Annu. Rev. Astron. Astrophys. 2003. 41: 191-239
Copyright © 2003 by . All rights reserved

Next Contents Previous


Abundance determinations in X-ray astronomy are confused by the ever-changing abundance of the sun. In much of the X-ray literature the "photospheric" iron abundance from Anders & Grevesse (1989) is explicitly or implicitly assumed whereas the currently preferred "~meteoritic" value 3.2 × 10-5 (by number) is lower by a factor 0.69 (McWilliam 1997; Grevesse & Sauval 1999). Likewise, the solar abundance of O, S and Ar have recently changed by factors of 0.79, 1.32 and 0.69 respectively (Grevesse & Sauval 1998) and this is offered as a new option in XPEC. To facilitate comparisons between any two X-ray observations, it is essential that authors prominently list the solar abundances assumed, but this has not always been done.

Evidently, at early times the hot gas associated with E galaxies was enriched in metals by Type II supernovae (SNII) that accompanied the formation of the dominant old stellar population. At later times further enrichment has been provided by Type Ia supernovae (SNIa) and material expelled from red giant stars. For a proper cosmic census of supernova enrichment there is no substitute for rich clusters (<kT> gtapprox 3 keV) that are massive enough to retain metal-enriched, SNII-driven galactic winds (e.g. Renzini 1997; 1999). In group-centered elliptical galaxies most of the iron produced by SNII has been dispersed by a galactic wind into distant group gas or nearby intergalactic regions where the gas density is so low that it cannot be observed even with XMM (Brighenti & Mathews 1999a). Groups are not closed boxes. Chandra observations by Martin, Kobulnicky & Heckman (2002) of the dwarf starburst wind in NGC 1569 verify that these winds transport almost all of the SNII enrichment into the local intergalactic medium (Vader 1986). Early epoch winds from both dwarf and giant galaxies were undoubtedly even stronger than those in local starburst galaxies. Ram-pressure stripping may also be relevant to hot gas enrichment (e.g. Toniazzo & Schindler 2001). Finally, ongoing metal-enriched galactic outflows energized by SNIa are energetically likely in low luminosity ellipticals, LB ltapprox LB, crit.

Yields from Type II supernovae yII (expressed in Modot) vary with initial main sequence stellar mass (e.g. Hamuy 2003) and must be averaged over the initial mass function. While yields for elements that can be identified in X-ray spectra are still quite uncertain (e.g. Gibson, Loewenstein & Mushotzky 1997; Dupke & White 2000; Gastaldello & Molendi 2002), there are three broad categories: the ratio yIa / <yII> is about 10 for iron, ~ 1 for silicon and sulfur (as well as argon, calcium and nickel), and ~ 0.1 - 0.2 for oxygen and magnesium. Note that yIa / <yII> is far from uniform among the alpha elements O, Mg, Ar, Ca, S, and Si. Ideally, it should be possible to account for any observed abundance with a linear combination of SNII and SNIa yields, but current observations may be too uncertain to permit this. Abundances differ according to the quality of the observation, the relative isolation of the Kalpha or other prominent lines for a given element, the assumptions made about the distribution of gas temperature at every radius in the flow (1T, 2T, etc.), and possibly also the plasma code (Gastaldello & Molendi 2002).

In principle, abundances in the hot gas within the stellar images of E galaxies can provide critically important information about the direction and magnitude of the systematic radial flow of hot gas, rho u 4pi r2, which is also sensitive to distributed cooling. The known stellar density (r) and metallicity z*(r) define the source function for metal ejection. However, to extract this flow information one must combine uncertain observations with uncertain supernova yields and with uncertainties in the fraction of metal-enriched gas from SNIa and stellar winds that ultimately goes into the hot phase.

Nevertheless, most X-ray observers currently agree that the hot gas within and near group and cluster centered E galaxies has been enriched by SNIa (e.g. Matsumoto et al. 1996; Fukazawa et al. 1998; Dupke & White 2000; Finoguenov & Jones 2000; Buote 2000a, 2002; De Grandi & Molendi 2001; Buote et al. 2003b; Lewis, Stocke & Boute 2002). The best observed E galaxy is M87 centered in the Virgo cluster (Matsumoto et al. 1996: Finoguenov et al. 2002; Gastaldello & Molendi 2002), but the presence of a central AGN and X-ray jet in M87 may have disturbed the normal evolution of the hot gas. Similar XMM observations of somewhat lower accuracy are available within a few 100 kpc of the central E galaxies in Abell 496 (T ~ 2 - 5 keV; Tamura et al. 2001), NGC 1399 (Buote 2002) and NGC 5044 (T ~ 0.7 - 1.2 keV; Buote et al. 2003b). No XMM abundances for NGC 4472 or other normal, group-centered E galaxies have been published as of the Fall of 2002. In all galaxies observed so far the hot gas iron abundance increases toward the central elliptical, rising from ~ 0.2 - 0.4 solar at ~ 100 kpc to ~ 1 solar or more at the center, similar to the stellar abundances there. [Older very subsolar central iron abundances determined with ASCA are now known to be wrong because the radial temperature variation was neglected (the "Fe bias": Trinchieri et al. 1994; Buote & Fabian 1998; Buote 1999).] The scale of the iron enhanced region, ~ 50 - 100 kpc, is larger than the half-light radius Re of the central E (cD) galaxy.

Because oxygen and magnesium are formed in the outer layers of massive stars during normal stellar evolution, their Type II supernova yields should be reliable apart from uncertainties in the IMF and metallicity. Consequently, one would expect O and Mg abundances to vary together. Although XMM abundances of oxygen and magnesium are less accurate in the hot gas, they typically increase very little toward the core of the central elliptical and often become flat or even decline near the very center (Tamura et al. 2001; Buote et al. 2003b; Johnstone et al. 2002; Gastaldello & Molendi 2002). NGC 1399 and NGC 5044 data show that Mg traces Si and S more closely than O; only O seems flat. Either the O abundance in the mass-losing stars is surprisingly similar to that in the ambient hot gas, or the O expelled from O-rich stars is not all going into the hot gas.

