The Mysterious Ring in the Open Cluster NGC 3572: Planetary Nebula or Photoevaporating Globule?

The mysterious nebula in NGC 3572 shows an intriguing morphology in the [N ii] image in Figure 1 and in the Hα image published by Phelps & Janes (1991). It appears as a limb‐brightened elliptical shell or ring elongated northeast‐southwest, along with a small, almost circular inner ring. This is reminiscent of some elliptical PNe with equatorial toroids or rings. However, no star has been detected within this ring, down to V∼20 (according to photometry obtained by R. L. P.). There is a highly reddened V∼14.8 mag star projected at the western limb of the nebula that shows an emission spectrum with bright Ca ii H and K and Balmer lines (Fig. 5e).

Our Fabry‐Perot images in Figure 2 provide the first kinematic information for this object and show the nebular structure with higher spatial resolution than previous images. Figure 2a shows an image made by summing over all velocities for [N ii] λ6583, but Figures 2b–2f show several channel maps at selected velocities. At blueshifted velocities in Figure 2b, one can see that the northwest rim of the inner ring is brighter than at other velocities, while at redshifted velocities in Figures 2d and 2e, one sees several faint knots or filaments outside the bright rim of the nebula (at radii around 15^''), especially toward the south and east.

Figure 4 shows a position‐velocity plot for a thin (2^'') slice through the data cube, along the diagonal dashed line marked in Figure 3 (velocities in Fig. 4 are plotted with respect to the centroid for the integrated emission from the whole nebula). Differential radial expansion velocities are quite slow—most emission is found within a range of 30 km s ^-1, comparable to our spectral resolution. Nevertheless, there is some evidence in Figure 4 for a weak bipolar expansion pattern, in the sense that the relative velocity centroid of the northeast limb (feature a) appears to be redshifted by +4 km s ^-1, while the southeast limb at the other extreme edge of the nebula (feature d) is blueshifted by a similar amount (−3 km s ^-1). In contrast, emission from portions of the inner ring show the opposite trend—the northeast edge (b) is blueshifted by −7 km s ^-1, while the southeast edge of the inner ring (c) is slightly redshifted by only +1 km s ^-1 in Figure 4. Thus, features b and c might be construed as part of an expanding equatorial ring inside a larger elliptical or bipolar shell with the southwest polar axis tilted toward us. However, long‐slit spectra with higher spectral resolution are needed because the expansion velocities are apparently too slow for us to resolve emission components from the front and back of an expanding shell, if they exist. For the same reasons, the nebula's dynamical age is ambiguous.

Zoom In Zoom Out Reset image size

At all positions in Figure 4, the [N ii] λ6583 profile shows a faint redshifted component at +40 to +60 km s ^-1. The redshifted component is somewhat broader than the main emission line. This is probably contamination from the bright Hα line in an overlapping spectral order.

An emission‐line blob was detected roughly 70^'' west of the ring's center, along the southern slit position in Figure 3; its spectrum is shown in Figure 5a. For the ring nebula itself, the two‐dimensional spectrogram revealed an obvious positional variation in excitation across the ring; two positions of different apparent excitation (high, "HEx," and low, "LEx") were sampled from the sections of the slit indicated in Figure 3, and the resulting spectra are shown in Figures 5b, 5c, and 5d (c and d are the same LEx spectrum shown with different intensity scales). Figure 5e shows the spectrum of the star at the west edge of the ring extracted from the aperture section shown in Figure 3, after subtracting adjacent nebular emission.

Zoom In Zoom Out Reset image size

3.2.1. Reddening

Correcting the Balmer decrement in the raw nebular spectra to the theoretical case B ratios (Hummer & Storey 1987) implies a reddening of E(B−V) ≈ 0.4 (A_V ≈ 1.24), somewhat less than the average reddening derived from the cluster photometry of E(B−V) = 0.46 ± 0.06 (Moffat & Vogt 1975). Values of E(B−V) for individual stars vary as a result of differential reddening across the cluster. For example, referring to the stars near the nebula numbered in Figure 1, star 20 has a reddening of E(B−V) = 0.38, stars 2 and 21 have 0.4, star 3 has 0.41, and star 10 has 0.5 (Moffat & Vogt 1975); all these stars are thought to be members of the cluster. This suggests that the nebula is indeed at roughly the same distance as the rest of the group. The flux shown in Figure 3 for each position represents the total dereddened flux included in the corresponding rectangular sections of the slit drawn in Figure 3. Table 2 lists relative dereddened line intensities.

