4.1. Origin of the X-ray emission in NGC 3923
The spectral analysis over the (0.5-10) keV band showed that the superposition of two thermal components of keV, and keV, is the most reasonable representation of the spectral data. Whether the softer component is actually a multi-temperature component, such as a cooling flow, cannot be investigated with these data. Another caveat concerns the hard component, because, as seen in Sect. 3.1, BC detected a few point sources over the ROSAT PSPC image of NGC 3923. Were all these sources background AGNs, they could harden the spectrum of NGC 3923, and alter the estimate of the hard flux from NGC 3923. Given the number counts involved (see Sect. 3.1), the effects on the derived spectral parameters are expected to be within their uncertainties; in particular, there could be a spuriuos increase in the hard flux of %.
The value of suggests as origin of the soft emission a hot gas, that comes from the accumulation of the stellar mass loss during the galaxy lifetime. The value of is close to that found for the hard emission of the low / galaxy NGC 4382, based on ASCA data (i.e., keV; Kim et al. 1996). Following Fabbiano et al. (1992, 1994), these authors suggest for the origin of the hard component the integrated emission of LMXBs, an interpretation similar to that given for the X-ray emission of the bulges of early-type spirals (see Sect. 1). In fact the spectra of LMXBs can be described by a thermal bremsstrahlung model with keV (van Paradijs 1998), the spectrum of M31 can be fitted by bremsstrahlung emission at a temperature keV (Fabbiano et al. 1987), and those of bulge-dominated spirals can be fitted with bremsstrahlung at keV (Makishima et al. 1989). We now examine more in detail the suggested origins for the two components of the X-ray emission of NGC 3923.
4.1.1. Origin of the soft component
How does compare with the possible average temperature of a gas flow in NGC 3923? A simple estimate of the kinetic temperature of the stars in NGC 3923 gives keV, when using the central stellar velocity dispersion in Table 1, and . This is close to found by the modeling, i.e., =0.39-0.45 keV, with some room left for other heating mechanisms in addition to the thermalization of the stellar random motions. These are especially needed when considering that 0.36 keV is the central temperature of the stars, and that this decreases outward, while would be the emission-weighted temperature of the hot gas. Possible heating mechanisms are supernova heating (from type Ia supernovae, hereafter SNIa) and/or compressional heating operated by the gravitational field 4. Compressional heating is produced when the hot ISM is in a global inflow. The amount of the emission in the soft component ( erg s-1) is actually quite lower than that typical of a global inflow, in a galaxy of an optical luminosity as high as that of NGC 3923. For example, when , of many times erg s-1 in the (0.5-4.0) keV band, and hot gas masses 5 of a few times , are predicted by the steady state cooling flow models, for various combinations of dark matter and SNIa rates (e.g., Sarazin & White 1987, Bertin & Toniazzo 1995). Besides eliminating SNIa's, various kinds of reductions of the high values in the framework of the cooling flow scenario have been suggested: by assuming, at fixed , reductions in the stellar mass loss rate, or in the efficiency of its thermalization, or a higher efficiency of thermal instabilities in the hot gas (Sarazin & Ashe 1989, Bertin & Toniazzo 1995), or by including the effect of the rotation of the galaxy (Brighenti & Mathews 1996), or of a lower abundance (Irwin & Sarazin 1998). None of these effects has proved to be able to reduce by a large factor. up to erg s-1, and gas temperatures higher than the stellar kinetic temperatures, are instead predicted for a galaxy like NGC 3923 from a gas that is experiencing an outflow or a partial wind; these can be driven by SNIa explosions at a rate comparable to the most recent optical estimates of Cappellaro et al. (1997) (Pellegrini & Fabbiano 1994, Pellegrini & Ciotti 1998). In addition, within this framework, global energy considerations and two-dimensional simulations showed that in general the flattening of the galaxy favors the loss of gas, while rotation has a minor role (Ciotti & Pellegrini 1996; D'Ercole & Ciotti 1998). This could be an explanation for the absence of a global inflow in NGC 3923, which is a considerably flat galaxy with no rotation (Sect. 2). A detailed modeling of the structure of NGC 3923, and hydrodynamical simulations of the hot gas behavior specific for this galaxy, are required for a definite answer about the gas flow state. There is a potential problem with a scenario that involves a substantial heating from SNIa's. The abundance of the soft component is not constrained by the BeppoSAX data for NGC 3923, but is extremely low at the best fit. Low abundances for the hot gas are not expected in presence of SNIa explosions, yet they have almost always resulted also from the analysis of ASCA data; there seems even to be a trend of decreasing abundances with decreasing / (Matsushita 1998). An unsolved puzzle is represented at present by the discrepancy between the low hot gas abundances and the abundances in the stellar mass loss which feeds the gas (but this discrepancy is narrowing for an increasing number of galaxies, after accurate re-analyses; Matsushita 1998, BF, Buote 1998), eventually further increased by the metals in the ejecta of SNIa's, which are seen to explode in E/S0s. Various solutions have been suggested (Arimoto et al. 1997, Fujita et al. 1996), but none has been recognized as the final one yet. We note that in NGC 3923 the stellar central iron abundance is [Fe/H]=0.2, or 1.6 solar, and that the mean abundance is likely a factor of 2 lower (the central Mg2 value, from Faber et al. 1989, has been converted into central and average stellar iron abundance following the detailed prescriptions of Arimoto et al. 1997). If we adopt the abundance value of for the hot gas found by BF (Sect. 3.2.2), in this galaxy there is room for enrichment by SNIa's exploding at the rate of Cappellaro et al. (1997).
