ABSTRACT
Using Spitzer Infrared Array Camera (IRAC) observations from the SAGE-SMC Legacy program and archived Spitzer IRAC data, we investigate dust production in 47 Tuc, a nearby massive Galactic globular cluster. A previous study detected infrared excess, indicative of circumstellar dust, in a large population of stars in 47 Tuc, spanning the entire red giant branch (RGB). We show that those results suffered from effects caused by stellar blending and imaging artifacts and that it is likely that no stars below ≈1 mag from the tip of the RGB are producing dust. The only stars that appear to harbor dust are variable stars, which are also the coolest and most luminous stars in the cluster.
Export citation and abstract BibTeX RIS
1. INTRODUCTION
Stellar mass loss is critical to the late stages of stellar evolution. Mass loss in red giant branch (RGB) stars determines cluster horizontal-branch (HB) morphology (e.g., Rood 1973; Catelan 2000) and the pulsational properties of RR Lyrae variables (e.g., Christy 1966; Caloi & D'Antona 2008). Strong mass loss on the asymptotic giant branch (AGB) can exceed the nuclear consumption rate, determining the timescale of subsequent evolution (e.g., van Loon et al. 1999; Boyer et al. 2009a). AGB winds are enriched due to a convective process that brings processed material to the stellar surface (the third dredge up; e.g., Renzini & Voli 1981), making AGB stars a major source of galaxy enrichment (Gehrz 1989). In the low-mass stars of old globular clusters (GCs; M < 0.8 M☉), more mass is lost on the RGB than the AGB (McDonald et al. 2009).
Mass loss can arise from acoustic and/or electromagnetic chromospheric activity (Hartmann & MacGregor 1980; Hartmann & Avrett 1984; van Loon et al. 2010), or from pulsationally enhanced, dust-driven winds (e.g., Bowen 1988; van Loon et al. 2008b). Radial pulsations levitate material to radii where the stellar atmosphere is cool, and pulsational shocks provide the density required for dust condensation. Radiation pressure on the grains and dust–gas momentum coupling drive the wind. However, it is uncertain whether the dust is the driving mechanism or is simply formed as a byproduct of the wind (van Loon et al. 2005; Ferrarotti & Gail 2006; Höfner & Andersen 2007; Höfner 2008).
With the launch of the Spitzer Space Telescope (Werner et al. 2004), circumstellar dust has become easily detectable, leading to a more complete picture of dust-driven mass loss: (1) dust forms only at or above the tip of the RGB (TRGB; e.g., Boyer et al. 2008, 2009b), (2) dust formation is not inhibited at low metallicity (e.g., Boyer et al. 2006, 2009a; Sloan et al. 2009), and (3) dust formation appears episodic, possibly tied to the pulsation cycle (e.g., Origlia et al. 2002; McDonald et al. 2009). The first of these points has been more controversial. Kalirai et al. (2007) speculate that strong mass loss in RGB stars might explain the presence of He white dwarfs in NGC 6791, although van Loon et al. (2008a) find that, along with an absence of dust, the RGB and HB luminosity function is not depleted. Origlia et al. (2007) find 93 stars with infrared (IR) excess as faint as the HB in 47 Tuc, and Boyer et al. (2006) find 12 such sources in M15 (see their Figure 3(b)). The presence of dusty stars so far down the RGB is surprising given low luminosities, warm photospheres, and a lack of pulsation. Indeed, more recent Spitzer observations show very few dusty stars in NGC 362 (Boyer et al. 2009b) and ω Cen (Boyer et al. 2008; McDonald et al. 2009).
This letter examines dust production in 47 Tuc, which is among the most massive (MV = −9.42 mag9; Harris 1996), metal-rich ([Fe/H] = −0.7; Thompson et al. 2010), and nearby GCs ((m − M)0 = 13.32; Ferraro et al. 1999). We present new Spitzer observations along with a re-analysis of the Origlia et al. (2007) data. We discuss how stellar blending affects the low-resolution 8 μm photometry and find that there is little, if any, dust below the TRGB in 47 Tuc.
