Next Contents Previous

6. IR STARBURST AND ULTRA-LUMINOUS GALAXIES IN THE LOCAL UNIVERSE

For a variety of reasons it is unlikely that star-formation (SF) in galaxies has proceeded quietly during the Hubble time. 'A posteriori' evidence has accumulated that a fraction of stars in stellar systems was produced during short-lived events (see e.g. the excellent review in Moorwood, 1996). These SF events are expected to be very luminous, either in the optical or in the IR, and are expected to contribute substantially to the global energetics from baryon thermonuclear reactions, to the synthesis of metals, and the generation of background radiations in the optical, IR and sub-mm. Also the origin of Es, S0s and of the bulges of spirals may have some relationship with luminous and ultra-luminous starburst events at high-z.

If the study of star-formation in high-redshift sources is a primary task for modern cosmology, it is obvious that relevant information for the interpretation of distant objects comes from a close up on local galaxies with enhanced SF. For this reason we consider in this Section a sub-class of local galaxies, the starburst galaxies and the IR luminous and ultra-luminous galaxies, including a small fraction (few %) of all local objects, but accounting for a large percentage of the present-day star formation in galaxies.

The discovery of the starburst phenomenon dates back to the 1970's and came almost simultaneously from two quite independent lines of investigation: from objective-plate (Markarian) surveys of UV-excess galaxies, and from the first pioneering IR observations of galaxies in the local universe. IR observations, in particular, revealed the existence of galaxies with IR luminosities and L/M ratios appearently too high to be sustained over their lifetimes (Harwit and Pacini 1975). This brought to the idea that some galaxies undergo a sudden burst of massive star formation, with dust reprocessing of UV photons emitted by the young stars interpreted as the source of the IR light.

From these observations it was clear that SF has a twofold appearence, a UV excess and an IR excess, which may be explained by the stocastical nature of the interaction between photons and dust in star-forming regions of galaxies (see above).

However, the abilities of UV and IR surveys to sample the starburst phenomenon are very different: while at low bolometric luminosities UV and IR surveys sample roughly the same kind of objects, at high luminosities the UV flux is no more a good tracer of the SF, which is better sampled by the IR emission. This effect is due to dust extinction of the UV-light by young stars becoming more and more relevant at the higher bolometric luminosities (Lbol > 1011 L$\scriptstyle \odot$, Sanders and Mirabel 1996). At the highest values of Lbol (> 1012 L$\scriptstyle \odot$) most (> 95%) of the flux comes out in the IR.

Lbol is also tightly correlated with the optical morphology: while at low-L there is a "natural" mix of various (mostly late) types, at the higher-L nearly all objects appear to be interacting galaxies, and at the highest-L they look as advanced mergers. Also, the correlation is in the sense that while in low-L objects the SF activity is spread over the galactic disk (enhanced in the spiral arms), at increasing luminosity the SF gets more and more concentrated in the nuclear regions.

In the higher-L objects in particular, it is often observed a concomitant stellar and nuclear non-thermal (AGN) activity, usually the latter occurring in the dynamical center of the galaxy and the former in a circum-nuclear ring (at $ \sim$ 1 kpc).

A basic difficulty encountered in studies of active galaxies is to disentangle between starburst-dominated and AGN-dominated energy sources of the IR-luminosity. In fact, the two astrophysical processes are quite often associated in the same object. Optical line ratios (high vs. low excitation, e.g. [OIII]5007 / H$ \beta$ vs. [NII]6583 / H$ \alpha$, the Osterbrock diagram) and line widths (few hundreds Km/s for starbursts, larger for AGNs) are sometimes useful indicators, even in the presence of dust.

Useful near-IR lines, accessible from ground, are the Hydrogen Br$ \gamma$2.166µ, HeI2.058µ, H2, but also higher atomic number species, [FeII] among others. The Br$ \gamma$2.166µ and HeI2.058µ, in particular, so close in $ \lambda$ that differential extinction is negligible, constrain the underline ionization spectrum.

However, the most reliable information is provided by mid- and far-IR spectroscopy by space observatories. Extremely promising in this field, in addition to ISO and SIRTF in the next few years, are the planned large space telescopes: NGST in the mid-IR and FIRST in the far-IR.

