ALMA CONTINUUM OBSERVATIONS OF A 30 Myr OLD GASEOUS DEBRIS DISK AROUND HD 21997

Circumstellar disks around stars older than 10 Myr are expected to be gas-poor. By that time, the massive gaseous primordial disks, natural by-products of star formation and potential birthplaces of planetary systems, are largely depleted. They evolve into tenuous debris disks, composed of second-generation dust grains produced from collisional erosion and/or evaporation of previously formed planetesimals (Wyatt 2008). Debris disks are thought to have an extremely low gas-to-dust ratio compared to primordial disks. Gas in these systems—if it exists at all—may also be second-generation, produced from icy planetesimals/grains (e.g., Vidal-Madjar et al. 1994; Grigorieva et al. 2007; Czechowski & Mann 2007). However, we know of two circumstellar disks—around the ∼30 Myr old, A3-type HD 21997 (Moór et al. 2011a) and the ∼40 Myr old, A1-type 49 Ceti (Zuckerman et al. 1995; Hughes et al. 2008)—that harbor detectable amounts of CO gas revealed by millimeter observations, while at the same time infrared observations show debris-like dust content. The origin of the gas in these systems is hotly debated. Zuckerman & Song (2012) proposed that the observed CO may be released from colliding comets thus it would have a secondary origin. By modeling the gas observations of 49 Ceti, Roberge et al. (2013) also suggested that the gas may be secondary material. Although the ages of these systems significantly exceed both model predictions for disk clearing and the ages of the oldest transitional disks, the gas may still be residual primordial material that survived longer in the outer disks than is usually assumed (e.g., Krivov et al. 2009).

Gas around HD 21997 has recently been discovered (Moór et al. 2011a). At a distance of 72 pc, HD 21997 belongs to the 30 Myr old Columba association (Moór et al. 2006; Torres et al. 2008). By modeling the spectral energy distribution (SED) and the single-dish CO measurements available at the time, we could not determine the spatial distribution of gas and dust unambiguously due to the unknown inclination and the degeneracy between grain size and location. To study the temperature and density structure of the circumstellar matter, the physical interaction between gas and dust, and the origin of the gas, spatially resolved observations are indispensable. In this Letter, we present the first spatially resolved continuum image of HD 21997 with the Atacama Large Millimeter/Submillimeter Array (ALMA) at 886 μm. These data were supplemented by far-infrared/submillimeter measurements obtained with the Herschel Space Observatory (Pilbratt et al. 2010). In order to study the gas properties in the disk, we also observed CO lines with ALMA, presented in Kóspál et al. (2013).

2.1. ALMA Observations

We observed the continuum emission from HD 21997 as part of our ALMA Early Science Cycle 0 program (PI: Á. Kóspál, project ID: 2011.0.00780.S). Observations were executed on 2011 November 4 in a compact configuration using 15 antennas, with low spectral resolution (time division mode, TDM) offering baselines from 18.3 m to 134.2 m. Four spectral windows with a bandwidth of 2 GHz each were defined around 345 GHz. The total on-source exposure time was ≈2 hr. J0403–360 served as bandpass calibrator, while Callisto was used to constrain the absolute amplitude scale. Our project also included high spectral resolution line observations of HD 21997 around 330–345 GHz in frequency division mode (FDM), described in Kóspál et al. (2013). In order to increase the signal-to-noise ratio, we extracted the continuum emission from the FDM data using line-free channels, and concatenated these with our TDM measurements using appropriate weighting. The effective wavelength of the combined data set is 886 μm. The TDM and FDM maps are consistent in terms of flux level. We performed imaging with the multi-frequency synthesis algorithm with the Common Astronomy Software Applications (CASA) v3.3 task CLEAN. The image (Figure 1(a)) has a typical rms noise of 4.5 × 10⁻² mJy beam⁻¹, a beam size of 156 × 126, and a beam position angle (P.A.) of −534.

Zoom In Zoom Out Reset image size

2.2. Herschel Observations

2.3. Spitzer/IRS Observations

HD 21997 was observed with the InfraRed Spectrograph (IRS; Houck et al. 2004) onboard the Spitzer Space Telescope on 2004 January 5. Small 2 × 3 maps were taken with the low-resolution IRS modules, covering the 5.2–38 μm range. We downloaded the data processed with the pipeline version S18.18.0 from the Spitzer archive and further processed them with the Spitzer IRS Custom Extraction Software (SPICE v2.5.0). We used the two central map positions and subtracted them from each other. Then, we extracted the positive signal from a wavelength-dependent, tapered aperture, and averaged them. The final spectrum is plotted in Figure 2(a).

