REPROCESSED UV PULSES FROM THE BINARY COMPANIONS * OF X-RAY PULSARS 1 1,2 1,3 Joseph F. Dolan , Robert J. Hill , Patricia T. Boyd , 1,2 4 5 Jeffrey M. Silvis , Edward L. Robinson , Jeffrey W. Percival , 5 6 Robert C. Bless and G. W. van Citters 1. Laboratory for Astronomy & Solar Physics, NASA Goddard Space Flight Center, Greenbelt, MD 20771; tejfd@stars.gsfc.nasa.gov 2. Hughes STX Corporation @ NASA Goddard Space Flight Center; hill@noether.gsfc.nasa.gov; silvis@stars.gsfc.nasa.gov 3. Universities Space Research Association @ Nasa Goddard Space Flight Center; padi@dragons.gsfc.nasa.gov 4. Department of Astronomy & McDonald Observatory, University of Texas, Austin, TX 78712; rob@emx.utexas.edu 5. Department of Astronomy & Space Astronomy Laboratory, University of Wisconsin, Madison, WI 53706; jwp@jerry.sal.wisc.edu, bless@jerry.sal.wisc.edu 6. Division of Astronomical Sciences, National Science Foundation, 4201 Wilson Blvd., Arlington, VA 22230; gvancitt@note1.nsf.gov Send offprint requests to: J. F. Dolan. * Based in part on observations with the Hubble Space Telescope obtained at the Space Telescope Science Institute, which is operated by AURA, Inc. under NASA contract NAS 5-26555. Received: _______________; Accepted: ______________ - 2 - Running head: J. F. Dolan et al.: UV Pulses in X-Ray Binaries ______________________________________________________________ Main Journal Section 6: Formation, structure and evolution of stars 08.02.1, 08.09.2 HD77581, 08.14.1, 08.16.7 4U0900-40, 13.21.5 - 3 - ABSTRACT: A search for reprocessed X-rays causing UV pulsations in the companion stars to X-ray pulsars was carried out in a 200 A wide bandpass centered at 1450 A using the High Speed Photometer on the Hubble Space Telescope. We observed the systems A0535+26 = HD245770, 4U0900-40 (Vela XR-1) = HD77581, A1118-61 = He3-640, and 4U1145-619 = HD102567. A positive detection of reprocessed X-rays occurred in 4U0900-40; we give upper limits on the pulsed fraction in our UV bandpass present in the other systems. We find evidence in archival IUE spectra that the UV radiation from HD77581 is emitted primarily in resonance emission lines from multiply ionized metals in the B0.5Ib star's outer atmosphere. The existence of two different pulse frequencies (UV and X-ray) in an X-ray binary system makes it dynamically equivalent to a double-lined spectroscopic binary. The mass of the neutron star can be derived from the orbital phase dependence of the two frequencies. Key words: binaries: close - stars: individual: HD77581 - stars: neutron - pulsars: individual: 4U0900-40 - ultraviolet: stars - 4 - 1. INTRODUCTION X-ray binary systems often contain neutron stars (NS) possessing strong magnetic fields which ultimately channel the mass accreted from the companion star (via a stellar wind or Roche lobe overflow) onto the regions surrounding the NS magnetic poles. The radiation pattern emitted is strongly directional, with the rotating NS often appearing to observers as an X-ray pulsar. Shortly after the discovery of X-ray pulsars in binary systems, it was realized that the illumination of the hemisphere facing the NS by the X-ray pulses should lead to pulsed radiation from the companion in the UV and visible (Basko & Sunyaev 1973; Avni & Bahcall 1974). The X-rays are absorbed in the upper levels of the companion star's atmosphere and their energy is expected to be reradiated on a timescale of seconds (Dahab 1974; Alme & Wilson 1974). The resulting pulses will have a different frequency from that of the X-ray pulses emitted simultaneously, being Doppler shifted by an amount corresponding to the relative velocity of the pulsar and the emitting region as projected on the line of sight (Groth 1974). The two different pulse frequencies in the system make it dynamically the equivalent of a double-lined spectroscopic binary, and the mass of the NS star can be derived from the orbital phase dependence of the two frequencies (Middleditch & Nelson 1976). A well-determined range of NS masses is needed to constrain the equation of state of NS matter and to indicate the - 5 - evolutionary tracks of stars which form NS (van Kerkwijk, van Paradijs, & Zuiderwijk 1995a). The detection of reprocessed optical pulses from binary systems containing X-ray pulsars offers the possibility, at least in theory, of determining masses for several different neutron stars accurately enough to guide the theoretical analyses of these problems. Optical pulses were first detected from HZ Her, the companion star to Her X-1, by Davidsen et al. (1972), although they did not attribute the pulses to reprocessed X-ray pulses. A complete analysis of the Her X-1 system was provided by Middleditch and Nelson (1976), who found a mass of 1.30 + 0.14 M for the NS (but cf. Chester [1978], who O found 1.2 + 0.4 M from a different analysis of the same O data). Since then, no pulses