A Multi-ringed, Modestly Inclined Protoplanetary Disk around AA Tau

The Band 6 and 7 continuum observations were concatenated and imaged using multi-frequency synthesis CLEAN, with a reference frequency of 271.6 GHz. A high sensitivity image of the AA Tau dust continuum (Figure 1, panel (a)) was created using Briggs "robust" weighting (Briggs 1995), with a robust parameter of 0, yielding a synthesized beam of 018 × 012 (26 × 17 au at 145 pc) at a PA of ∼27° and an rms noise of 44 μJy bm⁻¹. As seen in the figure, AA Tau hosts a system of nested dust rings with an apparent inclination of ∼59°, which we confirm through visibility modeling (see Section 3.2). This is substantially lower than the 71° ± 1° previously fit to infrared scattered light emission (Cox et al. 2013), and we discuss this discrepancy in Section 4.1. Figure 1, panel (b), highlights the location of the three rings in a deprojected and azimuthally averaged radial intensity profile, calculated from the image in panel (a) using the model-constrained inclination of 59° and PA of 93°. The third ring is marginally detected in this radial profile, but confirmed through our visibility modeling (see Section 3.2). The three rings are nearly evenly spaced, peaking at 034, 066, and 099 (49, 95, and 143 au).

Zoom In Zoom Out Reset image size

To better display structure in the inner two rings, we additionally imaged the data using super-uniform weighting (Figure 1, panel (c)), yielding a synthesized beam of 015 × 011 (22 × 16 au at 145 pc) at a PA of ∼33° and a slightly higher rms noise of 60 μJy bm⁻¹. A deprojected and azimuthally averaged radial profile is shown in Figure 1, panel (d). The higher resolution image shows that the inner ring has apparently symmetric azimuthal variations, and there is a "bridge" of emission across the central clearing. At the resolution of the current observations, it is not immediately clear whether this "bridge" is caused by an unresolved inner disk, spiral arms (e.g., Muto et al. 2012; Pérez et al. 2016), or dust streamers similar to those hinted at in HD 142527 (Casassus et al. 2013). We discuss these possibilities further in Section 4.2.

We have interpreted the continuum emission with a simple parametric model composed of three Gaussian rings and a Gaussian inner disk. For each ring, the peak radius, width (FWHM), inclination, position angle, integrated flux in each band, and central offset were allowed to vary. As the observations do not have sufficient spatial resolution to constrain the PA and inclination of the inner disk, we fix the PA to 93°, constrained by the PA of the observed jet (Cox et al. 2013), and the inclination to 75°, constrained by photopolarimetry modeling (O'Sullivan et al. 2005). The FWHM and flux of the inner disk were allowed to vary. Two nuisance parameters were added to constrain the central position of the inner disk, which was then treated as the reference point for the outer ring offsets. This 29 parameter model was fit to the observed visibilities using the MCMC routine emcee (Foreman-Mackey et al. 2013) and the visibility sampling routine vis_sample,¹ yielding a final best-fit model with a reduced χ² value of 1.02.

The best-fit values and 68% confidence intervals (1σ) for all model parameters are presented in Table 1. The retrieved ring radii are 48.7 ± 0.1, 94.9 ± 0.2, and 142.6 ± 0.6 au, nearly evenly spaced. The FWHM widths of the rings range from 22 to 29 au, not much larger than the beam along the major axis of the disk, suggesting the individual rings are, at best, marginally resolved, and they may contain further substructure. In addition to ring locations and widths, we constrain the ring inclinations to an average of 591 ± 03, significantly lower than the 71° ± 1° that Cox et al. (2013) fit to their scattered light observations. In contrast, the fit PAs of the rings are 92°–94°, in good agreement with Cox et al. (2013) and the predictions of Ménard et al. (2003) from AA Tau's polarization curve. We also find small (<5 au) mutual offsets between the ring centers, which may be indicative of geometries not considered by our simple model (e.g., eccentricity in the rings).

This parametric model replicates well the observed visibilities and azimuthally averaged radial intensity profile (Figure 2). When comparing the imaged data and simulated observations (using the CASA task simobserve) of our best-fit model, however, it becomes clear that an axisymmetric model does not fit the data perfectly (Figure 3). After subtracting the model visibilities from the data and imaging the residuals, we find structured residuals with a peak of ∼12σ (σ = 44 μJy bm⁻¹), suggesting that there is azimuthal structure that is not captured by our models. This is consistent with the symmetric azimuthal variations seen in the imaged data (Figure 3, left). Beam convolution often creates artificial bright spots on either side of an inclined ring, but the simulated observations show that the opposing "tails" of emission cannot be explained by this effect.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To address the origin of these residuals, we tested several variations of our simple concentric ring model. First, we fit a model with independent ring parameters for each of the Band 6 and Band 7 data sets. We found small differences between the bands for all parameters, but they remained broadly consistent with the values in Table 1, and residuals were not improved over the previously described model. As both models produced structured residuals, it is unclear whether the differences between the Band 6 and Band 7 parameters are real. Given this structure in the residuals, we also investigated if a combination of two point sources embedded in the disk near the innermost ring could replicate the observations. We added two variable strength point sources fixed at the innermost ring radius with a variable azimuthal location to the model, and fit it in an identical fashion to the first model. Residuals improved (∼4σ), but remained structured, suggesting that the responsible feature is resolved and not point-like.