6.1. The infrared-radio correlation

While there is no direct proof for the basic interpretation of the IR starburst phenomenon (i.e. being due to UV light from newly formed stars absorbed by dust and re-emitted in the IR), an indirect support comes from the well-known radio to far-IR luminosity correlation (de Jong et al. 1985, Helou et al. 1985). This, which is the tightest correlation involving global properties of galaxies, provides an important constraint on the physics ruling starbursts of any luminosities. It not only involves luminous active starbursting galaxies, but also many other galaxies, like quiescent spirals.

The correlation is parametrized by the ratio of the bolometric far-IR flux FFIR (in erg/s/cm2) to the radio flux S$\scriptstyle \nu$ (in erg/s/cm2/Hz):

Equation 6.14   (6.14)

which is observed to keep remarkably constant with Lbol ranging over many orders of magnitude, from low-luminosity spirals up to ultraluminous objects (Arp 220) [small departures from linearity appearing at the low- and high-luminosity ends].

The relation is interpreted as an effect of the ongoing star formation: the far-IR emission comes from dust heated by UV photons by young stars, which also heat the ISM producing free-free emission and generate SN originating high-energy e- and synchrotron flux mostly by interaction with the general galactic magnetic field. This same scheme explains the departures from linearity: e.g. q slightly increases at the low-luminosity end because LFIR is also contributed by the flux by old stars heating the dust. The radio emission tends to be less concentrated than the far-IR, because of fast e- diffusion.

6.2. Estimates of the star formation rate (SFR)

As the bolometric luminosity increases, the optical indicators of the SFR (e.g. the UV flux, or the EW of H$ \alpha$) become increasingly uncertain, as a larger and larger fraction of short-$ \lambda$ photons are extinguished. In such a situation, the IR luminosity (proportional to the luminosity by young stars) becomes the most reliable indicator of the SFR. A slight complication here is that older stars illuminating the diffuse cirrus dust in galaxies also contribute to the far-IR flux, particularly in low-luminosity inactive systems.

The SFR is estimated by Telesco (1988) from the energy released by the CNO cycle and assuming a Salpeter IMF (eq. 4.13):

Equation 6.14a

the former relation referring to the OBA star formation. A refined calibration is given by Rowan-Robinson et al. (1997):

Equation 6.14b

where $ \phi$ incorporates the correction from a Salpeter IMF to the true IMF ($ \phi$ $ \sim$ 3.3 going to a Miller-Scalo) and includes corrections for the cut in the IMF (e.g. $ \phi$ $ \sim$ 1/3 if only OBA stars are formed), $ \epsilon$ being the fraction of photons re-radiated in the IR.

Another mean of estimating the ongoing SFR exploits the radio flux (Condon 1992), by relating the SN rate to the rate of SF and using observations of the radio luminosity of the Milky Way to calibrate the relation. Since the synchrotron emission (proportional to the rate of SN remnant production) and thermal radiation (from HII regions heated by young OB stars) dissipate in 107 - 108 yrs, the radio flux provides a good measure of the instantaneous SFR. Operatively, one needs to estimate the fraction of stars with masses M > 8 M$\scriptstyle \odot$, progenitors of type-II SN, formed per unit time. The problem with faint radio-source observations is that the radio emission of stellar origin gets easily confused with non-thermal emission by a radio-loud AGN.

Finally, ISO observations indicate that also the mid-IR flux [dominated by hot dust and PDR emission] traces very well the SFR (see Sect. 12.1.2 below).

All these long-wavelength methods provide obvious advantages, in terms of robustness with respect to dust-extinction, compared with the optical ones, namely the relation of SFR with the UV continuum flux by Madau et al. (1996): SFR(all stars) = 5.3 10-10 L2800Å / L$\scriptstyle \odot$ [M$\scriptstyle \odot$/yr]; SFR(metals) = 1/42 SFR(stars), and that between the H$ \alpha$ line flux and the SFR (Kennicut 1998):

Equation 6.14c

Poggianti, Bressan & Franceschini (2000) and Franceschini et al. (2000) have shown that even after correcting for extinction the H$ \alpha$ flux using measurements of the Balmer decrement, the H$ \alpha$-based SFR is typically a factor $ \sim$ 3 lower than the appropriate value inferred from the bolometric flux in IR-luminous galaxies.

