8 Discussion

8.1 Clump summary

We have shown that the clumps are composed of many bright super star clusters, and that many of these are less than 6 Myr. This conclusion remains, despite the difficulties in deriving ages below 7 Myr due to nebular emission and unknown geometry. The nuclear region contains a bar of star clusters, some of which are certainly young, and others which may be old, but are likely also young and suffering from dust extinction.

Clump B is a luminous, high surface brightness, and very young region, which has some evidence for an older, underlying stellar population. The young star clusters have ages less than 6 Myr, with masses ranging from $5 \times 10^{4}~M_{\odot}$ to $5 \times 10^{5}~M_{\odot}$. For these 8 clusters, we have a combined mass of approximately $8 \times 10^{5}~M_{\odot}$.

We have shown in Sect. 7 that $33 \%$ of the flux at F255W is coming from these clusters, thus 2/3 of the massive star formation is below our detection limits. We can use this to estimate the SFR in this region over the last 6 Myr, if we assume a constant M/L. If in the 1/3 of the light that we observe we have $8 \times 10^{5}~M_{\odot}$, and we are missing $\sim$2/3 of the massive stellar flux, this implies that the total mass formed in this region is $\sim$ $2.4 \times 10^{6}~M_{\odot}$. With a radius of 0.4 kpc, Clump B then has a SF intensity of 0.8 $M_{\odot}$ yr^-1 kpc². This incredible SF density puts it in the league with the most vigorously star-forming galaxies in the universe. Lanzetta et al. (2002) found objects at $z \ge 3$ with similar SF intensities over significantly larger areas than that of Clump B. However, Clump B still stands out as a large and intense extra-nuclear star forming site that can provide insights into the consequences of intense star formation on relatively large spatial scales.

Figure 15: The two-filter cluster sample within the main galaxy body. Orientation is as for Fig. 4. From left of right: blue clusters ([555-814] < 0.4), neutral (0.4 < [555-814] < 0.7), and red ([555-814] > 0.7). The field of view for each panel is 40 arcsec, or 10 kpc at the distance of NGC 7673.

Clump C is undoubtably young judging from its blue overall [255-555] color and those of the composing clusters. It may have embedded clusters judging from the combination of H$\alpha$ emission and the anomalous cluster colors in our two-color diagrams.

Clump D is composed of fainter clusters, pointing to an older age and possibly less massive clusters. The derived age of the brightest cluster is between 9 and 13 Myr, and there are regions to the NE and SE with H$\alpha$emission, characteristic of young star formation regions, i.e. those with ages of less than 10 Myr. We suggest that this is evidence of propagating star formation, which should be followed up with high spatial resolution spectroscopy.

Clump F is not a clump but an H II region dominated by a single cluster with an age of 4-5 Myr. This object is extremely bright, and the disparity between the F255W and F555W structures indicates smaller scale structure than we can resolve associated with dust or multiple ages.

The clumps in NGC 7673 appear to represent a step beyond normal OB associations in terms of their hierarchy of compact, gravitationally bound objects. In a simple OB association, the rare massive stars, some of which are either binary or multiple systems, define the upper end of the hierarchy. They contain only a modest fraction of the stellar mass, and are surrounded by more numerous, less massive single and multiple stars, which contain most of the mass.

In starburst clumps, compact star clusters define the top of the compact, bound mass distribution, and contain a significant fraction (more than 16%) of the recently formed stars (see Sect. 7). These in turn are embedded in more diffuse fields of massive stars, which resemble rich examples of OB associations. Since high redshift galaxies also show signatures of starburst clumps, this mode of star formation, while rare in the current epoch, may have played an important role in building galaxies. As one of the nearest optically accessible clumpy starburst galaxies, NGC 7673 provides an excellent opportunity for further exploration of intense, large scale star formation.

8.2 Starburst picture: Age of the current burst

HG (1999) discussed the starburst trigger in detail, concluding that either the burst was triggered by an interaction with its companion, NGC 7677, or by the consumption of a small galaxy less than $10\%$ of its mass in a minor merger. Neither of these scenarios can be ruled out, but in both cases the model must be able to account for the following:

1. The H I map of this pair shows a few appendages to the disk of NGC 7673, and a small extension from NGC 7677 pointing toward its companion, but otherwise is surprisingly regular.

2. NGC 7673 has an outer optical shell. This type of feature is usually associated with merger candidate E, S0, and only a few Sa galaxies (Schweizer & Seitzer 1988), but has also been found around the starburst galaxy NGC 3310, which is thought to be triggered by a minor merger (Mulder & van Driel 1996). Theoretical work indicates, however, that interaction scenarios other than mergers can produce arc-like features in outer disks like the one seen in NGC 7673 (Hernquist & Quinn 1989; Hernquist & Spergel 1992; Howard et al. 1993).

