Gemini-LIGHTS: Herbig Ae/Be and Massive T Tauri Protoplanetary Disks Imaged with Gemini Planet Imager

Observations of the targets were taken at the Gemini South Observatory using the GPI (Macintosh et al. 2014; Poyneer et al. 2014; Larkin et al. 2014) in the J- and H-band filters. Observations were in the polarimetry mode using coronagraphic spots of sizes 184.7 mas and 246.7 mas for the J and H bands, respectively, and a pixel scale of 14.14 mas. Observations of the targets HD 100453 and HD 142527 were the only observations taken without a coronagraphic mask. Each image of a target measures the orthogonal polarization state of the 20 × 20 field of view (FOV). Between each image of a target, the half-wave plate is rotated 225. This creates a polarization set of four images at four different wave plate positions (0°, 225, 45°, 675). A typical observational epoch of a target observed eight sets of four polarization images producing 32 images per epoch. Exposure of the image for each target was adjusted for the targets brightness such that the PSF of the star does not saturate the images. Observations of the targets occurred between 2014 April and 2019 May with the majority of observations occurring between 2017 and 2019, and a list of the observations can be found in Table 4. Calibration files such as lamps and dark images were taken by the Gemini observatory staff every 2–4 weeks and accessed through the Gemini archive.

The images are flux calibrated using the four satellite spots in each of the PODC images. To increase the signal to noise of satellite spots, the PODC files are mean combined. For H-band images, we used the first-order satellite spots. For J-band images, we used the second-order spots to ensure the spots were further away from the central PSF core. We note that the conversion factor between the satellite spots was updated since version 1.4 of DRP that was utilized by Laws et al. (2020) and Monnier et al. (2019). We used the known flux of the star from the 2MASS catalog to calculate the flux conversion factor and apply it to the Q_ϕ and U_ϕ images, as shown in Figure 3. In our sample, we measure an average scale factor of 3.05 ± 0.57 for the J band and 3.09 ± 0.51 for the H band mJy ''⁻²/(ADU/s). We note that time (s) in the above scale factor is the total exposure time (ITIME×coadds).

Central stars flux can be variable with time; thus without contemporaneous photometry we have taken effort to investigate and correct target epochs with discrepant flux calibration values. These different scale values could also be due to seeing conditions and AO performance. We investigated 14 observations that had a scale value 1σ (0.57 for the J band, 0.51 for the H band) away from the average scale value of 3.05 in the J band (3.09 in the H band). Four of these epochs had similar values to observations taken on the same night suggesting that the flux scale deviation is a correction for poor seeing or AO performance. Another four of these epochs had similar scale values to previous epochs suggesting that the divergent high or low scale value might be due to an over or underestimate of the flux of the target. The other six targets: FU Ori, HD 104237, HD 142666, HD 45677, HD 144432, and PDS 66 are likely to be due to variability; thus we use the average flux scale of other targets in that epoch. Finally, HD 100453 and HD 142527 were observed without coronographic spots; thus we choose to use flux calibration scale factors from Long et al. (2017) 0.836 mJy ''⁻²/(ADU/s) for both observations.

Due to the design of the GPI instrument, flat-field corrections of the pixel-to-pixel variations in the raw frame are not currently possible. The current correction uses lamp flat observations to correct for low-frequency variations across the FOV. This issue was first investigated by Millar-Blanchaer et al. (2016). We observed the twilight sky to obtain sky flats in order to independently estimate the effects of the flat-field correction onto our data. We reduced the data using the same parameters as described above in Section 3. We plot an example of an individual Q and U frames as a % flux deviation from the average of the Q and U frames, as shown in Figure 4. We show that the low-frequency flat-field fails to correct for all large-scale flat-fielding variations, especially toward the edge of the detector. However, there are still sizable flux variations between 2%–4% at the center of the image. Thus we conclude that any azimuthal flux variations observed in the disk that are on the scale of 2%–4% may not be astrophysical but instrumental due to poor flat-fielding of the image.

Zoom In Zoom Out Reset image size

We estimate the uncertainty of our images by bootstrapping images from the fully reduced Stokes Q_ϕ and U_ϕ images. We performed this bootstrapping 100 times and these sets of bootstrapped images are used for all analysis and error propagation in Section 4 and utilized in calculating the uncertainties of the polarization statistics for each observed epoch as shown in Table 5.