in the visible attributed to reprocessed X-radiation have been detected from any other binary system. Systems in which reprocessed visible pulses have been sought without success include A0535+26 = HD245770, 4U0900-40 (Vela XR-1) = HD77581, and 4U1145-619 = HD102567 (Margon et al. 1977; Payne & Coe 1987). Steiner (1977) reported that the flux in the H line of HD77581 was modulated with 2% amplitude at the frequency of the X-ray pulses in the system, but because this modulation existed even during the phases of X-ray eclipse, it could not be caused by reprocessed X-rays. These H pulsations could not be confirmed in observations by Nelson et al. (1979) or Thomas et al. (1981), however. - 6 - The lack of success in observing reprocessed pulses in the visible from X-ray binary systems led us to consider observing these systems in the UV. The structure of the companion star's atmosphere at the depth at which the X-rays -2 are absorbed (a few g cm ) would be consistent with the re-radiation of this energy preferentially in recombination lines from multiply ionized species whose ionization state would be affected by the temperature variation caused by the variable energy input from the X-ray pulses. For OB stars, these emission lines fall primarily in the far UV. We chose a 200 A wide bandpass in the UV which includes the resonance lines of Si IV (1394 and 1403 A), ionization potential (IP) 45 eV; C IV (1551 A), IP 64 eV; and O V (1371 A), IP 113 eV. C II (1335 A), IP 24 eV, is at the edge of the bandpass. We observed 4U0900-40 and the X-ray transients A0535+26, A1118-61, and 4U1145-619. We report here the detection of reprocessed UV pulses from HD77581 (= 4U0900-40), an indication of UV pulses in A1118-61 eleven months after an X-ray outburst, and upper limits to the modulation of the UV flux from A0535+26 and 4U1145-619 in quiescence. We also give upper limits to the flux in our UV bandpass from the systems 1E1145.1-6141, GX304-1, and 4U1538-52. - 7 - 2. OBSERVATIONS All observations were obtained with the High Speed Photometer (HSP), one of the first-generation instruments on board the Hubble Space Telescope (HST). A description of the HSP and its method of operation is given by Bless et al. (1996). Photometric observations were obtained using the F145M filter, which defined a 200 A wide bandpass (FWHM response to a flat incident spectrum) centered at 1450 A (Bless et al. 1992). All observations used a 1.0" circular aperture. Flux densities were calibrated by observations of o BD +75 325, an O5p IUE spectrophotometric standard. The X-ray binaries observed are listed in Table 1, together with the starting time and duration of each continuous observation. The X-ray pulse period in each system is > 100 s; each observation used a sample time of 1.0 s. The uncertainty of the mean counting rate is derived from the variance of the counting rate about its mean, and does not include the systematic uncertainty in HSP photometry caused by the spherical aberration of the HST mirror in conjunction with milliarcsecond variations in target centering (Dolan et al. 1994). The orbital ephemerides from which the binary phases at the epochs of observation were calculated are given in Sec. 3 for each individual source. For transient sources, the binary phase is normalized to X-ray outburst at = 0; for 4U0900-40, the orbital phase is referred to superior conjunction of the X-ray source (the center of X-ray eclipse) at = 0. - 8 - 3. RESULTS 3.1. 4U0900-40 4U0900-40/HD77581 is an eclipsing binary system with a B0.5Ib primary and a NS secondary. We adopt for it the ephemeris of Sato et al. (1986), where the time of superior conjunction of the X-ray source is T = JD 2,445,785.28 + 0 0.08, and the orbital period is 8.96426 + 0.00018 d. The X-ray eclipse occurs centered at phase zero and extends from ~ 0.92 to ~ 0.08 (Dolan et al. 1981). The NS is an X-ray pulsar. Lutovinov et al. (1994) measured its rotational period to be 283.326 + 0.020 s on 1992 June 13. The X-ray pulse profile is essentially unchanging over a timescale of years; Fig. 21 of Lutovinov et al. shows a typical shape. The double-peaked structure occurring over one pulse period suggests that we are observing radiation from both magnetic poles of the NS. Boynton et al. (1986) and Deeter et al. (1987) found the NS orbit to be slightly elliptical, e = 0.089 + 0.003, and give (a /c) sin i = 112.7 x + 0.5 s based on X-ray pulse timing. We observed HD77581 on 1993 January 6. The observation started at an orbital phase near the onset of X-ray eclipse, = 0.921 + 0.011, where the uncertainty in the phase is propagated from the uncertainty in T and period given by 0 Sato et al (1986). An occultation of the X-ray source during the 48 minute observation would make any reprocessed X-ray pulses disappear at the same time. Hence, we searched the data for a periodic signal with a possible