3. There is a difference in timescale between the formation of this outer disk ripple and the very young starburst occurring in the inner disk. The outer disk "ripple'' is characteristic of the late phases of an interaction, yet the starburst remains quite active. For example, if the starburst is associated with a close passage of NGC 7673's disturbed neighbor, NGC 7677 at a projected distance of 95 kpc, then the triggering event probably took place several hundred Myr in the past.

That O stars are present through out the galaxy is clear from the prominence of H II regions, indicating star formation within the last 10 Myr (see HG). Duflot-Augarde & Alloin (1982) also found evidence for an underlying older stellar population within the nuclear region from a weak 400 nm continuum break and Balmer absorption underlying the strong emission lines. However, the digital spectrum of the nuclear region taken by Gallagher et al. (1989) suggests that the 4000 Å "break'' is probably due to a reddened younger population. Duflot-Augarde & Alloin also found weak evidence from the G band for the presence of an evolved stellar component in Clump B. This is supported by the fraction of flux at F555W from cluster as opposed to F255W. So the starburst history in this galaxy is complex.

Our two-color diagrams are useful in age-dating the brightest regions, although the rapid fading and sensitivity to dust of the F255W flux means we have only a small number of the entire cluster sample for this type of investigation. With this we find evidence for very young, <6 Myr star clusters in Clump B, Clump C, and the nuclear region.

The color magnitude diagram for the entire cluster population shows a broad range in [555-814] and V-I color, spanning the allowed model color range fully. It is difficult to constrain ages with only this color, although we do note that the mean is comparable to other starburst systems with burst ages less than 30 Myr (such as the Antennae and NGC 7252), and one as young as 4 Myr (NGC 1741).

An overview of the distribution of cluster ages is illustrated in Fig. 15. This shows the bluest star clusters in [555-814], whose intrinsic colors imply ages of <20 Myr, are concentrated in the blue blob, Clump B, with some presence in Clumps C and F. These objects avoid the inner parts of the starburst, yet Clump A also displays strong H$\alpha$ emission (see Fig. 5). The possible conclusions are that the H-ionizing stars in Clump A are distributed in the field rather than in compact star clusters, or dust is an important factor.

The middle panel of Fig. 15 shows star clusters with colors consistent with ages of about 20-30 Myr up to 1 Gyr if they are unreddened or moderately reddened younger star clusters. Note these are absent in Clump B, where we see little evidence for dust; Clump B apparently is a young feature. The combination of "blue'' and "neutral'' star clusters delineate the main features of NGC 7673. The strong luminosity bias that favors observation of younger star clusters, the coincidence between the locations of these clusters and H$\alpha$ emission, and the results from the F255W filter observations lend additional support to the view that the clumps in NGC 7673 are relatively young, with ages of $\le$50 Myr.

In this interpretation the upper age bound is set by Clump D, which has weak H$\alpha$ emission, faint star clusters with similar colors, and comparatively red global colors (Table 5). The third panel of Fig. 15 shows cluster candidates which either are highly reddened and young, luminous objects, or have colors of star clusters with ages of 1 Gyr or more. We are encouraged that this poorly defined sample of redder objects is not associated with the main clumps. Whatever the red objects are (possibly a mixture of background galaxies, clusters, and objects with large color errors), they do not seem to be associated with the ongoing starburst.

Using the clump colors in Table 4, we can estimate ages, which will lie between the values predicted by instantaneous and continuous SFR models. We again use the Starburst99 models. Clump B then must predate the red supergiant flash, and is younger than about 8-10 Myr. Clump C is 10-15 Myr old, while Clump D is 15-50 Myr in age. These ages refer to the mean stellar population age. They are not corrected for any internal reddening, and including these would reduce the estimated ages.

8.3 Evolution of the starburst clumps

Huge regions of active star formation, such as those seen in NGC 7673, presumably reflect the presence of supergiant complexes of gravitationally bound interstellar gas clouds. These can form when a galactic disk becomes Jeans unstable. The Jeans mass for a rotating gaseous disk scales as $M_{\rm J} \propto \overline{\sigma_{\rm g}}^{2}/\mu_{\rm g} \propto \mu_{\rm g}^3 Q_{\rm g}/ \kappa^3$, where $\mu_{\rm g}$ is the gas disk's surface density, $\overline{\sigma_{\rm g}}$ the velocity dispersion, $\kappa$ the orbital epicyclic frequency, and $Q_{\rm g} \sim 1$ the Toomre disk stability parameter (Elmegreen et al. 1993; Noguchi 1999). Galaxy interactions favor the formation of super cloud complexes in two ways: (1) They can increase the gas velocity dispersion $\overline{\sigma_{\rm g}}^{2}$ (Elmegreen et al. 1993). (2) They tend to drive gas inwards, either as a direct result of the interaction, or indirectly through the presence of tidally-induced bars (e.g., Noguchi 1987; Barnes & Herquist 1996). We might expect to find starburst clumps in disturbed, gas-rich disk galaxies, such as NGC 7673.

