ALMA OBSERVATIONS OF THE T TAURI BINARY SYSTEM AS 205: EVIDENCE FOR MOLECULAR WINDS AND/OR BINARY INTERACTIONS

In Figure 1 we show the observed (CLEANed) ¹²CO 2–1, ¹³CO 2–1, and C¹⁸O 2–1 zeroth-moment maps and 1.3 mm continuum images of the AS 205 system. Note the different gray scales and contour intervals for each row. Based on the continuum measurements, the separation of the two disks is 13, and the position angle of the binary is 214°. Previous works have found separations of 13–14 and position angles between 204° and 213° (Reipurth & Zinnecker 1993; Ghez et al. 1993; Koresko 2002; Prato et al. 2003; Eisner et al. 2005), with no coherent temporal evolution. Both sources appear to be detected in all three lines and in the continuum, and basic statistics for these detections are listed in Table 1.

Zoom In Zoom Out Reset image size

For all maps, the synthesized beam is approximately ∼07 × 05 with a position angle of ∼90°–100°. The continuum emission appears unresolved or barely resolved, with a spatial FWHM ranging from 063 to 076 and a position angle of 99°—consistent with that of the synthesized beam. The line emission is spatially extended in all three isotopologues and for both components of the system, although the spatial extent of the northern component is larger than for the southern component. The ¹²CO emission and ¹³CO emission from the northern component appear asymmetric, with a larger spatial extent toward the north and northwest.

Although all isotopologues are present in the zeroth-moment maps near AS 205 S and AS 205 N, the ¹³CO emission spatially associated with AS 205 S is actually dominated by gas at a similar velocity to AS 205 N. Therefore, some of this emission may instead be associated with AS 205 N. We also note that the C¹⁸O emission near AS 205 S is detected in the zeroth-moment map only, and at only the 3σ level. Therefore, we caution that this detection is tentative and could, for example, arise from errors in continuum subtraction.

Channel maps and first-moment (velocity field) maps for the northern component in ¹²CO and ¹³CO are shown in Figures 2 and 3. Channel maps and first-moment maps for the southern component in ¹²CO are shown in Figure 4. Line profiles for the three isotopologues are shown in Figure 5. ¹²CO and ¹³CO lines are computed for the entire system, including the northern and southern components and any extended emission, using a 648 high by 432 wide extraction box. The emission line for C¹⁸O was computed for the northern component only, using a small 168 high by 144 wide extraction box, to reduce the contribution from the noisy background. The fluxes and line widths are summarized in Table 1. The ¹²CO emission has a total flux (for both components combined) of 22.25 Jy km s⁻¹, consistent with the flux of 20.06 ± 0.13 Jy km s⁻¹ found by Öberg et al. (2011) if we allow for a 15% error in absolute flux calibration.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Using the mean and standard deviation of the line centers, we estimate v_LSR of AS 205 N to be 4.29  ±  0.18, consistent to within 1.6σ of the published stellar velocity of 1.97 ± 1.51 from Melo (2003), but inconsistent with the −0.25 ± 0.87 km s⁻¹ found by Eisner et al. (2005). The high and variable veiling in this source (Gahm et al. 2008) could contribute to large systematic errors in stellar velocities. For AS 205 S, we take the v_LSR for ¹²CO (0.80  ±  0.09 km s⁻¹) to be the most likely value, since we were able to separate the signals from AS 205 N and AS 205 S (see Table 1). Eisner et al. (2005) find stellar velocities for the components of this spectroscopic binary of 11.05 ± 0.46 and −6.36 ± 1.11 km s⁻¹, with an average of 2.35 ± 1.2 km s⁻¹, consistent with our result.

To aid with the analysis of the data, we constructed disk models to compare with the observed data. Dust radiative transfer was performed with RADMC (Dullemond & Dominik 2004), and line radiative transfer with RADLite (Pontoppidan et al. 2009). Simulated ALMA data sets were created using the CASA routines simanalyze and simobserve. All simulated observations use the antenna configuration from the observations on 2012 May 4 and an exposure time of 30 minutes. Simulated data sets are shown in Figures 1–4.