time-variable - 9 - power by using the Gabor transform (Boyd et al. 1995), a type of short-time-windowed Fourier transform. The Gabor transform of our data set is shown in Fig. 1. The frequencies of maximum power in our data set are near the harmonics of the X-ray pulse frequency for the first 600 s of the observation. The power at these frequencies then decreases rapidly to the level associated with the random variations about the mean of the data set. Power from reprocessed X-rays would be most intense at harmonics of the 283 s X-ray pulse period because the pulse profile consists of two different intensity pulses per NS rotation period, requiring the higher harmonic Fourier components to represent it. The temporal behavior of the power at these frequencies is consistent with the disappearance of the UV pulsation being caused by the occultation of the X-ray pulsar. To estimate the significance of the signal detected in the Gabor transform, we analyzed the first 600 s of our data set with an auto-correlation function (ACF) analysis (Percival et al. 1995). The data set was first pre-whitened to remove a small, long term (93 minute period) trend in HSP count rate caused by orbital heating effects on the secondary mirror truss (Taylor et al. 1993), but the results are independent of whether a pre-whitened or non-pre-whitened data set is used. The ACF up to 400s lag time (Fig. 2) shows its largest feature near 281 s, at the 5 level of significance. A peak as large as this (or larger occurs) in -7 the ACF of only 6 x 10 of all time series with the same - 10 - mean and variance as our data set when those time series have a random distribution of arrival intervals between counts with no periodic signal present. The exact period of the modulation is not well determined by the ACF because its time-resolution is only 1 s. We searched the first 600 s of our data for the most likely period of this signal using the Rayleigh test (Mardia 1972), a test which does not depend on the unknown phase of maximum of the variation. We investigated periods in the range p = -1 283.3 + 0.5 s, corresponding to velocities of +500 km s (or more than twice as large as any orbital or rotational velocity known to exist in the system), by summing the power in the periods of the fundamental and first 2 harmonics (i.e., at p, p/2, and p/3) (DeJager et al. 1988). We found a maximum in the Rayleigh test statistic at the 99.95% level of significance (= 3.5 ) at p = 283.295 + 0.040 s at the spacecraft. This maximum in the Rayleigh test statistic is broad in period space because only 2 cycles of the modulation are present in 600 s of data. One slightly larger maximum occurs in the range 280 s < p < 286 s, at 283.815 + 0.060 s at the spacecraft. When corrected to the heliocenter, this -1 would correspond to a velocity of +540 km s relative to the center of mass of the binary. We consider this period to be an alias of the 283.295 s period. The data folded modulo the 283.295 s period are shown in Fig. 3 with pulse phase zero arbitrarily corresponding to the start of our observation. The UV pulse profile closely resembles the typical pulse profile seen in X-rays, - 11 - particularly at energies above 6 keV (Lutovinov et al. 1994; Staubert et al. 1980). Both profiles consist of two peaks separated by ~ 0.5 in phase. One maximum has a higher intensity than the other; the minimum following the higher intensity peak is not as low as the other interpeak minimum. The lower peak also appears to have a longer rise to maximum in both X-ray and UV pulse profiles. Published X-ray pulse ephemerides do not allow us to establish the X-ray pulse phase of our observations and so we can not compare the relative alignment of the X-ray and UV peaks. If the UV signal we detect comes from 4U0900- 40, then correcting its observed period for the orbital velocity of the spacecraft and the Earth during our 600 s observation gives a heliocentric period of 283.317 + 0.040 s. If we define the pulsed fraction as PF = [ - S ]/, [1] 0 where is the mean counting rate per bin in Fig. 3 -1 (713.2 + 1.8 s ) and S is the minimum counting rate in any -1 0 2 bin (704.8 + 3.5 s ), then PF = 1.2 + 0.6 %. (The value of the data in Fig. 3 about their mean is 23.9 for 9 degrees of freedom, corresponding to the 2.8 level of significance that the pulse profile is not consistent with a constant value independent of bin; this is lower than the significance of the signal in the Rayleigh test because the 2 test assumes an arbitrary phase for each bin boundary. The PF is formally significant only at the 2 level because of the propagation of error from both and S .) 