Table 5. Polarization Flux and Point-source Detection Upper Limits

Target	Band	Epoch	Qϕ /Fstar	5σ Point-source Detection Upper limit
		(YYYYMMDD)	(per 1000)	Contrast at 02	Mass Sensitivity ( M⊙)
AK Sco	H	20180811	1.2 ± 0.2	1.6E-04	0.011
CU Cha	J	20170406	2.7 ± 0.9	3.9E-05	0.006
CU Cha	H	20180413	3.8 ± 1.0	2.5E-04	0.02
FU Ori	J	20180103	3.7 ± 0.6	1.2E-04	0.03
GW Ori	J	20180104	9.4 ± 0.9	1.6E-04	0.011
GW Ori	H	20180104	13.3 ± 1.1	5.9E-05	0.007
HD 34700 A	J	20180103	15.7 ± 1.8	1.3E-04	0.008
HD 34700 A	H	20180103	22.2 ± 1.8	5.3E-05	0.005
HD 36917	J	20171231	<0.02	3.6E-05	0.007
HD 36917	H	20190126	<0.03	8.6E-05	0.01
HD 37806	J	20171231	<0.04	3.8E-05	0.008
HD 38087	J	20180103	<0.12	6.1E-05	0.008
HD 45677	J	20171231	8.4 ± 1.1	5.5E-05	0.01
HD 45677	H	20180101	6.9 ± 1.0	3.0E-05	0.01
HD 50138	J	20180104	0.9 ± 0.2	6.0E-05	0.011
HD 85567	J	20180103	0.10 ± 0.08	1.0E-04	0.03
HD 85567	J	20180520	0.12 ± 0.07	4.7E-05	0.02
HD 85567	H	20190127	0.13 ± 0.11	2.5E-04	0.06
HD 95881	J	20180104	<0.09	9.6E-05	0.03
HD 98800	H	20190127	0.19 ± 0.14	⋯	⋯
HD 98922	J	20180321	0.21 ± 0.16	6.9E-05	0.03
HD 100453	J	20150410	2.9 ± 0.2	⋯	⋯
HD 100546	J	20170220	10.2 ± 0.6	8.7E-05	0.008
HD 101412	H	20180319	<0.04	1.7E-04	0.013
HD 104237	J	20170407	0.04 ± 0.03	9.2E-05	0.013
HD 104237	J	20180317	0.03 ± 0.02	1.6E-04	0.014
HD 104237	J	20180520	0.06 ± 0.03	2.2E-04	0.02
HD 104237	J	20190217	<0.06	1.4E-04	0.014
HD 139614	J	20170406	2.5 ± 0.7	8.9E-05	0.01
HD 139614	J	20190513	1.5 ± 0.6	6.7E-05	0.009
HD 139614	H	20180608	2.5 ± 0.7	3.1E-03	0.06
HD 141569	J	20180609	0.29 ± 0.28	1.5E-04	0.013
HD 141569	J	20190223	<0.3	4.8E-05	0.007
HD 142527	H	20140425	9.26 ± 1.17	⋯	⋯
HD 142666	J	20170703	0.19 ± 0.17	⋯	⋯
HD 142666	J	20180609	0.09 ± 0.08	⋯	⋯
HD 144432	H	20150709	<0.03	⋯	⋯
HD 145718	J	20180607	0.4 ± 0.3	4.4E-05	0.007
HD 145718	H	20180608	0.4 ± 0.2	1.1E-04	0.011
HD 158643	J	20180609	<0.01	2.4E-05	0.005
HD 158643	J	20190223	<0.04	⋯	⋯
HD 169142	J	20140425	4.5 ± 0.8	1.0E-04	0.008
HD 176386	J	20180817	<0.01	2.8E-05	0.005
HD 176386	H	20180607	<0.05	6.7E-05	0.006
HD 176386	H	20180816	<0.02	6.8E-05	0.006
Hen 3-1330	J	20190512	<0.03	4.8E-05	0.04
Hen 3-225	J	20190127	<0.04	7.3E-05	0.013
Hen 3-225	H	20190127	<0.06	6.7E-05	0.01
Hen 3-365	J	20170406	5.5 ± 0.6	3.3E-05	0.05
HR 5999	J	20170406	<0.02	3.3E-04	0.03
HR 5999	J	20170702	0.03 ± 0.02	7.8E-05	0.01
HT Lup	H	20190514	0.4 ± 0.2	5.6E-04	0.009
MWC 147	J	20171231	<0.07	8.4E-05	0.011
MWC 147	H	20190127	<0.1	3.3E-04	0.04
MWC 166	J	20170406	<0.06	6.0E-03	1.4
MWC 275	J	20140424	0.3 ± 0.2	7.3E-05	0.01
MWC 297	H	20180608	0.3 ± 0.2	8.5E-05	0.04
MWC 297	H	20180817	0.3 ± 0.2	1.0E-04	0.04
MWC 614	J	20180816	2.0 ± 0.3	1.5E-04	0.014
MWC 614	H	20180816	2.4 ± 0.2	1.7E-04	0.013
MWC 789	H	20181120	5.9 ± 1.5	6.3E-04	0.06
MWC 863	J	20140424	<0.3	1.1E-03	0.03
PDS 66	J	20160306	2.2 ± 1.1	5.9E-05	0.003
Ty CrA	H	20180608	<0.6	8.7E-04	0.3
Ty CrA	H	20180817	<0.44	1.5E-04	0.06
V1295 Aql	J	20180816	<0.02	2.5E-04	0.05
V1295 Aql	H	20180608	<0.02	1.0E-04	0.03
V1295 Aql	H	20180816	<0.04	2.9E-04	0.05
V921 Sco	H	20190513	0.06 ± 0.05	1.6E-04	0.13
WRAY 15-535	J	20180319	0.11 ± 0.07	1.0E-04	⋯
WRAY 15-535	J	20190222	<0.04	1.5E-05	⋯

Note. Mass estimate based on measured companion flux and the system age (see Table 2 and use models from Baraffe et al. (2015) for masses >0.01 M_⊙ and Phillips et al. (2020) for masses <0.01 M_⊙. The Q_ϕ /Fstar values utilized in this work are the weighted average of the epochs listed above.

Download table as:  ASCIITypeset images: 1 2

We identified point sources in the FOV by looking for bright companions in the intensity image and identifying dimmer point sources using an angular differential imaging (ADI) reduction technique. For the ADI reduction, we used the Karhunen–Loéve image projection (KLIP) algorithm for epochs that had FOV rotation that was sufficiently large to take advantage of an ADI reduction. Using the centered PODC files from the standard reduction discussed above, we utilized the pyKLIP ¹⁵ package output (Wang et al. 2015). In order to avoid self-subtraction issues, we used parameters of subsection = 1 and annuli = 1, which minimize the number of individual segments in the FOV used to calculate the PSF. We found that 16 systems have point sources identified by the pyKLIP reduction. Upon close examination, all of these point sources are visible in the intensity images. A further 6 targets have point sources that were identified in the total intensity image but did not have sufficient FOV rotation for the pyKLIP analysis.

We measured the flux and position of each point source by fitting a Gaussian to each point source in the intensity frame. The location of these point sources can be found in Q_ϕ and U_ϕ images in the online figure sets (e.g., Figure 2) and are marked by dotted black circles. The point-source location, flux, and estimated mass are shown in Table 6. For the companion mass estimates we used models from Baraffe et al. (2015) for masses >0.01 M_⊙ and the chemical equilibrium models from Phillips et al. (2020) for masses <0.01 M_⊙.