Finally, we tested if eccentricity could be responsible by adding two model parameters to the innermost ring: an eccentricity and an angle of perihelion. An ellipse with a Gaussian cross-section was used to describe the ring, with a1/r² term added to approximately account for pericenter glow (e.g., Wyatt et al. 1999). After fitting the model using the procedure previously described, we found that a slight amount of eccentricity was preferred for the inner ring, but the residuals barely improved (∼1σ) and remained structured. We therefore conclude that eccentricity alone cannot explain the observed azimuthal structure. Several other possible explanations are discussed in Section 4.2.

To trace molecular gas kinematics in the disk, we analyzed HCO⁺ 3–2 emission. The observations were continuum subtracted using uvcontsub and then imaged with CLEAN using natural weighting at a channel resolution of 0.4 km s⁻¹. A custom CLEAN mask was created for each channel to match the emission, and CLEANing was terminated when residuals reached 3σ. An integrated emission map (Figure 4, panel (a)) shows the HCO⁺ emission to be quite bright interior to the innermost ring, compared with the outer disk emission. This is especially clear in a deprojected and azimuthally averaged HCO⁺ radial profile (Figure 4, panel (b)), where the emission intensity increases by over a factor of three interior to the innermost ring. In contrast, ¹³CO 3–2 emission (Figure 4, panels (c) and (d), identically imaged to the HCO⁺) is not centrally peaked, and its radial profile is essentially flat within the inner dust ring. The position angle and inclination of the dust continuum were used for deprojection. The radial profiles show kinks near the locations of the inner two rings, but it is unclear whether this reflects gas density variations, dust opacity effects, or artifacts of deprojecting flared emission with a single inclination.

Zoom In Zoom Out Reset image size

In any case, the presence of sharply centrally peaked HCO⁺ emission and flat ¹³CO emission interior to the inner dust ring is interesting, as it implies that a high bulk gas density (traced by ¹³CO) interior to the ring is not responsible for the HCO⁺ emission intensity. The bright inner disk emission must therefore result from peculiar excitation effects or enhanced HCO⁺ formation chemistry (e.g., from a high-ionization environment). The HCO⁺ moment-1 map (Figure 5) is similarly intriguing, showing a twist interior to the inner dust ring, with the emission misaligned by ∼10° with respect to the continuum orientation. This feature is similar to previous observations of CO in HD 142527 and HD 100546 (Pineda et al. 2014; Rosenfeld et al. 2014) and HCO⁺ in HD 97048 (van der Plas et al. 2017; Walsh et al. 2016), and has been suggested to indicate either a disk warp or radial flow.

Zoom In Zoom Out Reset image size

A position–velocity diagram of the HCO⁺ emission (Figure 6) provides additional evidence for kinematics that cannot arise from a single Keplerian velocity field. The expected Keplerian velocity profiles for two inclinations (59° and 40°) and a stellar mass of 0.85 M_⊙ are overlaid in red and orange, respectively. Although the higher inclination of 59° is able to describe the HCO⁺ emission well in the outer regions of the disk, it is not consistent with the emission inside of the innermost ring, which is better fit by a lower inclination of 40°. Similar to the signature in the moment-1 map, this could be indicative of either a disk warp or a non-Keplerian radial flow (Pineda et al. 2014). We discuss possible interpretations of these kinematic signatures in Section 4.2.

Zoom In Zoom Out Reset image size

We have found that AA Tau hosts three nearly evenly spaced mm dust rings at an inclination of ∼59°, adding it to a growing list of substructured disks with rings and gaps. In contrast to HL Tau and TW Hya (ALMA Partnership et al. 2015; Andrews et al. 2016), however, which host power-law dust disks with numerous narrow gaps, the dust in AA Tau is distributed in rings with broad gaps (Figure 1, panel (c)), more similar to HD 163296 and HD 169142 (Isella et al. 2016; Fedele et al. 2017). If the dust gaps in AA Tau result from a planet–disk interaction, as suggested for HD 163296 and HD 169142, multiple massive planets might be involved (Pinilla et al. 2012; Picogna & Kley 2015), although Gonzalez et al. (2015) have shown that a single massive planet can also create multiple outer dust rings. More detailed modeling is necessary to interpret the observations in this vein, however, as our simple parametric model describes only the continuum surface brightness, rather than the disk surface density.

The inclination (591 ± 03) of the outer disk dust rings we observe significantly deviates from the inclinations of both the scattered light disk (71° ± 1°, Cox et al. 2013) and the inner disk (∼75°, O'Sullivan et al. 2005). Understanding the true disk inclination is imperative, as a close to edge-on viewing geometry underpins the warped inner disk explanation of AA Tau's short-term variability. This discrepancy suggests that the previous inclinations had larger uncertainties than reported, or the inner and outer disks are misaligned (e.g., Marino et al. 2015).

