4.1. The dense gas content of the W 3 GMC
Since star formation requires dense gas, the star formation efficiency of giant molecular clouds is determined in part by the fraction of the total cloud mass in dense cores. We have searched for such dense cores where the strong CO emission of the HDL has been detected (cf. Fig. 1). We found that strong emission (cm-3 , integrated temperatures K km s-1 ) fills 9% of our W 3 survey, and we have found that the strong emission is distributed into three dense cores. The sum of the virial masses for the three cores is 3300 . To compare the total virial mass to the total mass of the low density gas traced by the CO, we use the CO (1-0) data presented in Fig. 1. Converting this data from Heyer et al. (1998) to and using the standard conversion of N(H2) [cm-2 ] = dv [K km s-1 ] (Rohlfs & Wilson 1996), we calculate the total mass traced by CO in our surveyed region to be . Thus, for our surveyed region we find the dense cores to have 20% of the total mass traced by CO. This concentration of the dense gas into a few cores has two implications: first, that ongoing star formation is concentrated in a very small area of the GMC, and second, that the total star formation efficiency of the cloud is low in part due to the small fraction of the total gas mass in the dense cores.
One of the most extensive dense gas surveys to date is the CS (2-1) survey of the Orion B cloud by Lada et al. (1991). It covers a wide strip extending over from the dense star-forming regions NGC 2071, NGC 2086, and M78 in the north of the Orion B cloud to the star-forming cores NGC 2024, NGC 2023 and the Horsehead Nebula in the south. Thus, both the Orion B and W 3 HDL survey are similar in that they both cover a long and narrow strip which includes the strong CO emission toward known star-forming regions and the weak CO emission in between these regions. The Orion B survey covers four times the spatial area of our W 3 survey. The Orion B survey shows that the dense cores (cm-3 , integrated CS temperatures 0.5 K km s-1 ) cover only 10% of the surveyed area and % of the mass in this area. Although there may be significant uncertainties in the mass determinations, this is comparable to the area (9%) and mass fraction (20%) of dense gas we find for our surveyed region of the W 3 GMC.
Despite the apparent similarity of the two GMCs, the large number of compact and ultracompact H II regions detected toward W 3 suggests a much higher high-mass star formation rate than in the Orion B cloud, where only one compact H II region, NGC 2024, is observed. We suggest that the difference in the apparent formation rate of high-mass stars may be related to two significant differences in the global properties of these two GMCs:
Firstly, although the region surveyed in W 3 is a quarter of the size of the region surveyed by Lada et al. (1991) in Orion B (55 pc2 for W 3 vs. 220 pc2 for Orion B), the total CO mass in the two regions surveyed is comparable. Thus, the column density of molecular gas toward the W 3 GMC appears to be much higher. This may be the result of a higher average volume density in W 3.
Secondly, the total mass of the dense cores detected in toward W 3 and in CS toward Orion B is the same, but the mass is contained in fewer and more massive cores in W 3. The virial masses of the W 3 West and W 3(OH) cores ( ) are significantly higher than the virial masses of the most massive cores ( ) identified by Lada et al. (1991) toward NGC 2071, NGC 2068, M78, NGC 2024, and NGC 2023.
Since the maps of the Orion B cloud and W 3 GMC were made at different spatial resolutions and in different tracers, it is conceivable that the higher virial masses observed in the W 3 GMC may reflect an observational bias and not intrinsic differences in the core masses. The resolution of the W 3 GMC map is lower than the Orion B map (0.16 pc spacing for a beam of 0.21 pc FWHM toward Orion B vs. 0.22 pc spacing for a beam of 0.45 pc FWHM toward W 3); however, the most massive and actively star-forming cores in the Orion B cloud have diameters of pc and thus would be resolved by our survey if located at the distance of W 3. The higher masses of the cores may also be due to excitation. The inversion transition has a lower critical density ( cm-3) than the CS (2-1) transition ( cm-3); consequently may trace a larger, lower density, and more massive volume of gas. However, in W 3 West, observations in the C18O (2-1), C34S (3-2) and C34S (5-4) lines by Tieftrunk et al. (1995) found densities of cm-3 and a mass of 1400 . Thus, in the case of W 3 West, the appears to be primarily tracing gas with densities sufficient to excite strong CS (2-1) emission.