Table 6. Point Sources Detected in the Gemini-LIGHTS Sample

Target	Band	Epoch	Comp. #	Separation	PA	Flux	Saturated	Mass Est.
		(YYYYMMDD)		('')	(° )	(mJy)		( M⊙)
FU Ori	J	20180103	1	0.431 ± 0.003	163.3 ± 0.3	27 ± 4	TRUE	0.6 ${}_{-0.1}^{+0.1}$
HD 100453	J	20150411	1	0.909 ± 0.002	131.63 ± 0.09	15.4 ± 1.3	FALSE	0.08 ${}_{-0.04}^{+0.01}$
HD 101412	H	20180319	1	0.460 ± 0.002	149.9 ± 0.2	1.46 ± 0.11	FALSE	0.053 ${}_{-0.003}^{+0.03}$
HD 101412	H	20180319	2	0.153 ± 0.003	184.9 ± 1.0	5.0 ± 0.4	FALSE	0.19 ${}_{-0.01}^{+0.06}$
HD 104237	J	20170407	1	1.226 ± 0.004	255.43 ± 0.14	29 ± 9	FALSE	0.11 ${}_{-0.04}^{+0.04}$
HD 104237	J	20180317	1	1.214 ± 0.004	255.034 ± 0.13	19 ± 6	FALSE	0.065 ${}_{-0.02}^{+0.02}$
HD 104237	J	20180520	1	1.204 ± 0.003	255.1 ± 0.11	32 ± 6	FALSE	0.12 ${}_{-0.02}^{+0.02}$
HD 104237	J	20190217	1	1.198 ± 0.004	255.28 ± 0.11	34 ± 5	FALSE	0.13 ${}_{-0.02}^{+0.02}$
HD 144432	H	20150708	1	1.256 ± 0.004	6.07 ± 0.16	51 ± 7	TRUE	0.3 ${}_{-0.05}^{+0.03}$
HD 158643	J	20180609	1	0.118 ± 0.013	294 ± 5	30 ± 4	FALSE	0.07 ${}_{-0.02}^{+0.01}$
HD 38087	J	20180103	1	0.303 ± 0.003	183.7 ± 0.4	47 ± 8	TRUE	0.8 ${}_{-0.3}^{+0.5}$
HD 50138	J	20180104	1	0.729 ± 0.003	106.6 ± 0.2	10.1 ± 1.5	FALSE	*0.12
HD 98800	H	20190127	1	0.374 ± 0.010	9.8 ± 0.6	633 ± 9	TRUE	0.4 ${}_{-0.03}^{+0.03}$
HR 5999	J	20170406	1	1.251 ± 0.004	112.55 ± 0.11	141 ± 16	TRUE	0.6 ${}_{-0.2}^{+0.1}$
HR 5999	J	20170702	1	1.260 ± 0.005	112.4 ± 0.2	128 ± 19	TRUE	0.6 ${}_{-0.2}^{+0.1}$
HT Lup	H	20190514	1	0.161 ± 0.003	246.6 ± 0.7	70 ± 7	TRUE	0.13 ${}_{-0.01}^{+0.01}$
Hen 3-365	J	20170406	1	0.818 ± 0.003	101.08 ± 0.14	1.3 ± 0.2	FALSE	*0.17
Hen 3-365	J	20170406	2	0.446 ± 0.008	133.0 ± 0.8	0.6 ± 0.1	FALSE	*0.13
Hen 3-225	J	20190127	1	1.564 ± 0.019	303.4 ± 0.3	1.46 ± 0.02	FALSE	*0.12
Hen 3-1330	J	20190512	1	0.268 ± 0.017	49 ± 2	0.8 ± 0.3	FALSE	*0.13
MWC 147	J	20171231	1	0.137 ± 0.004	56 ± 1	30 ± 6	TRUE	0.6 ${}_{-0.1}^{+0.4}$
MWC 147	J	20190127	1	0.156 ± 0.003	54.1 ± 0.8	33 ± 5	TRUE	0.5${}_{-0.1}^{+0.4}$
MWC 166	J	20170406	1	0.587 ± 0.004	298.4 ± 0.3	282 ± 9	TRUE	* > 1.4
MWC 297	H	20180608	1	0.573 ± 0.003	86.5 ± 0.2	2.06 ± 0.09	FALSE	*0.04
MWC 297	H	20180817	1	0.576 ± 0.004	86.6 ± 0.2	1.74 ± 0.14	FALSE	*0.04
MWC 789	H	20181120	1	0.371 ± 0.002	216.8 ± 0.3	10.7 ± 0.8	FALSE	0.8 ${}_{-0.2}^{+0.1}$
MWC 863	J	20140421	1	0.975 ± 0.004	226.97 ± 0.14	154 ± 19	TRUE	0.8 ${}_{-0.1}^{+0.1}$
Ty CrA	H	20180608	1	0.129 ± 0.004	255.2 ± 1.3	26 ± 4	FALSE	>1.4
Ty CrA	H	20180817	1	0.112 ± 0.003	257.3 ± 1.0	41 ± 4	FALSE	>1.4
V921 Sco	H	20190513	1	0.475 ± 0.013	319.0 ± 1.1	0.09 ± 0.03	FALSE	0.03 ${}_{-0.01}^{+0.01}$
V921 Sco	H	20190513	2	1.122 ± 0.003	323.12 ± 0.09	0.96 ± 0.08	FALSE	0.15 ${}_{-0.01}^{+0.01}$
V1295 Aql	H	20180608	1	1.113 ± 0.003	40.52 ± 0.11	0.026 ± 0.002	FALSE	*0.010
V1295 Aql	H	20180816	1	1.135 ± 0.016	40.1 ± 0.5	0.055 ± 0.009	FALSE	*0.013
V1295 Aql	J	20180817	1	1.119 ± 0.005	40.6 ± 0.2	0.015 ± 0.002	FALSE	*0.010

Note. Point sources with their PSF cores saturated are noted in the table. Mass estimate based on measured companion flux and the system age (see Table 2). Mass estimates use models from Baraffe et al. (2015) for masses >0.01 M_⊙ and Phillips et al. (2020) for masses <0.01 M_⊙. A * denotes any target with ages <0.5 Myr that are younger than the youngest model in Baraffe et al. (2015) or Phillips et al. (2020).

