Irregular Dust Features around Intermediate-mass Young Stars with GPI: Signs of Youth or Misaligned Disks?

The presented work is part of Gemini-LIGHTS (Gemini Large Imaging with Gpi Herbig/T-tauri Survey), a systematic survey of intermediate-mass pre-main-sequence stars that are thought to host disks. We observe using the Gemini Planet Imager (GPI), an NIR instrument on the Gemini South telescope (Macintosh et al. 2014; Perrin et al. 2015). Like most instruments of its kind, GPI uses PDI with extreme adaptive optics (XAO) to counteract wavefront distortion from atmospheric turbulence and obtain images with resolution close to the diffraction limit (Macintosh et al. 2014). The survey aims to observe a representative sample of over 30 young objects to find the frequency of scattered light features, to link observed features with the objects' SED structures, and to search for trends in the disk properties with the stellar masses and cluster ages.

The sample was selected from the full list of ∼500 objects from the nearest stellar associations and other T Tauri stars, Herbig Ae/Be, young stellar objects, and pre-main-sequence stars from the literature. We selected our final survey targets from this list by considering observational constraints. The survey targets are nearby young objects with strong NIR and mid-infrared emission (i.e., debris disks are excluded) and with I-band magnitude brighter than 9 mag so that GPI's wavefront sensor can function well. Originally, we selected objects with a range of disk types—a roughly equal number of transition disks, pretransition disks, and full disks. The full SED shapes and features provide evidence of disk structures out to several tens of astronomical units, and in selecting our sample we simplified this by using a color–color diagram to probe disk structures within a few astronomical units. The targets were grouped using the color–color diagram according to the ratio of WISE2 – WISE4 to J – WISE2 magnitudes, where the dividing line between full and transition structures corresponds to a power-law SED for a flat spectrum in λF_λ space. The objects were selected for having significant excesses in the J – WISE2 and WISE2 – WISE4 color–color plane, and so likely harboring a disk. For the presented objects, these magnitudes are given in Table 1, and a color–color diagram is shown in Figure 1.

Zoom In Zoom Out Reset image size

During the course of the Gemini-LIGHTS survey, we published data on resolved ring structures and other scattered light detections from Herbig Ae/Be stars (Monnier et al. 2017), and recently we have also presented spectacular structures in the disk around HD 34700 including a bright ring with many spiral arms (Monnier et al. 2019). The observed structures typically span several hundred astronomical units.

Here we present five objects with significant scattered light, four of which are new detections. They represent some of the more spectacular examples from our sample due to their especially bright and extended features, and we present an in depth analysis of each. Note that the observations for the survey are now complete, and it will conclude with a statistical analysis of all of the targets in a forthcoming paper.

In Section 2, we review the literature on the five objects, followed by Section 3 with details of our observations and the data reduction process. Then, in Section 4, we present results including images of our reduced data, with focused analysis of each object and discussion of links between the observed scattered light features and literature predictions. We give our conclusions in Section 5.

FU Ori is the prototype of the FU Orionis class of objects that are characterized by variability as they undergo a phase of extremely high mass accretion, predicted to last decades to centuries (Herbig 1966, 1977; Hartmann & Kenyon 1985; also see reviews by Hartmann & Kenyon 1996; Audard et al. 2014). The distance to the object is 416.2 ± 8.6 pc (Gaia Collaboration et al. 2016, 2018), and it is composed of a 2 Myr old visual binary system where the masses of FU Ori and its companion FU Ori S are 0.3 M_⊙ and 1.2 M_⊙, respectively (Wang et al. 2004; Zhu et al. 2007; Beck & Aspin 2012). The two components have a separation of 050 (Wang et al. 2004). Comparing infrared excesses (see Table 2 and Figure 1) indicates that there is a full disk structure. The object has an inclination of 52°–58° (Malbet et al. 2005; Quanz et al. 2006).

Subaru-HiCIAO scattered light observations of FU Ori in the H band reveal an asymmetrical structure spanning ∼400 au with a large-scale stream and a bright northeastern arm extending ∼200 au, features that are likely a stream or filament in the envelope (Liu et al. 2016; Takami et al. 2018). Liu et al. (2016) conclude that the features may be caused by gravitational instability and that the companion star FU Ori S is also seen via scattering in its own circumstellar disk. K-band observations from the Keck Interferometer reveal that the asymmetric features continue down to the inner 1 au scale (Eisner & Hillenbrand 2011).

FU Ori has also been the subject of resolved submillimeter and radio observations. No substantial circumbinary ring was detected in ALMA band 7 continuum observations; however, each star could have its own unresolved disk of maximum radius 01 or 200 au (Hales et al. 2015; scaled to Gaia DR2 distance). The 33 GHz continuum observations of FU Ori with the Very Large Array (VLA) give a lower limit on the system dust mass of 8–16 × 10⁻⁵ M_⊙ with assumed brightness temperature ${210}_{-51}^{+81}$ K and opacity 0.07–0.13 cm² g⁻¹ (Liu et al. 2017). The respective dust mass lower limits of FU Ori and its companion FU Ori S, derived from ALMA continuum emission, are 2 × 10⁻⁴ M_⊙ and 8 × 10⁻⁵ M_⊙, respectively (Hales et al. 2015).

MWC 789 is a Herbig Be star of spectral type B9 located in the Gemini OB1 molecular cloud complex at a distance of ${697}_{-64}^{+94}$ pc and effective temperature of 8700 K (Carpenter et al. 1995; Hernández et al. 2004; Gaia Collaboration et al. 2016, 2018; Vioque et al. 2018). A stellar age of ${1.42}_{-1.09}^{+2.21}$ Myr has been derived from VLT spectra, broadly consistent with the age of ${2.56}_{-0.14}^{+0.30}$ Myr from fits to pre-main-sequence evolutionary tracks (Fairlamb et al. 2015; Vioque et al. 2018, respectively).

