Annu. Rev. Astron. Astrophys. 2004. 42: 603-683 Copyright © 2004 by Annual Reviews. All rights reserved

5. CENTRAL STAR FORMATION AND PSEUDOBULGE GROWTH

What is the rate at which star formation is building stellar mass density in pseudobulges? Is the picture of secular evolution in Sections 2 and 3 consistent via plausible formation timescales with the properties of pseudobulges in Section 4?

Figure 8 shows central gas disks in barred and oval galaxies that have radii and masses comparable to those of pseudobulges and that are intensely forming stars. They are a window on pseudobulge formation. We begin by discussing well-studied systems in which star-formation rates (SFRs), gas masses, stellar mass deposition rates, and hence, evolution timescales can be constrained accurately. We then review the broader body of observations of circumnuclear gas disks and SFRs. Star formation is ubiquitous in late-type galaxies. This means that it cannot be driven by episodic events such as mergers. It must be secular.

5.1. Case Studies: NGC 1326, NGC 1512, NGC 4314, and NGC 5248

These galaxies are excellent prototypes for studying circumnuclear disks, because each has been studied in depth using a combination of HST and ground-based imaging (Buta et al. 2000, Maoz et al. 2001, Benedict et al. 2002, Jogee et al. 2002). Also, the four objects span a representative range of host galaxy properties. They are all barred, and they all have similar luminosities (-19.0 M_B⁰ -20.3), but they cover a wide range of morphological types (SB0/a-SBbc) and environments. NGC 1326 is in the Fornax cluster, NGC 1512 is in an interacting pair with NGC 1510, and NGC 4314 and NGC 5248 are relatively isolated field galaxies located in loose groups. Nearly all (>80%) of the star formation in NGC 1326 and NGC 4314 is contained in the circumnuclear rings, whereas NGC 1512 and NGC 5248 have actively star-forming outer disks, with less than 40% of the total SFR near the center. The diversity in galaxy properties and environments already suggests that internal structure (e.g., bars) is more important than external influences in feeding the central star formation.

The central star-forming rings of NGC 1512, NGC 1326, and NGC 4314 are illustrated in Figure 8. At high spatial resolution, the rings of H II regions and young star clusters often are revealed to be pairs of tightly wound spiral arms. This is shown for NGC 1512 in Figure 3. The spiral structure is seen most clearly in red continuum images, where networks of dust features spiraling toward the center from within the star-forming rings can be seen. The continuum images also reveal large numbers of bright stellar knots (> 70 in NGC 4314; 500-1000 in the others). The luminosities and de-reddened colors of these knots indicate that they are not single stars but instead are luminous associations or star clusters. The brightest of these have stellar masses of order 10⁵ M, placing them in the class of populous blue clusters observed in the Magellanic Clouds (MCs) and other gas-rich galaxies. Some may be young progenitors of globular clusters. Many of the clusters are coincident with H II regions, but most are free of surrounding nebulosity, and these are probably older than the 5-10-Myr lifetimes of typical H II regions.

Current SFRs in these regions can be estimated from extinction-corrected H or Pa measurements converted using the SFR calibrations of Kennicutt (1998a). The resulting SFRs range from ~ 0.13 M yr^-1 in NGC 4314 (Benedict et al. 2002) to 1 M yr^-1 in NGC 1326 and NGC 1512 (Buta et al. 2000, Maoz et al. 2001) and ~ 2 M yr^-1 in NGC 5248 (Maoz et al. 2001 corrected to a distance of 15 Mpc). These values are probably accurate to within ± 50%, given uncertainties in the amounts and patchiness of the extinction and in the assumed distances to the galaxies. This is sufficient to characterize the evolutionary properties and physical conditions in these regions.