Download table as:  ASCIITypeset image

Using the pyKLIP routine, we also placed limits on the point-source companions we are sensitive to detecting in our sample. We calculated the 5σ flux limit of point sources at angular separations of 02 and 05 for all targets with sufficient field rotation. We define sufficient rotation where a point source would move >1 pixel due to field rotation, corresponding to a total field rotation of 4° at 02 and 1.6° at 05. We note that 6 systems, HD 98800 B, HD 100453, HD 142527, HD 142666, HD 158643, and WRAY 15-535, did not have sufficient field rotation at 02 to determine the contrast at a separation of 02. Based on the contrast and the systems age, we estimated the upper limit mass that would be detectable at 02 and 05. For angular separations of 02, in 9 systems we can detect down to Jupiter-mass companions, for 26 we can detect down to brown dwarf mass companions, and in 3 systems we can detect some stellar mass companions. For angular separations of 05, in all systems we can detect companions down to 1 M_jup mass companions. The results for each epoch can be found in the appendix in Table 5. For the companion mass estimate limits we used models from Baraffe et al. (2015) for masses >0.01 M_⊙ and the chemical equilibrium models from Phillips et al. (2020) for masses <0.01 M_⊙.

Here we establish which targets have detected scattered light and which targets are nondetections. One advantage in rotating the polarization I, Q, and U into Q_ϕ and U_ϕ is the ease of interpretation. The Q_ϕ flux should be dominated by photons that scattered only once and from one source, while the U_ϕ flux could be dominated by photons that scattered multiple times, or scattering that is not azimuthally symmetric with the assumed central star (i.e., presence of a binary). We compute the amount of Q_ϕ and U_ϕ flux within 04 (30 pixels) divided by the stellar flux for that given band. For sources with no close-in point source (e.g., within 04 of central star), for disks with very low Q_ϕ , the amount of U_ϕ will go to zero before Q_ϕ , as shown in Figure 5. As Q_ϕ and U_ϕ are drawn from the same observables, the noise in Q_ϕ is the same as U_ϕ . Thus we look at the standard deviation of U_ϕ for epochs <0.01% Q_ϕ and U_ϕ and removed binaries by removing epochs >3σ. We then took the standard deviation of the summed U_ϕ flux, which is also the uncertainty of the Q_ϕ flux. Using this metric, we establish that we do not detect polarized light around 18 of the 77 epochs using a 3σ deviation of U_ϕ of 0.002% in the J band and 0.002% in the H band. Targets that are nondetections include: HD 36917, HD 101412, HD 144432, HD 158643, HD 176386, Hen 2-225, Hen 3-1330, MWC 147, MWC 166. Two targets, V1295 Aql (20180608 H band, 20180816 H band) and TY CrA (20180608 H band) where not detected in some epochs but were detected in other epochs (V1295 Aql: 20180816 J band; TY CrA: 20180817 H band). Thus, we do not detect polarization light around 9 of the 44 targets and additionally do not always detect polarization light in 2 of the targets.

Zoom In Zoom Out Reset image size

We are categorizing our disks only by the observed disk structure in the Q_ϕ images. The categories include spiral armed disks, ringed disks, continuous disks, irregular disks, and undetermined disks structure. Our chosen categories are inspired by the disk structure categories defined by Garufi et al. (2018). We have deviated when appropriate to better describe our sample. The results of our classification are tabulate in the appendix in Table 2.

4.2.1. Spiral Armed Disks

We find there are four targets that host one or more spiral arms including HD 100453, HD 139614, HD 34700 A, and HD 142527, as shown in Figure 6. HD 100453 hosts symmetric spiral arms, while HD 34700A and HD 142527 most multiple arms that are not symmetric. HD 139614 has an arm structure on one side of the disk. All four of these systems have had the origin of their spiral arms investigated previously (HD 100453: Wagner et al. 2015; HD 139614: Laws et al. 2020; HD 34700 A: Monnier et al. 2019; A. S. E. Laws 2022, in preparation; HD 142527: Long et al. 2017).

Zoom In Zoom Out Reset image size

4.2.2. Ringed Disks

We define disks that have one or more concentric polarized light dust rings in their Q_ϕ image, as shown in Figure 7. We note that we deviate from the definition of Garufi et al. (2018) as they make a distinction between ringed and rimmed systems; however scattered light imaging lacks the information to robustly make the distinction due to the inner working angle. We find that seven targets show signatures of rings: HD 169142, HD 141569, MWC 275, CU Cha, HD 34700 A, HD 142527, and PDS 66. We note that two of the four spiral armed disks are also ringed disks (HD 142527, HD 34700 A).

Zoom In Zoom Out Reset image size

4.2.3. Continuous Disks

We define continuous disks as disks that are resolved and do not appear to have a gap or hole in their PDI or show a strong nonazimuthal structure, as shown in Figure 8. Some of these disks may have a gap or ringed structure, but it is not shown in the polarized light imagery due to the inclination (HD 145718: Davies et al. 2022) or inner working angle (IWA; MWC 614: Kluska et al. 2018). We note that this does not imply that these disks do not have gaps at all, but those gaps are not visible in our polarized light imaging. We find that 11 targets show signatures of rings: AK Sco, HD 45677, HD 50138, HD 100453, HD 100546, HD 139614, HD 142666, HD 145718, HT Lup, MWC 297, and MWC 614. We note that one of the spiral armed disks are also continuous disks (HD 139614).