2. DATA AND ANALYSIS
The small separation (≈2°) between 47 Tuc and the Small Magellanic Cloud (SMC) resulted in serendipitous Spitzer observations of the southeastern half of 47 Tuc (Figure 1) as part of the Spitzer Surveying the Agents of Galaxy Evolution Legacy program (SAGE-SMC). The SAGE-SMC observations consist of two epochs, separated by three months, of Infrared Array Camera (IRAC) maps at 3.6, 4.5, 5.8, and 8 μm. Here, we use photometry obtained by combining both epochs. Exposure times are short enough (12 s) to ensure that bright sources remain unsaturated, but long enough to obtain sufficient sensitivity for reasonably complete photometry to beyond the HB (Boyer et al. 2009b). Angular resolution with IRAC ranges from 1
7 at 3.6 μm to 19 at 8 μm. The SAGE-SMC MIPS observations (24–160 μm) did not cover 47 Tuc. For details regarding data acquisition and reduction for SAGE-SMC, see K. D. Gordon et al. (2010, in preparation).
Figure 1. DSS image of 47 Tuc showing the Spitzer coverage. The SAGE-SMC coverage used here is shown in black and the Rood coverage in white. The 3.6 and 8 μm coverages are represented by dashed and solid lines, respectively. We limited the SAGE-SMC coverage to R < 10' to minimize contamination from the SMC and to mimic the Rood coverage.
Download figure:
Standard image High-resolution image47 Tuc was also observed with IRAC as part of PID20298 (PI: R. Rood). We obtained these data from the Spitzer archive and reduced them independently (hereafter referred to as the Rood data). The Rood field of view at each IRAC wavelength reaches a cluster radius of ≈10', with an overlapping area of 5'× 9' centered on 47 Tuc (Figure 1). Two depths were achieved: 12 s pixel−1 and 936 s pixel−1, hereafter described as the shallow and deep data, respectively. Observation details are described in Origlia et al. (2007). Here, these data were processed with pipeline S18.7.0. The Basic Calibrated Data were reduced and mosaicked with the MOPEX reduction package (Makovoz & Marleau 2005) after applying an array distortion correction. We implemented the MOPEX overlap routine to match backgrounds between overlapping areas of the images and the MOPEX mosaicker for outlier elimination, image interpolation, and co-addition. The final mosaics have pixel sizes of 086 pixel−1.
3. PHOTOMETRY
K. D. Gordon et al. (2010, in preparation) contain details about SAGE-SMC point-source photometry.
Point-source extraction (>4σ) was performed on the Rood data with the DAOphot II package (Stetson 1987), with point-spread functions (PSFs) derived from >10 isolated stars in each mosaic. Extended sources and outliers broader or narrower than the PSF were rejected based on sharpness/roundness cutoffs. The final fluxes are color-corrected using a 5000 K blackbody (Spitzer Science Center 2006), an appropriate temperature for a typical RGB star. This correction differs by <1% from that for a 2000 K blackbody. A pixel-phase-dependent correction (Reach et al. 2005) was applied to the 3.6 μm photometry. Photometric errors include standard DAOphot errors and the IRAC absolute calibration errors (Reach et al. 2005). Magnitudes relative to α Lyrae (Vega) are derived using the zero magnitudes from the Spitzer IRAC data handbook (Spitzer Science Center 2006). False star tests including both the core and outskirts of the cluster indicate the photometry is ≳85% complete at the HB in both the deep and shallow mosaics at all four IRAC wavelengths. Bolometric corrections were determined using Two Micron All Sky Survey (2MASS) J- and Ks-band photometry (Skrutskie et al. 2006), the transformation from Montegriffo et al. (1998), and the reddening (E(B − V) = 0.04 mag) and distance modulus ((m − M)0 = 13.32 mag) used in Origlia et al. (2007).
The shallow Rood data are confusion limited at R ≲ 2' (Figure 2). Our analysis suggests that the 3.6 μm data deviate from a King profile at ≈2
2 (rcore = 24''; Howell et al. 2000), while the 8 μm data deviate near 18. Assuming the Origlia et al. (2007) analysis is limited by the 8 μm data (since they use high-quality K-band data instead of 3.6 μm), we might expect to have found up to 25% fewer sources in the inner 2' than were detected in their analysis due to greater incompleteness at 3.6 μm compared to 8 μm.
