Star formation in the CMZ seems to be rather different from that seen elsewhere in the Galaxy. For one thing, one finds in this region the most extreme young star clusters in the Galaxy. There are three remarkable, massive, short-lived star clusters known in this region: the Arches, the Quintuplet and the central parsec cluster. These clusters have ages < 10⁷ years and masses on the order of 10⁴ M_ּ. All three have a large number of unusually massive, windy stars, including a mix of Wolf–Rayet stars, Luminous Blue variables, Ofpe stars, and a sizeable population of OB stars (Krabbe et al 1995; Najarro et al 1997; Figer et al 1999a, b; Paumard et al 2001). Other exotic categories of very luminous stars are also present, all falling under the rubric of helium emission-line stars, after the presence of the 2.06 μm He emission line, an indicator of a substantial, strong wind. The precise mix of the more evolved stars the WR stars and LBVs is an indicator of the cluster age.

The Arches cluster is the youngest and most extreme (figure 1.1). It has ∼160 O stars, a total luminosity exceeding 10⁸ L_ּ, and a Lyman continuum production rate of ∼4 × 10⁵¹ s⁻¹ (Figer et al 1999b). With this radiation field, the Arches cluster dominates the local heating and ionization of the interstellar medium, and in fact the region surrounding the Arches cluster is the most luminous portion of the Galactic Center region at mid-infrared wavelengths (e.g. Shipman et al 1997), because much of the luminosity of the Arches is reradiated at mid infrared wavelengths. Furthermore, in the radio regime, the unusually large HII region known as the arched filaments linear, ionized features lying at the surface of a molecular cloud (Lang et al 2001, 2002) is apparently attributable to the Arches cluster.

The youth of these clusters is assured, because, by virtue of being within about 40 pc of the Galactic Center, they are subject to tidally induced evaporation  on timescales not much larger than their ages, ∼10⁷ years (Kim et al 1999, 2000; Portegies-Zwart et al 2002). The disintegration of these clusters is hastened by stellar evolution; the large rate of mass loss by the massive stars steadily reduces the cluster mass, which in turn reduces the tidal radius.

The cluster of young stars within the central parsec is not subject to tidal disruption because it is not a bound system. It consists of a grouping of massive stars having independent, phase-mixed orbits in the potential well created by the black hole and the central cluster of nuclear bulge stars. The youth of the emission-line stars in the central cluster raises a troubling question about where they were formed. It ordinarily takes a time far longer than the age of these stars to bring them individually into the central parsec from larger distances by

Figure 1.1. The Arches cluster, observed with HST/NICMOS (Figer et al 1999b). This false-color image was made by combining images made with three near-infrared filters. See also color section.

relaxation processes such as dynamical friction, or, equivalently, mass segregation (Morris 1993). However, in situ star formation is problematical because of the strong tidal forces exerted by the central black hole. At a distance of 0.25 pc, the typical distance of the luminous, He emission-line stars from Sgr A^*, the limiting Roche density is ∼10¹⁰ H atoms cm⁻³. This is 10⁴-10⁵ times denser than any gas presently observed near the Galactic Center, and there are serious problems with understanding how gas could be compressed to such high densities in such a warm, turbulent region, except possibly by the sudden release of an enormous quantity of mechanical energy, presumably by a dramatic accretion event onto the black hole. This scenario was considered by Morris et al (1999), who hypothesized a limit cycle of activity within the central parsec. Assuming that the CND is a long-lived configuration continuously fed from the outside, and noting that the natural evolution of the inner edge of the turbulent, magnetized CND is to move inward as a result of viscous evolution, Morris et al suggested that the collective winds from all the young stars in the central parsec exert a ram pressure on the inside edge of the disk which is sufficient to impede that inward migration (except perhaps for Rayleigh–Taylor instabilities such as these referred to earlier in the discussion of the arms of Sgr A West). However, when the most massive young stars in the central cluster finish their evolution, on timescales of ∼10⁷ years, the winds will die out and the inner edge of the CND will proceed inward. Eventually, the CND will converge upon the central black hole. If this leads to a sudden increase in the black hole's accretion rate, then the release of accretion energy will be explosive, and the portions of the CND near the central black hole will be strongly compressed. Whether that compression is sufficient to overcome the tidal forces and allow self-gravity of the compressed layer to form stars remains to be seen, but if it does happen like that, then the newly induced generation of stars and their stellar winds will establish a new dynamical equilibrium with the inside edge of the CND, initially joining with the outpouring of accretion energy from the black hole to evacuate the center of the CND. Thus the cycle would start anew, with a quasi-static equilibrium again resembling the current situation in the central parsec.