Another possibility to explain the gas mass content of NGC 3923 is that a substantial amount of hot ISM was lost as a consequence of the episod of interaction or merger which is at the origin of the system of shells shown in the optical. Actually, the very detailed modeling of the shell formation that has been made for this galaxy, plus observations of the galaxy colors and ISM content at other wavelengths (see Sect. 2), established that the interaction or merger involved a small galaxy, devoid of gas, and that significative star formation (in the form of a starburst with supernova explosions that could have heated the gas) did not take place. It could be, though, that the gas flow is very sensitive to perturbations in the potential, and that even small perturbations can help a significant portion of the hot ISM to escape the galaxy. Numerical simulations are needed to test whether this was the case for NGC 3923.
4.1.2. Origin of the hard component
is in good agreement with that of bulge-dominated spirals; what about the amount of the hard emission in NGC 3923? CFT had estimated the luminosity of the integrated contribution of LMXBs in the (0.5-4.5) keV band () by scaling it from the emission of the bulge of M31. By assuming a linear relation with , they had obtained log = 29.6 + log , where is in , and had estimated for any given galaxy to scatter by about a factor of 3 about this relation, since this is the observed scatter in the X-ray to optical luminosity ratio for subclasses of spiral galaxies. This relation has been recently confirmed (both in shape and normalization) using ASCA data by Matsumoto et al. (1997). The present analysis gives, in the (0.5-4.5) keV band, log , i.e., is close to the value predicted for by CFT (it is just 12% higher, so well within the quoted uncertainties). An interpretation in terms of stellar sources of the hard emission can be judged also by inspecting Fig. 2, where the MECS surface brightness profile is compared with the distribution of optical light. The V-band profile of NGC 3923 has been derived from Kodaira et al. (1990); since this extends out to a radius of , it has been extrapolated out to the radius of the X-ray emission with a fit. The optical profile has then been convolved with the MECS response, appropriate for the distribution of the counts in the different energy channels. A good agreement between the profile over (1.7-10) keV and the convolved optical profile is found, which would support the hypothesis of the origin of the hard emission in stellar sources. We note here that, consistently with the finding of an amount of hard emission 12% larger than the predicted , and with the estimate of a possible spurious contribution up to 14% of the MECS counts from hard foreground/background sources, some excess of hard emission is also shown by the X-ray profile, with respect to the convolved stellar profile, at radii , i.e., for (the excess at is likely due to a foreground/background source, see Sect. 3.1). We cannot give a great significance to the detailed shape of the X-ray profile, because of the MECS moderate spatial resolution (Sect. 1); we just note that this hard excess cannot be produced by hot gas, neither belonging to the galaxy (at =0.4 keV), nor to a possible intragroup medium, because the low value of the velocity dispersion of the group (Sect. 2) corresponds to a kinetic temperature that is even lower than . An X-ray profile of NGC 3923 with a superior spatial resolution has been derived in the soft ROSAT band (0.4-2) keV by BC. For radii this is in good agreement with the de Vaucouleurs extrapolation of the R-band profile within , while it is flatter than the optical one for radii . BC conclude that not all the hard emission is to be attributed to stellar sources, while some fraction of it could come from another phase of the hot gas. In line with the modeling done to interpret the ROSAT data of another galaxy which showed an X-ray profile centrally flatter than the optical one (NGC 4365; Pellegrini & Fabbiano 1994), we suggest the possibility that the hot gas has quite an extended distribution in the central regions, i.e., flatter than the optical one within 10 arcsec (consistently with what can be derived also by the BeppoSAX data, BC estimate that just 35% of the total 0.4-2 keV emission is due to a hard component; so this cannot fully determine the total shape). The hypothesis of some peculiarities in the hot gas distribution can be supported also by the consideration of the past galaxy history, where a merging occurred.