Altogether, with these calibrations, moderate luminosity IR starbursts have SFR$ \sim$3-30 [M$\scriptstyle \odot$/yr], (corresponding to $ \sim$ 105 O stars present during a typical burst). The most luminous objects, if indeed powered by SF, have SFR up to 1000 [M$\scriptstyle \odot$/yr]. Bolometric flux and SFR are correlated with the broad-band IR to optical luminosity ratio: LIR/LB $ \sim$ 0.1 in inactive galaxies (M31, M33), LIR/LB $ \sim$ 3 - 10 in luminous (L $ \sim$ 1011 L$\scriptstyle \odot$) SBs, LIR / LB $ \sim$ 100 in ultra-luminous objects (L > 1012 L$\scriptstyle \odot$, e.g. Arp 220).

6.3. Gas reservoirs, depletion times, starburst duration

The duration of the starburst is critically related with the mass fraction of stars produced during the event and to the available gas reservoir. Assuming that the SB dominates the spectrum on top of the old stellar population emission, an estimator of the SB duration is the EW of the Br$ \gamma$ line, which is a measure of the ratio of the OB stellar flux (the excitation flux) to the red supergiant star flux (evolved OBA stars). The EW is then expected to evolve monotonically with time. Also, the comparison of the Br$ \gamma$ line with the CO NIR absorption lines is an age indicator (Rieke et al. 1988). Moorwood et al. (1996) find in this way ages of 107 to $ \sim$ 108 yrs.

However, the most direct way to estimate at least an upper limit to the burst duration is the comparison of the total mass of molecular material in the galaxy nucleus with the estimated SFR, which is also a measure of the efficiency of SF. The gas mass is usually estimated from mm-wave CO line emission and from mm continuum observations of dust emission (assuming suitable conversion factors for H2/CO and dust/gas). Chini et al. (1995) have found that the two independent evaluations of the molecular mass provide consistent results, showing that luminous IR galaxies are very rich in gas (2 × 109 to 2 × 1010 M$\scriptstyle \odot$). The ratio LIR / Mgas assumes enormously different values in different stages of galaxy activity: in normal inactive spirals LIR / Mgas $ \sim$ 5 (L$\scriptstyle \odot$/M$\scriptstyle \odot$) (e.g. M31), in moderate starbursts LIR / Mgas $ \sim$ 20 (M82, NGC253), in ultra-luminous IR galaxies LIR / Mgas $ \sim$ 200 (Arp 220), in quasars LIR / Mgas $ \sim$ 500.

A limit to the SB duration is then given by tdepletion = 1010 Mgas / LIR yrs, ranging from typically several Gyrs for inactive spirals down to a few 107 yrs for the more active SBs.

6.4. Starburst-driven super-winds

There are several evidences that extremely energetic outfows of gas are taking place in starbursts: (a) from optical spectroscopy, evidence for Wolf-Rayet lines indicative of very young SBs (< 107 yrs) and outflow of ionized gas, with velocities up to 1000 Km/s (Heckman, Armus & Miley 1990; Lehnert & Heckman 1996); (b) from optical imaging there are evidences of bubbles and cavities left over by large, galactic-scale explosions; (c) from X-ray spectroscopy, evidence for plasmas at very high temperatures (up to few KeV), far in excess of what the gravitational field could explain (e.g. Cappi et al. 1999).

These highly energetic processes are interpreted as due to radiative pressure by massive stars, stellar winds and supernovae explosions occurring in a small volume in the galaxy core, able to efficiently energize the gas and to produce a dynamical unbalance followed by a large scale outflow of the remaining gas.

This phenomenon has relevant implications. It is likely at the origin of the huge amounts of metals observed in the Intracluster Plasma (ICP) in local galaxy clusters and groups. It should be noted that the estimated Fe metallicity of the outflowing plasma ($ \sim$ 0.2 - 0.3 solar, Cappi et al. 1999) is similar to the one observed in the ICP. Also the higher abundances (1.5-2 solar) observed in the 2 archetypal starbursts for $ \alpha$ elements (Si, O, Mg) may indicate that type-II SN (those produced by very massive stars, M > 8 M$\scriptstyle \odot$) are mostly responsible (Gibson, Loewenstein & Mushotzky 1997). Similar properties are observed in the hot halo plasmas around elliptical galaxies, also rising the question of a possible relationship of the hyper-luminous IR galaxy phenomenon with the formation of early-type galaxies. Therefore, the enriched plasmas found in local clusters and groups may represent the fossile records of ancient starbursts of the kind we see in local luminous IR starbursts.