Our nominal disk model for the northern component uses most of the same stellar and disk parameters derived by Andrews et al. (2010) from a combined fit to the spectral energy distribution and millimeter visibilities—R_⋆ = 3.7 R_☉, M_⋆ = 1 M_☉, T_⋆ = 4250 K, R_in = 0.14 AU, M_disk = 0.029 M_☉, and H ∼ 20 AU at 80 AU, with H/R∝(R/80 AU)^0.11. However, in contrast to Andrews et al. (2010), who use the similarity solution for surface density, Σ, we use a broken power law Σ, ∝(R/AU)^−0.9 out to R_out = 80 AU, and ∝(R/AU)⁻¹² beyond. We also use a simple grain opacity with 80% small amorphous astronomical silicates and 20% carbon grains. Less information is available in the literature for the southern component, and so for simplicity we utilize many of the same model parameters as for AS 205 N. However, we assume R_⋆ = 4.2 R_☉, M_⋆ = 0.6 M_☉, and T_⋆ = 3450 K, values derived by Prato et al. (2003) but with a correction for the fact that the southern source is a spectroscopic binary (Eisner et al. 2005). We find a good match to the observed continuum data with a disk mass of 0.006 M_☉.

The CO 2–1 emission is dominated by disk regions with densities well above the critical density for this line (e.g., Bruderer et al. 2012), and even the more tenuous winds we are investigating here are predicted to have densities at or above the critical density out to large distances from the star (Panoglou et al. 2012). Therefore, all models assume local thermodynamic equilibrium conditions. We assumed a gas-to-dust ratio of 100 and used a parameterized gas temperature increase in the disk atmosphere, following the procedure described in Zhang et al. (2013). In particular, we set T_gas − T_dust = 300 × e^−r/30 AU above the τ = 1 disk surface. We also assumed an H₂/CO ratio of 5000 (Lacy et al. 1994) and a CO freeze-out temperature of 17 K. For isotopologue abundances, we assumed ¹²CO/¹³CO =60 and ¹²CO/C¹⁸O =500. Since our limited information would result in strong model degeneracies were we to try to exactly match the observed data, we simply made small adjustments to the total flux produced by the models to approximately match the data. For AS 205 N, we multiplied the continuum models and the ¹²CO, ¹³CO, and C¹⁸O continuum-subtracted models by factors of 1.3, 1.4, 0.4, and 0.2. For the southern source models, we multiplied by factors of 1, 1.2, 0.5, and 0.3. None of the conclusions in this work depend on the absolute flux of the models.

We find that an inclination of ∼15° ± 5° is required to fit the observed ¹³CO velocity field for AS 205 N. Previous spectro-astrometric results, which also resolved the binary, found a disk inclination of ∼20°. Using the ¹²CO velocity field, we find a disk inclination of 25° ± 10° for AS 205 S. These inclinations and error bars are determined via by-eye examination of the channel maps and residuals, and based on a combination of the relative fluxes in the low- and high-velocity channels, as well as the shapes of the maps. We also find that the disk around AS 205 S must be significantly smaller than AS 205 N, and we set R_out = 35 AU. We discuss this result in more detail in Section 3.3. There is some uncertainty in the literature about the mass of AS 205 S. Eisner et al. (2005) derive a larger mass than assumed in our nominal model, 1.28 M_☉, along with temperatures for AS 205 N and S of 3800 and 4000 K, respectively, and a total luminosity of 0.7 L_☉. With this alternative choice of parameters, we find an inclination for AS 205 S of ∼15°. Given the limited resolution of our observations compared with the size of the AS 205 S disk, spatial information cannot be used to break the degeneracy between stellar mass and inclination.

Figure 3 shows that the AS 205 N disk model deviates from the ¹³CO data in some respects. In particular, the high-velocity emission is slightly overpredicted. However, decreasing the inclination of the model would produce maps less elongated than the data at low velocities. A similar discrepancy was noted by Rosenfeld et al. (2012) for TW Hya and attributed to a warp in the TW Hya disk. In our case, the discrepancy could instead be related to the other evidence for non-Keplerian motions in our data set, which is discussed in much greater detail throughout this work.