0 - 12 - 3.2. A1118-61 A1118-61 is a recurrent X-ray transient, first observed in outburst on 1975 December 20. The X-ray flux was pulsed with a period of 405.3 + 0.6 s (Ives et al. 1975). The flux decayed to background over the following two weeks. The optical counterpart of the system is the Be star He3-640 = Wray 793, of spectral type O9.5 IV-Ve (Janot-Pacheco et al. 1981; Coe & Payne 1985). X-ray transients in a Be star system with pulse periods > 100 s are probably caused by accretion onto a rotating NS secondary. The accretion occurs when the NS passes through the Be star's dense wind (or extended atmosphere) near periastron (Motch et al. 1988). The only other detected outburst of A1118-61 started 1992 January 1 (Brandt et al. 1993). This outburst was larger than the first; the x-ray source was still detectable on 1992 March 9 (Coe et al. 1994). The X-ray pulse period was originally 406.57 + 0.05 s but gradually decreased to 406.34 + 0.02 s over a month and then fluctuated around that period (Coe et al. 1994). The X-ray pulse profile conisted of a single peak with FWHM 0.37 in phase. The pulsed fraction in X-rays is very large, ~ 85% (Ives et al. 1975). We observed He3-640 (= A1118-61) on 1992 December 29, almost one year after the onset of the second X-ray outburst. If the two outbursts correspond to successive periastrons of the NS, then the period of the system is 17.0 years and our observations occurred at ~ 0.05 after periastron. (Note that this hypothetical phase is not referred to the superior - 13 - conjunction of the X-ray source.) We searched the 2,880 s long data set for a periodic signal using both the ACF and power-spectrum function (PSF) techniques. A peak appeared in the ACF at the 2.9 level of significance at a lag time of 409 s; the ACF at half this lag (205 s) was significantly above the mean ACF, at the 2.4 level of significance. Neither peak can be considered as a significant detection under the usual assumptions of time-series analysis. The PSF showed excess power at a period of 206 s (14 cycles in 2,880 s), but only at a level which is exceeded by ~ 40% of all random distribution with no signal in them having the same mean and variance as our data set. The Rayleigh test showed a maximum at the 2.5 level of significance when the data were folded modulo p = 204.51 + 0.05 s at the spacecraft. If this were a signal from A1118-61, its heliocentric period would be 204.52 + 0.05 s. This is ~1/2 the period of the X-ray pulsar. It might be argued that the X-ray period is the true rotational period, and that we see two UV pulses per NS rotation because the Be star intercepts the X-ray beam from both NS magnetic poles while the Earth intercepts only one. Although the pulse profile of our data folded modulo 204.5 s shows a sinusoidal shape, that folded modulo 409.0 s shows no clear pattern. Both "pulse profiles" are consistent with a random 2 distribution about the mean under the assumptions of the test. We interpret our results as indicating that < 4.0 % of the flux from A1118-61 in the F145M bandpass is pulsed with a - 14 - period near that of the X-ray pulses during our observation. We derived this upper limit by adding an artificial signal with p' = 300 s to our data set. The signal's mean count -1 rate was 0.124 s , or 4 % of the mean count rate from 2 A1118-61. The input pulse shape was of form sin , where is the 300 s phase. ( = i/300, where i is the 1 s duration time series bin.) The single pulse was 150 s wide at zero maximum; no counts were added from = 0.5 to 1.0. 2 (The FWHM of the sin curve is 0.25 in .) The probability of bin i receiving a count was pr(i) = 2 150 2 N sin / sin , where N is the total number of counts i=1 per 300 s period in the pulse. No attempt was made to replicate Poisson statistics with regard to bins containing -1 more than one count. For A1118-61, N = 300 s (0.124 s ) = 37. p' = 300 s was chosen for A1118-61 to be similar to the X-ray pulse period, but not near it. We detected this signal at the 3.0 level of significance in the PSF at a frequency of 10 cycles per 2,880 s, and at the 1.6 level of significance at the frequency of the first harmonic, 19 cycles per 2,880 s. Assuming a 3 level of detection to be significant gives the upper limit we quote. 3.3. A0535+26 A0535+26 is a repetitive X-ray transient whose X-ray flux is pulsed with p ~ 104 s (Rosenberg et al. 1975). Its optical counterpart is HDE245770, a B0Ve star (Wade & Oke 1977). Priedhorsky and Terrell (1983) find a periodicity of 111.0 + 0.4 d between outbursts, presumably the orbital period of a NS about the primary. We observed HDE245770 on - 15 - 1993 January 5. An X-ray outburst was observed from the source on 1993 July 8 (JD 2,449,177), and lasted for longer than one week (Wilson et al. 1993); its pulse period in 20 - 120 keV X-rays was 103.377 + 0.005 s. Using JD 2,449,177 as the time of onset of an outburst, we observed A0535+26 at = 0.34 + 0.01, where the phase is referenced to the onset of X-ray outburst. (Using the original T of Priedhorsky and Terrell (1983) gives = 0 0.45 + 0.24.) No significant feature was found in the ACF or PSF of our data near the pulse period of A0535+26 (or near its harmonics). We interpret this to mean that < 0.7% (detection at the 3 level of significance) of the flux in the F145M bandpass from HDE245770 was pulsed with any period near 104 s during our observation. We derived this upper limit using the same technique of adding an artificial signal to our data set described above. 