These rates are modest when compared with the total SFRs in giant spiral galaxies, which typically range from 0.1-1 M yr^-1 in normal Sa galaxies to 1-10 M yr^-1 in Sb-Sc galaxies (Kennicutt 1983, 1998a). However, they are quite exceptional considering the physical compactness of the star-forming regions. The star-forming rings have radii of 500-700 pc, so the surface densities of star formation are of order 0.1 to 1 M yr^-1 kpc^-2. This is 1 to 3 orders of magnitude larger than the typical disk-averaged SFR densities in normal galaxies, and it approaches the SFR densities seen in some infrared starburst galaxies (Kennicutt 1998a, b). If these rates were to persist over a Hubble time, they would produce bulges with stellar masses of 10⁹-10¹⁰ M. Thus, while the total amounts of star formation in these regions are not unusual by galactic standards, the character of the star formation is quite distinct.

The distinctive character of the star formation is underscored by the large populations of luminous young star clusters. Their extinction-corrected absolute magnitudes range from M_V⁰ = -13 to M_V⁰ ~ -8. For clusters fainter than this, HST photometry becomes very incomplete. The corresponding masses, corrected for the ages of the clusters, are ~ 10³ to 10⁵ M. These are similar to the masses of giant OB associations such as those in supergiant H II regions like 30 Doradus in the Large Magellanic Cloud (LMC) and to the masses of the populous blue star clusters found in the LMC and other gas-rich galaxies (e.g., Kennicutt & Chu 1988). The luminosity functions of the knots are well fitted by a power law with slope dN / dm ~ -2. They are consistent with the luminosity functions of H II regions and their embedded OB associations (e.g., Kennicutt, Edgar & Hodge 1989, Bresolin & Kennicutt 1997). They are also similar to the young star cluster populations in merger remnants such as NGC 4038-39 (e.g., Zhang & Fall 1999). However, no examples are found in these galaxies of the so-called super star clusters (SSCs) with M_V < -15 that are often seen in merger remnants and luminous starburst galaxies. This may be a reflection of the lower total amounts of star formation in these rings rather than any sign of a different cluster mass spectrum. Even if the power-law cluster mass spectra extend to the realm of the SSCs in these objects, the number of SSCs that we expect to observe at any one moment is less than one, based on the total size of the populations observed. We need to observe more central star-forming rings to determine whether they can form SSCs.

The star clusters can be age-dated using multicolor photometry and synthesis models (e.g., Leitherer et al. 1999). These measurements provide a powerful probe of the star-formation histories in the circumnuclear regions. In all four of these galaxies, multiband HST imaging in different combinations of U, B, V, I, H, and H have been used to derive reddening-corrected colors, luminosities, and hence age distributions. The galaxies all show a spread in cluster ages from zero to 200-300 Myr. The age distributions are heavily weighted toward younger clusters, but this is readily accounted for by dimming with age and by dynamical disruption effects. When corrections are applied for these processes, the age distributions are generally consistent with a roughly constant cluster formation rate over the past 200-300 Myr (Maoz et al. 2001). However, more sporadic histories cannot be ruled out.