Zoom In Zoom Out Reset image size

4.2.4. Irregulars

We define irregular disks as disks that host structures that have strong nonazimuthal features and disks that are very large in size (>300 au). There are four disks in our sample that exhibit these features: MWC 789, FU Ori, GW Ori, and Hen 3-365, which are shown in Figure 9. GW Ori irregularity is thought to be due to disk tearing as investigated by Kraus et al. (2020).

Zoom In Zoom Out Reset image size

4.2.5. Undetermined

There are some targets that have a significant amount of polarized light detected (see Section 4.1); however the polarized light in the image does not extend far enough away to reliably categorize the disk into one of the above categories. Objects that fit this description include: HD 37806, HD38087, HD 85567, HD 95881, HD 98922, HD98800, HD 104237, HR 5999, MWC 863, TY CrA, V921 Sco, V1295 Aql, and WRAY 15-535, as shown in Figure 10. We note this is a similar to the "small disk" category utilized by Garufi et al. (2018). However, our sample is not strictly distance limited; thus some objects have disks that do appear to be small (e.g., HD 104237 disk radius <21 au), while other objects such as V921 Sco appear to be distant and have rather large disks (<300 au). Thus we choose to classify these objects as undetermined.

Zoom In Zoom Out Reset image size

4.2.6. Categorization Summary

Utilizing the polarized disk morphology described above, we now investigate if we see any trends with the disk morphology. We plotted all 44 targets on an HR diagram and an infrared color–color diagram in Figure 11. We note that targets can be in multiple categories. The first trend is that ringed systems only occur in systems with masses <3 M_⊙. This could partially be due to a distance effect where the ringed structure is within the IWA of the observations, though less likely as there are six systems with resolved disks that do not exhibit any ringed structures. Second, irregular systems appear to be younger; however, this is not a definitive trend as only three objects on the HR diagram are classified as irregular.

Zoom In Zoom Out Reset image size

Next, we plot the complete sample on a color–color diagram as shown in Figure 11. We generally find that full disks are unresolved, undetected, or continuous disks. Pre-transitional disks are mostly continuous disks with some irregulars and ring disks. Finally, transitional disks are dominated by ringed disks, as expected.

Four of our targets have been identified as FS CMa targets, which are a subtype of B[e] stars: HD 45677, HD 50138, HD 85567, and HD 98922 (Miroshnichenko et al. 2007; Vioque et al. 2020). These stars are potentially post-main-sequence stars. We note that HD 45677 is also known as FS CMa, the prototype for this classification. We have the first resolved images of the dust around HD 45677, HD 50138, HD 85567, and HD 98922, as shown in Figure 12. All objects have significant Q_ϕ flux that is detected out to sizable distances of ∼250 au (HD 50138), ∼700 au (HD 45677), ∼300 au (HD 85567), and ∼180 au (HD 98922). Additionally, each of the objects also show significant U_ϕ flux, which is suggestive of multiple scatterings due to optical depth effects or multiple illumination sources due to a binary. The U_ϕ flux for both HD 45677 and HD 50138 is likely to be due to optical depth effects as the pattern is within the location of high Q_ϕ flux similarly seen in other optically thick sources such as HD 34700 A, GW Ori, and HD 100546. HD 50138 does have a point source located within 10 of the central star but is not co-located with the concentric Q_ϕ flux around the central star. There is excess polarized flux north of the main disk that could be faint additional structure (e.g., back of the disk, spiral arm), but we are unable to definitively determine its origin. HD 85567 is very distant (1047 pc); thus only the very outer portion of the system is imaged. Finally, HD 98922 exhibits a centro-symmetric Q_ϕ pattern around the inner working angle.

Zoom In Zoom Out Reset image size

We found 24 point sources within the FOV of the PDI images with 21 targets hosting these point sources. Point sources are identified in Q_ϕ and U_ϕ images in the online figure sets (e.g., Figure 2) with a dotted black circle marking the companions locations. The values are also tabulated in Table 6 located in the appendix. We assessed the likelihood that a given point source was a background star from the measured separation and brightness from the target star. We measured the local star density by using the number of stars per half magnitude within 1 deg of the target from the 2MASS catalog. We then calculated the probability that a background star would be located within the projected distance of the target. We found that 22 of the 24 point sources are likely not background stars within 3σ. Two point sources found around V921 Sco are less than a 2σ probability not being background stars. Based on the likelihood that all of the point-source companions are gravitationally bound; for the purposes of this work we will assume that all of the 24 point sources are binaries. However, future follow-up observations, especially V921 Sco, are needed.

We estimated the masses of each of the companions using the age of the system and the companions measured flux and extrapolated the mass from evolutionary models from Baraffe et al. (2015) for masses >0.01 M_⊙ and Phillips et al. (2020) for masses <0.01 M_⊙. We estimated the mass uncertainties using the uncertainty in the systems age and flux. We note that some of our targets have ages less than 0.5 Myr, which is younger than the youngest age by Baraffe et al. (2015). In these cases, we estimated their mass based on the youngest age in the Baraffe et al. (2015) models to give an idea of their mass but note these masses should not be used for future analysis as they are not reliable. Estimated masses are tabulated in Table 6 located in the appendix.

We compare the estimated companion mass and the projected companion separation to the stellar mass, as shown in Figure 13. Of our 24 point sources, one point source around V1295 Aql has a mass consistent with a super Jupiter, four have masses consistent with brown dwarf masses (HD 101412, HD 158643, MWC 297, and V921 Sco), 12 systems have companions with masses consistent with M dwarfs (0.08–0.57 M_⊙), and seven have masses >0.57 M_⊙. Further, we note that more massive stellar objects (>6 M_⊙) have larger projected separations due to these objects being more distant. For these objects with separations >1000 au, future observations are necessary to determine if these objects are comoving and if they are gravitationally bound.