6.5. Starburst models

More precise quantifications of the basic parameters describing the SB phenomenon require detailed modelling. The first successful attempt accounting in some detail for the observed IR and radio data was by Rieke et al. (1980), who demonstrated that the remarkable properties of M82 and N253 are consistent with SB activity.

Since then, a number of groups elaborated sophisticated models of SBs. These successfully reproduce SB properties assuming exponentially decaying SFRs with burst durations of 107 to 108 yrs, whereas both instantaneous and long duration bursts are excluded.

An important issue addressed by these models is about the stellar IMF during the burst: Rieke et al. found that assuming for M82 a Salpeter IMF with standard low-M cutoff at 0.1 M$\scriptstyle \odot$ resulted in a stellar mass exceeding the limit implied by dynamical mass evaluations. The problem was resolved by assuming that formation of stars with masses less than a few M$\scriptstyle \odot$ is strongly suppressed. This result however is not univocally supported by more recent studies of M82: e.g. Leitherer and Heckman (1995) solution is for a 1 to 30 M$\scriptstyle \odot$ IMF.

Interesting constraints on the IMF come in particular from the analysis of CO line kinematics in Arp 220: Scoville et al. (1996) indicate that the dynamical mass, the Lyman continuum, the SFR and the burst timescale can be reconciled by assuming a IMF truncated outside 5 to 23 M$\scriptstyle \odot$, with a SFR $ \sim$ 90 M$\scriptstyle \odot$/yr for stars within this mass range. Altogether, there seem to be fairly clear indications for a "top-heavy" mass function in the more luminous SBs, as compared with quiescent SF in the Milky Way and in spirals. This has relevant implications for the SFR history in galaxies, the cosmic production of light and of heavy elements.

A very detailed modellistic study of starbursts was given by Leitherer & Heckman (1995) and Leitherer et al. (1999), incorporating all up-to-date improvements in the treatment of stellar evolution and non-LTE stellar atmospheric models. The model successfully explains most basic properties of starbursts, as observed in the optical. Model predictions for a continuous SF over 108 yrs and a 1-30 M$\scriptstyle \odot$ Salpeter IMF, normalized to a SFR=1 M$\scriptstyle \odot$/yr are: bolometric luminosity = 1.3 1010 L$\scriptstyle \odot$; number of O stars = 2 × 104; ionizing photon flux ($ \lambda$ < 912Å) = 1.5 1053 photons/sec; SN rate = 0.02 yr-1; K = -20.5 mag; mass deposition rate = 0.25 M$\scriptstyle \odot$/yr; mechanical energy deposition rate = 6 1041 erg/sec. Important outcomes of these papers are predictions for the EW of most important line tracers of the SF (H$ \alpha$, Pa$ \beta$, Br$ \gamma$), as a function of the time after the onset of SF and of IMF shape.

Models of dusty starbursts have been discussed by Silva et al. (1998), Jimenez et al. (2000), Siebenmorgen, Rowan-Robinson & Efstathiou (2000), Poggianti & Wu (2000), Poggianti, Bressan & Franceschini (2001). The latter two, in particular, address the question of the classification and interpretation of optical spectra of luminous IR starbursts: they find that the elusive class of e(a) (emission+absorption) spectra, representative of a large fraction (> 50%) of all IR SBs, are better understandable as ongoing active and dusty starbursts, in which the amount of extinction is anti-correlated with the age of the population (the youngest stars are the more extinguished, see also Sect. 4.2.3), rather than post-starburst galaxies as sometimes have been interpreted.

6.6. Statistical properties of active galaxy populations

Statistical properties of SB galaxies provide guidelines to understand the origin and triggering mechanisms of the phenomenon. A fundamental descriptor of the population properties is provided by the Local Luminosity Function (LLF), detailing the distribution of space density as a function of galaxy luminosity in a given waveband.

While the faint luminosity end is important for cosmogonic purposes (providing constraints on the formation models, its flattish shape being roughly similar at all wavelengths), SBs and their complex physics dominate at the bright end of the LLF. Indeed, the latter is observed to undergo substantial changes as a function of $ \lambda$: if the optical/near-IR LLF's display the classical "Schechter" exponential convergence at high-luminosities (essentially tracing the galaxy mass function), LLF's for galaxies selected at longer wavelengths show flatter and flatter slopes (see Fig. 8 below). This flattening is progressive with $ \lambda$ going from the optical up to 60 µm. When expressed in differential units ( Mpc-3L-1), the bright-end slope of the 60 µm LLF is $ \propto$ L60$\scriptstyle \mu$-2, according to the extensive sampling by IRAS (Saunders et al. 1990). Note that this flattening is not due to the contribution of AGNs at 60 µm, which is modest here and quite more important instead at 12 µm. What is progressively increasing with $ \lambda$ up to $ \lambda$ = 60 µ is the incidence of the starburst contribution to the luminosity: it is the starbursting nature of 60µ selected galaxies responsible for the shape of the LLF.

