2.2. Galactic -ray Continuum Emission

Confusion and point sources

The analysis of the galactic diffuse emission can be seriously complicated by unresolved galactic point sources which may have a sky distribution similar to that of interstellar gas. Because of six objects already detected, pulsars are the most likely input from discrete sources. Many authors have addressed this problem on the basis of pulsar emission models (e.g. Yadigaroglu and Romani 1995; Sturner and Dermer 1996) and consistently estimated the contribution of pulsars to the diffuse -ray intensity, above 100 MeV integrated over the whole sky, to be a few percent. Another strategy is to base the analysis only on the observed properties of the six identified -ray pulsars, which also allows an inspection of the spectrum of the unresolved pulsars (Pohl et al. 1997). It is found that pulsars contribute mostly at -ray energies above 1 GeV, and preferrentially exactly in the Galactic plane where they can provide more than 20% of the observed emission for a reasonable number of directly observable objects.

Estimates for the contribution of discrete sources other than pulsars are very uncertain due to the lack of clear identification of -ray sources with any known population of Galactic objects. It is interesting to see that roughly ten unidentified EGRET sources can be associated with supernova remnants (SNR) or with OB associations, or with both (SNOBs) (e.g. Sturner and Dermer 1995; Esposito et al. 1996; Yadigaroglu and Romani 1997). Obviously these sources may also be radio-quiet pulsars or highly dispersed radio pulsars.

The spatial distribution of -ray emission

Observations of the Magellanic Clouds with EGRET have finally settled a long-standing debate on whether cosmic rays in the GeV energy range are Galactic or extragalactic. The -ray flux of the Large Magellanic Cloud is weakly less (Sreekumar et al. 1992), and that of the Small Magellanic Cloud is strongly less (Sreekumar et al. 1993) than expected, if cosmic ray protons were uniformly distributed in space. Therefore the bulk of the locally observed protons at GeV energies must be Galactic, and we have to think about which Galactic accelerators are capable of producing cosmic rays with a source power of ~ 10⁴¹ erg/sec.

The spatial distribution of diffuse Galactic -rays is usually described as ``the gradient'', that is a plot of the decline of -ray emissivity per H-atom in the Galactic plane versus the galactocentric radius. This approach implicitely assumes that gas interactions (i.e. ⁰ production and bremsstrahlung) dominate over inverse Compton scattering in the Galactic disk. To investigate the -ray emission originating from ⁰-decay and bremsstrahlung, we need some prior knowledge of the distribution of interstellar gas in the Galaxy. This includes not only HI but also H₂, which is indirectly traced by CO emission lines, and HII, which is traced by H and pulsar dispersion measurements. Even in case of the directly observable atomic hydrogen we obtain only line-of-sight integrals, albeit with some kinematic information. Any deconvolution of the velocity shifts into distance is hampered by the line broadening of the contribution from individual gas clouds and by the proper motion of clouds with respect to the main rotation flow.

Different authors use different models of the 3D gas distribution in the Galaxy and thus calculate different gradients (e.g. Strong and Mattox 1996; Erlykin et al. 1996a). Detailed analysis of isolated gas clouds in the solar vicinity shows that both the -ray emissivity and the CO line flux to molecular gas mass conversion factor, X, can vary from place to place in the Galaxy (Digel et al. 1995; Digel et al. 1996; Erlykin et al. 1996b). Any comparison of gradients with the Galactic distribution of putative cosmic ray sources should therefore be made with care. It may be safe to say, however, that the cosmic ray intensity decreases somewhat from the inner Galaxy to the outer Galaxy.

The Galactic diffuse -ray spectrum at low energies

The OSSE (Purcell et al. 1996) and COMPTEL (Strong et al. 1994, 1996) instruments have provided evidence that the diffuse Galactic continuum emission extends down to photon energies below 100 keV, as shown in Figure 2. In an analysis of Galactic plane observations made with OSSE (Purcell et al. 1996), it was found that when the contribution from prominent point sources monitored during simultaneous observations with SIGMA is subtracted from the Galactic center spectrum measured with OSSE, the residual intensity is roughly constant over the central radian of the Galaxy, but is lower by a factor 4 at l 95° (Skibo et al. 1997). Estimates based on the luminosities and number-flux distributions of Galactic sources indicate that the point source contribution to the hard X-ray emission from the Galactic plane is less than 20% (Yamasaki et al. 1997; Kaneda 1997). The residual source-subtracted spectrum of this emission changes from a photon index = 1.7 at energies above 200 keV (Strong et al. 1994), to a photon index = 2.7 at lower energies (Purcell et al. 1996). Thus the soft -ray continuum from the Galactic plane is more intense than the extrapolation of the higher energy emission. Observations of the Galactic ridge in the hard X-ray range with GINGA (Yamasaki et al. 1997) and RXTE (Valinia and Marshall 1998) indicate that the soft spectrum below 200 keV extends down to about 10 keV energy, though the best spectral fit between 15 keV and 150 keV gives a photon index of = 2.3.