The object's Hα polarization has a position angle of 175°, which is unstable with a variation of 65°, and the continuum polarization is observed to vary on a month timescale (Jain & Bhatt 1995; Vink et al. 2005). This could be linked to the system's rotating accretion disk and fast stellar rotation shown by the short-term variability of the Ca ii K line (Catala et al. 1991; Vink et al. 2002). The ratio of infrared excesses indicates a disk that is on the border between the predicted transition disk and full disk structure (see Figure 1). An SED fit, based on the model grid of Robitaille et al. (2007), gives a total disk mass log M_disk = −1.02 ± 0.13 log M_⊙ (Liu et al. 2011). There are no published scattered light images of this object, but H-band NIR interferometry with VLTI/PIONIER has been used to derive a disk inclination of 52° and position angle 132° (Lazareff et al. 2017). However, we note that observations of CO rovibrational lines and the narrow OH ²Π_3/2 P4.5 (1+, 1−) line using CRIRES on the VLT indicate a more face-on configuration with inclination 8°–15° (Fedele et al. 2011; Hein Bertelsen et al. 2016).

There have been 1.3 mm continuum emission nondetections of both the circumstellar dust of the star and its associated globule CB 39 with the IRAM 30 m and SEST 15 m telescopes, giving a gas mass <4.0 × 10⁻¹ M_⊙ (Launhardt & Henning 1997; Henning et al. 1998). This is consistent with the Liu et al. (2011) disk mass estimate. Liu et al. (2011) also report detections of emission lines of ¹³CO (2–1), CO (2–1), and CO (3–2). The object also displays Brackett−γ emission, indicative of an accretion rate of 1.6 ± 0.4 × 10⁻⁸ M_⊙ yr⁻¹ (Donehew & Brittain 2011). Ultraviolet absorption line observations have been used to estimate an H₂ column density of ${1.81}_{-1.09}^{+0.89}\times {10}^{19}\,{{\rm{cm}}}^{-2}$ (Bouret et al. 2003) with radial velocity measurements suggesting it is bound to the star. We note that this H₂ column is consistent with the line of sight to MWC 789 that probes a low-density envelope rather than a disk, suggesting the object has a low inclination.

HD 45677 is a variable Herbig Be star exhibiting the B[e] phenomenon (Allen & Swings 1976; The et al. 1994). The evolutionary status of HD 45677 is contested owing to its peculiar properties such as forbidden line emission (Allen & Swings 1976; Lamers et al. 1998). These have led to the creation of the FS CMa classification of B[e] stars, of which HD 45677 is the prototypical object (Miroshnichenko 2007). Its location on the Hertzsprung–Russell diagram points to it being on/near the main sequence (Miroshnichenko 2007), but its classification as a Herbig Be star suggests that it is pre-main sequence (The et al. 1994). The SED is consistent with a full disk structure, which may imply that the object is at an early evolutionary stage. An alternate interpretation of HD 45677 is that it is an evolved close binary, as the irregular emission spectra variations match those from hot winds from later evolutionary stages (Miroshnichenko et al. 2015). HD 45677 is also isolated from star-forming regions and has been considered unlikely to be pre-main sequence on this basis (Herbig 1994). The latest Gaia data release gives a distance to HD 45677 of 620 pc (Gaia Collaboration et al. 2016, 2018) and Vioque et al. (2018) used this with pre-main-sequence stellar tracks to derive an age of 0.6 Myr. This young age favors the interpretation that HD 45677 is a pre-main-sequence object. In Section 4.3, we perform our own analysis of the stellar photometry of this object in an attempt to better constrain the stellar properties.

The innermost region of the disk has been modeled using a ring with radius <10 mas in order to fit H-band IOTA interferometeric visibilities (Monnier et al. 2006). The ring is inclined at 43° and has a position angle of 70° with the westward side of the ring facing toward the observer (Lazareff et al. 2017).

Various studies have pointed toward HD 45677 having a bipolar outflow of material. An outflow is consistent with complex Balmer lines (Swings 1973; Israelian et al. 1996), is required for the continuous production of small grains responsible for the object's gradual color change (de Winter & van den Ancker 1997), and is necessary to explain the variation with wavelength of the degree and position angle of polarization (Schulte-Ladbeck et al. 1992). Patel et al. (2006) proposed that HD 45677 has a polarization angle of 164° ± 3° and an outflow with a position angle of $175^\circ \pm 1^\circ$.

Hen 3-365 is a system containing a massive star that exhibits B[e] phenomenon and is thought to have a mass of ∼25 M⊙ (e.g., Oudmaijer et al. 1998). K- and H-band image reconstructions from VLTI/AMBER reveal a companion with a separation of ∼34 ± 0.5 mas (Millour et al. 2009). The two stars are surrounded by an oxygen-rich dusty envelope or circumprimary dust ring at  6 au, with an inner dusty disk radius ∼2.5–3.0 au (Millour et al. 2009).