3.4. 4U1145-619 4U1145-619 is a repetitive X-ray transient whose X-ray flux is pulsed with p ~ 292 s (Cook & Warwick 1987). Its optical counterpart is HD102567, a B1Ve star (Bianchi & Bernacca 1980). The X-ray source undergoes outbursts every 186.5 d (Cook & Warwick 1987), which is the orbital period ascribed to a NS secondary. Outbursts typically last 10 d. We observed HD102567 on 1993 August 8. An X-ray outburst was observed from the source with onset 1994 March 12.5 (JD 2,449,424.0); its barycentric pulse period during this outburst was 293.4464 + 0.0016 s in 20 - 40 keV X-rays - 16 - (Wilson et al. 1994). Using JD 2,449,424.0 as T , we observed 4U1145-619 at = 0 0.84, where the phase is referenced to the onset of X-ray outburst. (The jitter in the phase of outburst given by Cook and Warwick (1987) is typically +0.03.) No significant feature was found in the ACF or PSF of our data near the pulse period of 4U1145-619 (or near its harmonics). We interpret this to mean that < 0.2% (3 upper limit) of the flux in the F145M bandpass from HD102567 was pulsed with any period near 293 s during our observation. 3.5. Other Sources We also observed the optical counterparts of the X-ray pulsars GX304-1 (= 4U1258-61) on 1993 June 3, starting JD 2,449,142.484; 1E1145.1-6141 on 1993 October 5, starting JD 2,449,265.934; and 4U1538-52 on 1993 September 21, starting JD 2,449,251.543. No flux above background was detected from any of these systems in the F145M bandpass. The corresponding 3 upper limit on the flux densities from these stars at the epoch of our observation are given in Table 2, together with the flux densities observed from the four detected stars. - 17 - 4. Discussion The signal present in the F145M bandpass during our observation of HD77581 has the following characteristics expected from reprocessed X-ray pulses from 4U0900-40: - the UV pulse period is near the X-ray pulse period observed by Lutovinov et al. (1994) six months earlier. The UV and X-ray heliocentric periods differ by 0.009 + 0.045 s; - the UV and X-ray pulse profiles are similar in shape; - the pulsed fraction of the UV pulse profile is similar to that predicted for reprocessed X-radiation from 4U0900-40 (Chester 1979); - the UV pulses disappeared at the approximate phase of immersion of the NS into eclipse. 4.1. Signatures of Reprocessed Pulses 4.1.1. _U_V _P_u_l_s_e _P_e_r_i_o_d. The period of the UV pulses emitted by the primary is Doppler shifted from the observed X-ray pulse period by three separate effects: (i) a non-circular orbit of either star around the center of mass (COM) of the binary system introduces a radial velocity component of the NS relative to the primary at most phases. Deeter et al. (1987) find e = 0.089 and (a /c) sin i o x = 112.7 s. Using i = 76 (Dolan & Tapia 1988) and P = 8.964 d -1 for the NS orbit gives a mean NS velocity of 280 km s about the COM, and an maximum radial velocity relative to the - 18 - -1 primary of 32 km s . This velocity would produce a maximum shift in the 283 s pulse period of 0.030 s. Whether the period would be red-shifted or blue-shifted depends on the phase of the observation relative to the phase of periastron. The X-ray pulse period of 4U0900-40 also exhibits both secular variations and short-term fluctuations in either direction (spin-down or spin-up) (Nagase et al. 1984). The two effects combined can change the period by as much as 0.04 s over 6 months. Hence, the X-ray period itself during our UV observations may have been slightly different from that observed 6 months earlier by Lutovinov et al. (1994). (ii) the velocity of the COM of the primary in its orbit introduces a Doppler shift into the UV pulse period. For the purpose of calculating the order of magnitude of this effect, assume that M = 25 M and M = 1.4 M . Then o O x O the instantaneous velocity ratio of the two stars v /v = -1 o x M /M , or = 16 km s . At = 0.92, this produces a x o o -1 radial velocity of the primary of -7 km s . This radial velocity would blue-shift the UV pulse by 0.007 s. (iii) the rotation of the primary about its axis also introduces a Doppler shift into the UV pulse period. If the rotation of the primary is direct, the period is red-shifted at = 0.92; if retrograde, the period is blue- -1 shifted. v sin i of the primary is ~ 130 km s (Dolan rot et al. 1981). Depending on the shape of the area of the visible hemisphere illuminated by the X-ray beam, the center- - 19 - of-light velocity of the illuminated lune from which the UV pulses are seen at = 0.92 is probably < 0.7 of this value, -1 or < 90 km s . This corresponds to a Doppler shift in the 283 s period of < 0.085 s. The observed difference between the period of the detected UV signal and the X-ray pulse period 6 months before (-0.009 + 0.045 s) is within the range allowed for reprocessed X-ray pulses. 