Broad-Band X-Ray Spectroscopy of Cen X-3

     The high mass X-ray binary pulsar Centaurus X-3 was observed outside of eclipse by RXTE between September 10 and September 27 1996, with pointings during five different binary orbits. We present some preliminary results from our analysis. We observed the source's intensity drop by a factor of 40, accompanied by changes in the pulse profile. We find a possible broad cyclotron scattering resonance feature between 20 and 30 keV.
    

Centaurus X-3 is an X-ray binary pulsar with a pulse period of 4.8 s. The neutron star orbits its O-type companion every 2.09 days. Gradual X-Ray eclipses occur between orbital phases -0.2 and 0.2 (with phase 0 at mid-eclipse). The X-ray spectrum of Cen X-3 in the 1 to 10 keV range can be modeled by a power law with iron K emission and interstellar absorption. The fluorescent iron emission feature between 6 and 7 keV has been found to pulsate with an amplitude ~ 50% of the mean intensity (Day et al. 1993). The iron line has been resolved into components at 6.4, 6.7, and 6.9 keV (Audley et al. 1996; Ebisawa et al. 1996). The 6.4 keV line is due to fluorescence of iron in low ionization stages while the others originate in the extended stellar wind of the companion (Nagase et al. 1992).
     Above about 10 keV the spectrum falls off faster. This high energy roll off may be approximated by the factor exp(-[(E-E_co)/(E_f)]) (White, Swank & Holt 1983) with a cut off energy, E_co of ~ 15 keV and an e-folding energy E_f of ~ 8 keV. Nagase et al. (1992) obtained a better fit with a roll off of the form exp(-[(t(WE/E_c)²)/((E-E_c)²+W²)]). This form is similar to that of opacity due to cyclotron scattering where E_c is the cyclotron resonance energy, t is the optical depth, and W is the width of the resonance. They estimated a resonance energy of 30±2 keV from the shape of the roll off. This is at the upper end of Ginga's bandpass so the identification of the roll off with cyclotron scattering opacity is not conclusive.
    

The pulse shape is variable. We see the usual asymmetric single peak during orbit #1 when the source is in its high state (eg. Audley et al. 1996). There is also a less prominent subsidiary peak which is separated from the main peak by about 180^°. The subsidiary peak has a softer spectrum than the main peak. The 5-8 keV flux is strongly correlated with the subsidiary peak but not with the main peak. This is true for all of the pointings. If we regard the 6.4 keV fluorescent line as a tracer of X-ray reflection this means that the subsidiary peak contains a component due to reflection of X-rays by matter within a few light seconds of the pulsar while the main peak is mainly due to emission from one of the pulsar's magnetic poles. As the luminosity decreases we see the profile become more double-peaked, resembling the structure observed by Nagase et al. (1992). The main and subsidiary peaks also appear to change places.
    

This behavior is similar to that observed in EXO 2030+375 (Parmar, White, & Stella 1989). Parmar et al. attributed this luminosity dependence of the pulse profile to intensity dependent changes in the structure of the accretion column. They suggested that as the intensity decreased the beaming pattern changed from a fan beam with emission normal to the magnetic field to a pencil beam with emission along the magnetic field.
     This may be understood in terms of a model in which infalling matter is decelerated in a shock above the neutron star's surface. The height of this shock increases with luminosity resulting in a narrow column atmosphere for high luminosities. This column radiates from its sides in a fan beam pattern. At lower luminosities accreting matter is decelerated close to the neutron star and the atmoshpere is a thin slab which radiates predominantly along the magnetic field giving rise to a pencil beam emission pattern.
    

The best fit to the PCA data with a simple continuum model is obtained from a cut off power law with a fluorescent iron line and interstellar absorption (which is hardly constrained by the PCA). When a CSRF is added c²_n goes from 9.64 for 72 degrees of freedom to 5.37 for 69 degrees of freedom. A better fit to the PCA data is obtained with Mihara's NPEX continuum model (Mihara 1995). This has the functional form I(E^-a₁+E²)exp(-[E/kT]) and approximates unsaturated Comptonization. Using this continuum model with a CSRF gives c²_n=3.35 for 69 degrees of freedom, while c²_n=16.6 without the CSRF. Thus the fit to the PCA data is significantly improved by the addition of a CSRF between 20 and 30 keV. A similar feature is also required by the HEXTE data.
    

However there are some caveats. The PCA fits are not formally acceptable. We believe that this is because of systematic effects. The 90% errors quoted for the CSRF parameters on the PCA figure were obtained after adding a 0.5% fractional systematic error to the model. The Xenon edge around 35 keV may be affecting the results. The HEXTE fits are formally acceptable with c²_n close to 1 and appear to be consistent with the presence of a first harmonic around 50 keV. However HEXTE's effective area also has a sharp feature between 30 and 40 keV. The CSRF parameters obtained by fitting the HEXTE cluster 0 data between 15 and 45 keV with the same continuum model are E_c=23.0±1.5 keV, t=1.1±0.4, and W=2.4±1.3. These differ slightly from the simultaneous PCA values. With a power law continuum in the energy range 15-120 keV the best-fit values are E_c=22.1±0.6 keV, t=0.42±0.07, and W=7±3. This model includes a harmonic at 2E_c with t₁=2.0±0.7, and W₁=5±2.
     If the apparent absorption feature around 24 keV is in fact due to cyclotron opacity this would imply that the magnetic field is about 2×10¹² G. Audley et al (1996) used the beat frequency mass accretion model of the 40 mHz quasi-periodic oscillations to predict a surface magnetic field of about 3.4×10¹² G. If the scattering region extends far above the neutron star surface the field derived from a CSRF would be lower than this.
    

We have observed the intensity of Cen X-3 vary by a factor of 40 in a few days.

     We have seen variations in the pulse profile which may be caused by luminosity-induced changes in the structure of the accretion column.

     There is a possible cyclotron scattering resonance feature between 20 and 30 keV. This result may change as better PCA responses become available.
    

     Audley, M. D., Kelley, R. L., Boldt, E. A., Jahoda, K. M., Marshall, F. E., Petre, R., Serlemitsos, P. J., Smale, A. P., Swank, J. H., & Weaver, K. A. 1996, ApJ, 457, 397

     Day, C. S. R., Nagase, F., Asai, K., & Takeshima, T. 1993, ApJ, 408, 656

     Ebisawa, K, Day, C. S. R., Kallman, T. R., Nagase, F., Kotani, T., Kawashima, K., Kitamoto, S., and Woo, J. W. 1996, PASJ, 48, 425

     Mihara, T. 1995b, Ph.D. Dissertation, University of Tokyo

     Nagase, F., Corbet, R. H. D., Day, C. S. R., Inoue, H., Takeshima, T., Yoshida, K., & Mihara, T. 1992, ApJ, 396, 147

     Parmar, A. N., White, N. E., & Stella, L. 1989, ApJ, 338, 373

[an error occurred while processing this directive]