The position angles of the AS 205 N and AS 205 S disks are observed to be ∼135° ± 5° and ∼320° ± 10°, respectively, and are therefore misaligned by nearly 180°. The relative alignments of the disk rotation axes cannot be unambiguously determined unless the near and far sides of both disks are known. The disks are either oriented in the same direction with respect to the plane of the sky, with rotation vectors nearly anti-aligned, or are oriented in opposite directions with respect to the plane of the sky, with rotation vectors having a misalignment of 40°(25° + 15°).

The ¹³CO emission from AS 205 N shows a distinct NE/SW asymmetry, with NE regions being brighter. This is especially apparent in the 4.40 km s⁻¹ channel in Figure 3, for example. If attributed solely to disk emission, this effect is naturally produced if the NE side is the far side of the disk, owing to a combination of projection and optical depth effects (e.g., Pontoppidan et al. 2009; Rosenfeld et al. 2013). This orientation is consistent with the spectro-astrometric results of Pontoppidan et al. (2011). However, this interpretation of the asymmetry is not definitive, as our data set includes several features that are not well explained by a disk model alone. It is not possible to determine the near/far-side orientation of the disk around AS 205 S with these data.

In this work, our disk models assume a broken power-law surface density that declines sharply at some radius R_out. For AS 205 N, the data are consistent with R_out ∼ 80 AU. This is demonstrated in Figure 6, which shows a normalized spatial profile along a cut in the zeroth-moment map of ¹³CO emission. We have chosen to extract an E–W oriented profile to minimize contamination from the extended emission observed toward the NW of the image, as well as from the southern source. The FWHM of the observed spatial profile is ∼12. We show that the observed ¹³CO spatial profile is well reproduced by our nominal disk model.

Zoom In Zoom Out Reset image size

The line emission from AS 205 S, however, is more compact. Because the zeroth-moment ¹²CO map is dominated by emission from AS 205 N, we consider in this case a single channel map, shown in Figure 6, which does not suffer any contamination. In this case, the FWHM of the emission is only 09, and a model with R_out = 80 AU does not fit the data. Instead, we find that models with R_out ∼ 30–40 AU provide a reasonable fit. In principle, the spatial extent of the image also depends on the disk inclination and position angle; however, we find the spatial extent of the model to be relatively insensitive to position angle and inclinations in the range of 15°–35°.

This truncation can be compared with expectations from binary disk truncation models. For example, the parametric fit to truncated disk sizes from Pichardo et al. (2005) predicts a relative size ratio for the two disks of a binary that depends only on the binary mass ratio (assuming that the disks are close enough to each other to be truncated). For this system, with M₁ = 1.0 M_☉ and M₂ = 0.6 M_☉, the predicted ratio of disk sizes would be ∼1.3. Our best-fit disk models have ratios near two—slightly larger than predictions. With the masses assumed by Eisner et al. (2005), M₁ = 1.2 M_☉ and M₂ = 1.28 M_☉, the two disks would be expected to be nearly the same size, which is not consistent with our observations.

3.4.1. Basic Characteristics of Extended Emission

3.4.2. Mass and Optical Depth

3.4.3. Gas/Dust Ratio

3.4.4. Non-Keplerian Velocity Field

The ¹²CO and ¹³CO first-moment maps also show a distinct "S"-shaped warp in the velocity field, which is different from expectations for a Keplerian disk. Figure 7 shows deviations of the observed velocity fields from our nominal disk model. The observations deviate from expectation by as much as ∼1 km s⁻¹.

Zoom In Zoom Out Reset image size

One possible explanation for our observations is that we are seeing small-scale structure in the surrounding molecular cloud, whose emission is primarily resolved out by the interferometer owing to its large spatial scale. The Ophiuchus cloud region was mapped in ¹²CO 1–0 by de Geus et al. (1990). AS 205 N lies to the north of most of the prominent filaments in the region and so is less likely to suffer from cloud contamination than regions closer to the core of Ophiuchus. More importantly, the local cloud velocity is ∼3 km s⁻¹, and, as shown in Figure 5, this velocity is far from the observed emission-line center. Additionally, much of the spatially extended emission we observe is redder than the emission-line center, further separating it from the cloud velocity. Velocities ≳4.5 km s⁻¹ are only observed in the ¹²CO 1–0 cloud maps at large distances (≳ 5°) from AS 205 N. Therefore, we do not believe that molecular cloud emission can be the cause of the observed spatially extended emission.