The hot gas abundance zi (by mass in solar units) of element i should vary as

Equation 7 (7)

where yIa, i / Msn is the fraction of mass in element i ejected by Type Ia supernovae, zi, odot is its solar abundance and 1.4 is the ratio of total to hydrogen mass at solar abundance. This equation is quite general, but it is interesting to apply it to subsonic galaxy/group inflows or outflows in which the gas density rho(r) remains essentially constant for several Gyrs. We assume that all stellar and supernova ejecta goes into the hot gas. In a Lagrangian frame moving slowly with the gas velocity u, the abundance of element i in the gas increases from some initial value z0, i to an asymptotic limit

Equation 8

This equilibrium abundance is reached after teq = (rho / rho*) / alpha* = 2.80 × 108 rkpc1.18 yrs, using NGC 4472 densities from Figure 2a.

For iron enrichment (yIa, Fe = 0.7 Modot; Msn = 1.4 Modot; SNu(tn) = 0.06 SNu; zodot, Fe = 1.83 × 10-3 by mass) we find zeq, Fe = z*, Fe + 2.35 solar. This value is similar to the gas abundance in the isolated solution shown in the first column of Figure 5 in which zFe decreases from ~ 3 solar at the center to ~ 2 at large radii where z*, Fe becomes small. It is significant that this flow solution overestimates the iron abundance observed in NGC 4472, particularly for r gtapprox 10 kpc. The correct iron abundance gradient in 10 ltapprox r ltapprox 50 kpc is, however, nicely fit by the solution that includes circumgalactic gas (dashed line, first column, Fig. 5). This agreement is only possible if the low metallicity circumstellar (or cluster) gas is flowing inward, diluting the entire flow, i.e. zFe increases toward zeq, Fe as the gas flows in. Of course inflow is only possible if some (distributed) cooling occurs. If the global radial flow is outward, the circumgalactic gas (with low zFe) would be pushed outward and the Fe abundance would approach ~ zeq, Fe within r approx 3tGyr0.85 kpc. In this case the hot gas iron abundance zFe(r) would be at least as large as that of the isolated flow solution (column 1, Fig. 5) which clearly exceeds the iron abundance observed in NGC 4472. The observed iron abundance gradient is therefore a powerful argument for systematically inflowing hot gas, similar to the inflow required to explain the gas temperature profile (dT / dr > 0).

The abundance of oxygen in the hot gas, which comes mainly from stellar mass loss (yIa, O approx 0), should approach the stellar abundance, zeq, O = z*, O in time teq. The O abundance in M87 rises gradually with , which is consistent with some stellar enrichment, but it is curious that the O/Fe ratio is almost constant with galactic radius (Gastaldello & Molendi 2002). The small O abundance zO ~ 0.3 observed in NGC 5044 and M87 (Buote et al. 2003b; Gastaldello & Molendi 2002) suggests that only a fraction of the gas lost from stars may go into the hot phase. Alternatively, the stellar O abundance in E galaxies may be subsolar, as suggested by observations of forbidden emission lines in E galaxy planetary nebulae (Jacoby & Ciardullo 1999; Walsh et al. 1999), although nebular forbidden lines are known to give lower O abunances than the recombination lines. But O and Mg are expected to vary in a similar manner and stellar Mg is known to increase with (Trager et al. 2000). If the O abundance is systematically low in E galaxies, this may indicate the presence of hypernovae in which oxygen is burned into heavier elements (Loewenstein 2001).

One of the most remarkable peculiarities in the abundance profiles are the central minima observed for almost all elements in M87 (Gastaldello & Molendi 2002), in NGC 4636 (Loewenstein & Mushotzky), in AWM 7 (Johnstone et al. 2002) and elsewhere. For ellipticals that are resolved by ROSAT within 5 kpc, the observations compiled by Buote (2000a, c) indicate that the iron abundance is often flat (NGC 1399) or decreasing (NGC 4472; NGC 5846; NGC 4636) within 5 kpc, assuming no intrinsic absorption. Only NGC 4649 shows an increase. Subsolar central hot gas iron abundances, ~ 0.8 solar, for r ltapprox 3 kpc have been confirmed with XMM RGS observations (Xu et al. 2002; Sakelliou et al 2002). It has been realized for some time that the central abundances could appear lower if the prominent X-ray resonance lines emitted there were optically thick and diffused outward before escaping (Gil'fanov et al. 1987, Tawara et al. 1997, Shigeyama 1998; Böhringer et al. 2001; Mathews et al. 2001; Sazonov et al. 2002). However, the XMM RGS spectra of Abell 496 (Tamura et al. 2001) and M87 (Gastaldello & Molendi 2002) show central reversals even for non-resonant lines. In M87 the optical depth may be too low for line radiative transfer to be important, particularly if the hot gas is mildly turbulent (Mathews et al. 2001; Sakelliou et al. 2002). Morris & Fabian (2003) describe a transient enrichment process that produces a central minimum but this feature would probably disappear if the source function were constant with time. There is at present no fully satisfactory explanation of these strange abundance profiles, assuming they are real.

We thank David Buote for reading an early draft of this review and for his helpful suggestions. Studies of the evolution of hot gas in elliptical galaxies at UC Santa Cruz are supported by NASA grants NAG 5-8409 & NAG 5-9956 and NSF grants AST-9802994 & AST-0098351 for which we are very grateful. FB is supported in part by grant MURST-Cofin 00.

Next Contents Previous