If the rings have been forming stars at the current rate for 0.2 to 0.3 Gyr, the total mass of stars formed is (2-6) × 10⁸ M in NGC 1326, NGC 1512, and NGC 5248 and about 2 × 10⁷ M in NGC 4314. We can check the consistency of these results by comparing the SFRs with the masses of the circumnuclear gas disks. Millimeter measurements of CO emission in the centers of NGC 1326, NGC 4314, and NGC 5248 have been reported by Garcia-Barreto et al. (1991), Combes et al. (1992), Benedict et al. (1996), Sakamoto et al. (1999), Jogee et al. (2002), and Helfer et al. (2003). The corresponding molecular gas masses range from 0.7 × 10⁸ M in NGC 4314 to (5-12) × 10⁸ M in NGC 1326 and NGC 5248. Several authors (e.g., Wilson 1995, Paglione et al. 2001, Regan 2000) have advocated using a lower conversion factor of CO intensity and H₂ column density for these metal-rich environments; if these were used this would reduce the masses by factors of up to 2-3. The gas masses are comparable to the masses of stars already formed in the central disks during the current star-formation burst. This is what one would expect if observations are made at random times during the burst. Combining the gas masses with the SFRs also shows that there is sufficient fuel to feed the current circumnuclear SFRs for another 0.2-1 Gyr. By the time the gas is exhausted, central stellar disks with masses of 10⁸ to 10⁹ M will have formed. Of course, the masses are even larger if gas from the galaxies' bars continues to feed the star formation. In the cases of NGC 1326, NGC 1512, and NGC 5248, these masses are several times higher than the mass in stars (~ 10⁸ M, see below) formed in the main exponential disks if the parameters of these disks are extrapolated to the center. In fact, the stellar disks being formed by the star-forming rings have characteristic masses and sizes that are comparable to those of pseudobulges. Thus in these four galaxies, we are almost certainly observing the formation of pseudobulges, or the continued growth of pre-existing pseudobulges.

5.2. General Properties of Circumnuclear Regions

The circumnuclear star-forming rings discussed above are not extreme cases; even higher SFRs are observed at the centers of NGC 1097 and some other nearby galaxies. Nevertheless, before we attempt to characterize the global rates of star formation in these objects, it is important to review the properties of circumnuclear star-forming rings and disks in general. This subject was reviewed by Kennicutt (1998a), with emphasis on the most luminous starburst galaxies. These are nearly always associated with major mergers of gas-rich galaxies that are forming high-mass bulges and elliptical galaxies (Sanders & Mirabel 1996; Kennicutt, Schweizer & Barnes 1998, and references therein). Here, we focus exclusively on the central regions of normal spiral galaxies, where the circumnuclear activity is fed by the kinds of secular processes discussed in this review.

The frequency of occurrence of dense central gas disks and vigorous star formation can be estimated from two independent lines of evidence, surveys of central star formation in the ultraviolet, visible, or mid-infrared, and CO surveys of central molecular gas. Prominent circumnuclear rings like those in Figure 8 are easily identified. From an ultraviolet imaging survey of 110 nearby spirals, Maoz et al. (1996) estimate that approximately 10% of Sc and earlier-type spirals contain such strong circumnuclear star-forming regions. This is roughly consistent with the frequency of circumnuclear hot-spot galaxies in the survey of Sérsic (1973), which was based on blue photographic plates. Most of these galaxies are barred. This includes all of the galaxies that have ultraviolet-bright rings identified by Maoz et al. (1996), 88% of the hotspot galaxies in the Sérsic (1973) compilation, and 81% of all galaxies with peculiar nuclei identified by Sérsic. Galaxies with strong circumnuclear star formation tend to have Hubble types between Sa and Sbc (Devereux 1987, Pogge 1989, Ho et al. 1997), although there are examples outside of this type range. Altogether, the frequency of circumnuclear rings among the core population of massive, intermediate-type barred galaxies is of order 20%.

Quantifying the star-formation statistics in less spectacularly star-forming galaxy centers is more difficult. In early-type galaxies, the typical levels of extended disk star formation are relatively low (Kennicutt 1998a and references therein). Any nuclear star formation stands out. However, in the gas-rich, late-type spirals that dominate the total star formation in the local universe, it can be difficult to distinguish central star formation that might be associated with pseudobulge growth from the background of general disk star formation.

CO interferometer surveys give a clearer picture. Some of the more comprehensive aperture synthesis CO surveys include studies by Sakamoto et al. (1999), the BIMA Survey of Nearby Galaxies (SONG), (Regan et al. 2001, Helfer et al. 2003, Jogee 1998, Jogee et al. 2004). Studies have also been made by Kenney et al. (1992), Schinnerer et al. (2002, 2003) and many others. Larger samples of galaxies have been observed in the ¹²CO (1-0) and (2-1) rotational lines using single-dish telescopes, with typical beam diameters of 11-50" (Young & Devereux 1991; Braine et al. 1993; Böker, Lisenfeld & Schinnerer 2003a). A survey in HCN that includes a large subsample of normal galaxies is given in Gao & Solomon (2004).