Zoom In Zoom Out Reset image size

Two targets, MWC 147 and TY CrA, have polarized flux that is adjacent to the target, as shown in Figure 14. The flux plotted is flux that is 1σ per pixel above the noise in the image. Both targets are known to be associated with reflection nebulae; thus the flux is likely material that is part of the larger reflection nebulae. We also note that the flux may be an instrumental effect due to known issues with flat-fielding (see Section 3.2). However, this explanation is unlikely as the features appear in the same location over multiple epochs and multiple bands, and no other sources show these features taken on the same night.

Zoom In Zoom Out Reset image size

Six of our systems have stellar masses >8 M_⊙ including Hen 3-365, Hen 3-1330, MWC 166, MWC 297, V921 Sco, and WRAY 15-535. We do not detect polarized light around two of the six systems (Hen 3-1330 and MWC 166), and we classified the morphology of the rest as one irregular (Hen 3-365), one continuous (MWC 297), and two undetermined (V921 Sco, WRAY 15-535). Several of these objects have conflicting classifications between young stellar objects (YSOs) and post -main-sequence stars. Hen 3-365 is discussed by Laws et al. (2020) who noted that it is inconsistent with being a massive supergiant due to the large H I column density, but parallax distances are consistent with being an evolved star (Oudmaijer et al. 1998; Maravelias et al. 2018). Hen 3-1330 has previously been classified as a WR+O binary system (Richardson et al. 2011). V921 Sco has been classified as both a supergiant and as a Herbig Be star (Kreplin et al. 2020). WRAY 15-535 has been classified as a supergiant B[e] star (Domiciano de Souza et al. 2007). Finally, previous work investigating MWC 166 and MWC 297 has shown that they are consistent with being Herbig Ae/Be stars (Manoj et al. 2007; Wichittanakom et al. 2020). We leave the polarized light investigations of these individual targets to future work.

We calculate the total amount of Q_ϕ flux within the image relative to the stellar flux Fstar. To mitigate noise being added into the images, we sum all of the Q_ϕ light between the IWA of the target and where the radial profile of the Q_ϕ is within 3σ of zero flux using error propagation from the bootstrapped images discussed in Section 3.3. The summed Q_ϕ flux is then divided by the flux of the star using the 2MASS J- and H-band flux. The Q_ϕ /Fstar ratio can be found in the appendix in Table 5. We note that this simplified methodology does not account for the r² flux loss. However, this method can easily be applied to our entire sample and is not dependent on knowing the disk geometry (i.e., disk inclination), which is not possible for some systems (i.e., unclassified, irregular; see Section 4.2). As there are multiple epochs of some targets, we will use the weighted average of Q_ϕ /Fstar ratio for all analysis.

We first compare the Q_ϕ /Fstar ratio to the color–color diagram shown in Figure 15, which we utilized for our target selection described in Section 2. The size of the circle corresponds to the Q_ϕ /Fstar ratio on a logarithmic scale. We find that the targets with the largest Q_ϕ /Fstar ratio are commonly found in the middle of the color–color plot between 3< W2–W4 <6 mag and 2 < J–W2 <4 mag. This region coincides with the transitional disk region as outlined in Figure 1. Low and nondetected Q_ϕ /Fstar ratios are scattered throughout the diagram but have a higher concentration at the bottom with W2–W4 <3 mag. This lower-right portion of the diagram coincides with the location of the full disks shown in Figure 1. Interpretation of these results will be discussed in Section 6.

Zoom In Zoom Out Reset image size

We first compare the Q_ϕ /Fstar ratio to the W2–W4 color shown in Figure 16. We find that bluer targets are more likely to have lower Q_ϕ /Fstar ratios than redder targets. By our group definition above, Group II targets are more likely to have lower Q_ϕ /Fstar ratios, while Group I targets are more likely to have higher Q_ϕ /Fstar ratios. While there is a positive trend between the Q_ϕ /Fstar ratio and the W2–W4 color, there is a large spread of values in the trend especially noting the logarithmic plots in Figure 16. We attribute this large spread due to the presence of binaries. When the binaries are removed from the sample, as shown in the upper right in Figure 16, there is not as significant of a spread in values. This suggests that binarity complicates the correlation between polarized flux from the disk versus infrared color. Very bright binaries can make it difficult to remove the SIP, as discussed in Section 3, possibly resulting in a larger spread in Q_ϕ /Fstar ratio values. However, this is unlikely to be the main cause of the spread as the majority of binaries in these systems are not bright enough to leave the quadrupole residual in Q_ϕ and U_ϕ .

Zoom In Zoom Out Reset image size

There are two clear exceptions to this trend, TY CrA and HD 176386. Both of these targets have no detected Q_ϕ /Fstar ratios; yet both are very red (W2–W4 >6 mag). TY CrA and HD 176386 are both located in an extended reflection nebulae. Images from unWISE of TY CrA exhibit extended emission offset from TY CrA (Lang et al. 2016). Also, HD 176386 is most likely blended with CrA-54 in WISE bands, where CrA-54 is a K7 star that is known to host a disk (Cazzoletti et al. 2019). Thus we conclude that the W2–W4 color for TY CrA and HD 176386 are likely contaminated from their surrounding environment (reflection nebulae, close-by stars, dust clumps), not originating from the protoplanetary disk dust around these stars. However, we lack observations of adequate spatial resolution to provide accurate WISE-band fluxes; thus we will keep these targets in our analysis but label them as suspicious WISE fluxes.

We next compare the summed Q_ϕ /Fstar flux to the ages of the systems. Stars will evolve toward the zero-age main sequence (ZAMS) at different rates with 8 M_⊙ stars evolving much quicker than 1.4 M_⊙ stars. Thus we plot the ratio of the current age of the stars divided by the target ZAMS for a given estimated stellar mass, as shown at the bottom left of Figure 16. For reference, the ZAMS is also plotted on an HR diagram for the sample in Figure 11. There are no systems with Q_ϕ /Fstar ratios <0.2 that are on the main sequence (>1.0 Age/ZAMS). We do not see a correlation between the system age/ZAMS and the Q_ϕ /Fstar ratio.