Zoom In Zoom Out Reset image size

3.1. ALMA Image

3.2. Herschel PACS and SPIRE Maps

3.3. Modeling of Dust Distribution

As preparation for modeling the morphology of the dust disk, we compiled the SED of the source using data from the literature and our new Herschel and ALMA observations. For the fitting process, the IRS spectra were sampled in 11 adjacent bins using the same method as in Moór et al. (2011b). Table 1 summarizes these photometric data. A stellar photosphere model with a luminosity of 11.3 L_☉ was taken from Moór et al. (2011a). The SED in Figure 3(a) shows that at most wavelengths the flux densities measured by different instruments are consistent; however, our ALMA flux diverges from the trend delineated by non-interferometric submillimeter observations. The 886 μm ALMA flux is only ∼30% of the flux measured with SCUBA at 850 μm (with an FWHM of 145) and of the 886 μm flux extrapolated from the SPIRE observations. This is not a calibration issue, because our ALMA line observations agree well with our previous APEX measurements (Kóspál et al. 2013), and the continuum maps created from FDM and TDM data, obtained on different epochs, also provide fluxes consistent within 8%. Since there are no additional sources in the vicinity of HD 21997 in the PACS and ALMA images, we suspect that this flux loss occurred because some extended emission was filtered out by the ALMA interferometer.

Zoom In Zoom Out Reset image size

3.3.1. Simple Model

We first deprojected the ALMA visibilities using the disk inclination and position angle derived in Section 3.1 and then radially averaged in 7 kλ wide bins. Figure 3(b) displays the real and imaginary components of the binned fluxes. The DEBris disk RAdiative transfer code (DEBRA; Olofsson et al. 2012), developed for optically thin radiation, was used to model the visibilities. The code was extended with a ray-tracer to calculate images for ALMA simulations. HD 21997 has a featureless mid-infrared spectrum, providing no information about dust composition. Dust particles were assumed to be a mixture of amorphous silicates with olivine stoichiometry and amorphous carbon. The optical constants of silicate and carbon were taken from Dorschner et al. (1995) and Jäger et al. (1998), respectively, while mass absorption and scattering coefficients were calculated using Mie-theory. The dust size distribution was assumed to be a power law with an exponent of −3.5, characteristic of collisionally dominated systems. The maximum grain size (s_max) was fixed at 3000 μm, the minimum grain size (s_min) was varied by setting 4, 6, 8, and 16 μm. The model disk had a power-law surface density distribution between an inner (R_in) and outer radius (R_out). We calculated a grid of models varying R_in, R_out, and the surface density exponent (p). The best-fit solution and the corresponding errors were derived using a marginal likelihood estimator, calculated from the χ² of the fit of individual models. We determined best-fit models for each dust size distribution and compared the model SEDs with the observations. This comparison excluded s_min = 4 μm for its incompatibility with mid-infrared data. Our best-fit solution, plotted in Figures 3(a) and (b), has R_in = 62 ± 13 AU, R_out = 150 ± 47 AU, p = −2.4 ± 4.5, s_min = 6 μm, and M_dust = 0.09 ± 0.03 M_⊕.

3.3.2. Extended Model

As Figure 3(a) shows, our best-fit model underestimates the observed fluxes at most wavelengths. Moreover, the simulated PACS images—calculated by convolving our model with the appropriately rotated PACS PSFs—are less extended than the measured ones. These discrepancies hint at a cold outer extension of the disk. To test this hypothesis, we fitted the SED, the PACS profiles, and the ALMA visibilities simultaneously. The ALMA flux was discarded from the SED fitting since it suffers from spatial filtering. We used DEBRA with the same modeling setup as above. We combined χ² values of the SED and ALMA visibilities with equal weight and performed a consistency check on the PACS images. The best-fit solution was reached with R_in = 55 ± 16 AU, p = −1.6 ± 0.8, M_dust = 0.27 ± 0.11 M_⊕, and s_min = 6 μm. R_out cannot be constrained by the ALMA visibilities, and models that reproduce the SCUBA flux and having R_out  >  490 AU are all consistent with the data within the uncertainties. Our best-fit SED and visibility models with R_out  =  490 AU are plotted in Figures 3(a) and (b). The fit reproduces both the long wavelength data points and visibilities and, moreover, between ∼55 and 150 AU it is consistent within uncertainties with our previous model. Furthermore, consistently with the IRS spectrum, our model does not show silicate features at mid-infrared wavelengths. Thus, our modeling suggests the existence of a cold extended component, but its physical parameters are only weakly constrained by our data.

A special feature of the HD 21997 disk is its substantial molecular gas content. Using ALMA, Kóspál et al. (2013) resolved the disk in J = 2–1 and J = 3–2 transitions of ¹²CO and ¹³CO and in J = 2–1 transition of C¹⁸O. They found that the observed CO brightness distribution could be reproduced by a simple disk model with inner and outer radii of <26 AU and 138 ± 20 AU. The inner radius of the dust disk, R_in = 55 ± 16 AU is larger than the quoted 1σ upper limit of the gas inner radius at a 1.8σ level, implying that there probably is a dust-poor inner region and gas and dust may only be co-located between ∼55 AU and ∼140 AU. The differences in the inner disk morphologies can also be visualized by comparing our Figure 1(a) with Figure 2 (left) in Kóspál et al. (2013). Our extended modeling also implies the existence of an outer dust halo.