4.1.2. _U_V _P_u_l_s_e _P_r_o_f_i_l_e_s. The similarity in shape of the UV pulse profile (Fig. 3) with that seen at X-ray energies has been discussed in Sec. 3.1. In theory, the UV pulse profile can differ significantly from the X-ray pulse profile because the hemisphere of the primary may intercept a different (or at least, larger) fraction of the X-ray beam from the pulsar than the Earth. For example, the UV pulse might show two peaks per X-ray pulse period (representing X-rays from both magnetic poles), while the X-ray pulse profile shows only one; or the relative intensity and shape of the two peaks might be different in the UV and X-ray. This does not seem to be the case for 4U0900-40, where the Earth appears to intercept the X-ray pulse from both magnetic poles. The pulsed fraction of 4U0900-40 in X-rays is dependent on energy, reaching a value as large as 0.59 + 0.14 between 21 and 33 keV (Dolan et al. 1981). Chester (1979) estimates the pulsed fraction in the continuum around 4000 A from HD77581 at 0.2% - 0.001%. As discussed below, the pulsed - 20 - fraction in the emission lines in the far UV, where we observed, should be larger than that in the continuum in the visible. The UV pulsed fraction we see (1.2 + 0.6 %) thus seems consistent with the original theoretical estimate. 4.1.3. _U_V _P_u_l_s_e _E_c_l_i_p_s_e. If the X-ray eclipse is caused by the geometric occultation of the pulsar by the primary (cf. Dolan et al. 1981), then no X-ray illuminated part of the primary should be visible from Earth during X-ray eclipse. The UV pulses we observed disappeared at the phase of immersion of the NS into eclipse within the uncertainty of the orbital ephemeris. The disappearance of UV pulses may not be related to the X-ray eclipse, however. The (3 - 9) keV pulses from 4U0900-40 have been observed to disappear (i.e., the counting rate dropped to < 0.1 that of the previous 15 minutes) over a timescale < 1 minute during observations with the TENMA satellite (Inoue et al. 1984) at ~ 0.68. The X-ray pulses did not reappear over at least the next 18 minutes. They were present at their normal intensity when 4U0900-40 was re-observed an hour later. This occurrence was not related to the geometric eclipse of the X-ray source by the primary. The disappearance of the UV pulsations after the first 600 s of our observation is consistent with an X-ray eclipse, but the occultation of the X-ray source is not the only mechanism which can cause the disappearance of reprocessed X-ray pulses. - 21 - 4.2. Origin of the UV Pulses Davidsen, Margon, & Middleditch (1975) attributed the optical pulses they detected from HZ Her at the 1.24 s X-ray pulse period to pulsed emission from emission lines such as He II 4686 and N III 4640. The amplitudes of pulsation in these lines would have had to exceed 25% to reproduce the pulsed fraction they observed in their bandpass. In theory (Chester 1979), most of the X-ray energy is absorbed at the top of the primary's atmosphere (i.e., at small continuum optical depth) by photoelectric absorption. Radiative recombination to the original ionization state of the absorbing atom will then produce line emission on a time scale much less than 1 s (Davidsen, Margon, & Middleditch 1975; Chester 1978). Chester (1979), however, calculated the pulsed fraction around 4000 A only in the continuum because Nelson, Chanan, & Middleditch (1977) reported that the pulses from HZ Her have the same spectrum as the unpulsed spectrum and arise primarily from heating of the stellar surface, not from emission lines. The mechanism generating pulses in the far UV in 4U0900-40 may be different from that generating continuum radiation pulses in the visible in Her X-1, however. Several strong resonance lines of abundant metals occur in the F145 M bandpass, and these emission lines can act as efficient radiators of any temporary input of excess energy. We examined 49 archival IUE spectra of HD77581 which included the 1350 - 1550 A region of its spectrum taken at many different phases and epochs (cf. van Kerkwijk et al. 1995b). - 22 - 8 of these spectra were taken at different phases during two consecutive orbits (Sadakane et al. 1985). Strong resonance lines of Si IV (1394 and 1403 A), C IV (1551 A), O V (1371 A) and C II (1335 A) lie within (or in the wings of) the F145M bandpass. The ionization potential (IP) of these ions ranges from 24 to 113 eV. Their profiles are P Cygni in nature, indicating that a major part of the emission arises in the wind associated with the primary. INSERT