Finally, in addition to the three dense cores, we have detected weak emission extending over one fourth of the surveyed region. This weak emission is concentrated in regions of strong to moderate CO emission (cf. Fig. 1). Our analysis of one such region suggested that the emission from these regions is weak simply because it has a lower column density and hence, a lower volume density than the cores. This suggests that these regions are tracing gas with densities just sufficient to excite the inversion transition ( cm-3 ). However, the small angular size of some of the weak emission regions suggests that beam dilution may also contribute to the weakness of the emission. Observations in other molecular gas tracers (13CO, CS) are needed to elucidate the properties of these emission regions and determine if they trace bound clumps (as suggested by our analysis of one example region) or transient density enhancements in a turbulent cloud.
4.2. Star formation and ammonia cores toward W 3
A comparative study of the global distribution of dense gas and young stars in the Orion B cloud was discussed by Lada (1992). Lada concluded that the majority of young stars detected in the near-infrared survey of Lada et al. (1991) were found in clusters associated with three of the five most massive cores detected in the CS (2-1) survey of Lada et al. (1991). The study implied that most stars form in clusters and that these clusters occur in the most massive cores.
A key goal of our survey was also to study the relationship between the dense gas in the W 3 GMC as traced by and the distribution of known sites of recent or ongoing star formation as identified by H II regions, stellar clusters, and nebulosities. First, we note that all these tracers of star formation toward W 3, except the more evolved and optically visible IC 1795 H II region and its cluster, are found within 1 pc of one of the three dense cores.
On scales smaller than 1 pc, closer examination of W 3 Main (Fig. 3) shows a more complex picture. The sites of recent star formation, as distinguished by ultracompact H II regions and NIR sources (or OH and H2O masers, see Harris & Wynn-Wiliams 1976), appear to be anticorrelated with the cores. In the case of W 3 West, the H II regions and NIR sources are found along the perimeter of the core, which appears as a dark cloud in the K' mosaic (cf. Fig. 4). The core W 3 SE also appears in a region of obscuration in the -band data. As we noted in Sec. 3.2, the detection of the "jet" toward W 3 SE indicates the presence of at least one site of ongoing star formation. However, toward W 3 SE there are no H II regions or clusters (the -band data shows several stars toward W 3 SE, but it is not clear whether these are embedded stars or unrelated stars in the line of sight; cf. Fig. 4).
A simple explanation for the anticorrelation is that the stars embedded in these cores are hidden by the enormous extinctions implied by the derived H2 column densities. These stars then become visible once the dense gas is dispersed, leading to the appearance of a stellar cluster. Although this is a pausible explanation for the anticorrelation observed in the NIR data, we believe that the high extinction cannot be the only reason for the anticorrelation for two reasons:
Firstly, the VLA maps (cf. Fig. 3) should not be affected by extinction, but toward the center of W 3 West and toward W 3 SE there is no evidence for recently formed O-type or early B-type stars (Tieftrunk et al. 1997). In contrast, toward the W 3 East core, which does not exhibit strong emission, a cluster of ultra- and hypercompact H II regions signals the presence of several recently formed OB stars (Tieftrunk et al. 1997).
Secondly, in the case of the W 3 East core, which shows a rich cluster of young stars and H II regions, the anticorrelation between the NIR cluster and ammonia appears to be the result of a lower relative abundance and not the dispersal of the dense gas. Although W 3 East does not show strong emission, the presence of dense gas in the W 3 East cluster has been inferred from a number of molecular tracers (e.g. in 13CO: Roberts et al. 1997; in C18O : Tieftrunk et al. 1995, Oldham et al. 1994; in C34S : Tieftrunk et al. 1995; in CO Hasegawa et al. 1994). Toward W 3 East, a rough correlation between the column density of dense gas and the observed stellar density of detected K-band sources is found by Megeath et al. (1996). This correlation is evident despite ten magnitudes of extinction at 2.2 µm through the W 3 East core, indicating that the stellar cluster is apparent in the -band data due to the high density of embedded stars in the core (Megeath et al. 1996).