4.2. The nature of the X-ray emission in medium and low / galaxies
A fundamental diagnostic of the X-ray emission from early-type galaxies is the - plane. The large scatter in of more than two orders of magnitude at fixed shown by this plane is not an artifact of distance errors [see Pellegrini & Ciotti (1998) for a more detailed discussion]. The explanation of this scatter is a largely varying quantity of hot gas within the galaxies (e.g., Matsumoto et al. 1997), but it is still a controversial issue how these variations are established. Are they fundamentally a consequence of environmental differences, or of different dynamical phases for the hot gas flows (provided that it was not possible to reproduce the observed scatter with various adjustments to the cooling flow scenario, see Sect. 4.1.1)? The first hypothesis can affect only galaxies in clusters or groups; actually this is the case for the majority of E/S0s. Then accretion of external gas can explain the extremely X-ray bright objects (Renzini et al. 1993, Mathews & Brighenti 1998), while in the X-ray faint ones the hot gaseous halos should have been stripped by ambient gas, if it is sufficiently dense, or in encounters with other galaxies (White & Sarazin 1991). It is not clear yet whether the primary stripping agents would be other galaxies or the ambient gas. The effectiveness of the stripping by an ambient gas has been explored theoretically, and it turned out to depend largely on various factors (shape of the orbit, velocity and internal dynamics of the galaxy, density of the environment), for which the observed range is wide (e.g., Portnoy et al. 1993). Observationally, evidence of stripping by the intracluster medium is the famous plume shown by the hot halo of the Virgo elliptical M86. ROSAT though showed that the sample of early-type galaxies of the Coma cluster, that is richer than Virgo, has the same average / as that of Virgo (Dow & White 1995); but that instead the X-ray luminosities are on average lower in the rich cluster A2634 (Sakelliou & Merrifield 1998). In the Pegasus I group and in the poor cluster Cancer A, where a medium has been detected, the X-ray image shows also many clumps that could be the X-ray halos from individual galaxies with a `normal' / (Trinchieri et al. 1997). For what is concerning the interactions among galaxies, observationally there is an indication that lower / galaxies occur across the whole range of galaxy densities, while the higher / ones are mostly confined at low densities (Mackie & Fabbiano 1997). Theoretically, galaxy interactions could produce some scatter in / as follows: group-dominant ellipticals may acquire dark matter and hot gas by mergers or tidal interactions early in their evolution, and then become very X-ray bright; the other E/S0s, in which the gas is in a global inflow, may be tidally truncated in their dark matter and hot gas halo, at different radii, and so end up with different sizes and different / (Mathews & Brighenti 1998). The problems with explaining the / plane only with environmental factors are that: 1) low or medium / values are also shown by galaxies that do not reside in a high density medium [e.g., NGC 5866 (Pellegrini 1994), NGC 3923], and by galaxies that reside in a region where the galaxy density is not particularly high, and where also galaxies of high / are found (Mackie & Fabbiano 1997); 2) the effect of merging and tidal interaction on the hot gas flow (in various dynamical states) is still quite conjectural, as is the evolution of the hot gas in galaxies that undergo these phenomena. The only models available so far, those of Mathews & Brighenti (1998), are not aimed at reproducing all the / variation, down to the lowest / values observed (log /, with and in erg s-1), but stop at log /. Probably it is not possible to reproduce the lowest / values by simply truncating global inflows, because one continues to obtain galaxies quite rich in hot gas. We note here also that NGC 3923 is a group-dominant elliptical, but does not show a very large hot gas content (log ).
The second way of explaining the scatter, through different dynamical phases of the gas flows, regulated by internal agents, has the advantage of being a general explanation, i.e., of applying to all the galaxies, regardless of their environment (accretion is always needed for the extremely X-ray bright galaxies). At fixed , any of the flow phases 1 ranging from winds to subsonic outflows to partial and global inflows, can be found at the present epoch, depending on the various depths and shapes of the potential well of the galaxies (Ciotti et al. 1991, Pellegrini & Ciotti 1998). In this way the large scatter in is easily accounted for: in the X-ray bright galaxies the soft X-ray emitting gas dominates the emission, being in the inflow phase, that resembles a cooling flow; in the X-ray faint galaxies the hard stellar emission dominates, these being in the wind phase; in intermediate / galaxies, the hot gas is in the outflow or partial wind phase, and the amount of soft emission varies from being comparable to that of the stars, to being dominating. In this scenario a crucial role is played by the SNIa explosions, that heat the flow in an extent sometimes large enough to drive all or part of the gas out of the galaxies, and by the evolution of the explosion rate 6. Numerical simulations using the updated rate given recently by Cappellaro et al. 1997, which is reduced with respect to that used by Ciotti et al. (1991), show that the partial wind phase is the most frequent, and that a large scatter in the stagnation radius corresponds to a large scatter in the amount of hot gas (Pellegrini & Ciotti 1998). The problem with this scenario is represented by the puzzle of the extremely low hot gas iron abundances revealed by ASCA (Sect. 4.1.1).
© European Southern Observatory (ESO) 1999
Online publication: March 1, 1999