The distance to Hen 3-365 is uncertain and contributes to the unknown evolutionary status of the star. The extinction and kinematics indicate that Hen 3-365 is more likely an evolved supergiant at a distance of several kiloparsecs than a close young star (Oudmaijer et al. 1998), and Maravelias et al. (2018) classify Hen 3-365 as a massive supergiant. However, the H I column density is far too high for Hen 3-365 to be a supergiant (Oudmaijer et al. 1998). Vioque et al. (2018) derived a distance to Hen 3-365 of 2010 pc using a prior, as the Gaia DR2 distance to Hen 3-365 is unreliable, due to a poor parallax calculation (potentially caused by confusion with the object's close companion). But Vioque et al. (2018) noted that if Hen 3-365 is an evolved object, it would be inappropriate to use pre-main-sequence tracks as they have. This could partially explain the large uncertainties on the other derived parameters for Hen 3-365 in Table 3.

Assuming that Hen 3-365 is a pre-main-sequence object, the SED of Hen 3-365 is best modeled using a transition disk structure as the mid-infrared excess is greater than the NIR excess (see Table 1, Figure 1). The system displays 1.0–2.5 μm NIR excess from thermal dust emission with a hot and cold dust mass of 1.6 × 10⁻⁷ M_⊙ and 3.7 × 10⁻³ M_⊙, respectively (Allen 1973; McGregor et al. 1988). CO band emission is observed in a ring ∼200 au from the star and is best fit with a model ring with a near pole-on inclination of 74 (Maravelias et al. 2018).

The Hen 3-365 system contains a fast polar wind, a slow disk wind, and a large nebula that impact the disk properties (Oudmaijer et al. 1998). The reflection nebula is seen in cool dust emission and has broken structures that could be due to periodic mass transfer between the two stars if Hen 3-365 is an evolved system (McGregor et al. 1988; Millour et al. 2009). The geometry of these nebula components matches the smaller-scale polarization angle of the asymmetric disk seen with spectropolarimetry (Oudmaijer et al. 1998). The reflection nebula is likely caused by mass loss via an outflow as seen from P Cygni profiles of Ca ii H and K and Balmer lines (Surdej et al. 1981). The outflow velocity is variously described as having a lower limit of 1200 km s⁻¹ or a velocity of ∼800 km s⁻¹ (Surdej et al. 1981; de Freitas Pacheco et al. 1985). Baines et al. (2006) describe two outflows, one symmetric in the EW direction and one asymmetric in the N direction, that could have been caused by a bow-shock structure in the north.

HD 139614 is an A7-type star in the Sco OB2-3 association whose age estimates range between 9 and 14.5 Myr (Acke & van den Ancker 2004; Alecian et al. 2013; Matter et al. 2014; Vioque et al. 2018) with a stellar mass of 1.8 M_⊙ and accretion rate 10⁻⁸ M_⊙ yr⁻¹ (Garcia Lopez et al. 2006; Alecian et al. 2013). The BANYAN Σ association finder places HD 139614 in the Upper Centaurus-Lupus group with 99.9% likeliness (Gagné et al. 2018). This is a group with stars of ages 12–15 Myr (de Geus et al. 1989), and Vioque et al. (2018) derived an age of HD 139614 of 14.5 Myr from pre-main-sequence tracks. The object has a transition disk structure based on its infrared excesses (see Table 1 and Figure 1). The warm dust mass is 2.9 × 10⁻⁸ M_⊙ (Meeus et al. 1998; Mariñas et al. 2011), and the disk contains polycyclic aromatic hydrocarbons with a mass within 10 au of the star of 1.83 × 10⁻⁶ M_Earth (Seok & Li 2017). The NIR-emitting region of the disk spans 0.2–2.5 au with a dust scale height of ∼0.01 au at a radius of 0.2 au (Matter et al. 2016). There is a gap between an inner halo component and the dust disk from 2.5 to 5.7 au with a 10³ factor of depletion compared with the outer disk (Matter et al. 2014, 2016; Menu et al. 2015), which continues from 5.6 au to an outer radius of 75–100 au (Panić and Hogerheijde 2009; Labadie et al. 2014).

The object's very low v sin i (24 ± 1 km s⁻¹; Dunkin et al. 1997b; Alecian et al. 2013) is indicative of a near pole-on inclination of ∼5°. This is supported by the object's low linear polarization (Yudin et al. 1999). Disk model fits to NIR VLTI visibilities reveal that the inner disk is discontinuous, although there are no significant brightness asymmetries (Matter et al. 2014, 2016), and the inner rim is optically thin and not puffed up, allowing the star to warm the inner wall of the outer disk (Meeus et al. 2001; Matter et al. 2014). Warm dust in the flared outer disk is thought to cause the disk's mid-IR emission in the SED (Meeus et al. 2001).

Hydrodynamical simulations suggest that a planetary companion of mass ∼3M_Jup may be able to open the gap at 4.5 au from the star (Matter et al. 2016). A planetary companion at 4 au with mass <2M_Jup is also consistent with model fits to CO lines (Carmona et al. 2017).

There has been an H-band scattered light nondetection by Subaru-HiCIAO using XAO and a coronagraph but without polarimetry (Fukagawa et al. 2010). Other Subaru-HiCIAO H-band PDI observations found no companion candidates within 400 au (Uyama et al. 2017).

Our observations were taken using GPI (Macintosh et al. 2014), an instrument on the Gemini South telescope, during 2017–2019. The details including observation dates and data quality are in Table 4. The program IDs containing our observations are GS-2017A-LP-12, GS-2017B-LP-12, GS-2018A-LP-12, GS-2018B-LP-12, and GS-2019A-LP-12.

All observations used GPI in polarimetry mode and wavebands J or H. For most targets, individual exposures had an integration time of 29.10 s, with the exception of Hen 3-365 and HD 139614 (without a spot), with shorter times of 14.55 and 2.19 s, respectively, to prevent or minimize saturation. Our objects typically had 32 usable exposures, resulting in approximately one hour total observing time per target including overheads. Files were determined to be unusable if they suffered from poor AO performance. All of our data were taken in 70th percentile image quality, the second-best possible tier, and translating to seeing of 04–07 for the J band and 04–05 for the H band.