Based on a variety of evolutionary indicators, AS 205 N is a Stage II source, with a disk but no envelope. According to the classification scheme of Greene et al. (1994), Class II objects (believed to represent Stage II) have 2–24 μm spectral indices, α = (dlog λS_λ/dlog λ), between −1.6 and −0.3, while AS 205 N has a 2–24 μm spectral index in the range of −1 to −0.4. AS 205 N also has a bolometric temperature between 1200 and 1400 K⁶, while Class II objects are defined to have T_bol > 650 K (Chen et al. 1995). Furthermore, AS 205 has no detectable envelope in the millimeter maps obtained by Andre & Montmerle (1994) (with a beam size of 12''), while all of the Class 0/I objects in their sample do show resolved envelope emission. In addition, the total 1.3 mm single-dish flux density of 450 ± 90 mJy beam⁻¹ (assuming a 20% flux uncertainty; Andre & Montmerle 1994) is consistent with our measured system flux density of 441 mJy, suggesting little to no unseen envelope material. Attributing the 9 mJy difference to an envelope (assuming T = 30 K and κ = 0.02 cm² g⁻¹, which assumes a gas/dust ratio of 100) would imply a mass of 0.001 M_☉—at least an order of magnitude lower than typical envelope masses in Ophiuchus (Andre & Montmerle 1994), although the large error bars on the single-dish measurement could allow a mass up to 0.01 M_☉. Therefore, AS 205 sits in the Class II regime, although the presence of a very low mass envelope cannot be excluded.

However, these ALMA observations represent a new resolution regime for observations of low-J CO emission from protoplanetary disks, and studies at similar resolution have so far focused on more evolved transition disks (e.g., Andrews et al. 2012; Casassus et al. 2013). Instead, AS 205 N's somewhat high spectral index, as well as its relatively high accretion rate (7 × 10⁻⁷ M_☉ yr⁻¹; Eisner et al. 2005) and disk mass, could imply that it sits at the evolutionarily younger end of Stage II. Therefore, in principle this extended emission could represent the first observation of a small-scale, low-mass remnant of the natal envelope. However, this hypothesis suffers from stability arguments. Cloud material at a distance of 2'' should not be stable in a system with a 13 binary; this distance is too far to be stable circumprimary material and too close to be stable circumbinary material (Artymowicz & Lubow 1994).

We consider here whether the observations might be explained as accretion flows from a circumbinary disk. The arc-like structures seen in the extended ¹²CO emission near AS 205 N appear similar to those observed around other ∼1'' separation binaries, such as UY Aur, GG Tau, and SR 24 (Duvert et al. 1998; Close et al. 1998; Roddier et al. 1996), structures that are posited to represent accretion streamers from circumbinary disks. Since this scenario requires the existence of a circumbinary disk, we discuss its possible existence below.

Based on the nondetection of a circumbinary ring in ¹²CO emission, AS 205 cannot have a circumbinary disk similar to those around either GG Tau or UY Aur. GG Tau has a circumbinary disk extending from ∼1'' to 2'' from a 024 binary (White & Ghez 2001), with a total mass of ∼0.1 M_☉ (Piétu et al. 2011). The UY Aur circumbinary disk has a radius of ∼500 AU and a mass of ∼10⁻³–10⁻² M_☉ (Duvert et al. 1998). Assuming a face-on circumbinary ring extending from 2'' to 6'' (250–750 AU), an excitation temperature of 30 K, an H₂:CO ratio of 5000, and using a 3σ upper limit derived from our zeroth-moment ¹²CO map (0.03 Jy km s⁻¹ beam⁻¹), we derive an upper limit to the mass of a circumbinary disk around AS 205 of 10⁻⁴ M_☉—significantly lower in mass than that observed around either GG Tau or UY Aur. Note that we use a CO abundance consistent with the ¹³CO abundance utilized by Duvert et al. (1998), and so the choice of CO abundance does not affect the comparison between the two disks. In addition, we note that our observations are sufficiently sensitive to detect disks with similar surface densities as those observed around GG Tau or UY Aur.

