In summary, problems with the simple cooling flow picture emerged from two fronts. The first is the enormous implied mass deposition in some objects where for example ~ 1000 M yr-1 (e.g. PKS 0745-191; Fabian et al. 1985) which is then not seen in cooled form (the 'mass sink' problem). The central galaxy of this, and many other brightest cluster galaxies where the surrounding gas have tcool < 3 Gyr, has much excess blue light, optical/UV/IR emission-line nebulosity (Crawford et al. 1999 and refs. therein) and even molecular gas (Edge et al. 2001). The total star formation rate and gas mass is often one to two orders of magnitude below such a large value of operating for several Gyr. The flow could be somewhat intermittent but not overly so or short central cooling times would not be so common (Peres et al. 1998). A non-standard IMF for the stars is one possibility (Fabian et al. 1982), but has to be extreme (Prestwich et al. 1997).
The second problem is that a simple cooling flow spectrum is a poor fit to the data in the soft X-ray band (White et al. 1991, Johnstone et al. 1992, Fabian et al. 1984). What was usually done was to fit the model spectrum of an isobaric cooling flow (Johnstone et al. 1992) plus an isothermal spectrum to represent the luminosity from gravitational work done, to the data from the inner regions. A deficit in soft X-ray emission was found in the data when compared with the model. This was then modelled as photoelectric absorption due to an absorber intrinsic to the central regions. The lack of any obvious absorption in very soft ROSAT spectra was attributed to emission from the intervening gas in the cluster (Allen & Fabian 1997).
Photoelectric absorption intrinsic to the cluster does not however appear in the spectra of the jet in M87 (Böhringer et al. 2002) or the nucleus of NGC 1275 (Churazov et al. 2003). Nor is it apparent in the detailed XMM/RGS spectra of cool cluster cores (Peterson et al. 2001, Tamura et al. 2001a, Kaastra et al. 2001, Sakelliou et al. 2002, Peterson et al. 2003). Absorption has generally been abandoned as an explanation, although no detailed test has been made of a model in which the absorption is intimately linked with the coolest cooling gas clouds. Note though that the presence of warm emission-line nebulosities, dust-lanes and clear optical absorption lines (Carter et al. 1997, Sparks et al. 1997) show that at some level there must be distributed intrinsic absorption.
Curiously, the 'missing' soft X-ray luminosity (i.e. the above soft X-ray deficit) is close to the luminosity emitted from the emission-line nebulosity commonly found in these systems (Fabian et al. 2002a, Soker et al. 2004). This has led to the suggestion that the gas is cooling but not radiatively once its temperature has dropped to ~ 1 keV or so. It can be cooled by mixing (Fabian et al. 2002a) or conduction (Soker et al. 2004) with cold gas. A possible picture would be that the gas cools from 100 to within 20 kpc radiatively as a single-phase flow by which radius its temperature has dropped by a few. It then shares its temperature with cooler gas and thereby rapidly drops down to below 106 K. The cooler gas is close to the peak of the cosmic cooling curve and radiates the energy in the UV and optical bands (and IR if there is molecular gas and dust). Such a solution to the soft X-ray deficit still suffers from the mass sink problem.
A further situation which could give a soft X-ray deficit is if the metals in the ICM are highly inhomogeneous. They are presumably introduced in a very localized manner and at high abundance by stars and supernovae. If they do not mix but after time just reside at a similar temperature to their surroundings then when bremsstrahlung cooling dominates (which it does above about 1 keV for Solar abundances) then all the gas cools together. But when it has cooled so that line cooling dominates, the highly enriched clumps would then cool rapidly and drop out (Fabian et al. 2001, Morris & Fabian 2003). This has been considered in part by Böhringer et al. (2002), although no conclusive test has been performed.
The main reason that complete cooling flows, in which gas cools from the cluster virial temperature to well below 106 K, are often now considered to be ruled out is the high spectral resolution XMM/RGS data (Peterson et al. 2001, Tamura et al. 2001a, Kaastra et al. 2001, Sakelliou et al. 2002, Peterson et al. 2003). Chandra spectra lead to a similar conclusion (David et al. 2001, Allen et al. 2001a). These spectra have been covered in depth earlier in this review. We summarize here to note that most data are consistent with the ICM temperature dropping within the cooling radius by a factor of about 3 (sometimes significantly larger values are found e.g. A2597 (Morris & Fabian 2005) and Centaurus (Fabian et al. 2005a). Data from all wavelengths are consistent with a residual flow ranging from a few M yr-1 to a typical value of tens M yr-1 up to ~ 100 M yr-1 in some cases (e.g. Bayer-Kim et al. 2002; Wise et al. 2004; McNamara et al. 2004). These values are generally compatible with the ongoing star formation seen.
6.1. Cooled gas and star formation
The brightest galaxy in X-ray peaked clusters often has excess blue light indicating massive star formation and IR/optical/UV emission lines (Crawford et al. 1999, Donahue et al. 2000, Hicks & Mushotzky 2005). The regions are often dusty so the determination of the star formation rate depends on uncertin dust corrections. It can be tens M yr-1 to about 100 M yr-1 in the case of A1835. In a typical BCG it is one to ten per cent of the mass cooling rate inferred from the hotter X-ray emitting gas assuming it cools completely. UV-excess BCGs probably have the highest rate of star formation of early-type galaxies at the present epoch. The brightest galaxy in non-X-ray-peaked clusters do not have this emission.
The excitation of the emission-line nebulosity, which often has a filamentary structure (e.g. the Perseus cluster: Conselice et al. 2000: the Centaurus cluster: Crawford et al. 2005; the Virgo cluster: Sparks et al. 2004) is likely related to the star formation but in detail there are problems. Hotter emission is required than expected from the stars observed and there are no stars obvious within the outer filaments. The gas, even in the outer filaments, contains dust and molecular hydrogen (Hatch et al. 2005a, Jaffe et al. 2005). Simple models for the molecular gas imply higher pressures than for the surrounding gas and very short radiative cooling times (Jaffe et al. 2001, Wilman et al. 2002). The filaments may have been pulled out from a central reservoir of cold and warm gas by the action of radio bubbles (Fabian et al. 2003b, Hatch et al. 2005b). Where the central gas comes from is not clear although radiative cooling is likely. A residual cooling flow is therefore taking place in many clusters.
Gas at intermediate temperatures of about 105.5 K is seen through OVI emission with FUSE. Oegerle et al. (2001) found such emission from A2597 and recently it has been found in A426 and A1795 (Bregman et al. 2005). The inferred cooling rates in the 30 arcsec FUSE aperture are 30-50 M yr-1.