Our observations use PDI (e.g., Perrin et al. 2015). A half-wave plate (HWP) is used to restrict the polarization position angle of light that passes through the instrument. We use four HWP angles (0°, 225, 45°, 675); note: the 2017 April observations use an additional four angles where 90°, 1125, 135°, and 1575 are treated as equivalent to 0°, 225, 45°, and 675 respectively. After the HWP, the light passes through a polarizing Wollaston prism and is split into two beams with orthogonal polarization angles that fall onto the detector simultaneously. Frames from four HWP angles can be combined into a Stokes data cube (stokesdc—see Section 3.2) using singular value decomposition and double differencing (e.g., Perrin et al. 2015).

The observations use an Apodized-Pupil Lyot Coronagraph with spot radius 0085 for J-band and 0111 for H-band images (Macintosh et al. 2014). A coronagraph was used in all cases, except for observations of HD 139614 in the J band from 2017, where a set of observations was taken both with and without a coronagraph. The pixel scale of the detector is 00141, giving each stokesdc a maximum field of view of 27 × 27.

Our data reduction follows the process in Monnier et al. (2017, 2019) using the GPI Data Reduction Pipeline (DRP) version 1.4 routines to convert between file types—raw data files, polarization data cubes (podc), and Stokes data cubes (stokesdc; Maire et al. 2010). We also perform stellar polarization corrections and combine multiple data cubes using original routines written in Python to manipulate output files from the GPI DRP. The data were calibrated using dark and flat frames taken near the respective observation dates, which in turn were calibrated using the DRP as per the documentation. We note that we did not calibrate using low spatial frequency flat fields. When reducing the data, we did not include the system Mueller matrix when converting from podc to Stokes in the Combine Polarization Sequence primitive.

We follow conventions for the Stokes parameters Q and U and the radial Stokes parameters Q_ϕ and U_ϕ given in Monnier et al. (2019, Appendix A). In short, Q and U follow the IAU standards (Contopoulos & Jappel 1974) and Q_ϕ, U_ϕ use a modified version of that defined in Schmid et al. (2006), where now azimuthally oriented polarization is positive and polarization in the radial direction is negative. As such, the azimuthal Stokes maps Q_ϕ and U_ϕ have the following form:

$$\begin{eqnarray}&&{Q}_{\phi }=-U\sin (2\phi )-Q\cos (2\phi )\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&{U}_{\phi }=+Q\sin (2\phi )-U\cos (2\phi )\end{eqnarray} \tag{ 2 }$$

with polar angle

$$\begin{eqnarray}&&\phi ={\tan }^{-1}\left(\displaystyle \frac{Y-{Y}_{0}}{X-{X}_{0}}\right)+\gamma \end{eqnarray} \tag{ 3 }$$

for the pixel grid of X and Y coordinates with center pixel (X₀,Y₀) and the position angle γ.

The basic steps of the reduction process are as follows:

• 1.  
  DRP: raw data converted to podc.
• 2.  
  DRP: each group of four podc converted to stokesdc.
• 3.  
  Python routines: stellar polarization correction for each stokesdc.
• 4.  
  Python routines: all stokesdc combined into one final stokesdc.

To begin the reduction, the raw data files were converted into podc containing two orthogonally polarized images using the standard DRP recipe "Simple Polarization data cube Extraction." A dark field was subtracted from each image, then a flat field was used to find the expected locations of light on the detector. Finally, the images were destriped, had bad pixels interpolated, and were assembled into a podc.

The podc were split into groups of four podc that contained a full HWP cycle. The podc were checked individually by eye for evidence of poor AO performance, and cycles containing poor files were removed. Podc with poor AO performance had unusually extended point spread functions (PSFs) and very weak AO spots. These frames typically have larger rms wavefront error values compared with frames with better AO of the same object. For example, FU Ori frames that were removed for poor performance had larger rms wavefront values of ∼260 compared to frames with better AO performance of ∼216. When a poor podc was identified, the whole HWP cycle containing the podc was removed. We removed the final two cycles for FU Ori, the first cycle for Hen 3-365, and the final two cycles of HD 139614 from 2017 with the coronagraph. These changes are reflected in Table 4 along with the average rms wavefront error values for that observing epoch.

Each group was reduced separately into a stokesdc using the standard DRP recipe "Basic Polarization Sequence," without the step to subtract mean stellar polarization. In this recipe, the polarization pairs are cleaned using double differencing. Then all images were rotated to north and aligned, and converted to a stokesdc. We use the term "interim stokesdc" to refer to these stokesdc that are formed from only one HWP cycle. Later in the reduction process, the multiple interim files will be combined into one final stokesdc.

