Recovery of 29 second oscillations in Hubble Space Telescope eclipse observations of the cataclysmic variable UX Ursae Majoris

Low-amplitude (similar or equal to 0.5%) 29 s oscillations have been detected in Hubble Space Telescope Faint Object Spectrograph eclipse observations of the nova-like cataclysmic variable UX UMa. These are the same dwarf nova-type oscillations that were originally discovered in this system by Warner & Nather in 1972. The 29 s oscillations are seen in one pair of eclipse sequences obtained with the FOS/PRISM in 1994 November but not in a similar pair obtained with the FOS/G160L grating in August of the same year. The oscillations in the PRISM data are sinusoidal to within the small observational errors and undergo an approximately -360 degrees phase shift during eclipses (i.e., one cycle is lost). The amplitudes are highest at preeclipse orbital phases and exhibit a rather gradual eclipse whose shape is roughly similar to, although perhaps slightly narrower than, UX UMa's overall light curve in the PRISM bandpass (2000-8000 Angstrom). Spectra of the oscillations have been constructed from pre-, mid, and posteclipse data segments of the November observations. The spectra obtained from the out-of-eclipse segments are extremely blue, and only lower limits can be placed on the temperature of the source that dominates the modulated flux at these orbital phases. Lower limits derived from blackbody (stellar atmosphere) model fits to these data are greater than or equal to 95,000 K (greater than or equal to 85,000 K); the corresponding upper limits on the projected area of this source are all less than 2% of the white dwarf (WD) surface area. By contrast, oscillation spectra derived from mideclipse data segments are much redder. Fits to these spectra yield temperature estimates in the range 20,000 K less than or similar to T less than or similar to 30,000 K for both blackbody and stellar atmosphere models and corresponding projected areas of a few percent of the WD surface area. These estimates are subject to revision if the modulated emission is optically thin. We suggest that the ultimate source of the oscillations is a hot, compact region near disk center, but that a significant fraction of the observed, modulated flux is due to reprocessing of the light emitted by this source in the accretion disk atmosphere. The compact source is occulted at orbital phases near mideclipse, leaving only part of the more extended reprocessing region(s) to produce the weak oscillations that persist even at conjunction. The highly sinusoidal oscillation pulse shape does not permit the identification of the compact component in this model with emission produced by a rotating disturbance in the inner disk or in a classical, equatorial boundary layer. Instead, this component could arise in a bright spot on the surface of the WD, possibly associated with a magnetic pole. However, a standard intermediate polar model can also be ruled out since UX UMa's oscillation period has been seen to change on timescales much shorter than the minimum timescale required to spin up the WD by accretion torques. A model invoking magnetically controlled accretion onto differentially rotating WD surface layers may be viable, but needs more theoretical work.