The circumbinary disk could potentially remain hidden if it were significantly more extended (and therefore resolved out by the interferometer). We are sensitive to circumbinary structure up to ∼8''. Since the observed ¹²CO emission extends no more than ∼2'', it cannot represent an accretion streamer from an 8'' wide disk. Models of this process (e.g., Günther & Kley 2002; Ochi et al. 2005; Hanawa et al. 2010) show continuity of the accretion streamers out to the inner edge of the circumbinary disk.

It is also possible that significant freeze-out of CO could hide a circumbinary disk from our view. However, single-dish flux measurements for this source place upper limits on the total mass of the system. AS 205 has a single-dish 1.3 mm flux consistent with its interferometric flux (as discussed in Section 4.2) and therefore consistent with little or no unseen material. Nevertheless, the large error bars on the absolute value of the 1.3 mm flux could allow for a mass of up to ∼0.01 M_☉ (where this implicitly assumes a gas/dust ratio of 100). We note that this situation is in contrast to the binary SR 24, which shows features in scattered light that have been attributed to accretion streamers (Mayama et al. 2010) and, therefore, could be considered an analog to AS 205. SR 24 has a single-dish flux three to four times higher than its interferometric flux (Andrews & Williams 2005), consistent with a significant amount of extended emission—a mass of ∼0.02–0.03 M_☉, assuming κ_1.3 mm = 0.02 cm² g⁻¹ and T = 30 K.

In conclusion, we show that observations of AS 205 place very firm limits on the possible mass of a CO-rich disk (≲ 10⁻⁴ M_☉), and the single-dish flux is consistent with no unseen dust, although a total disk mass of up to 0.01 M_☉ (assuming a gas/dust ratio of 100) is allowed by the large error in absolute flux. With no evidence of a large reservoir to supply mass to the circumstellar disks, we believe that it is unlikely that we are observing accretion streams from a circumbinary disk. Additional single-dish measurements would be helpful to place firmer limits on any possible unseen material in this system.

Another possible explanation for our observations is tidal stripping by AS 205 S. This explanation has been previously invoked by Cabrit et al. (2006) to explain the morphological and kinematic signatures in the disk around the T Tauri star RW Auriga A, which bears some interesting similarities to our observations. RW Aur is a 14 binary whose ¹²CO 2–1 emission shows a redshifted "tail"-like feature extending to 2'', although at somewhat higher velocities, up to ∼7 km s⁻¹—equivalent to 2–2.5 km s⁻¹ if adjusted to the 15° inclination of AS 205 N.

We first evaluate the relevance of tidal effects by computing the tidal radius of the secondary. For M₁ = 1.0 M_☉ and M₂ = 0.6 M_☉, the binary mass ratio is μ = M₂/(M₁ + M₂) ≈ 0.38. The corresponding Lagrangian point L₁ (Murray & Dermott 1999) is located at ∼0.55r_sep ∼ 89 AU from M₁, that is, reaching just outside the ∼80 AU outer radius of the AS 205 N disk. Consequently, some perturbations to the disk symmetry are to be expected.

Tidal effects will be strongest if the encounter is resonant or "quasi-resonant" (D'Onghia et al. 2010). Roughly speaking, as in a forced oscillator model, tidal interactions are most pronounced when there is frequency matching between the internal modes of the victim (in this case, the disk around AS 205 N) and the external forcing frequency exerted by the perturber, M₂. In other words, maximum response occurs at the disk radii R such that the internal orbital frequency $\Omega (R)=\sqrt{GM_1/R^3}$ is an integer fraction of $\Omega _\mathrm{orb}=\sqrt{G(M_1+M_2)/p^3}$, where p is the pericenter distance.⁷ Note that the resonance also requires a prograde orbit of the perturber with respect to the victim's spin.