DISCUSSION OF PULSED FLUX RATIOS HERE The wide range of IP's among the species producing the resonance lines in this bandpass, and the effectiveness of resonance line radiation from metals in cooling a heated gas, implies that reprocessed UV line radiation in this bandpass will be prominent at every phase at which the X-ray pulsar illuminates a significant part of the primary's visible hemisphere. We also note that reprocessed pulses were seen in the optical from HZ Her at the same orbital phase at which we detected UV pulses from 4U0900-40 (Middleditch & Nelson 1976). - 24 - 4.3. X-Ray Transients No UV signal with a period near that of the X-ray pulses was found in our observations of the X-ray transients A1118-61, A0535+26, and 4U1145-619. Because these systems are usually modeled as a NS + Be star binary with an eccentric orbit, their X-ray outbursts should occur near periastron. The lack of UV pulses from the optical star is then naturally explained by the low X-ray luminosity at the orbital phases at which we observed (Table 1). The one exception may be A1118-61, which was observed at = 0.05 if one assumes the 17 years between its two known X-ray outbursts is an orbital period. Its X-ray luminosity may have been non-zero 12 months after the onset of the outburst (cf. Coe et al. 1994). It is only speculation, however, to link the non-significant UV signal we detected in A1118-61 with this possibility. Because reprocessed X-ray pulses are not seen over most orbital phases in these binaries, X-ray transients are poor candidates for systems in which to determine the mass of the NS by using the double-lined spectroscopic binary technique. The binary companions of non-transient X-ray pulsars are more likely to exhibit UV pulses at most orbital phases (those at which the X-ray pulses are visible at Earth) and should be more suitable candidates for investigations attempting to determine the mass range of NS. (A list of X-ray pulsar systems is given by Nagase (1989), but note the misprint in the HD number of the optical counterpart of 4U1145-619 in his Table 1.) - 25 - 5. Conclusions We have detected reprocessed X-ray pulses from the 4U0900-40 system in the far UV. The 1350 - 1550 A bandpass includes resonance lines of several different metals having a wide range of ionization potentials. The energy absorbed from the X-ray pulses is re-radiated primarily in the resonance emission lines of metals. The pulsed emission in the UV has a larger pulsed fraction than the continuum emission in the visible which results from the heating of the stellar photosphere. UV pulses should be visible at all orbital phases outside of X-ray eclipse. The determination of NS masses in X-ray binary systems using the Doppler shift of reprocessed pulses in the UV appears to be feasible. Acknowledgements: The HSP Investigation Definition Team is supported by NASA contract NAG 5-1613 to the University of Wisconsin. RJH acknowledges support from NASA contract NAS 5-32350 to Hughes STX Corp. PTB acknowledges support from NASA contract NAS 5-32484 to USRA. - 26 - REFERENCES Alme, M. L., Wilson, J. R., 1974, ApJ, 194, 147 Avni, Y., Bahcall, J. N., 1974, ApJ, 191,221 Basko, M. M., Sunyaev, R. A., 1973, Ap&SS, 23, 117 Bianchi, L., Bernacca, P. L., 1980, A&A, 89, 214 Bless, R. C., Percival, J. W., Walter, L. E., White, R. L., 1992, High Speed Photometer Handbook, STScI, Baltimore Bless, R. C., Richards, E. E., Percival, J. W., et al., 1996, in preparation Boyd, P. T., Carter, P. H., Gilmore, R., Dolan, J. F., 1995, ApJ, 445, 861 Boynton, P. E., Deeter, J. E., Lamb, F. K., Zylstra, G., 1986, ApJ, 307, 545 Brandt, S., Castro-Tirado, A. J., Lund, N., et al., 1993, A&AS, 97, 257 Chester, T. J., 1978, ApJ, 222, 652 Chester, T. J., 1979, ApJ, 227, 569 Coe, M. J., Payne, B. J., 1985, Ap&SS, 109, 175 Coe, M. J., Roche, P., Everall, C., et al., 1994, A&A, 289, 784 Cook, M. C., Warwick, R. S., 1987, MNRAS, 225, 369 Dahab, R. E., 1974, ApJ, 187, 351 Davidsen, A., Henry, J. P., Middleditch, J., Smith, H., 1972, ApJ, 177, L97 Davidsen, A., Margon, B., Middleditch, J., 1975, ApJ, 198, 653 - 27 - Deeter, J. E., Boynton, P. E., Lamb, F. K., Zylstra, G., 1987, ApJ, 314, 634 DeJager, O. C., Raubenheimer, B. C., North, A. R., Nel, H. I., Van Urk, G., 1988, ApJ, 329, 831 Dolan, J. F., Ellison, D. C., Crannell, C. J., et al., 1981, ApJ, 250, 335 Dolan, J. F., Tapia, S., 1988, A&A, 202, 124 Dolan, J. F., Boyd, P. T., Wolinski, K. G., et al., 1994, ApJ, 432, 560 Groth, E. J., 1974, ApJ, 192, 517 Hammerschlag-Hensberger, G., et al., 1980, A&A, 85, 119 Hutchings, J. B., Crampton, B., Cowley, A. P., 1981, AJ, 86, 871 Hyland, A. R., Mould, J. R., 1973, ApJ, 186, 993 Inoue, H., Ogawara, Y., Ohashi, T., et al., 1984, PASJ, 36, 709 Ives, J. C., Sanford, P. W., Bell-Burnell, S. J., 1975, Nature, 254, 578 Janot-Pacheco, E., Ilovaisky, S. A., Chevalier, C., 1981, A&A, 99, 