was greatest early in the mission, with roughly half of the total decline occurring prior to 100 hours MET, and 75% of the total decline occurring by 138 hours MET. A family of correction curves was calculated by fitting three-term Chebyshev polynomials to the ratios of late to early observations of HZ 43 and GD 394. These correction factors are shown in Figure 3. A correction curve for any given MET can be obtained by interpolating among the curves.
Neither the wavelength dependence nor the time dependence of this degradation matches that of the detector exposure: the total flux per unit area on the detector has a broad maximum over 1100--1350 Å and drops rapidly towards either end of the detector, and it increases with time during the mission primarily during observations of a relatively small number of bright stars. Both, however, are consistent with possible degradation of the primary mirror. Two possible mechanisms for this are condensation of contaminants onto the mirror, and oxidation of the SiC coating by energetic atomic oxygen. One of the witness mirrors removed from the telescope shortly after the flight shows similar degradation, and is being investigated further.
An internal consistency check on the white dwarf models and on our data reduction procedures can be made by comparing the observed flux with the model predictions for other hot DA white dwarf stars spanning a range of effective temperatures and surface gravities. We observed GD 71, GD 153, RE 0512--004, GD 50, Wolf 1346, and RE 1738+669 through the full aperture as part of the Guest Investigator program of D. Finley, D. Koester, and R. Kimble. These spectra were reduced and compared with models in the same manner as described above for HZ 43. As shown in Figure 4, for GD 71, GD 153, RE 0512--004, and GD 50 the observed spectra agree with the models to better than 5%. The parameters for the models shown are: GD 71: = 32300 K, log g = 7.73, V = 13.032; GD 153: = 38500 K, log g = 7.67, V = 13.353; RE 0512--004: = 32000 K, log g = 7.4, V = 13.80; and GD 50: = 41000 K, log g = 9.2, V = 14.00. All of the model parameters were derived from fits to Balmer line profiles and from V-band photometry ([Finley, Koester, & Basri 1995]). For Wolf 1346 the observed spectrum agrees with the model to within measurement errors, except for wavelengths 1080--1150 Å\ (between the broad merging absorption wings of Lyman and Lyman ), where the data exceed the model by as much as 18%. The observed flux for RE 1738+669 is consistent with the IUE flux; comparisons with the model flux will require corrections for interstellar extinction and molecular hydrogen that are still in progress.
The instrument throughput for door configurations other than fully open was determined by taking ratios of spectra obtained through more than one door state on the same object. Since the various door states illuminate different regions of the mirror, grating, and detector (and illuminate the detector at different angles), the corresponding throughputs differ by more than just the relative exposed areas of the mirror.
The throughput for the half--open door state was determined by obtaining spectra of both HZ 43 and GD 153 through both the full and half open door states. The two sets of spectra gave essentially similar ratios of half to full door flux, so the two ratios were averaged. The resulting throughput ratio rises from 0.475 at 912 Å to 0.50 at 1400 Å then declines gradually to 0.415 at 1830 Å.
The 750 door state was calibrated by two different means: by observing GD 394 through this door as well as through the full aperture, and by direct observation of the hot DA white dwarf G191--B2B. The data for G191--B2B were reduced in the same manner as described above for HZ 43. The model atmosphere parameters used for G191--B2B were = 60000 K, log g = 7.4, and V = 11.79. This model differs by 2% from that used to define the HUT calibration for Astro-1. In order to keep any model dependence in the calibration the same for all door states however, the effective area curve for this door state was determined by fitting a 2--piece cubic spline to the ratio of the G191--B2B-derived effective area and the HZ 43-derived full aperture effective area, and then multiplying by the full aperture effective area. This retains the large--scale wavelength dependence of the G191--B2B effective area determination, but should remove any additional small--scale artifacts introduced by small errors in the model parameters, mismatches between the model and the actual spectral resolution, etc. The observations of GD 394 gave results that were consistent with the G191--B2B calibration, but with considerably poorer statistics.
The 200 door state was also calibrated by an observation of G191--B2B. The data reduction procedures were the same as for the 750 observation. The ratio of the two spectra was nearly independent of wavelength and was equal to 0.273, quite close to the nominal ratio of geometric areas (0.267).
The 50 door state was calibrated by observing the hot sdO star BD+75 325 through both the 200 and 50 doors. Longward of 1025 Å the ratio of the two spectra is independent of wavelength and is equal to 0.273; shortward of 1025 Å the ratio drops smoothly to about 0.24 at 912 Å. The effective area for the 200 door derived from the G191--B2B measurement was scaled by this ratio to determine the 50 door effective area. Since the calibration for these two door states is the furthest removed from the HZ 43-derived full aperture calibration, a test of our data reduction and calibration procedures can be made by comparing our spectrum for BD+75 325 with that obtained by the FOS (kindly provided by R. Bohlin). The HUT and FOS spectra for this star are shown in Figure 5, for the wavelength region common to both instruments. The two spectra agree to within 5% at most wavelengths and to within 10% everywhere.