Figure 2. Top: source density profile in SAGE-SMC data. The SMC begins to dominate beyond R ≳ 15'. We restricted SAGE photometry to R < 10' (see the text). Bottom: source density profiles for the shallow Rood data. Dashed lines are King profiles with Rcore = 24''. Note that the data deviate from the King profiles due to source confusion, inside 22 at 3.6 μm and 18 at 8 μm.
Download figure:
Standard image High-resolution imageThe SAGE-SMC 3.6 and 8 μm mosaics cover the southeastern side of the cluster with 3' ≲ R < 10'. We include only sources within 10' to match the Rood coverage (Figure 1) and limit contamination from the SMC, which begins to dominate the source density at R ≳ 15' (Figure 2).
4. DISCUSSION
4.1. Dust Excess
The color–magnitude diagrams (CMDs) in Figure 3 include SAGE-SMC and Rood data. RGB star bolometric magnitudes peak in the near-IR, and warm circumstellar dust is detectable as photometric excess at 8 μm. The envelopes we attempt to detect are optically thin at IR wavelengths (those that are not would stand out); attenuation of starlight can thus be neglected, while dust emission is insignificant at 3.6 μm. Hence it does not matter whether the K − [8] color is used, as in Origlia et al. (2007), or the [3.6] − [8] color. Indeed, in the re-reduced Rood data, all stars with red Ks − [8] colors on the upper half of the RGB (where photometric scatter is small) also have red [3.6] − [8] colors. We also note that the mean Ks − [3.6]SAGE color for 47 Tuc is a mere 0.08 mag, insignificant compared to the range in dust excess (0 < K − [8] < 1) suggested by Origlia et al. (2007). We prefer to use the [3.6] − [8] color because it does not mix different origins of data, and we can confidently apply the stellar blending test described in Section 4.4.
Figure 3. CMDs of 47 Tuc. Left: SAGE-SMC; middle: shallow Rood data; right: deep Rood data. Note that we see far fewer red sources here than seen by Origlia et al. (2007) also in the same original data. The blue line in the left panel is an isochrone from Marigo et al. (2008), with Z = 0.0024 and tage = 11.3 Gyr; the dashed portion marks the AGB. The black dot-dashed line marks the TRGB luminosity, as indicated by the isochrone. Saturated sources are excluded from the deep CMD. Variable stars LW10 and V26 may trace the AGB, while the other variable stars trace the RGB.
Download figure:
Standard image High-resolution image4.2. SAGE-SMC CMD
The SAGE-SMC CMD is presented in Figure 3. The blue line represents an isochrone from Marigo et al. (2008) with the appropriate age (11.3 Gyr) and metallicity (Z = 0.0024) for 47 Tuc from Thompson et al. (2010); the solid and dashed lines represent the RGB and AGB, respectively. The isochrone indicates that the TRGB is located at Mbol ≈ −3.5 mag, agreeing with the luminosity marking the onset of dust production in 47 Tuc (L ≈ 2000 L☉; Lebzelter et al. 2006).
Aside from three variable stars within 0.5 mag of the TRGB (Lebzelter & Wood 2005), only one source ≈1 mag brighter than the HB is red, with [3.6] − [8] ≈ 0.25 mag (R.A. = 00h25m39
4, decl. = −72°08'210, J2000). This source is not identified as variable by either Weldrake et al. (2004) or Lebzelter & Wood (2005), but its 3.6 μm magnitude varies by 0.14 mag between the two SAGE epochs. This source is isolated, and is not affected by stellar blending or other imaging artifacts, so the IR excess is likely real. However, the source is in the outskirts of the cluster, in the direction of the SMC, and may not belong to 47 Tuc. If at the distance of the SMC, its absolute 3.6 μm magnitude is M3.6 = −7.75 mag, which is >1 mag brighter than the TRGB and consistent with a typical AGB star (Bolatto et al. 2007). Thus, we conclude that significant IR excess in 47 Tuc is limited to <1 mag from the TRGB in the SAGE-SMC data, similar to what is seen in ω Cen, NGC 6791, and NGC 362 (Boyer et al. 2008, 2009b; van Loon et al. 2008a).