271 Lutovinov, A. A., Grebinev, S. A., Syunyaev, R. A., Paulinskii, M. N., 1994, Astron. Lett. (AZh), 20, 538 Mardia, K. V., 1972, Statistics of Directional Data, Academic Press, New York, p. 133 Margon, B., Nelson, J., Chanan, G., Thorstensen, J. R., Bowyer, S., 1977, ApJ, 216, 811 Middleditch, J., Nelson, J., 1976, ApJ, 208, 567 - 28 - Motch, C., Janot-Pacheco, E., Pakull, M. W., Mouchet, M. 1988, A&A, 201, 63 Nagase, F., Hayakawa, S., Kunieda, H., et al., 1984, ApJ, 280, 259 Nagase, F., 1989, PASJ, 41, 1 Nelson, J. E., Chanan, G. A., Middleditch, J., 1977, ApJ, 212, 215 Nelson, J., Cordova, F., Middleditch, J., 1979, ApJ, 229, 294 Parkes, G. E., Murdin, P. G., Mason, K. O., 1978, MNRAS, 184, 73P Parkes, G. E., Murdin, P. G., Mason, K. O., 1980, MNRAS, 190, 537 Payne, B. J., Coe, M. J., 1987, MNRAS, 225, 985 Percival, J. W., Boyd, P. T., Biggs, J. D., et al., 1995, ApJ, 446, 832 Priedhorsky, W. C., Terrell, J., 1983, Nature, 303, 681 Rosenberg, F. D., Eyles, C. J., Skinner, G. K, Willmore, A. P., 1975, Nature, 256, 628 Sadakane, K., Hirata, R., Jugaku, J., et al., 1985, ApJ, 288, 284 Sato, N., Hayakawa, S., Nagase, F., et al., 1986, PASJ, 38, 731 Staubert, R., Kendziorra, E., Pietsch, W., et al., 1980, ApJ, 239, 1010 Steiner, J. E., 1977, A&A, 61, L35 Taylor, M., Nelson, M. J., Bless, R. C., et al., 1993, ApJ, 413, L125 - 29 - Thomas, R. M., Murdin, P., Morton, D. C., 1981, MNRAS, 195, 915 van Kerkwijk, M. H., van Paradijs, J., Zuiderwijk, E. J., 1995a, A&A, 303, 497 van Kerkwijk, M. H., van Paradijs, J., Zuiderwijk, E. J., et al., 1995b, A&A, 303, 483 Wade, R. A., Oke, J. B., 1977, ApJ, 215, 568 Wilson, R. B., Finger, M.H., Gibby, L., Fishman, G. J., 1993, IAU Circ. 5833 Wilson, R. B., Finger, M. H., Stollberg, M., et al., 1994, IAU Circ. 5955 - 30 - Table 1 Journal of Observations System Start Duration F145M Mean Binary JD 2,440,000+ (s) Counting Rate Phase(*) -1 (s ) A0535+26 8993.1709 2617 146 + 0.2 0.34 4U0900-40 8993.7815 2880 714 0.5 0.921 - 0.926 A1118-61 8986.0671 2880 3.1 0.03 0.05: 4U1145-619 9207.8250 2880 1061 0.6 0.84 * See text for explanation. - 31 - Table 2 Flux Density in the F145M Bandpass Star X-ray m Spectral E(B - V) Flux Density Pulsar V Type (mJy) HD245770 A0535+26 9.1 B0Ve 0.8(a) 53 + 1 HD77581 4U0900-40 6.7 B0.5Ib 0.76(b) 261 6 He3-640 A1118-61 12.1 O9.5IV-Ve 0.9(c) 1.15 0.03 HD102567 4U1145-619 9.2 B1Ve 0.45(d) 387 8 - 1E1145.1-6141 13.1 B2Iae 2.0:(e) < 0.003 - GX304-1 14.7 B2Vne 1.7(f) < 0.003 QV Nor 4U1538-52 14.4 B0Ie 2.4(g) < 0.004 (a) Wade & Oke 1985 (b) Hyland & Mould 1973 (c) Coe & Payne 1985 (d) Hammerschlag-Hensberge et al. 1980 (e) Hutchings, Crampton, & Cowley 1981 (f) Parkes, Murdin & Mason 1980 (g) Parkes, Murdin & Mason 1978 - 32 - FIGURE CAPTIONS Fig. 1. The Gabor transform of the 4U0900-40 counting rate in the F145M bandpass using a 600 s wide time-window. Contours of equal power are shown, The observation was 2,880 s long; the data was padded to 4,096 s with random numbers (gaussian white noise) having the same mean and variance as the data, so the data set extends between 608 s and 3488 s on the time axis. The time evolution of the 41 integral frequencies between 15 and 55 cycles per 4,096 s (p = 273 s to 74 s) is shown. The strong signal in the first 600 s of data (608 s to 1208 s on the time axis) peaks at higher harmonics of the 283 s X-ray period (i.e., around 29 and 43 cycles per 4,096 s) because the pulse profile consists of two different intensity peaks per NS rotation period. Fig. 2. The Auto-correlation Function (ACF) of the counting rate of 4U0900-40 in the F145M bandpass during the first 600 s of our observation. The ACF corresponding to the first 400 lag times at the 1 s sample rate is shown. The maximum at 281 s lag time is the largest excursion of the ACF from its mean and corresponds to a periodic signal being present in the data at the 5 level of significance. This maximum is associated with a broad feature surrounding it in lag time which indicates the pulse is many tens of seconds wide. Fig. 3. The counting rate of 4U0900-40 in the F145M - 33 - bandpass during the first 600 s of our observations, folded modulo the 283.295 s period found using the Rayleigh statistic. The +1 uncertainty on the counting rates is shown for three typical points. The starting time of our observations was arbitrarily assigned pulse phase 0.95. The data was pre-whitened to remove a long-term trend in the counting rate caused by heating effects on the secondary truss of HST. The UV pulse profile closely resembles the pulse profile seen in X-rays. The data is repeated twice for clarity. Fig. 4. The ratio of the integrated flux at the center of the Si IV 1403 line P Cyg absorption profile between 1395 A and 1405 A, relative to the flux during X-ray eclipse. The dispersion in values of the ratio at a single phase is representative of the uncertainty on a single measurement.