Figure 2. The current best estimate of the diffuse high energy continuum from the inner radian of the Galactic plane. The measured fluxes have been mul tiplied by 2.

A hadronic origin for the hard X-ray/soft -ray continuum via inverse and secondary bremsstrahlung is excluded by the stringent observational limits on the flux of nuclear -ray lines and ⁰-decay -rays from the inner Galaxy (Pohl 1998). Therefore the -ray continuum emission in this energy band is most likely electron bremsstrahlung in the interstellar medium. The power required in low energy (< 10 MeV) cosmic ray electrons to produce a given amount of bremsstrahlung is a fixed quantity that depends only on the energy spectrum of the radiating electrons and weakly on the ionization state of the interstellar medium. Attributing this power input to injection in cosmic ray electron sources, it has been estimated that, integrated over the whole Galaxy, a source power of about 4 x 10⁴¹ erg s^-1 (Skibo and Ramaty 1993) or, if the bremsstrahlung emission extends down to photon energies of 10 keV, up to ~ 10⁴³ erg sec^-1 (Skibo et al. 1996) in low energy MeV) electrons is required, to retain sufficient electrons in the face of severe Coulomb and ionization losses. This electron power exceeds the power supplied to the nuclear cosmic ray component by at least an order of magnitude. The energy losses of the required large population of low energy electrons would be more than adequate to account for the observed hydrogen ionization rate in the interstellar medium (Valinia and Marshall 1998). Proving the truly diffuse nature of the galactic continuum emission below 1 MeV is of utmost importance in pinning down the most relevant particle acceleration process and to understand the interstellar medium ecosystem.

Recently, the extension of the bremsstrahlung continuum emission to these low energies has been attributed to the existence of in-situ stochastic electron acceleration by the interstellar plasma turbulence (Schlickeiser 1997; Schlickeiser and Miller 1998), rather than to the existence of a second electron source component. This turbulence with a measured energy density of 4 x 10^-14 erg cm^-3 (Minter and Spangler 1997) is an important additional energy source of cosmic ray particles.

The -ray spectrum at high energies

The spatial and spectral distributions of the diffuse emission within 10° of the Galactic plane have recently been compared with a model calculation of this emission which is based on realistic interstellar matter, photon distributions and dynamical balance (Hunter et al. 1997). The distribution of the total intensity above 100 MeV agrees surprisingly well with the model predictions. However, at higher energies, above 1 GeV, the model systematically underpredicts the -ray intensity. If the model is scaled up by a factor 1.6, the model prediction and the observed intensity above 1 GeV agree well. This deficit can be explained neither by a possible miscalibration of EGRET, nor by spectral changes in the nucleonic ⁰-decay emission component (Mori et al. 1997), nor by unresolved point sources like pulsars (Pohl et al. 1997b).

The diffuse model deficit above 1 GeV is visible also at higher latitudes, e.g. in the plots of observed intensity versus Galactic diffuse model shown in the paper of Sreekumar et al. (1998). Uncritical use of the nominal Galactic diffuse model may therefore lead to apparent -ray excesses at higher latitudes, which then may be mistaken as evidence for a -ray halo of exotic origin.

Thus the model displays a deficit of ~ 40% of the total observed emission which depends, if at all, only weakly on location. One feature of the models is the relatively soft electron injection spectral index of s = 2.4 (Skibo 1993), which is required to account for the local electron spectrum above 50 GeV. Consequently at energies above 1 GeV, around 90% of the model intensity is due to ⁰-decay (i.e. hadronic processes) and only 10% is due to interactions of electrons.

The recent detections of non-thermal X-ray synchrotron radiation from the four supernova remnants SN1006 (Koyama et al. 1995), RX J1713.7-3946 (Koyama et al. 1997), IC443 (Keohane et al. 1997), and Cas A (Allen et al. 1997), and the subsequent detection of SN1006 at TeV energies (Tanimori et al. 1998) and flux levels according to theoretical predictions (Pohl 1996), support the hypothesis that Galactic cosmic ray electrons are accelerated predominantly in SNR. It has been shown that, if this is indeed the case, the local electron spectra above 30 GeV are variable on time scales of about 10⁵ years (Pohl and Esposito 1998). This variability stems from the Poisson fluctuations in the number of SNR in the solar vicinity within a certain time period. While the electron spectra below 10 GeV are stable, the level of fluctuation increases with electron energy, and above 100 GeV the local electron flux is more or less unpredictable.

Considering this time variability, an electron injection index of s = 2.0 is consistent with direct particle measurements if SNR are the dominant source of cosmic ray electrons. While being entirely consistent with the local electron flux, and with the radio synchrotron spectrum towards the North Galactic Pole, the leptonic contribution to the diffuse Galactic -ray emission above 1 GeV in the Galactic plane would increase to 30-48% of the total observed intensity for an injection index of s = 2.0, depending on the assumed spatial distributionof SNR and on whether some dispersion of injection spectral indices is allowed (Pohl and Esposito 1998). An electron injection index of s = 2.0 may therefore explain the bulk of the observed -ray excess over that predicted by the Hunter et al. model.