3.2.2. Excitation and Nebular Diagnostics

3.2.3. Elemental Abundances

The limited wavelength coverage of the spectra in Figure 5 allows only crude estimates of the nebular abundances. The only firm conclusion is that we find no significant evidence for overabundance of any element compared to cosmic abundances; this is perhaps most relevant regarding nitrogen in certain evolutionary scenarios. The He ⁺ abundance can be inferred from He i recombination lines using the relation

where λ is the wavelength of a He i line and the α^eff are the appropriate case B recombination coefficients (see Osterbrock 1989). The flux‐weighted average derived from λ4471, λ5876, and λ6680 in the LEx region is N(He ⁺)/N(H) ≈ 0.08, and that in the HEx region is about 0.12. Much of the helium may be neutral in the LEx region of the nebula, and no He ii λ4686 is seen in the HEx spectrum, so 0.12 is probably a better estimate.

Estimating chemical abundances in the gas from collisionally excited optical lines is difficult given the limited information in Table 1; such lines are sensitive to temperature, and we have been able to derive only T_e from the [N ii] line ratio. Since spectra in the two distinct regions of our target show different excitation, it is likely that derived ionic abundances would be sensitive to ionization fractions, rather than differences in chemical composition. For instance, using the NEBULAR package in IRAF to calculate ionic abundances from observed line ratios, we find S ⁺ underabundant by factors of roughly 4–10 compared to normal cosmic abundances; it is likely that much of the sulfur is S ⁺⁺, since He i lines are seen. Abundances for N and O ions are probably more reliable in the LEx region, where we can measure lines of [N i], [N ii], [O i], [O ii], and [O iii]. For the LEx region, we find that the total nitrogen (N⁰, N ⁺) and total oxygen (O⁰, O ⁺, O ⁺⁺) abundances are comparable to those in the Orion Nebula (Esteban et al. 1998).

3.2.4. Mass of Ionized Gas

Using our velocity‐integrated [N ii] λ6583 image (Fig. 2a), we can estimate the fraction of the total flux included in the spectroscopic slit apertures used to measure the spectrum in Figures 5b–5d. Since the Hα image published by Phelps & Janes (1991) looks identical to our [N ii] image (allowing for differences in spatial resolution), we can infer a total Hβ flux of F_Hβ ≈ 2 × 10^-12 ergs s ^-1 cm ^-2 for the LEx region and 1.2 × 10^-12 ergs s ^-1 cm ^-2 for the outer HEx zone. With electron densities for these regions derived above, the total mass in each can be estimated using

where μ is the mean atomic weight per electron, taken to be ∼1.4 (see above), m_H is the proton mass, D = 2.8 kpc is the heliocentric distance, hν_Hβ is the energy of an Hβ photon, and α^eff_Hβ is the Hβ recombination coefficient, which is 3.02 × 10^-14 cm³ s ^-1 for T_e ≈ 10⁴ K and n_e ≈ 10² cm ^-3. Estimating the mass separately for the two different excitation regions and adding them together, we find a total visible ionized gas mass of roughly 1.4 M_⊙ for the ring nebula in NGC 3572.

Figure 6 shows the broadband infrared spectral energy distribution at the position of the ring nebula as measured by MSX and IRAS; fluxes are given in Table 1. (An MSX image of the region is shown in Fig. 7, where the object discussed here is identified by an arrow.) For comparison, we show a model for thermal continuum emission from two blackbodies with a λ⁻¹ emissivity law, appropriate for optically thin emission from normal interstellar grains at mid‐ to far‐infrared wavelengths. These fits are crude because they neglect contamination from emission features. In any case, the blackbody emission is a good first approximation, and it is likely that dust makes a substantial contribution to the observed emission. The dust mass at some temperature T required to produce the observed luminosity L can be expressed as

