Time-dependent structure in ultraviolet absorption lines of the rapid rotators HD 64760 (B0 Ib) and HD 93521 (O9.5 V)

Cross-correlation provides an effective line-averaging function for spectra containing many line features, a result which can be exploited in order to perform time-series and spatial-domain analyses of absorption-line variability in data of (relatively) poor quality. We apply this method to high-resolution IUE spectra. For the known non-radial pulsator HD 93521 (09.5 V), time-series analysis of the cross-correlation function recovers two periods that are confirmed in independent optical data (P-1 = 1.77 h, P-2 = 2.90 h); there is no statistically significant excess power at these frequencies in lines formed in the stellar wind. By comparing phase information from the time-series analysis with results from pulsation models we estimate l=10+/-1 and 6+/-1 as the harmonic degrees for P-1 and P-2, respectively, with l+m less than or similar to 2 for each mode (where pn is the azimuthal order of the mode). We also present evidence for absorption-line variability in HD 64760 (HR 3090; B0.5 Ib), finding marginally significant signals with P-1 = 8.9 h (or, possibly, 14.2h) and P-2 = 29 h. The longer period is present, with a strong signal, in wind-formed lines. We consider possible circumstellar and (quasi-)photospheric origins for P-2, and conclude that this signal probably does not arise through rotational modulation (with the corollary that the stellar-wind signal also does not arise in corotating structures, contrary to previous suggestions). The phase behaviour of the signals is consistent with non-radial pulsation models characterized by l=5+/-1 (P-1) and 2+/-1 (P-2), with l+m less than or equal to 1, supporting a previous suggestion by Baade that both low- and higher-order modes are present; the wind modulation at P-2 may then result from leakage of pulsation energy into the supersonic outflow. The lack of significant photometric variability is a serious difficulty for this model (any sinusoidal photometric variability at P-2 has semiamplitude <2 mmag at lambda(eff) = 1575 Angstrom, with 95 per cent confidence), but this may itself be a consequence of wave leakage.