A single passage by the perturber can excite prominent two-armed spirals (similar in morphology to what we observe) in the victim provided that the resonance occurs within the victim's disk's radial extent (see, e.g., Barnes & Hernquist 1996; Mihos & Hernquist 1996; D'Onghia et al. 2010). However, if the pericenter distance, p, is equal to the binary projected separation of ∼160 AU or larger, low-order near-resonance is not achieved for radii inside the primary's disk with our assumed μ = 0.38. To achieve low-order resonance, AS 205 S would need to be more massive. As discussed in Section 3.2, there is some uncertainty about the mass of AS 205 S. With the AS 205 S mass estimate derived by Eisner et al. (2005), ∼1.3 M_☉, resonant forcing would occur at ∼70 AU, just inside the outer radius of the AS 205 N disk.

However, even if AS 205 S is massive enough to have low-order resonant interactions with the disk of AS 205 N, a major drawback of this explanation is that bound pairs are expected to be already devoid of gas at the lowest-order resonances that create strong two-armed features, since those resonances would be quickly cleared out after a few subsequent pericenter passages (Paczynski 1977; Artymowicz & Lubow 1994). Therefore, we would need to be observing AS 205 during its first, or nearly first, encounter. This is unlikely if AS 205 is a bound pair, as the orbital period is approximately a few thousand years, while the estimated age of AS 205 is 10⁵ yr (Prato et al. 2003).

It is also very unlikely that this is a first (flyby) encounter between two stars. Thies et al. (2005) suggest that close stellar encounters may be likely during the ∼6 Myr lifetime of disks around stars in massive young clusters; with encounter probabilities of 10%–30% for encounter distances of 60–200 AU (assuming a cluster mass of 500 M_☉ and a size of 0.3 pc), one can expect ∼100 close stellar encounters throughout the lifetime of the cluster. However, Ophiuchus is significantly less massive and dense than this model presumes, with a stellar mass of ∼53 M_☉ in the central 2 pc² of the ρ Oph cloud core (Kenyon et al. 1998), and, furthermore, AS 205 lies in a less dense region at the edge of the Ophiuchus cloud (de Geus et al. 1990). Assuming 53 M_☉ in 2 pc², and using the equations found in Thies et al., we find encounter probabilities of a few × 10⁻⁵ out to 300 AU—equivalent to an expected number of collisions of a few ×  10⁻³. Finally, we must compute not the likelihood of an encounter, but the likelihood of observing an encounter in progress. If we assume an encounter velocity of 1 km s⁻¹ and an impact parameter of 500 AU, the encounter lasts ∼2000 yr —a factor of 500 shorter than the cluster age of at least 1 Myr (e.g., Wilking et al. 2005); therefore, the probability of viewing an encounter is 500 times lower than the probability of an encounter occurring. These factors combined suggest that the probability of observing a flyby encounter with an impact parameter of a few hundred AU in this region is <10⁻⁵.

Although lower-order resonances do not appear to explain the features we observe, in the long run, the evolution of circumprimary and circumbinary disks is dominated by higher-order, weaker resonances. As with the lower-order resonances, the resonant forcing and the internal damping of the disk reach a quasi-stationary state, in which the disk size is nearly in equilibrium. A balance is reached between the viscous and the tidal torques, eventually defining the truncation radius of both circumprimary and circumsecondary disks (Artymowicz & Lubow 1994). However, one subtlety in the analysis by Artymowicz & Lubow (1994) is that the equilibrium truncation radius is reached only after a viscous relaxation timescale that depends on the binary eccentricity, the final disk radius, the disk aspect ratio and the disk viscosity. Therefore, for very young systems, transient responses on the disks are still possible during every pericenter passage, especially for very eccentric binaries. Unfortunately, at this point we are not able to pursue this analysis further owing to the largely unconstrained orbital properties of this system and the uncertain mass of AS 205 S.

Our observations might also be explained by some form of wind. Thermal (photoevaporative) winds are unlikely to explain our observations, as the driving radiation for these flows tends to destroy molecules (e.g., Gorti & Hollenbach 2009). On the other hand, MHD disk winds, which are believed to be (at least partly) responsible for the larger-scale bipolar jets and outflows emanating from young stars (e.g., Ferreira et al. 2006; Pudritz et al. 2007; Ray et al. 2007, and references therein), can produce relatively low-velocity winds with a significant molecular content (Panoglou et al. 2012).