The total CO mass was estimated to be 0.04–0.08 M_⊕ based on optically thin C¹⁸O observation (Kóspál et al. 2013). If the gas is of secondary origin, produced from icy planetesimals/grains, then the total gas mass is of the same order of magnitude as the CO mass. This would yield a gas-to-dust mass ratio of ∼0.4 in the co-located region. However, in Kóspál et al. (2013) we found that this scenario could not explain satisfactorily the difference in gas/dust spatial distribution and would require an unreasonably high gas production rate, implying that the gas is more likely of residual primordial origin. In this case, H₂ is also expected in the disk, and assuming a canonical CO/H₂ ratio of 10⁻⁴, the total gas mass would be 30–60 M_⊕. This gives a gas-to-dust mass ratio of ∼300 in the co-located region. One reason for this high value could be that most of the dust ended up in planetesimals that cannot be detected at submillimeter wavelengths.

Our extended model gives a fractional luminosity of 4.8 × 10⁻⁴ and a dust temperature range between 20 and 100 K. Thus, the fundamental properties of the HD 21997 disk are similar to those of other young debris disks. The primordial gas scenario, however, raises the question of whether the dust grains are really secondary, since gas could dampen impact velocities between dust particles, thereby increasing their lifetime in regions where both gas and dust are present. Gas in such a disk is orbiting at sub-Keplerian velocities due to a radial pressure gradient. Dust grains also orbit with sub-Keplerian velocity but their pace depends on size and composition. The velocity difference between gas and dust induces grain migration within the disk (Takeuchi & Artymowicz 2001; Krivov et al. 2009). Grains with angular velocities larger/smaller than that of the gas lose/gain angular momentum and drift inward/outward until they are in corotation with the gas. Based on their models, Krivov et al. (2009) found that even in gaseous debris disks with a gas-to-dust mass ratio of 100, the grains' lifetime will be very limited—typically less than 2 × 10⁴ yr—because collisions occur between radially drifting grains with unequal sizes. Applying these findings for the case of HD 21997, we conclude that the grains must be replenished and are of secondary origin. The lack of spectral features from small particles in the IRS spectrum—unlike in Herbig Ae and transitional disks where grains are mainly primordial—and the dominance of large grains (≳6 μm, Section 3.3) similarly to other debris disks also supports this conclusion.

Emission at (sub)millimeter wavelengths predominantly comes from large grains with a size >100 μm. In a gas-free debris disk, such large grains are little affected by radiative forces, thus, they trace the distribution of the parent planetesimals (Wyatt 2008). In the secondary gas scenario of HD 21997 the models of Takeuchi & Artymowicz (2001) predict only weak coupling between gas and grains of >100 μm, thus, grain migration is negligible. This implies a radially broad planetesimal belt in the HD 21997 system. For destructive collisions between planetesimals, collision velocity must exceed a critical value that requires a dynamically excited (stirred) environment. In the self-stirring model (Kenyon & Bromley 2008), the formation of >1000 km sized planetesimals can initiate a collisional cascade in those disk regions where they appeared. Since the formation of such oligarchs takes longer at larger radii, the site where active dust production occurs is expected to propagate outward. Based on formulae from Kenyon & Bromley (2008), even in a disk with a surface density 10 times higher than the minimum-mass solar nebula, the formation of large oligarchs can spread out only to ∼90 AU within 30 Myr. The dust disk (and the planetesimal belt) of HD 21997 extends out to ≫90 AU, which cannot be explained via self-stirring. In this case, planetary stirring (Mustill & Wyatt 2009) may be a viable alternative stirring mechanism to explain the dynamical excitation in the outer regions. If a large amount of primordial gas is present (our favored scenario), the gas–dust coupling is stronger and it might induce radial migration of grains. Therefore, in the primordial gas scenario, it is possible that HD 21997 has a narrower planetesimal ring, and the extended morphology of the dust is explained by efficient grain migration. However, the details of grain migration (e.g., the magnitude of migration) cannot be reliably constrained by the available data, therefore we cannot judge the viability of this scenario.

Our study indicates that the dust content of the disk is of secondary origin while Kóspál et al. (2013) claims that the gas component is more likely primordial. We know of several young circumstellar disks that dominantly contain primordial gas and dust, but in their innermost region may have debris dust as well (e.g., Grady et al. 2009; Roberge et al. 2005; Eisner et al. 2006). HD 21997 may be a more evolved hybrid disk where the dust component is mainly made of second generation material.

We thank our anonymous referee whose comments improved the manuscript. This Letter makes use of the following ALMA data: ADS/JAO.ALMA#2011.0.00780.S. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada) and NSC and ASIAA (Taiwan), in cooperation with the Republic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO, and NAOJ. This research was partly funded by the Hungarian OTKA grants K101393/K104607 and the PECS-98073 program of the European Space Agency (ESA). A.M. and C.K. acknowledge the support of the Bolyai Fellowship. K.G. acknowledges support from the Hungarian OTKA grant NN102014.

Facilities: ALMA - Atacama Large Millimeter Array, Herschel - European Space Agency's Herschel space observatory