where a is a characteristic grain radius, ρ is the grain density, σ is the Stephan‐Boltzmann constant, and 〈Q〉 is a suitable average emissivity at the appropriate wavelengths. The warmer component at 180 K has a total luminosity of ∼80 L_⊙ (assuming a distance of 2.8 kpc) and requires an emitting dust mass around 2.3 × 10^-5 M_⊙, if we assume a∼0.01 μm, and we take 〈Q〉∼0.003 as a suitable average in the 8–30 μm range (see Draine 1985). Similarly, the 30 K component has a higher luminosity around 500 L_⊙ and requires a much higher emitting dust mass on the order of 0.07 M_⊙ for the same grain size and density, and 〈Q〉∼3 × 10^-4 from 60 to 100 μm. Assuming a conventional gas‐to‐dust mass ratio of 100, this would indicate a total mass for the nebula of roughly 5–10 M_⊙, nearly all of it in the cooler component.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 6 also shows a power law drawn through the infrared fluxes, which provides a fit as suitable as the blackbodies, given the limited spectral coverage. It is worth noting that if extended toward shorter wavelengths, this power law would pass through the observed optical continuum from the star at the edge of the nebula shown in Figure 5e. This provides circumstantial evidence that the star and nebula may indeed be associated, since hot stars obscured by their own heated circumstellar dust typically show power‐law spectra at red to mid‐infrared wavelengths (a good example is η Carinae; Rodgers & Searle 1967). This is due to the combined effects of the Rayleigh‐Jeans tail of the star, reddening from dust, and thermal emission from dust with a range of temperatures at various radii from the central engine.

In optical images, the ring in NGC 3572 shows a morphology reminiscent of PNe: it has an elliptical ring, which could be a limb‐brightened outer shell, and what looks like an inner equatorial ring. In our Fabry‐Perot images, it shows some outer condensations that are redshifted, and these resemble jets and ansae seen in many PNe (e.g., Balick 1987; Frank, Balick, & Livio 1996). However, aside from the optical morphology, evidence supporting the PN hypothesis is scarce, and severe contradictions arise.

The strongest and most obvious objection is that a PN in a young open cluster with O‐type stars contradicts conventional wisdom, as noted by Phelps & Janes (1991). PNe are the evolved remnants of low‐mass stars and would not reside in a cluster that is only a few Myr old. We have shown that the nebula is indeed within the cluster, because it is ionized from the outside by cluster members (see § 3.2.2). Therefore, we would need to accept the coincidence that this PN has drifted into the core of the open cluster at a time when we are lucky enough to see it there—this is a tall order and makes alternative explanations worth pursuing.

The nebula has lower characteristic excitation than other planetary nebulae, and it appears to be externally ionized, which is certainly not expected for a PN excited by a central hot white dwarf. No star is seen at the center of the ring down to about V = 20 mag, although one might expect a normal white dwarf to be that faint at a distance of ∼2.8 kpc and with interstellar extinction. There is a star within the nebula, but it is at the western edge, which would be quite unusual for PNe. Figure 4 indicates radial expansion velocities of ≲5 km s ^-1, while PNe characteristically expand radially at ∼20 km s ^-1. The mass derived from the Hβ flux is somewhat large, but not unreasonable for a PN; however, the total mass of roughly 5–10 M_⊙ implied by the infrared luminosity is far too high. These difficulties make the PN hypothesis hard to accept, but they are all consistent with an alternative interpretation, as discussed below.

Instead of the PN hypothesis, several clues hint that the peculiar nebula in NGC 3572 may be a photoevaporating molecular globule—a leftover fragment of the original molecular cloud that formed the open cluster. It could be analogous to Thackeray's globules in IC 2944 (Thackeray 1950; Reipurth et al. 1997) and to the many molecular globules seen in the Carina Nebula (Cox & Bronfman 1995; Brooks et al. 2000; Smith 2002). One of these objects in Carina is shown in Figure 8; in addition to its similar size and brightness, this object may be a particularly relevant comparison because it is irradiated from two sides—from the east by the Trumpler 16 cluster and from the west by the Trumpler 14 cluster—like the nebula in NGC 3572. While the object in NGC 3572 resembles some PNe, the molecular globule alternative cannot be ruled out simply by its apparent morphology.