MHD disk winds have launching velocities that are similar to the sound speed (e.g., Kurosawa et al. 2006); therefore, tracers of the wind-launching region can have relatively low velocities. Temperatures in the launch region range from a few thousand K in the inner disk down to <100 K in the outer disk, with associated velocities in the range of ∼1–3 km s⁻¹. Observed launching velocities will be even lower if the streamlines are not directed toward the observer. However, poloidal velocities increase with distance along a streamline, reaching velocities as high as approximately a few hundred km s⁻¹ in existing models (Pudritz et al. 2007), which are much higher than we observe here. Another obstacle for disk wind models may be whether there is sufficient overall gas density to explain the extended emission. In the models of Panoglou et al. (2012), CO densities of ∼0.1 cm⁻³ result in a theoretical column density of ∼10⁻⁹ g cm⁻², assuming a column with vertical height 20 AU. This is three orders of magnitude below the observed lower limit to the column density, ∼10⁻⁶ g cm⁻². However, if lower terminal velocities are possible, then mass conservation will correspondingly increase the density in the wind, allowing models to match observations. We consider this in Section 4.5.2.

4.5.1. Relationship to Prior Observations of Outflows

4.5.2. Comparison with Model Predictions

A simple parameterization of a disk wind was implemented by Pontoppidan et al. (2011) to fit observed rovibrational emission lines and incorporated into the radiative transfer code RADLite (Pontoppidan et al. 2009). We describe this model and our choice of nominal model parameters in detail in the Appendix.

We find that the nominal wind model does not produce any appreciable extended emission, beyond that expected for a disk, owing to the low predicted densities at distances of hundreds of AU from the central star. However, the model does reproduce some of the other features of our observations. As shown in Figure 8, it produces additional flux at higher velocities compared with a disk-only model. The wind model also produces deviations from a Keplerian velocity field, including a curved first-moment map, demonstrated in Figure 9. The curvature in the model is due to two effects. At small radii, the dominant effect of the wind is that it makes the azimuthal velocities sub-Keplerian. At larger radii, the poloidal velocities are larger and begin to dominate the differences, causing the far side of the disk (in this case, toward the northeast) to be preferentially redshifted and the near side (toward the southwest) to be preferentially blueshifted. There are some similarities between the modeled and observed first-moment maps, in that both show a complex "S"-like shape, although it is clear that the observed moment map is more complex than the modeled map.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

A twisted velocity field can also arise from radial inflow, as suggested by Rosenfeld et al. (2014) to explain the velocity field of the transitional disk around HD 142527. Inflow and outflow can only be distinguished if the near and far sides of the disks are identified. In detail, HD 142527 is not an analog to AS 205 N; the fast radial inflow in HD 142527 is proposed to reconcile its high accretion rate and inner clearing, while AS 205 N has a full disk with no inner clearing. In addition, radial inflow alone would not explain the very extended ¹²CO emission we observe. However, some wind models themselves predict that accretion may be dominated by restricted regions of the disk producing fast flows (Bai & Stone 2013). The resultant signature in the velocity field would then depend on the relative contributions of inflowing and outflowing gas to the observed emission and could vary throughout the disk, creating a complex velocity field. Unfortunately, disk-wide simulations of this process do not yet exist for comparison with observations.

Although the nominal wind model does not produce any appreciable extended emission, the density of the wind depends on the assumed mass-loss rate and terminal velocity. Figure 8 shows predicted channel maps for an alternative wind model, with a low terminal velocity (2 km s⁻¹) and high mass-loss rate (2 × 10⁻⁷ M_☉ yr⁻¹), which shows appreciable extended emission. The detailed structure does not obviously match our observations. It should be noted, however, that dynamical timescales for the binary orbit and the wind are similar, and so AS 205 S should cause perturbations to the wind structure. Therefore, our observations may be explained by a combination of disk winds and binary interaction.

In conclusion, wind models are in some ways consistent with the observations of AS 205 N. In particular, models produce a curved first-moment map with some similarities to the observed ¹²CO velocity field. However, existing wind models do not readily produce significant extended emission. Extended emission instead requires that the models be modified to have lower terminal velocities and/or higher mass-loss rates. Although the geometry of the resulting emission does not precisely match our observations, the detailed features we observe cannot be produced by an axisymmetric model. They could, however, be produced by interactions between the wind and the stellar companion, AS 205 S.