Supporting the third possibility, we find an opposite absolute disk orientation compared with that reported in Cox et al. (2013). They report an inclined disk with a northern near side, while we find a southern near side, with both orientations being fairly secure. Their orientation is derived both from the scattered light and from the observed jet, which is presumably aligned with the stellar axis and inner disk. In contrast, our orientation is derived from the HCO⁺ and CO emission geometry in their channel maps (not shown). In moderately inclined disks, Rosenfeld et al. (2013) have shown that emission arising from a vertically flared τ = 1 surface directly traces the absolute disk orientation. Misaligned inner and outer disks therefore appear to be the simplest explanation of all data sets, illustrated in a schematic diagram in Figure 7. Such an orientation would remain consistent with the warped inner disk explanation of AA Tau's periodic photometric variability (Bouvier et al. 1999), and the deviation between the mm and scattered light inclinations could be explained by shadowing from the inner disk (Dong et al. 2015).

Zoom In Zoom Out Reset image size

The continuum observations show azimuthal variations in the innermost ring that are resolved and not explained by eccentricity alone. Several non-mutually exclusive possibilities could explain these observations. First, shadowing from a misaligned inner disk could affect the dust temperature, and therefore emission of the inner ring. Second, spiral arms may be present in the disk (e.g., Muto et al. 2012; Pérez et al. 2016). Third, gap-crossing streamers could be present in both dust and gas around AA Tau. Dust streamers have been previously suggested in two disks (Casassus et al. 2013; Dutrey et al. 2014), although the former was not substantiated in further observations (Fukagawa et al. 2013; Muto et al. 2015). Beam convolution could cause these streamers to manifest as a non-axisymmetric contribution to the inner ring. Due to the suggestive HCO⁺ gas kinematics observed, we consider here the observable effects of this third scenario.

In general, the need for gap-crossing flows is observationally motivated, as accretion onto the central star continues to be observed even when gaps are present in disks. The small circumstellar disk will be rapidly depleted by accretion unless it is replenished (e.g., Verhoeff et al. 2011), implying that material must cross the inner gap. Models suggest that planets can drive dynamical instabilities that allow material to funnel into gap-bridging filaments (e.g., Dodson-Robinson & Salyk 2011). Observational evidence for these radial flows is mostly indirect (Beck et al. 2012; Rosenfeld et al. 2014; Zhang et al. 2015b; van der Plas et al. 2017; Walsh et al. 2016), but ALMA has begun to allow direct imaging (Casassus et al. 2013; Dutrey et al. 2014).

Radial gas flows in the AA Tau disk have previously been suggested as an interpretation of infrared CO absorption measurements (Zhang et al. 2015b). Our HCO⁺ observations may provide further indirect evidence for such a flow. The "kink" in the HCO⁺ moment-1 map and the shape of the P-V diagram have both been previously invoked as signatures of radial gas flows and disk warps (Pineda et al. 2014; Rosenfeld et al. 2014). Furthermore, the extreme brightness and broad line-width of emission within the innermost ring suggests possibly enhanced HCO⁺ formation in an ionizing environment (e.g., an accretion shock).

If AA Tau does host a radial flow, the central "bridge" and azimuthal asymmetry in the inner ring might then be explained by the presence of dust streamers. The addition of dust streamers to our best-fit model, shown in the bottom panels of Figure 7, is consistent with the observed continuum emission within the inner cavity. Such streamers, which shear off from the walls of the inner ring and spiral into the inner disk, are additionally able to replicate the twisted azimuthal variations of the inner ring. Future higher resolution (∼005) ALMA Cycle 4 observations (Figure 7, bottom right) will be able to distinguish between this scenario and a cavity with only an inner disk and no streamers.

As previously noted, AA Tau is the archetypal source for a class of stars with similar photometric variability. Recent observations have shown that a number of such "dipper stars" (Ansdell et al. 2016a) host millimeter dust disks at a wide range of inclinations (Ansdell et al. 2016b), at odds with the warped edge-on inner disk explanation of their variability. As suggested in Ansdell et al. (2016b), misaligned inner and outer disks might explain this apparent dilemma, and our observations provide the first evidence for such a geometry in one of these systems. Furthermore, a misaligned inner and outer disk in AA Tau would present the interesting possibility of gap-crossing material periodically intersecting our line of sight (Figure 4). This provides a physical motivation for the non-axisymmetric over-density suggested by Bouvier et al. (2013) and Rodriguez et al. (2015) to explain the long-duration dimming of AA Tau. If an inner disk radius of several au is assumed (consistent with our observations), then a direct path for a streamer between the inner ring and inner disk would suggest a LOS crossing distance between 5 and 10 au, consistent with previous estimates of >8 au (Rodriguez et al. 2015). Higher resolution observations of AA Tau and similar objects will be needed to test both of these hypotheses.