Zoom In Zoom Out Reset image size

The clear advantage of this interpretation is that it circumvents the uneasy coincidence of finding an old PN near the center of a young open cluster. There is still some diffuse gas adjacent to the cluster; in addition to the diffuse Hα noted by Noumaru & Ogura (1993), a wide‐field MSX image of the region (Fig. 7) reveals diffuse thermal infrared emission. This image is in the MSX band A filter, which traces the 8.6 μm PAH emission feature commonly seen from photodissociation regions on the surfaces of irradiated molecular clouds. Therefore, the suggestion that remnants of the original molecular cloud are still associated with the cluster is not unfounded. Furthermore, Figure 7 shows an interesting protrusion from the brightest region of diffuse emission, and it points directly at the infrared source associated with the ring nebula (identified by the arrow in Fig. 7). This would be suspicious if the object in question were a PN, but not if it were a molecular condensation that used to be at the head of a "dust pillar" or "elephant trunk" like those commonly seen in H ii regions.⁴ This young molecular globule hypothesis is consistent with several additional observations that seem more problematic for the PN hypothesis:

• 1.  
  No star is detected at the center of the ring, but there is a star at the western rim of the nebula. This star has a highly reddened emission‐line spectrum of a typical T Tauri star (Fig. 5e). It would not be surprising to find a young T Tauri star associated with molecular gas. Furthermore, this star's location at the western rim implies that the O8‐type star that dominates the cluster (located ∼25 west) may have shaped the globule.
• 2.  
  The nebula seems to be externally ionized by OB‐type stars in the cluster, since it has a high‐excitation halo around the low‐excitation interior. This is typical for an externally ionized photoevaporative flow. If the low‐ionization skin of the globule marks an ionization front in material evaporating from a reservoir inside the globule, the electron density n_e at the ionization front should be sufficient for recombinations to balance the flux of ionizing photons Φ such that

  

  where d is the distance between the ionization front and the UV source, R is the radius of the globule, and α_B = 2.6 × 10^-13 cm³ s ^-1 is the case B recombination coefficient for hydrogen at 10⁴ K. This is analogous to the treatment of photoevaporating protoplanetary disks in the Orion Nebula (see Johnstone, Hollenbach, & Bally 1998). As mentioned earlier, the southeast side of the globule in NGC 3572 is probably ionized by a pair of O9/B0‐type stars, each with Φ ≈ 5 × 10⁴⁷ s ^-1. These are located about 1^' away in projection, so we take d ≈ 3.5 × 10¹⁸ cm. If the globule's radius is ∼12^'' or R ≈ 5 × 10¹⁷ cm (corresponding to the outer ring in images), then we expect an electron density at the globule's surface of roughly 400 cm ^-3. This can be compared to the value of ∼300 cm ^-3 that we measure from [S ii] lines. This argument does not necessarily favor the evaporating globule hypothesis, but it is at least consistent. With such densities at the evaporating surface, and expansion speeds on the order of 3 km s ^-1 (Fig. 4), the globule would have a mass‐loss rate of M˙ ≈ 10^-5 M_⊙ yr ^-1. If we accept infrared mass indicators, the globule's lifetime will be ∼10⁶ years—a significant fraction of the cluster's age.
• 3.  
  As noted earlier, the expansion speeds shown in Figure 4 are quite slow for a PN but are comparable to the sound speed in ionized gas in transition between a photodissociation region and an H ii region. The marginally bipolar expansion pattern noted earlier does not offer a serious challenge to the photoevaporating globule hypothesis, as it is partly subjective.
• 4.  
  The thermal infrared spectral energy distribution in Figure 6 shows two temperature components. In the PN hypothesis, we might be tempted to attribute the warmer component to the inner ring, and the cooler component to the outer shell. However, it would be difficult to rectify this interpretation with the disparity in luminosity and mass of the two components, when their ionized gas seen in optical emission lines does not show such a wide discrepancy. In the molecular globule hypothesis, we could understand the two components as coming from the exposed skin of the globule (warm) and the shielded inner parts (cool).
• 5.  
  Much more important is the mass of 5–10 M_⊙ implied by the infrared emission, which is far too high for a PN. However, masses around 10 M_⊙ are typical for the similar photoevaporating globules in the nearby Carina Nebula (Cox & Bronfman 1995; Smith 2002), including the one shown in Figure 8.

To resolve the question of whether this object in NGC 3572 is an anomalous PN or a normal molecular globule with an interesting morphology, images with higher spatial resolution and long‐slit spectra with higher dispersion are needed. Thermal infrared images with higher spatial resolution than the MSX data would be useful to see if PAH emission comes mostly from the surfaces of the putative globule, and near‐infrared spectra could detect fluorescent emission from molecular hydrogen at these same surfaces, if present.

We thank Ted Williams (Rutgers University) and the CTIO staff for assistance with the Fabry‐Perot and R‐C Spectrograph observations. This research has been partially supported by NASA grant NAG 5‐12279 to the University of Colorado.