These are the highest-resolution observations of rotational CO emission from AS 205 to date. AS 205 was observed in ¹²CO 3–2 by Andrews et al. (2009) with a beam size of 219 × 160 and in ¹²CO 2–1 by Öberg et al. (2011) with a beam size of 31 × 20. In both cases, the beam size was not sufficient even to resolve the binary. Since the extended structure we observe is only a few arcseconds in size, a resolution of ∼1'' or better is required.

Only a few disks have been observed in molecular lines at a resolution of ∼1'' or better to date, as pre-ALMA observations were limited to the brightest disks. Therefore, it is difficult to say whether the phenomenon we observe is a common or unusual feature for protoplanetary disks. Interestingly, molecular line imaging at or near this resolution often seems to reveal surprises—for example, an extended CO disk around TW Hya (Andrews et al. 2012), an asymmetric, perhaps tidally induced "arm" near RW Aur (Cabrit et al. 2006), and asymmetric spiral arms in the disk of AB Aur (Corder et al. 2005; Piétu et al. 2005). ALMA observations of a few disks have already revealed azimuthal asymmetries and non-Keplerian flows (Casassus et al. 2013; Christiaens et al. 2014; Rosenfeld et al. 2014), as well as a helical disk wind (Klaassen et al. 2013). Therefore, it seems likely that the resolution and sensitivity afforded by ALMA will continue to reveal many new small-scale asymmetric structures.

If the extended structure we observe here is a common feature for protoplanetary disks, this could have implications for the interpretation of lower-resolution observations. For low-resolution observations, the observed CO flux could incorrectly be attributed solely to a disk. Although it is difficult to determine the precise contribution to the flux from the wind, the box shown in Figure 1(c) contributes 5% of the total observed ¹²CO 2–1 line flux. Also, if the wind has a different temperature than the disk, interpretation of observations of multiple CO isotopologues could incorrectly determine the gas temperatures. Since gas temperatures are in turn used to determine gas masses (e.g., Panić et al. 2008), this could have important consequences for our understanding of the planet-forming capabilities of disks.

If we are actually observing a wind, with some simple assumptions, we can estimate the mass-loss rate. For this calculation, we use the CO observed in the large box in Figure 1(c) (see Section 3.4.2) and multiply by four to account for all directions of the wind. We obtain M_gas ∼ 8 × 10⁻⁵ M_☉ for an excitation temperature of 30 K, or M_gas ∼ 2 × 10⁻³ M_☉ for 1000 K. The mass-loss rate is $\dot{M}_\mathrm{loss}\approx M_\mathrm{gas} v_\mathrm{gas}/D_\mathrm{gas}$, where v_gas and D_gas are the actual velocity and length of the outflow, respectively. Both D_gas and v_gas are different from the observed distance and velocity, where D_obs = D_gassin θ, v_obs = v_gascos θ, and θ is the angle between the outflow and the observer. Thus, $\dot{M}_\mathrm{loss}\approx ({M_\mathrm{gas} v_\mathrm{obs}\sin {\theta }}/{D_\mathrm{obs}\cos {\theta }})$. Taking D_obs ∼ 300 AU, v_obs ∼ 1 km s⁻¹ and assuming θ = 15° (equal to the best-fit disk inclination), we find $\dot{M}_\mathrm{loss}\sim 10^{-8}\hbox{--}10^{-7}\, M_\odot \ \mathrm{yr}^{-1}$. A different assumption about the inclination of the flow can either raise (for i > 15°) or lower (for i < 15°) the mass-loss rate. The disk + wind models described in Section 4.5.2 also required a rate of ∼10⁻⁷ M_☉ yr⁻¹ to produce significant extended emission.

Using observations of CO rovibrational line shapes and spectro-astrometry along with disk+wind radiative transfer models, Pontoppidan et al. (2011) estimated a similar or slightly lower mass-loss rate for AS 205 N—9 × 10⁻⁹ M_☉ yr⁻¹. Both results therefore suggest that $\dot{M}_\mathrm{loss}/\dot{M}_\mathrm{acc}\sim 0.01\hbox{--}0.1$.