It is interesting to consider that almost the same slopes $ \propto$ L-2 are found for all known classes of AGNs, from the luminous radio-galaxies (Auriemma et al. 1977; Toffolatti et al. 1987), to the optical and X-ray quasars (Miyaji, Hasinger & Schmidt 2000; Franceschini et al. 1994b). Also to note is the evidence that the L-2 slope for AGNs keeps almost exactly the same at any redshifts, in spite of the drastic increase of the source number-density and luminosity with z due to evolution.

There should be a ruling process originating the same functional law in a wide variety of categories of active galaxies and remarkably invariant with cosmic time, in spite of the dramatic differences in the environmental and physical conditions of the sources. This remarkable behaviour may be simply understood as an effect of the triggering mechanism for galaxy (AGN and starburst) activity: the galaxy-galaxy interactions (either violent mergers between gas rich objects or encounters triggering a slight increase of the activity).

The physical mechanism ruling the process is the variation in the angular momentum $ \Delta$J/J of the gas induced by the interaction, and the consequent gas accretion $ \Delta$m/m in the inner galaxy regions (Cavaliere & Vittorini 2000). Starting for example from a $ \delta$ -function-shaped LLF, the starburst triggered by the interaction produces a transient increase of L which translates into a distortion of the LF towards the high-L's. Assumed $ \Delta$m/m is ruled by the probability distribution of the impact parameter b, it is simple to reproduce in this way the LLF's observed asymptotic shape at the high luminosities.

All this points at the interactions as ruling the probability to observe a galaxy during the active phase.

6.7. Starburst triggering

In normal inactive spirals the disk SFR is enhanced in spiral arms in correspondence with density waves compressing the gas. This favours the growth and collapse of molecular clouds and eventually the formation of stars. This process is, however, slow and inefficient in making stars (also because of the feedback reaction to gas compression produced by young stars). This implies that very long timescales (several Gyrs) are needed to convert the ISM into a significant stellar component.

On the contrary, because of the extremely high compression of molecular gas inferred from CO observations in the central regions of luminous starburst galaxies, SF can proceed there much more efficiently. Both on theoretical and observational grounds, it is now well established that the trigger of a powerful nuclear starburst is due to a galaxy-galaxy interaction or merger, driving a sustained inflow of gas in the nuclear region. This gas has a completely different behaviour with respect to stars: it is extremely dissipative (gas clouds have a much larger cross-section and in cloud collisions gas efficiently radiates thermal energy generated by shocks). A strong dynamical interaction breaks the rotational symmetry and centrifugal support for gas, induces violent tydal forces producing very extended tails and bridges and triggers central bars, which produce shocks in the leading front, and efficiently disperse the ordered motions and the gas angular momentum. The gas is then efficiently compressed in the nuclear region and allowed to form stars.

These concepts are confirmed by numerical simulations of galaxy encounters. Toomre (1977) was the first to suggest that ellipticals may be formed by the interaction and merging of spirals. This suggestion is supported by various kinds of morphological features (e.g. tidal tails, rings) observed in the real objects and predicted by his pioneering numerical simulations.

Much more physically and numerically detailed elaborations have more recently been published by Barnes and Hernquist (1992), who model the dynamics of the encounters between 2 gas-rich spirals including disk/halo components, using a combined N-body and gas-dynamical code based on the Smooth Particle Hydrodynamics (SPH). Violent tidal forces act on the disk producing extended tails and triggering central bars, who sweep the inner half of each disk and concentrates the gas into a single giant cloud. The final half-mass radii of gas are much less than those of stars: for an M* galaxy of 10 11 M$\scriptstyle \odot$, $ \sim$ 109 M$\scriptstyle \odot$ of gas are compressed within 100-200 pc, with a density of 103 M$\scriptstyle \odot$/pc3 (Barnes & Hernquist 1996).

Various other simulations confirm these finding. SPH/N-body codes show in particular that the dynamical interaction in a merger has effects not only on the gas component, but also on the stellar one, where the stars re-distribute following the merging and violent relaxation of the potential.