We then apply the stellar polarization correction to each of the multiple interim stokesdc. We found that the correction method used could give noticeably different resulting radial Stokes maps, particularly in the lower-quality U_ϕ maps. Similarly to Monnier et al. (2019), we expect effects from instrumental polarization to be nondominant (with systematics on the order of 1% of total intensity and a few percent for Q_ϕ), and so we have not taken further steps to remove these effects outside the following. We compared results from the DRP with other methods including U_ϕ minimization (e.g., Avenhaus et al. 2018) and fractional total intensity subtractions in Appendix A. Most of the methods could not completely remove the contribution from stellar polarization, which causes a "butterfly" pattern in Q_ϕ and U_ϕ that can obscure structures and change their sign. For the objects presented in this paper, we have elected to use the method that consistently removes the background stellar polarization while introducing few structures in Q_ϕ and U_ϕ compared with the DRP reduction. The method works as follows: in each interim Stokes cube, we examine a ring-shaped region near the detector's edge with radius 70 ≤ r ≤ 80 pixels (∼1.0 ≤ r ≤ 11). Using all pixels centered in this region, we calculate ratios of Stokes Q and U to total intensity I, f_Q and f_U, respectively:

$$\begin{eqnarray}&&{f}_{Q}=\displaystyle \frac{{\rm{mean(}}Q{\rm{)}}}{{\rm{mean(}}I{\rm{)}}},\end{eqnarray} \tag{ 4 }$$

$$\begin{eqnarray}&&{f}_{U}=\displaystyle \frac{{\rm{mean(}}U{\rm{)}}}{{\rm{mean(}}I{\rm{)}}}.\end{eqnarray} \tag{ 5 }$$

These ratios are used as scaling factors across the full map to calculate the weighted Stokes maps Q^⋆ and U^⋆:

$$\begin{eqnarray}&&{Q}^{\star }=Q-I\times {f}_{Q},\end{eqnarray} \tag{ 6 }$$

$$\begin{eqnarray}&&{U}^{\star }=U-I\times {f}_{U}.\end{eqnarray} \tag{ 7 }$$

Then, the corrected stokesdc is saved, containing the images [I, Q^⋆, U^⋆, V]. This gives radial Stokes maps with this form:

$$\begin{eqnarray}&&{Q}_{\phi }=-{U}^{\star }\sin (2\phi )-{Q}^{\star }\cos (2\phi ),\end{eqnarray} \tag{ 8 }$$

$$\begin{eqnarray}&&{U}_{\phi }=+{Q}^{\star }\sin (2\phi )-{U}^{\star }\cos (2\phi ).\end{eqnarray} \tag{ 9 }$$

Finally, all corrected interim stokesdc were combined into one final stokesdc. This combination used the mean of all maps for each Stokes parameter, e.g., Q_final = 〈Q₁, Q₂, ..., Q_n〉 for n interim stokesdc. Conversely, the default DRP method creates one stokesdc from all n × 4 input files and performs the stellar background removal once on the resulting stokesdc. Combining the interim stokesdc in this way results in an improved signal-to-noise ratio and works to counter the effects of the rotation of the sky relative to the fixed detector, which can cause any unobscured stars on the frame (e.g., companions) to appear smeared out over an hour-long set of exposures. Reducing the files in smaller batches essentially limits the exposure to a matter of minutes and allows less time for the companion star to move relative to the detector. This effect is particularly pronounced for companion stars owing to their much higher flux relative to background levels on the frame. Because the data is averaged without correcting for sky rotation, the PSFs add together in the detector frame of reference.

Flux calibration was performed using standard GPI DRP routines to measure the satellite spot flux, as in Monnier et al. (2019). For each object, all podc were stacked using the primitives "Accumulate Images" and "Combine 3D data cubes" to create a podc made from the mean of all other cubes without rotating to north to prevent effects from the rotation of the detector with respect to the sky, e.g., the satellite spots would no longer be in a predefined area and may have spread out depending on the duration of all frames, as can be seen in Figure 2 where the spots of HD 139614 in the J band appear stretched compared with those in FU Ori. The satellite spot fluxes were then measured in this mean cube only as it has higher signal to noise than the separate podc. For J-band images, we used second-order satellite spots that are more free of speckles. These second-order satellite spots contain 25% less flux than the first-order spots. This factor is unaccounted for by the pipeline, resulting in the returned calibration factors being too large. To counteract this, we apply an additional factor of 0.75 to the final flux scales for J-band objects, a change that is reflected in Table 5.

Zoom In Zoom Out Reset image size

The scaling factor is calculated from the known ratio of the stellar magnitude from Table 2 and the satellite spot flux using the following primitives: "Measure Star Position for Polarimetry," "Measure Satellite Spot Flux in Polarimetry," and "Calibrate Photometric Flux in Pol Mode." The scaling factor X follows that used by Wolff et al. (2016), where the DRP default units (Analog-to-Digital Units (ADU)/coadd) are equal to X mJy s⁻¹ arcsec⁻². The scale factor X is found by multiplying the initial scaling factor returned by the DRP by each frame's exposure time in seconds and dividing by the area of one pixel on the GPI detector, 0.0141 arcsec². The uncertainty in the flux scale values is dominated by the systemic error, 20% for the J band and 13% for H (see Monnier et al. 2019), once these values are combined in quadrature with the percentage error returned by the DRP.

Table 5 gives the scale factors for our observations. The scale factor for FU Ori is inconsistent for this sample as it is double the mean of the other J-band values. This is likely due to an inaccurate or outdated 2MASS magnitude, which does not reflect the gradual fall in the intensity of FU Ori since its initial flare-up in 1936 (Herbig 1966). Therefore, we have taken the scale factor for FU Ori to be this mean value, 2.2 ± 0.5.

FU Ori shows a radically asymmetrical dust distribution. Figure 6 highlights some of the features seen in Q_ϕ, including the companion FU Ori S (A), a bright arm that arcs from the east to the northern sides of the disk (B) before culminating in an infrared-dark stripe pointing north (C), a faint tail of material stretching toward the west (D), and a diffuse rounded region (E). The bright arm (B) spreads around 200 au (05) north then 200 au west and is bright enough to be faintly visible in the total intensity image (Figure 2). B resembles infalling streams of material observed around other objects with ALMA line emission (e.g., L1489 IRS, Yen et al. 2014; HL Tau, Yen et al. 2019).

