|Annu. Rev. Astron. Astrophys. 1992. 30:
Copyright © 1992 by Annual Reviews Inc. All rights reserved
5.2 A New Class of Sources
The simplest interpretation of the data obtained so far is that the XRB is due to some new class of sources, or at least a new face on an already known class. One important fact has to be kept in mind: The spectral break at 40 keV can indeed be reproduced by a suitable superposition of broken power Jaws or any other spectra, but it would be surprising if it has no relation to something more fundamental. However, it might be just a second cosmic conspiracy in the origin of the XRB.
The flatness of the 3-10 keV spectrum of the XRB, with an energy index there of less than 0.5, is incompatible with emission from an optically-thin source of electrons which cool by the emission of the observed radiation. Either the electrons do not cool (as in the hot IGM model) or the source is optically thick, either to electron scattering (so Comptonization is important), or to photoelectric absorption. Note again that there is no observed class of extragalactic X-ray sources that has the spectrum of the XRB.
One further constraint on the properties of any source that makes most of the residual background is the smoothness of the residual XRB. ACF results require that the XRB originate in objects less correlated than clusters or quasars. When the effects due to the contribution of those objects are removed, the residual intensity must be significantly smoother.
A strong candidate has been the starburst galaxy (Bookbinder et al 1980, Stewart et al 1982, Fabbiano et al 1982, Weedman 1987, Griffiths & Padovani 1990, Helfand et al 1989). Supernovae or X-ray binaries in young, star-forming galaxies might produce large amounts of X-ray emission. Support for this view has been obtained from the high ratio of soft X-ray to optical luminosity of some nearby starburst galaxies, particularly where the metal abundance is low (e.g. Stewart et al 1982).
The main problems with the starburst hypothesis have been the uncertain evolution (see De Zotti et al 1989 for estimates; IRAS counts suggest steep evolution) and the apparent softness of the spectrum of at least one promising starburst galaxy (NGC 5408; G.C. Stewart, private communication). A further problem might be the lack of any obvious population of starburst candidates in the deep ROSAT images [this statement is based on the high proportion of the objects identified by Shanks et al (1991) with quasars; they identify very few sources with galaxies]. Jahoda et al (1991) obtained a strong limit on the local X-ray emissivity of the Universe from their cross-correlation study of X-ray surface brightness and bright galaxies, which does not allow a significant component of the XRB to arise from starbursts (or anything else) at the current epoch.
Perhaps the best candidate has been an earlier phase in the life of AGN. Leiter & Boldt (1982) suggested that a high accretion phase when the central engines were rapidly gaining mass might lead to qualitatively different X-ray spectra than that observed now from AGN which might be past their prime (in the sense of the ratio of the actual accretion rate to that corresponding to the Eddington luminosity). Zdziarski (1988) has produced a specific model which accounts for the overall spectrum in a physical way.
Alternatively, the 40 keV break can be seen as an important feature that any successful model should generate in a very basic manner. It should not be due to some arbitrary combination of, say, optical depth and temperature, but should involve some basic physics. Only then will many sources - presumably with different luminosities, redshifts, etc - create a relatively sharp break in the observed XRB spectrum. Some energy above 40 keV (when the source redshifts are taken into account) must therefore be identified. The electron rest mass is probably the relevant source, but not through the production of electron-positron pairs unless a redshift of 20 (with very little dispersion) for the sources is acceptable (Fabian et al 1988). A simpler way of involving the electron mass is through Compton recoil. A continuous hard X-ray spectrum incident onto a thick slab of matter gives rise to a ``reflection'' spectrum (i.e. the spectrum scattered back) which has a break at about 150 keV due to Compton recoil (Lightman & White 1988). A hard spectrum is produced in many AGN and there is a large slab of matter there in the form of the accretion disc. Consequently, a reflection spectrum of the right shape is expected from AGN at z ~ 2 (Fabian et al 1990; see also Rogers & Field 1991a, Terasawa 1991).
The reflection spectrum is now observed in the X-ray spectra of many Seyfert galaxies and the geometry is confirmed by the accompanying iron fluorescence line (Pounds et al 1990, Matsuoka et al 1990). It appears to be the cause of the flattening of the spectra of AGN above 10 keV. Photoelectric absorption in the reflection slab causes there to be few photons reflected at lower energies. (To some extent, the reflection model for the XRB is here similar to the absorption models discussed in Section 4.) The main problem with the model is that it requires less than about 10% of the direct hard spectrum to be visible. (The fraction depends on the index of the underlying power law.) This may be due to the geometry of the central engine or to anisotropic emission of the hard X-ray source. The X-ray emission may, for instance, originate from some funnel within a thick accretion disk (e.g. Madau 1988), or the emission may be anisotropic due to the position of the hard X-ray source above a soft X-ray emitting disk. The inverse Compton process then causes much of the energy of the hard X-ray emitting electrons to be scattered back onto the disk (Ghisellini et al 1991).
The reflection mechanism is inefficient in producing the XRB and the model predicts a large extreme ultraviolet background, since much of the hard X-ray emission is absorbed by the surrounding gas and is reradiated as quasi-thermal radiation, The relatively small extent of the so-called inverse effect in the QSO Lyman forests (i.e. the relative lack of Lyman absorption lines near the background QSO) and the lack of intergalactic neutral hydrogen also indicate the existence of such a large background (e.g. Steidel & Sargent 1989).
The 3-100 keV XRB in this model is predominately due to young AGN at z ~ 1-3. The Ginga fluctuation limit then requires a surface density of 300 sources per deg2, which is comparable to the integrated surface density of all AGN (about 1% of all galaxies). The X-ray luminosity of the individual sources is then ~ 3 x 1044 erg s-1.
Perhaps the most promising feature of reflection models is that there are no free parameters [unless the -ray background has to be explained with the same sources (Rogers & Field 1991b)] and that the 40 keV break (redshifted to z ~ 2) is intrinsic to the model. The major problem is, however, the high covering fraction of the cold reflecting material that is needed.