4.3. Spitzer Archive CMDs
Photometry extracted from the Rood data is noticeably noisier than the SAGE-SMC photometry (Figure 3, middle, right panels). This could partially be due to differences in data acquisition and reduction, but it appears that the scatter is caused mainly by inclusion of the inner 3' of the cluster (light gray dots in Figure 3), where crowding is an issue. The deeper photometry suffers from saturation at 3.6 μm already more than a magnitude below the TRGB. For clarity, saturated sources are excluded from the CMD in Figure 3.
We see immediately that the new photometry has resulted in far fewer IR-excessive sources than found by Origlia et al. (2007). That study found >90 IR-excessive stars (R < 2') distributed evenly along the entire RGB, while we see only a handful of red sources only near the upper RGB, or in increasing number and redness toward magnitudes fainter than ≈2 mag below the TRGB. In the shallow data (R < 10'), we find 24 sources with >3 σ IR excess ([3.6] − [8] ≳ 0.17 mag) and Mbol < 0 mag, four times fewer than found by Origlia et al. (2007) in the same area of their CMD. This is consistent with an AKARI analysis of 47 Tuc presented by Ita et al. (2007), where it can be seen in their Figure 2 that most dusty stars reside near the TRGB.
Thirteen of these 24 red stars are variable stars identified by Lebzelter & Wood (2005). The brightest variables, V26 and LW10, may be AGB stars, while the others trace the RGB. Curiously, both show no excess in 8–13 μm spectra (van Loon et al. 2006), but do show moderate excess in the Rood data ([3.6] − [8] ≈ 0.2 mag), which were taken 1 month prior to the spectral observations. The difference could be due to episodic dust production during the pulsation cycle. The variables V1 and V8 are the reddest stars in the shallow Rood data, agreeing with the dust excess seen in their mid-IR spectra (Ramdani & Jorissen 2001; Origlia et al. 2002; Lebzelter et al. 2006; van Loon et al. 2006).
When excluding dusty variable stars, we find only 11 sources with IR excess in the shallow Rood data, all but one more than ≈2 mag fainter than the TRGB. These sources all fall within 3' of the cluster center, suggesting that stellar crowding may be leading to inflated fluxes, as measured in the lower resolution 8 μm images.
In the deep data, we find 11 other sources with [3.6] − [8] excess >3 σ. However, excluding the inner 2' of the cluster as done in Origlia et al. (2007) leaves only one source with IR excess.
4.4. Stellar Blending and Imaging Artifacts at 8 μm
A re-analysis of the photometry has revealed only 24 red sources, missing 74% of the 93 red sources from Origlia et al. (2007). This discrepancy cannot be entirely due to the 25% incompleteness of the 3.6 μm data described in Section 3 since that would require red sources to be preferentially undetected compared to non-red sources. That scenario is unlikely since the red sources are evenly distributed in the cluster and since optically thin circumstellar dust emission is minimal at 3.6 μm. To investigate whether the 24 red sources are truly harboring circumstellar dust, we have performed a stellar blending test, similar to that done for ω Cen in Boyer et al. (2008). We convolved the 3.6 μm image with the 8 μm PSF and extracted point-source fluxes from the resulting low-resolution 3.6 μm image using the same procedure described in Section 3. In the ideal scenario, where all stars are isolated, [3.6]real − [3.6]convolved should always be near zero. Any sources that show [3.6]real − [3.6]convolved > 0 mag are affected by stellar blending, which inflates the convolved 3.6 μm flux. Stars that are truly red should have [3.6]real − [8]>0 mag and [3.6]real − [3.6]convolved ≈ 0 mag. Figure 4 shows the results of this test.
Figure 4. Upper panels: source blending and confusion at 8 μm above the HB. See Section 4.4 for a description of the test. Lower panels: CMDs showing the locations of stellar blends, truly red sources, and truly blue sources. The mean 3 σ excess ([3.6] − [8]>0.17 mag) is marked by a dotted line. Note that most red stars are confined to within 1 mag of the TRGB, and most stellar blending occurs among the faintest stars.
Download figure:
Standard image High-resolution image
In the crowded environment of a GC, blending can be complicated. The test described above may also result in blue colors, caused by stellar confusion within compact groups of stars. With many overlapping PSFs, it is no surprise that flux measurements can be uncertain. We choose a conservative cutoff of |[3.6]real − [3.6]convolved|>0.05 mag to define stars with affected photometry.
