Since there are no ultraviolet polarimetric standards in the sky, a detailed calibration of polarimetric efficiency was undertaken using a portable facility mounted on top of WUPPE at Kennedy Space Center (KSC) (Figure 1) . The light source is a Deuterium hollow-cathode lamp with an all-reflective condenser. After the condenser pinhole, a rotatable three-mirror (Al-MgF2-Al) Brewster-angle polarizer provides 100% linearly polarized light from 150-330 nm. This illuminates a 15 cm collimator which can be rotated into one of three stations separated by 120 degrees around the axis of the 50 cm WUPPE primary mirror. The calibration optics and WUPPE are then purged with argon gas, giving usable transmission to 150 nm. Data collected in this configuration were used to validate a model of the halfwave plates and Lyot analyzers and to calibrate the internal scattered light. The scattered light was evaluated by examining the residual signal at the ends of the spectrum where the sensitivity drops to zero: it was modelled as a linear ramp from 0.1% of the mean signal at the blue end to 0.9% at the red end. Figure 2 illustrates the apparent polarimetric efficiency of the halfwave mode, before and after correction for waveplate retardation wavelength dependence, beam convergence effects and scattered light.
The wavelength dependence of the flux response curve was calibrated using stars of known ultraviolet flux from observations by the International Ultraviolet Explorer (IUE) and (for wavelengths greater than 320 nm) the Orbiting Astronomical Observatory (OAO) . The relative flux calibration includes a correction for second-order contamination of the 280-330 nm spectrum by 140-165 nm light. The relative flux calibration is repeatable to approximately +/-10%. The absolute detector response was calibrated by evaluating the noise in the polarimetric signal for fainter targets, to give an effective quantum efficiency (since the detector is analog, photon counts are not available). The resulting absolute flux response curve is shown as a system effective area (summing the two orthogonal spectra) in Figure 3 . The feature near 240 nm is due to a detector flaw appearing in the "A" polarimetric channel. The effective area curve agrees well with prelaunch predictions. Due to pointing system problems the absolute fluxes of some targets observed through smaller apertures are low by up to a factor of two.
The halfwave polarimetric efficiency calibration was checked using highly polarized stars observed nearly simultaneously in the visible at Pine Bluff Observatory (PBO) (Figure 4) . Since the polarization at 320 nm is very sensitive to the calibration, the agreement seen is very satisfactory. The polarimetric efficiency calibration is good to 1 part in 20. The relative position angles of the analyzer filters were calibrated in the ground calibration. The absolute position angle was checked using the overlap region of highly polarized stars observed at PBO. The position angle calibration is repeatable to better than +/- 1 degree.
The ultraviolet instrumental polarization was calibrated using stars known to be unpolarized in the visible. The primary standards were Gamma Gem and Alpha Aur, with the red portion of Alpha Hyi and the central portion of Psi Vel used as a check. The latter two stars were observed in an attempt to detect the onset of polarimetric effects of rapid rotation, which had been predicted to appear only below 180 nm. All four stars have measured visible-wavelength polarization < 0.01% +/- 0.006%. Figure 5 shows the three highest signal/noise observations (two of Gamma Gem, one of Alpha Hyi) along with the adopted instrumental polarization (smooth curve), averaging about 0.05%. The U Stokes parameter observations agree to about 0.05%, while the Q parameter shows about 0.1% system noise. The remaining system noise is thought to be due to the residual effects of pointing variations.
-taken from Nordsieck et al. 1994, SPIE, Vol. 2010, p.2.