These surveys show that central molecular disks are common but not universal. The disks are found more frequently in barred spirals, and their masses tend to be higher in barred systems (Sakamoto et al. 1999, Helfer et al. 2003). Although a few systems show centrally peaked distributions that might be similar to the exponential profiles of stellar disks or pseudobulges, the predominant structures are barlike distributions, bipolar twin-peak distributions (Kenney et al. 1992), circumnuclear rings, spiral arms, or combinations of these structures. In many of these, the gas is unlikely to be in steady-state equilibrium, and the interpretation is complicated by the likelihood that variations in temperature influence the CO emission distributions. We can conclude only that the gas disks have radii that are characteristic of central bars, bulges, and pseudobulges.

The first CO observations that spatially resolved galaxies showed that the distribution of molecular gas often follows the starlight (e.g., Young & Scoville 1991). Recent observations confirm this result (Regan et al. 2001, Böker et al. 2003a). This occurs even when the stellar brightness rises steeply toward the center, above the extrapolated profile of the outer exponential disk. Five examples are shown in Figure 20. They are all excellent examples of objects in which a bar (top row), oval (middle row), or global spiral structure that reaches the center (NGC 4321 in the bottom row) provides an engine for inward gas transport. For all of these, the molecular gas is very centrally concentrated. Because star formation rates increase faster than linearly with gas density (Figure 21), the observation that the molecular gas density follows the starlight guarantees that star formation will further enhance the density contrast between the (pseudo)bulge and the outer disk. We have discussed several of these objects as typical pseudobulges. The exception in Figure 20 is NGC 7331, a galaxy that contains a probable classical bulge.

Figure 20. Radial profiles of CO and stellar K-band surface brightness from the BIMA SONG (adapted from Regan et al. 2001). CO surface brightness is in magnitudes of Jy km s-1 arcsec-2 with zeropoint at 1000 Jy km s-1 arcsec-2. The stellar surface brightness profiles have been shifted vertically to the CO profiles. Morphological types are from the RC3. NGC 2903 and 3627 are clearly barred in the K-band images shown in Regan et al. (2001). NGC 2903 and 4736 are oval galaxies (Section 3.2). NGC 4736 contains a prototypical pseudobulge; it is also illustrated in Figures 2, 8, and 17. NGC 4321 is an unbarred galaxy with no ILR (discussed in Section 3.4). All galaxies in this figure except NGC 7331 have structures that are expected to cause gas to flow toward the center. NGC 7331 is included to show the very different CO profile in a galaxy with a probable classical bulge.

Figure 21. Correlation between SFR surface density and total gas surface density for 20 circumnuclear star-forming rings ( filled squares with error bars) compared to disk-averaged values for 61 spiral galaxies ( filled circles) and the centers of these galaxies (open circles). The circumnuclear data are compiled in this paper, and the comparison data are from Kennicutt (1998b). The solid diagonal lines show constant gas consumption timescales (increasing downward) of 0.1, 1, and 10 Gyr.

When the data from the aperture synthesis surveys are combined with small-beam, single dish measurements (Braine et al. 1993, Böker et al. 2003), the gas masses show a large range, from ~ 10⁶ M to 2 × 10⁹ M. Here a standard CO-to-H₂ conversion factor, X_CO = 2.8 × 10²⁰ cm^-2 (K km s^-1)^-1, has been used. If instead we use a variable, metallicity-dependent conversion factor (e.g., Wilson 1995, Paglione et al. 2001, Boselli et al. 2002), then this range narrows to ~ 10⁷-10⁹ M (Böker et al. 2003a). Pseudobulges are expected to grow to masses at least as large as these. More massive pseudobulges would result if gas continues to be added to the nuclear disks.