Zoom In Zoom Out Reset image size

The companion of FU Ori (A), FU Ori S, is located roughly 05 (200 au) from the central star at a position angle of ∼165°, in agreement with Wang et al. (2004) and Reipurth & Aspin (2004). FU Ori S appears with a negative signal in Q_ϕ, i.e., the region around the companion is radially polarized with respect to the central star. This polarization position angle is also seen in the Subaru-HiCIAO H-band image (Takami et al. 2018). The companion appears distinctively in the radial profile of Figure 5, where there is an excess at 05 in total intensity and in U_ϕ but not in Q_ϕ. It is possible that this companion is bright enough to cause the U_ϕ structures east of the spot in Figure 4 in the same area as (B). In this case, U_ϕ would not indicate multiple scattering of light from the central star FU Ori, but a non-centrosymmetric polarization pattern due to light from FU Ori S. This would explain why U_ϕ is so strong in the dusty region nearest FU Ori S, in the bright arm to the east of FU Ori. However, we do not rule out the possibility that the U_ϕ structures are a result of multiple scattering from the central star.

Figure 6 also shows the Subaru-HiCIAO image of FU Ori in scattered light from Takami et al. (2018). Generally, we see the same features in both GPI and HiCIAO images, though the GPI image more clearly shows the northern dark stripe (C) and the small round structures just southwest of the spot (E). With GPI, we recover the same faint western stream (D) but at a much reduced fraction of the maximum arm brightness. We do not, however, recover structure F to the west of the spot in the Subaru-HiCIAO image, identified as a gas clump by Liu et al. (2016), and instead see a gap between the stream and the dust immediately to the west of the spot. We consider it unlikely that the nondetection of the HiCIAO feature F in our GPI image is due to temporal evolution in the density structure in the intervening 39 months or due to the different observing bands (H band and J band). Instead, we consider it more likely that feature F might represent residual speckle noise in the HiCIAO image. There is also the possibility that this temporal change is due to shadowing from the inner disk as in TW Hya (Debes et al. 2017) and LKHa 330 (Uyama et al. 2018), and the shadow is present in the GPI image but not the HiCIAO image. The speckle also appear in the same region as the faint feature E, which is indiscernible in the HiCIAO image.

There are just over three years between the Subaru-HiCIAO observation in 2014 October and the GPI observation in 2018 January. In this time there is no clear movement of FU Ori S with respect to the central star. Cross-correlation analysis is unable to overlap the positions of the star in the two images by first translating and then rotating one image with respect to the other. Comparing separations and position angles indicates that the companion FU Ori S has moved by a few 001 and/or a few degrees, but this is less likely a physical movement of the companion than an inconsistency in the calculated central star position. Having an incorrect central star position by even one pixel can cause changes to the separation and position angle on the order of the inconsistency between the Subaru-HiCIAO and GPI images. We calculate a back-of-the-envelope estimate for the orbital period T of FU Ori S assuming that the observed separation 204 ± 4 au is r, the radius of a circular orbit, and that the FU Ori system has a mass 1.5 M_⊙. Using Kepler's third law (T² ∝ r³), we find T ∼ 2400 yr. Therefore, in the three years between the HiCIAO and GPI observations the companion should have changed position angle on the order of 05, corresponding to a displacement of ∼1.6 au or 4 mas—about 30% of the size of one pixel in the GPI image. Any change in the companion position would be impossible to distinguish given that this value is on the order of our uncertainties.

In one theory of gravitational instability in disks, the growth timescale is at least as long as the orbital period at the outer disk edge (Adams et al. 1989). This would lead to prohibitively long timescales for growth, and thus this scenario is unlikely to be responsible for the observed dust structures. However, this picture ignores how disks are built up by infall from the protostellar envelope, and calculations including infall show that spiral structures can develop in this case on shorter timescales (Bae et al. 2014) or even gravitationally fragment into blobs (Vorobyov & Basu 2015). Such models are more relevant as it is generally thought that FU Ori outbursts occur during the protostellar infall phase (Hartmann & Kenyon 1996). In any case, the timescales of outburst decay show that the mass that is being accreted must come from inner disk regions regions ≲1 au (Zhu et al. 2007, 2010), well within our coronagraphic spot.

Given the similarity between the bright arm B and other infalling streams and the presence of the companion FU Ori S, it seems more likely that the vast majority of the observed structures are due to a complicated dusty environment rather than misaligned disks.

We have discovered a stellar companion candidate and fine dust structures around MWC 789. In the total intensity image (Figure 2), we see a companion candidate and a disconnected arc to the northwest near the detector edge. This arc is shown more clearly in the polarized Q_ϕ (Figure 3)—it is so extensive that it falls off the detector edge. The bright central region is surrounded by small bright tails spreading outwards, and a trail of dust appears to link the companion candidate with the central region. There is also an arm of material to the northeast of the spot that falls off the detector edge.

Because we cannot see beyond the eastern side of the detector, it is hard to say how the outer arc connects to the inner eastern arm. The two arcs could meet in the east to form one large curved arm, or alternatively, there could be a transition disk-like structure as the outer arc shows an illuminated inner wall of an outer disk region with a dust clearing. However, the latter scenario would require an unusually large disk with radius ≫700 au. Both scenarios could be owing to the companion candidate, either in generating the large arm or in causing the clearing. The companion candidate could also be the cause of the smaller bright trails near the occulting spot, particularly the trail that seems to connect the companion candidate to the central star. Future analysis could include SED fitting with a large clearing to see whether this option is feasible.