The boxes in Figure 4 show the mean 3σ [3.6] − [8] excess in the x-direction and the cutoff for blended sources in the y-direction. The shaded regions to the left and right of these boxes show the approximate locus of truly red or blue IR colors. Stars above or below the shaded region are suffering from blending and/or source confusion. Stars shown to have real red or blue [3.6] − [8] colors are marked in red and blue, respectively, and sources affected by blending and/or confusion are marked in green. The locations of these stars on the CMDs are shown in the lower panels. In the shallow data, the horizontal plume of stars at the TRGB is striking and very well distinguished from the vertical dispersion, which is a result of stellar blending and source confusion. Note that in the shallow data, only one red star with Mbol ≲ −3 mag is blended: variable LW8.
Of the 11 sources in the shallow data (identified in Section 4.3) brighter than the HB and not identified as variable stars, we find a total of six unblended, red sources below the TRGB, all with Mbol ≳ −2 mag. In the deep data, we find five such sources. Tellingly, none of these sources are red in both the shallow and deep data.
Closer inspection of the shallow 8 μm mosaic shows that all six of the faint, red, unblended stars fall directly on image artifacts caused by internal optical scattering inside the detector array (known as banding; Spitzer Science Center 2006), causing an artificial increase in their 8 μm fluxes. Therefore, there are no truly red sources below the TRGB in the shallow data. Of the five faint, red, unblended stars in the deep data, two are affected by banding, and two lie immediately adjacent to a saturated star, which may affect both their 3.6 and 8 μm fluxes. In the deep data, we therefore find only one potentially dusty source below the TRGB. Again, we note that this source is not red in the shallow data, indicating that the IR excess is not real. Therefore, there are no truly dusty sources fainter than ≈0.5 mag from the TRGB, consistent with the SAGE-SMC data.
4.5. Implications for Mass Loss and Dust Production
To be consistent with the analysis from Origlia et al. (2007), we would have detected up to 70 dusty sources, even after considering incompleteness in the 3.6 μm data compared to their K-band data. However, through close inspection of stellar blending and imaging artifacts in SAGE-SMC data and the re-reduced Origlia et al. (2007) data, we find that the number of truly dusty stars in 47 Tuc below about a magnitude fainter than the TRGB is consistent with zero. We therefore find that the mass-loss law derived by Origlia et al. (2007) over-predicts mass loss along the RGB—estimating nearly 2 orders of magnitude more mass loss among the faintest RGB stars than Reimers' law (Reimers 1975) and attributing ≈25% of the total mass loss to stars fainter than Mbol = −1.5 mag. With no dusty mass loss along the RGB, these faint RGB stars together contribute closer to 1% of the total cluster mass loss (based on Reimers' law). We do see a large number of red sources at or above the TRGB (Figure 3), as expected if the stellar structure only becomes favorable to dust production if the star is either close to the occurrence of core helium ignition (i.e., near the TRGB), or is experiencing helium-shell flashes (i.e., on the thermally pulsing AGB).
Very little, if any, dust is forming below the TRGB, but that is not to say that no mass is lost on the RGB. Indeed, the morphology of the HB requires that mass is lost on the RGB. Chromospherically driven winds are indicated in several RGB stars by asymmetries and coreshifts in chromospheric spectral lines, with mass-loss rates up to 10−8 M☉ yr−1 (e.g., Mauas et al. 2006; McDonald & van Loon 2007; Dupree et al. 2009), which is almost exactly what is required to explain the HB morphology.
In addition to only residing near the TRGB, the majority of the stars showing evidence for circumstellar dust are also variable. This supports the idea that dust production coincides with pulsationally enhanced stellar winds.
We thank the anonymous referee for his/her thorough review. This work was supported by Spitzer via JPL contracts 1309827 and 1340964 and makes use of data products from 2MASS, a joint project of the University of Massachusetts and IPAC/Caltech, funded by NASA and the NSF.
Footnotes
- 9
The Harris (1996) catalog was updated in 2003: http://www.physics.mcmaster.ca/~harris/mwgc.dat.