Studies of the SFRs in individual systems are too numerous to be listed here, but representative studies include Kennicutt, Keel & Blaha (1989), Pogge (1989), Phillips (1993), Maoz et al. (1996, 2001), Elmegreen et al. (1997, 2002), Usui et al. (1998, 2001), Buta et al. (2000), Colina & Wada (2000), Inoue et al. (2000), Alonso-Herrero & Knapen (2001), Ryder et al. (2001), Benedict et al. (2002), and Knapen et al. (2002). In these studies a variety of star-formation tracers have been used, including measurements of ultraviolet and infrared continua, and H, P, Br, and other hydrogen recombination lines (see Kennicutt 1998a). The Spitzer Space Telescope will have a strong impact on this subject by providing spatially resolved maps of the thermal-infrared dust emission from these regions.

SFRs measured by different authors are generally consistent at the factor-of-two level; this is comparable to the uncertainties that are typically quoted for these highly dust-attenuated regions. Although this limits the reliability of SFRs for any individual object, good measurements are available for ~ 40 galaxies, and this is sufficient to characterize the range of star-formation properties. The absolute SFRs within circumnuclear rings and disks range over a factor of about a thousand, from 0.01 to 10 M yr^-1. This covers the range of SFRs observed in our case studies and is comparable to the range observed in the integrated SFRs of normal spiral galaxies (Kennicutt 1998a and references therein). The central star formation accounts for 10-100% of the total SFR of spiral galaxies. The highest fractions occur in early-type galaxies, which typically have low SFRs in their outer disks (Kennicutt 1983, 1998a).

5.3. Constraining Evolution Timescales and Pseudobulge Growth

The data on SFRs and gas contents of the central regions of galaxies can be combined to constrain the evolutionary timescales and formation rates of pseudobulges. We first consider the prominent circumnuclear star-forming rings, which represent only the high-luminosity extreme of this activity, but for which we can derive relatively hard constraints. A search of the literature reveals 20 galaxies with circumnuclear star formation and reliable data on the SFRs, central gas masses, and sizes of the star-forming regions. For each galaxy, we derived the mean molecular gas surface density (using standard CO-H₂ conversion factors) and the mean SFR surface density within the circumnuclear regions. These are plotted as filled squares with error bars in Figure 21. The large error bars reflect considerable uncertainties in the SFRs owing to dust extinction and possible AGN contamination and uncertainties in the CO-H₂ conversions that provide the gas masses. Ignoring atomic gas introduces another uncertainty, but this is expected to be of order 10% or less. Figure 21 also shows the disk-averaged SFR and total gas densities for 61 spiral galaxies (solid circles), and the same data for the centers of those galaxies when spatially resolved data are available (open circles).

Figure 21 clearly shows that the circumnuclear rings populate a unique regime of molecular gas density and SFR. They extend the Schmidt SFR power law that is seen in the other galaxies (Kennicutt 1998b). For the sake of consistency, we adopted the same standard CO-H₂ conversion factor for all of the points; adopting a lower conversion factor for the centers would move the filled squares and open circles to the left by as much as 0.3 to 0.5 dex. This would increase the best-fitting Schmidt-law slope from N ~ 1.4 to N ~ 1.5.

The same diagram can be used to constrain the timescale on which the circumnuclear gas disks turn into stellar disks. Figure 21 shows lines of constant gas consumption times of 0.1, 1, and 10 Gyr. The outer star-forming disks of these galaxies are characterized by mean star-formation efficiencies of approximately 5% per 10⁸ yr and gas depletion times of ~ 2 Gyr on average. However, the star-formation efficiencies in most of the circumnuclear disks are much higher, of order 10-50% per 10⁸ years. Gas consumption timescales are 0.2-1 Gyr. If we assume that we observe the average disk at the midpoint of a gas accretion and starburst episode, this means that the typical formation timescales for these pseudobulges is approximately 0.4-2 Gyr, consistent with the (luminosity-averaged) star cluster ages of 0.0-0.3 Gyr inferred in the HST studies cited earlier. If a lower CO-H₂ conversion factor is appropriate in these regions, we overestimate the molecular masses, and the inferred timescales would be lowered by a comparable amount.