Finally, we compare the stellar mass to the ratio of Q_ϕ /Fstar shown at the bottom right of Figure 16. We find that the less massive stars have the largest Q_ϕ /Fstar. The one key result is looking at systems with stellar masses between 4 < M_star < 8 M_⊙, where only five systems have significantly detected Q_ϕ /Fstar. Four of these five systems are also classified as FS CMa stars. We will discuss this point further in Section 6.6.

We found that, in our sample, protoplanetary disks hosting ringed systems only occur in systems with stellar masses <3 M_⊙ (see Section 4). There are several potential causes that could explain this trend. First, we note that dust ringed systems have largely been theorized to exist due to the formation of exoplanets creating gaps in-between these dusty rings. Recent evidence for this theory has been provided in the case of PDS 70b,c (Keppler et al. 2018). One explanation is that the physical scenarios that make it possible for exoplanets to form visible rings at lower stellar masses (<3 M_⊙) do not exist for higher-stellar-mass systems. One possibility is the increased binarity for stellar masses >3 M_⊙ disrupts the formation of rings and gaps, which was similarly observed for millimeter dust grains as part of the DSHARP sample (Kurtovic et al. 2018). However, in our sample we see the same fraction of binaries around more massive stars (42% for 3 M_⊙ < stellar mass <8 M_⊙) than we do around lower-mass stars (40% for Stellar Mass <3 M_⊙) though we are not complete to inner and outer binaries as noted in Section 5. Second, the temperature of the higher-mass disks may prevent them from forming rings and gaps. Disks around higher-mass stars may be hotter, which can prevent the most volatile ice species, like CO, from freezing out onto dust grains, as found recently in a comparison of carbon depletion in T Tauri versus Herbig AeBe disks (van der Marel et al. 2021; Sturm et al. 2022). If the dust rings are created by the accumulation of millimeter-sized grains at the edges of gaps, which may in turn be carved by planets, reduced amounts of CO or CO₂ ice in the warmer disks could inhibit planetesimal formation or grain growth. This could inhibit outer disk ring formation for the higher-mass disks. Garufi et al. (2018) found that polarized light rings only appeared around older stars (>5 Myr); thus what we may be observing is that stellar systems with masses >3 M_⊙ do not have the time to develop polarized light rings before the dust has been removed from the system. Also, theoretical predictions suggest that the mass of a planet needed to open a gap in a disk is proportional to the mass of the star (see Equation (1) in Matsumura & Pudritz 2003). Thus the lack of gaps could be representative of more massive planets being needed to open up gaps around more massive stars.

Finally, the lack of detected ringed systems for stellar masses <3 M_⊙ could be an observational bias. In our sample, more massive targets are more likely to be more distant. Thus these more distant objects may host ringed structures but are located inside the IWA of our observations. Of the 20 systems with stellar masses >3 M_⊙ (see Figure 11), all rings in the two systems with stellar masses <3 M_⊙ could be detected, and an additional 11 systems are close enough that at least one ring could be detected. Our sample has a dust ring occurrence rate of 30% for systems with <3 M_⊙ stars; thus we would expect to detect at least one system with stellar mass >3 M_⊙ hosting rings. However, the detectability of rings around more massive systems is complicated because the central stars are also more luminous, increasing the outer disk brightness. Therefore ringed structures around systems with masses <3 M_⊙ that are too dim to be detected by our imaging surveys, such as those found around MWC 275 (330 au; Rich et al. 2019) and CU Cha (341 au; Ginski et al. 2016), would be brighter and detectable around more luminous systems. If we assume the distance of the furthest known ring (341 au), we could detect such a ring around 15 of 20 systems with stellar masses >3 M_⊙. Our survey is the first survey to search for trends in protoplanetary disks with stellar masses >3 M_⊙; thus future work needs to verify these findings and search for massive systems that host ringed and gaped structures.

We find a correlation between the polarized flux and the infrared colors, as shown for the entire sample in Figure 15. This trend replicates previous studies by Garufi et al. (2017, 2020), which found a similar correlation. For the entire sample, we find that objects associated with full disks are more likely to have a lower Q_ϕ /Fstar ratio, as shown in Figure 15. The lower Q_ϕ /Fstar ratio is expected as full disks are thought to self-shadow, causing the outer portions of the disk imaged with GPI to be dimmer in polarized light as compared to transitional disks where self-shadowing is not occurring.

A similar trend can be seen for our limited Herbig Ae/Be sample discussed in Section 5. The strong infrared flux of Group I objects has often been interpreted as due to strong disk flaring but could also be due to disk cavity, in either case leading to easily detected scattered light disk flux (Maaskant et al. 2013; Garufi et al. 2017, 2018). The bluer Group II objects can be explained either by strong self-shadowing by the inner disk or a flatter geometry due to dust growth/settling (e.g., Muro-Arena et al. 2018). This would broadly correspond to larger scattered light flux from Group I objects and less from Group II objects. Our observations shown in Figure 16 broadly replicate this trend though there is a large spread in values. This is primarily due to the presence of binaries. Further, we observe very few binaries that are in redder objects (Group I for Herbig's) and most binaries in our sample are associated with bluer objects, as shown in Figure 16. This strongly suggests that a system hosting a binary plays a significant role in the Group I versus Group II classification. One potential cause is that the presence of binaries truncate the outer disks causing systems with binaries to have less polarized fluxes or not be resolved. This would match with previous studies suggesting that Group I and Group II systems are two distinct evolutionary pathways (Maaskant et al. 2013; Garufi et al. 2018).