Thus, in the case of W 3 East, the ammonia gas appears to have been preferentially "thinned out" without the dispersal of the dense molecular gas. We note that similar abundance anomalies have been measured toward many other IR sources with associated outflows; these anomalies were most recently discussed by Davis & Dent (1993) based on observations of HH34 IRS in L1640. As in the case of W 3 East, Davis & Dent (1993) find that the ammonia does not trace the hot dense gas core to- ward the IR source, which is traced in HCO+ or CS, but rather that the relative abundance is reduced by an order of magnitude toward the IR source when compared to emission regions away from the source.
The difference between the W 3 East region and W 3 West is also evident in a compilation of infrared and submillimeter observations by Ladd et al. (1993). This data shows the presence of strong mid-infrared emission indicative of hot dust toward W 3 East. In contrast, the center of the W 3 West region (SMS 3 in the nomenclature of Ladd et al. 1993) is very luminous in the submillimeter, but has not been detected at wavelengths shorter than 100 µm, indicating a lack of hot dust. The peak of the emission toward W 3 West is coincident with this SMS 3 submillimeter core. The lack of hot dust indicates that there are not as many young stars heating W 3 West or perhaps, that the embedded stars are still deeply enshrouded in their natal cocoons.
The W 3 Main cloud appears to have undergone an extended period of star formation, generating the extensive cluster of H II regions and stars observed in our images. Many of the compact H II regions in W 3 Main appear to be ionization bounded regions which have dispersed their natal dense cores and are confined by the pervading low to moderate density gas traced by the CO (1-0) emission (Tieftrunk et al. 1997). We speculate that star formation in the W 3 GMC has been occurring in small bursts. Previous bursts may have completely disrupted their natal cores resulting in the compact H II regions and the "halo" of stars discussed in Megeath et al. (1996). An ongoing burst appears to be occurring in W 3 East. The remaining dense cores W 3 SE, with its jet, and W 3 West, with its luminous submillimeter source, are probably forming stars at either a slower rate or over a shorter time interval than W 3 East. These cores may be at the early stages of a burst or they may be inherently more quiescent regions.
The W 3(OH) region shows a much different morphology than the W 3 Main region. Here we find an extended molecular core which appears to connect all the apparent sites of star formation. The three clusters in the -band mosaic are located in two clearly separated regions. As in the case of W 3 Main, there is an anticorrelation between the newly detected clusters and the emission: the two newly detected stellar clusters are found at the edge of the plume, and strong emission is not found toward the cluster centers. The cluster associated with the H II region W 3(OH) is found displaced from the peak molecular core emission toward a notch in the contour map (cf. Fig. 6). Photometric data is necessary to show how deeply the stars in this cluster are embedded, but the presence of an ultracompact H II region (pc) and masers toward the core seems to indicate that the cluster may be just emerging from its natal cloud (Guilloteau et al. 1985). Since we do not yet have complementary C18O and CS data, it is unclear whether the anticorrelation toward the newly detected clusters and the notch toward the central cluster are the result of the dispersal of the dense gas, or the preferential "thinning out" of
In summary, although we find that most sites of star formation are associated with dense molecular cores, on a sub-parsec scale we find evidence for an anticorrelation between these sites of star formation and the peaks of the cores. We expect that this is in part due to the dispersal of the dense gas by winds and UV radiation. However, we have found one case, W 3 East, where the ammonia has been "thinned out", but the dense core is still detected in other tracers. Since W 3 East is distinguished by the amount of star formation apparent in the core, we propose that the winds and radiation from the embedded stars and the expanding H II regions have had a significant impact on the chemistry of this region and that an enhanced rate of destruction may have resulted in the observed underabundance. Since we do not have C18O maps to- ward W 3(OH), we do not know if there are other examples of regions with depleted abundances in W 3. Nevertheless, our work suggests that surveys for dense star-forming cores in may be influenced by large changes in the relative abundance which appear to be correlated with the amount of recent star formation. Large scale surveys in multiple tracers, such as CS, HCO+ , C18O and/or HC3N, are needed to distinguish between variations in cloud density and differences in chemistry.
© European Southern Observatory (ESO) 1998
Online publication: July 27, 1998