The companion candidate of MWC 789 appears at a distance of 0426 ± 0018 (297 ± 34 au) and a position angle of 2163 ± 03. This companion candidate has not been reported in the literature previously. Similarly to FU Ori, the companion candidate is easily seen in the radial profile in Figure 5 as an excess at 04 in total intensity and in U_ϕ. This location also marks an increase in the sizes of the error bars for Q_ϕ, which indicates that some surface brightness caused by the companion candidate appears to a lesser degree in Q_ϕ.

The H-band polarization fraction of 0.60% ± 0.09% (Table 7) is on the same order as the I-band polarization fraction of 0.95 ± 0.48 and is consistent with the trend of falling fraction with increasing wavelength in visible light (Jain & Bhatt 1995).

It seems unlikely that misaligned disks could be responsible for the complex strands near the companion candidate and coronagraphic spot. Therefore, the circumstellar environment of MWC 789 likely has complex small-scale structures.

There are no previous high-resolution images of MWC 789 in the data archives of other extreme AO instruments or ALMA. To constrain the dust distribution in the outer regions, MWC 789 should be observed with a larger field of view to discover how the northern arc is linked to the central star.

The scattered light in HD 45677 reveals a symmetrical bipolar structure that resembles an edge-on view of an hourglass. Aside from this large bipolar structure, we do not see the protoplanetary disk itself as it appears at far too small a size scale and is fully obscured by the coronagraph. Our analysis here will focus on the broad shape and symmetry in the scattered light, and the interpretation of finer details will be discussed in a future paper. This object lacks any sharp-edged shadows that could result from misaligned disks.

As noted above, the evolutionary state of HD 45677 is still rather uncertain. We have performed our own analysis using the UV photometry of Thompson et al. (1978) and the BVRI data compiled in Anderson & Francis (2012). We fitted Kurucz model atmospheres (with log g = 4) to the photometry, using a grid search to determine the best-fit stellar radius, reddening, and effective temperature while fixing the system's distance at 620 pc (as in Table 3). Our best fit has T_eff = 23,000 K, R = 3.3 R_⊙, and A_V = 0.6. The fit is relatively poor, with a reduced chi-squared of 3, perhaps unsurprising, given that the object displays photometric variability of over a magnitude on timescales of years and that the observations adopted were not contemporaneous. The best fit has a considerably higher T_eff than that determined by Vioque et al. (2018), T_eff = ${{\rm{16,500}}}_{-80}^{+3000}$ K, but is more in line with its early-B spectral classification. This luminosity and temperature are consistent with an 8 M_⊙ mass star at an age of 10⁶ yr according to the pre-main-sequence evolutionary models of Marigo et al. (2017), and rather more massive and older than the Vioque et al. (2018) estimate.

This analysis and the recent Vioque et al. (2018) fits suggest that HD 45677 is a pre-main-sequence object. In this case, the observed bipolar emission structure would be a protostellar envelope. This structure is common in objects in star-forming regions (Tamura et al. 1991). However, the emission nebula does not rule out HD 45677 being a more evolved object, e.g., a planetary nebula. Some bipolar planetary nebulae bear strong resemblance to the general shape of HD 45677 (e.g., the "Squid Nebula," Acker et al. 2012; "Hubble 12," Kwok & Hsia 2007). There is still the evidence for HD 45677 having a bipolar outflow (see Section 2), in which case the observed emission nebula structure could mark the walls of an outflow cavity. The evolutionary status of this object is still ambiguous and makes HD 45677 a prime candidate for follow-up modeling and follow-up observations.

Figure 7 shows the combined J+H image with the position angle of the hourglass structure, which we find by assuming that the underlying structure is axisymmetric. The right-hand side of the image is flipped horizontally and laid over the left-hand side, then the flipped half is subtracted from the original values. A symmetrical image gives a weaker residual structure remaining in the image after subtraction. The strength of the residual structure is measured with the sum of the absolute value of the residuals, with the sum using only the central half of the image to reduce the effect of values near the detector edge. The original image is then rotated and the residual sum calculated again in the same way. We use the built-in scipy.optimize.minimize function with the "Nelder–Mead" method to minimize the residual sum and so find the rotation angle that produces the most symmetry. Using this method with the combined J+H Q_ϕ image, we determine the symmetry axis position angle to 158°. For completeness, the values determined from the J-band and H-band images separately are 157° and 158°, respectively.

Zoom In Zoom Out Reset image size

Assuming stronger forward scattering, the brighter northwestern half of the emission nebula shape must be inclined toward the observer. Alternatively, the difference in brightness could be owing to an inner disk structure obstructing light from the farther lobe (Tamura et al. 1991). The second scenario would match the inner ring model from Monnier et al. (2006), which is highlighted in Figure 7 and has a similar minor axis position angle to the angle of symmetry in the hourglass structure—160° and 158°, respectively. Both position angles are comparable with the polarization and blowout angles of 164° and 175°, respectively (Patel et al. 2006), which establishes a connection between the inner ring and the larger-scale structure.

Figure 8 shows the ratio of the J and H Q_ϕ images. The emission is somewhat redder along the major axis of the unseen disk and bluer in the region of the bipolar structure. These features are clearer nearest the center of the image, and the underlying structure of the bipolar nebula is lost to the background scatter at larger radii. Near the center, the difference in surface brightness is distinctive and follows the bipolar shape enough to indicate that there is some systematic difference between the scattering in the J and H bands. The reddening could be caused by extinction or different grain sizes in the different regions. Separately, both wavelengths reveal the same finer structures such as distinct wisps or streams leading out of the emission nebula shape.