Our study of NGC 7673 suggests it contains two classes of clumps. Clump A, the nuclear region, lies near the middle of an offset stellar bar. It is a special place, where dynamical friction and dissipation can pile up material that has been transported inwards, possibly leading to the formation of a bulge (Noguchi 1999). The other clumps are located beyond the bar. A key question is whether the outer clumps can survive for sufficiently long to carry their material inwards to make a bulge upon their disruption near the center of the galaxy, as hypothesized by Noguchi (1999).

We take Clump D, with $M_B \approx -$16.3 as an example. Using the Starburst99 models and a Kroupa stellar IMF (2002), we estimate that (M/L) $_V \approx 0.03$ for a constant SFR age of 30 Myr. Clump D then has $M \ge 10^7~M_{\odot}$ or about 0.25% of the NGC 7673 dynamical mass; this mass is a lower limit as we have not included any gas. Taking the surface density of Clump D to be 5 times that of the mean disk, the Noguchi (1999) model then predicts the dynamical friction time scale for Clump D to reach the center of the galaxy is a few orbital periods, or >100 Myr.

While the non-nuclear clumps have a range in ages, with Clump B being the youngest, none shows compelling evidence for star clusters with ages of $\ge$100 Myr. This in part reflects the shallow nature of this initial survey of NGC 7673, which means we will miss any but the most massive older star clusters (see Sect. 6). However, we have seen that the integrated V-I colors of the outer clumps also imply ages of <100 Myr. Furthermore, we do not see evidence for inward migration of older clumps; both the young Clump B and older Clump D are located at similar radii. Our data therefore do not reveal the presence of long-lived clumps in NGC 7673, and so possibly, as also discussed by Noguchi (1999), this is a case where the clumps are marginally bound and less durable than in protogalaxy where more of the mass is in the form of gas.

However, we also recognize that we are observing NGC 7673 well after whatever interaction triggered its starburst. Perhaps earlier generations of clumps were more robust and already have accreted into the central region of the galaxy? After all, more than 100 Myr are likely to have passed since the starburst was triggered, an adequate amount of time for dynamical friction to act. In this case we would expect Clump A to contain an unusually wide spread in star cluster ages from the previously dissolved clumps. Unfortunately, the interpretation of star cluster colors in the nuclear region is complicated by a combination of crowding and dust. Despite these complications, Fig. 15 shows that the redder clusters are not overly concentrated in Clump A, as might be expected if a few large clumps had dissolved in the not too distant past.

8.4 Final fate of NGC 7673

What will ultimately happen to NGC 7673? In their study of LBCGs, Pisano et al. (2001) calculate $\tau$ $_{{\rm gas}}$, the H II mass divided by the star formation rate, to give a simple estimate of the gas depletion timescale. Using their H$\alpha$ luminosity as the SFR indicator, they find SFR =  $23.5~M_{\odot}$  ${\rm yr}^{-1}$ with the Kenicutt (1983) calibration, yielding $\tau$ $_{{\rm gas}}$ = 0.18 Gyr. If we use a more modern SFR calibration from combining a Kroupa parameterization of the IMF with Starburst99 Lyman continuum luminosity predictions, then we find SFR  $\approx 12~M_{\odot}$ yr^-1 from the Pisano et al. L(H$\alpha$). Using the $L_{{\rm FIR}}$ from Table 1, we get a similar result, and conclude that the current SFR is in the 10-20 $M_{\odot}$ yr^-1 range. Adopting $M_{{\rm gas}} = 1.3~M_{{\rm HI}}$ to allow for the presence of helium, we get an only slightly more conservative $\tau_{{\rm gas}} \approx$0.5 Gyr. This is short, and if NGC 7673 continues to form stars at the current rate, it will consume the rest of the available gas within the next 0.5 Gyr.

If it consumes all of its gas during its fast and furious starburst phase, then NGC 7673 will become a nearly gas-free disk galaxy, probably a small Sa or S0 system. If, on the other hand, the starburst is beginning its decline, this could allow a substantial fraction of the remaining $4 \times 10^{9}~M_{\odot}$ of H I to form stars more sedately. The resulting system would then appear to be a late-type galaxy. H I masses of $\sim$10 $^9~M_{\odot}$ are typical of small star-forming disk galaxies, and since NGC 7673 has been collisionally perturbed, we believe it would most likely have a somewhat thickened disk, such as those found in Magellanic spirals.