We can now construct a rough picture of the growth rates of pseudobulges in present-day spirals. The results above imply a typical formation timescale for the central disks of ~ 1 Gyr. When we combine this with a typical SFR range of 0.1-10 M yr^-1, gas accretion episodes will form pseudobulges with masses of order 10⁷-10¹⁰ M. Typical systems like the examples in Figure 8 fall in the 10⁸ to 3 × 10⁹ M range. Circumnuclear star-forming rings of this type are seen in ~ 10% of intermediate-type spiral galaxies (Sérsic 1973, Maoz et al. 1996). As discussed above, lookback studies suggest that strong bars first formed at least 5 Gyr ago. Combining these numbers suggests that about half of unbarred spirals and nearly all barred spirals may have formed pseudobulges in this mass range. Of course, this rough calculation is subject to a chain of possible systematic errors. However, it demonstrates the plausibility of a scenario in which pseudobulges are a common or even ubiquitous constituent of intermediate Hubble-type, massive spiral galaxies.

So far, our results are based solely on the occurrence of the most prominent circumnuclear star-forming rings in barred galaxies. Is there independent evidence based on the statistics of central molecular gas disks for lower levels of star formation in central pseudobulge disks? To make such an estimate, we used the BIMA SONG survey (Helfer et al. 2003) to derive the median central molecular gas surface density in their unbiased sample of 44 nearby spiral galaxies. This is ~ 200 M pc^-2, for a standard CO-H₂ conversion factor. This gas density already exceeds the central stellar surface density in most local spiral disks, which, from surface brightness measurements and M/L_B 1, is ~ 130 M pc^-2. Consequently, if the current gas disks are converted into stars, the central surface brightness of the disk more than doubles. If the gas infall continues for a few Gyr, a still brighter stellar component should form.

We can make a similar calculation by using Figure 21 to estimate the typical SFR densities in the centers of barred galaxies, and combine it with the typical star-formation timescales derived earlier to estimate the total surface density of stars formed. This calculation is not entirely independent, because the star-formation timescales are partly derived from the measured molecular gas densities. However, there are independent constraints on the star-formation timescales from HST studies of star clusters in circumnuclear starbursts and from photometric constraints derived from integrated light. For a typical SFR density of 0.1-1 M yr^-1 kpc^-2 (Figure 21), we expect to build up central stellar densities of 50-500 M pc^-2 for star-formation lifetimes of 0.5 Gyr, and 500-5000 M pc^-2 if the feeding of gas from the bar persists for 5 Gyr. This compares to Freeman (1970a) disk central densities of ~ 100-250 M pc^-2 for M/L_B = 1-2. The total masses in these components are of the same order as the observed molecular gas disks in the centers of these galaxies, 10⁷-10⁹ M, if there is no continued feeding of the nuclear disks, and may be up to 5 times larger if the gas feeding persists for 5 Gyr at a rate that is sufficient to replace the mass lost from star formation.

We reiterate that there are large uncertainties in these numbers. The most important uncertainties are the total duration of the inward gas transport in the bars, the CO-H₂ conversion factors used to estimate the molecular gas masses, and uncertainty in separating psudobulge star formation from steady-state disk and/or nuclear star formation. However, we believe that in a typical barred spiral, the total central star formation that results from secular gas inflow can easily exceed that in the underlying disk. By the same token, even the high end of the mass ranges described here falls 1 to 2 orders of magnitude short of the massive bulges that are typical of giant S0-Sab and elliptical galaxies.