Zoom In Zoom Out Reset image size

Our results show the first resolved high-contrast imaging of Hen 3-365. Hen 3-365 shows another remarkably asymmetric circumstellar environment. Immediately west of the occulting spot, we detect a rounded bright region, and there are additional arcs of material that are disconnected from this brighter region. The brightest, in the northwest, lies in a straight line ∼2000 au long at an angle almost tangential to the connecting line to the central star. This region also is distinguishable in total intensity and the U_ϕ image, which suggests there could be multiple scattering effects. The structure appears least significantly in the radial profile of total intensity (see Figure 5) but is clearly visible in both Q_ϕ and U_ϕ profiles. There are no obvious structures that correspond to the EW and N outflows as in Baines et al. (2006).

In addition, there is a dark region immediately to the east of the coronagraphic spot. This region is negative in Q_ϕ and is also distinguishable in U_ϕ. It appears to be physical in origin. The effect could have been exacerbated during the reduction process due to butterfly patterns from a subpar stellar polarization removal (see Appendix A), but this dark region does appear to some extent regardless of reduction method. The physical size of this dark region is so large that it is unlikely to be caused by a misaligned disk structure that would be much smaller in scale. The dark region could be an imprint of the companion candidates if they have cleared this region, because the three features appear in the same range of position angle.

Hen 3-365 shows two faint point sources that we designate B and C. Their fluxes, separations, and position angles are given in Table 6. The source Hen 3-365-B is considerably fainter than Hen 3-365-C and the companion candidate star around MWC 789, at only ∼10%–20% of the latter objects' fluxes. The two point sources appear with negative U_ϕ and are located away from the central stars, with distances on the order 1000 au. Because the point sources are so small and faint, it is unlikely that they have distorted the Q_ϕ, U_ϕ images to create false evidence of multiple scattering. In particular, they are far away in physical and projected distance from the northwestern arc that shows structure in U_ϕ and are unlikely to be responsible for this structure. Any association of the point sources and the central star should be investigated with follow-up observations to see if the point sources are comoving.

For the HD 139614 J-band observations, the 2019 data were used in all cases except for examining arm structures in Figure 9, where the 2017 data without the coronagraphic spot are presented using the 2017 data with the coronagraph for flux calibration.

Zoom In Zoom Out Reset image size

HD 139614 shows faint emission in Q_ϕ out to ∼150 au in all directions around the object (particularly on the western side) and a brighter region within 50 au of the center, consistent with predictions that the outer disk extends to 75–100 au (Patel et al. 2006; Labadie et al. 2014). Depending on the positioning of the coronagraphic spot, the structure at 50 au could correspond either to a ring or a tight spiral arm. Our three independent observations show that this structure lies consistently to the north of the star, which strongly supports the tight spiral arm interpretation. The spiral arm originates from the north and continues counterclockwise around the image. Its inner edge is just visible in data taken without coronagraphy shown in Figure 9. It is impossible to flux-calibrate an image without coronagraphy using the DRP methods described in Section 3.2, so we use the flux scale from contemporary data taken with the coronagraph. Both stars in the HD 139614 binary system appear saturated in the image without the coronagraph and cause a distorted band near the center of the image. It is not possible to locate the two stars behind this distortion.

Multiple faint arms to the northwest are easier to see in Figure 9 where we present a deprojected version of the Q_ϕ image that is translated into polar coordinates. To make this image, we created a list of coordinates that are regularly spaced in (r, ϕ) space—every 1.0 pixel or 00141 in radius from the image center to detector edge, and every 1° in position angle ϕ—and sample the values in these coordinates from the Cartesian image. In addition, the images in this figure are distance-normalized, i.e., the surface brightness values are scaled by the distance from the image center r in arcseconds to more clearly show the fainter outer arms.

The deprojected image shows that the arm structures generally increase in distance from the image center. This is most consistent with the arms being spiral structures. The outermost arm begins to decrease in radius again at a position angle of 300°, which makes it more likely that it is instead part of an elliptical ring or other irregular feature. The apparent gap between the two arm structures could be due to a shadow cast by the inner arm onto the outer arm. There is a break in the bright innermost arm (radius 013, PA 300° in Figure 9) that is likely due to a shadow cast by a misaligned disk. This break appears in all of our HD 139614 data sets and so is not due to any observing effects such as saturation of the central star or the affected region being covered by the coronagraph. Notably, there is little structure in U_ϕ (Figure 4) for this object compared with the other objects presented here, which suggests that multiple scattering is negligible in the resolved arm structures.

The two highlighted arm structures are too faint to be seen outside a certain range of position angle ∼210°–0°, i.e., the polarized emission is brighter on the western side. In the simple first approximation, this brightness asymmetry is consistent with a strong forward-scattering regime and the western side inclined toward us. This matches an inner disk model with an inclination 20° and position angle 112° as fitted to MIDI visibilities (Matter et al. 2014). The VLT/CRIRES prediction of a close-in planet could result in a skewed, tilted inner disk (Carmona et al. 2017), leading to the observed asymmetry in scattered light. A more robust estimate of the inclination of the object and its scattering phase function will be left to future analysis.

The inner disk and gap predicted by the literature lie under the occulting spot in our images. The Subaru-HiCIAO scattered light nondetection (Fukagawa et al. 2010) used a coronagraph with 10 spot diameter coronagraph, which is much larger than the order 01 diameter spot used in our GPI observations. As such, it is likely that even the fainter regions we detect are covered by the coronagraph in the Subaru-HiCIAO observation.