We observe a lack of systems with ages in the main sequence (>1.0 Age/ZAMS) and low Q_ϕ /Fstar ratios, as discussed in Section 5. This could be an indicator that systems with low Q_ϕ /Fstar ratios and older ages (>1.0 Age/ZAMS) do not have sufficient infrared excess and are thus not in our sample. However, the brightest systems with large Q_ϕ /Fstar ratios may have long-lived disks. This is similar to the conclusion found by Garufi et al. (2018), where the most massive disks are long-lived and easily observed in polarized light to older ages. Next, we do not find the correlation between Q_ϕ /Fstar and age, which had previously been observed by Garufi et al. (2018). However, our sample contains a larger range in stellar mass of 0.3–20 M_⊙, whereas the Garufi et al. (2018) study only contained targets with stellar masses <3 M_⊙. Thus, the correlation between polarized light and age may not be present for higher-mass stars. Finally, age results are difficult to interpret for more massive stars due to the lack of intermediate-mass T Tauri stars in our sample (see Section 2). Future work is needed to include intermediate-mass T Tauri stars in polarized light imaging surveys to assess age trends in Herbig Ae/Be systems.

We find that only 5 of the 12 systems with stellar masses between 4 M_⊙ < M_star < 8 M_⊙ have significantly detected Q_ϕ /Fstar ratios. There are four possible explanations for this trend. First, our more massive stars tend to be further away; thus, we may not be resolving as much of the polarization flux compared to closer systems. Second, these more massive stars tend to have more outer binaries (5 of 12) than less massive systems (<4 M_⊙). These binaries could be truncating or stripping the disks from the central star. The lack of polarized light may be caused by stellar evolution, where the quickly evolving bright star photoevaporates the disks so they are no longer visible. Finally, the systems may host large disks that our imaging would resolve. However, no polarized light is detected because the inner disk shadows the outer disk. This mechanism has previously been invoked to explain nongaped protoplanetary disks having little polarized light detected (Garufi et al. 2017; Muro-Arena et al. 2018).

We compare our point-source findings to those found in the literature. We limit our discussion to the substellar mass companions and notable stellar companions. Our survey has imaged five substellar point sources (V921 Sco, HD 158643, MWC 297, HD 101412, and V1295 Aql). We found two companions around V921 Sco at 05 (0.03 M_⊙) and 11 (0.15 M_⊙). The first companion with a mass consistent with a brown dwarf at 05 was previously discovered by Ubeira-Gabellini et al. (2019) with a similar mass estimate (0.06 M_⊙) but has not previously been confirmed until this work. They did not observe the second companion at 11; however this is expected as the second companion is located outside of the FOV of Ubeira-Gabellini et al. (2019) images. The brown dwarf object (0.07 M_⊙) HD 158643 has not been directly imaged previously. However, Kervella et al. (2019) measured a variation in the proper motion of the system consistent with an object with a mass of 0.05 M_⊙ normalized to a separation of 1 au. These values are potentially consistent, but further analysis is necessary. Ubeira-Gabellini et al. (2020) previously confirmed that MWC 297 B was comoving and measured the companion mass to be ${0.25}_{-0.15}^{+0.25}$ M_⊙. This is much larger than our estimated mass of 0.03 M_⊙. However, they were also able to measure a local high extinction of A_V ∼ 11.9 mag from the spectral slope of the planet, which is sufficient to match our findings. Finally, there is no known previous discovery of the planetary mass companion around V1295 Aql or a brown dwarf companion around HD 101412. Follow-up observations to confirm that these objects are comoving is necessary.

Ty CrA is a known quadrupal system with Chauvin et al. (2003) observing the fourth stellar companion around Ty CrA. The Chauvin et al. (2003) work was unable to verify if the object was comoving. Given the 16 year difference between our observations and Chauvin et al. (2003) observations and assuming the proper motion of Ty CrA, we find that if the fourth component was not comoving, the object would have moved by 052, while we measure a positional difference of 021. Given the large timescale and that the fourth component of Ty CrA is at a projected separation of ∼18 au, it is likely we are observing the orbital motion of the fourth component of Ty CrA. However, confirmation of the object being bound is necessary given the systems complex orbital dynamics.

For the four systems that are FS CMa candidates in our sample, HD 45677, HD 50138, HD 85567, and HD 98922, we observe polarized flux at large projected distances (>200 au), as shown in Figure 12. Additionally, these polarized signals are bright, and in the case of HD 45677 and HD 50138 there is significant U_ϕ polarized flux suggesting multiple scattering events due to the optically thick nature of the material.

Previous studies have detected extended structures around B[e] stars via H-alpha (Marston & McCollum 2008; Liimets et al. 2022). These structures are typically shell-like or indicative of bipolar outflows. We do not observe similar structures around our four FS CMa candidates, which appear to be closer to typical protoplanetary disk structures. However, previous studies were at a larger spatial scale (>1') and emission, while our observations are at smaller spatial scales and from a polarized source. Finally, our targets have a lower mass than those studied by Marston & McCollum (2008) and Liimets et al. (2022).

Stars that exhibit the B[e] phenomenon are a heterogeneous group that include pre-main-sequence, main-sequence, and evolved systems. Additionally, many of these stars exhibit similar features making classification difficult. FS CMa as defined by Miroshnichenko et al. (2007) hosts emission-line spectra containing hydrogen lines; large infrared excess that peaks at 10–30 μm, located outside of a star-forming region; and if it has a secondary companion it is either fainter and cooler than the primary or degenerate. However, even given these parameters, distinguishing between B[e] classifications can be difficult. For example, a thorough investigation by Varga et al. (2019), which studied HD 50138, concluded that the evolutionary state of HD 50138 could not be unambiguously determined through mid-infrared spectroscopy. We conclude that the structures observed around the FS CMa candidate stars in our Q_ϕ and U_ϕ images are more similar to the expected structures of protoplanetary disks rather than outflow from FS CMa or another type of evolved star. Future investigations into the resolved circumstellar material around stars are necessary including resolved ALMA observations to test the kinematics of the gas in the systems. The classification of these systems to be protoplanetary is extremely important as they represent four of the five systems with 4 M_⊙ < M_star <8 M_⊙ in our sample that have detected scattered light flux.