As an alternative, Gerhard (2001) recently explored the hypothesis that the young stars in the central cluster formed as part of a massive, Arches-type cluster originally located a few tens of parsecs away from the center. Because the timescale for spiraling inward to the central parsec as a result of dynamical friction is inversely proportional to the mass of the cluster, the cluster will move into the central parsec on a sufficiently short timescale if it is massive enough. Gerhard found that a mass as large as 10⁶ M_ּ is needed to account for the central young cluster (plus its parent cloud, if that cloud accompanies it most of the way into the center) if it starts as far out as the Arches cluster. While this mass exceeds that of the Arches cluster by two orders of magnitude, it is not unprecedented: super star clusters evidenced in starburst galaxies of various kinds have masses of 10⁵–10⁶ M_ּ (e.g. Ho and Filippenko 1996; O’Connell et al 1994, 1995; Tacconi Garman et al 1996; Turner et al 2000; Maoz et al 2001). However, in the Galactic Center, there is no evidence yet for the stellar tidal debris of young stars that would be left behind at radii beyond a parsec as the massive cluster migrated inward. This point is emphasized by Kim and Morris (2002), who use an N-body code to model the dynamics of a massive cluster at the Galactic Center. They confirm that it is possible to bring the remnant core of a cluster into the central parsec of the Galaxy if the cluster starts out massive enough, but, in general, the process should leave a halo of tidally stripped young stars throughout the inner several parsecs. Currently, there is no evidence for a population of young stars beyond the inner parsec.

The three remarkable clusters in the Galactic Center region, and the presence of super star clusters in the nuclear regions of starburst galaxies, suggest a particular mode of star formation that differs from that usually found in the Galactic disk (Morris 2001). Unless we live at a peculiar time, we must imagine that compact massive clusters like the Arches and Quintuplet form often and represent an important channel for populating the nuclear bulge. If other such clusters are currently forming, their formation sites have not yet been identified, although the star-forming core of the molecular cloud Sgr B2 is currently forming a fairly massive star cluster which may qualify for being related to the existing clusters (Mehringer et al 1993; Gaume et al 1995). The compactness and high mass of the Arches cluster or any other starburst cluster raises the issue of the timescale over which the cluster must have formed. The violence implied by the formation of hundreds of O stars within a few tenths of a parsec, including protostellar jets and winds and ionized gas flows at ionization fronts, is likely to quickly shut off star formation once the process begins. Indeed, these clusters may begin formation on the scale of a Jeans mass (∼10⁵ years), and then fragment hierarchically to stellar masses on a free-fall timescale. If so, then there is little room for the persistence of straggler gas clumps; the released gravitational energy rushing outward from the star formation cataclysm will commit any gas clump to immediate collapse or to a quick oblivion via ionization and Kelvin–Helmholtz instabilities. Massive starburst clusters must be quite close to coevality, so their use as probes of the IMF (initial mass function, i.e. the distribution of stellar masses when they form) should be little affected by a spread in stellar ages.

The core collapse of compact, massive, young clusters could provide a means of producing intermediate mass black holes via stellar coalescence. Over time, dynamical friction would bring such black holes into proximity and eventual coalescence with the central black hole, providing a means of growth for black holes in Galactic nuclei (Ebisuzaki et al 2001).

Of course, massive star clusters are not the only way that stars form in the Galactic Center region, and probably not even the dominant way. Many individual compact HII regions and emission line stars have been identified and studied there (Morris 1993; Figer et al 1994; Liszt and Spiker 1985: Lis et al 1994; Zhao et al 1993), although no generalizations about the properties of the stars have yet been offered.