6.8. Ultra-luminous IR galaxies (ULIRGs)

Defined as objects with bolometric luminosity Lbol $ \simeq$ LIR > 1012 L$\scriptstyle \odot$, they are at the upper rank of the galaxy luminosity function. A fundamental interpretative problem for this population is to understand the primary energy source, either an extinguished massive nuclear starburst, or a deeply buried AGN.

A systematic study of this class of sources was published by Genzel et al. (1998), based on ISO spectroscopy of low-excitation and high-excitation IR lines, as well as of the general shape of the mid-IR SED (the intensity of PAH features vs. continuum emission; see also Lutz et al. 1998). While the general conclusion of these analyses is that star-formation is the process dominating the energetics in the majority of ultraluminous IR galaxies, they have also proven that AGN and starburst activity are often concomitant in the same source. This fact is also proven by the evidence (e.g. Risaliti et al. 2000; Bassani et al. 2000) that many of the ULIRGs classified by Genzel et al. as starburst-dominated also show an hidden, strongly photoelectrically absorbed, hard X-ray spectrum of AGN origin. Soifer et al. (2000) have also found that several ULIRGs show very compact (100-300 pc) structures dominating the mid-IR flux, a fact they interprete as favouring AGN-dominated emission. The relative role of SF and AGN in ULIRGs is still to be quantified, hard X-ray (spectroscopic and imaging) observations by CHANDRA and XMM, as well as IR spectroscopy by space observatories (SIRTF, FIRST) will provide further crucial information.

6.9. Origin of elliptical galaxies and galaxy spheroids

As pointed out for the first time by Kormendy & Sanders (1992), the typical gas densities found by interferometric imaging of CO emission in ultra-luminous IR galaxies turn out to be very close to the high values of stellar densities in the cores of E/S0 galaxies. This is suggestive of the fact that ULIRG's have some relationships with the long-standing problem of the origin of early-type galaxies and spheroids.

Originally suggested by Toomre (1977), the concept that E/S0 could form in mergers of disk galaxies immediately faced the problem to explain the dramatic difference in phase-space densities between the cores of E/S0 and those of spirals. Some efficient dissipation is required during the merger, which can be provided by the gas. Indeed, the CO line observations in ULIRG's, also combined with those of the stellar nuclear velocity dispersions and effective radii, show them to share the same region of the "cooling diagram" occupied by ellipticals.

Detailed analyses of the H2 NIR vibrational lines in NGC 6240 and Arp 220 (van der Werf 1996) have provided interesting information about the mass, kinematics, and thermodynamics of the molecular gas. The conclusion is that shocks, the fundamental drivers for dissipation, can fully explain the origin of the H2 excitation. The evidence that the H2 emission is more peaked than stars, and located in between the two merging nuclei, is consistent with the fact that gas dissipates and concentrates more rapidly, while stars are expected to relax violently and follow on a longer timescale the new gravitational potential ensuing the merger.

A detailed study of Arp 220 by van der Werf (1996) has shown that most of the H2 line emission, corresponding to $ \sim$ 2 × 1010M$\scriptstyle \odot$ of molecular gas, comes from a region of 460 pc diameter, the gas mass is shocked at a rate of $ \sim$ 40 M$\scriptstyle \odot$/yr, not inconsistent with a SFR $ \sim$ 50 - 100 M$\scriptstyle \odot$/yr as discussed in Sect. 6.5. Compared with the bolometric luminosity of Arp 200, this requires a IMF during this bursting phase strongly at variance with respect to the Salpeter's one (eq. 4.13) and either cut at Mmin > > 0.1 M$\scriptstyle \odot$ or displaying a much flatter shape.

In support of the idea that ellipticals may form through merging processes there is evidence coming from high-resolution K-band imaging that the starlight distribution in hyper-luminous IR galaxies follows a de Vaucouleurs r-1/4 law typical of E/S0 (Clements & Baker 1997).

Also proven by simulations, after the formation of massive nuclear star clusters from the amount of gas (up to 10 10 M$\scriptstyle \odot$) collapsed in the inner Kpc, part of the stellar recycled gas has low momentum and further contracts into the dynamical center, eventually producing a super-massive Black Hole with the associated AGN or quasar activity (Norman and Scoville 1988, Sanders et al. 1988).

Next Contents Previous