Large Binocular Telescope Search for Companions and Substructures in the (Pre)transitional Disk of AB Aurigae

L'-band (λ₀ = 3.7 μm, Δλ = 0.58 μm) observations of AB Aur were obtained on the nights of 2014 February 8 and February 13. On February 8, observations of AB Aur were taken only with the left (SX) side of the telescope, while both sides were used (in noninterferometric mode) on February 13. Individual exposure times were 0.175 s; five exposures were co-added within the hardware before transferring from the camera. Observations were taken at two nod positions with nodding every 1–2 minutes. On February 13, we also observed HD 39925 (K5, K = 3.899, L = 3.52) as a PSF reference star, with identical AO settings to allow for referential differential imaging (RDI). We switched between the science target and the reference twice. Because AB Aur transits within 2° of zenith at LBT, the very rapid field rotation caused the AO loop to be unstable for a few minutes. In particular, the left (SX) side AO loop on February 13 was completely open between parallactic angles −61 to 54°. The right (DX) side obtained continuous observations in that time but with nonoptimal AO correction. Conditions varied from good seeing conditions on February 8, to poor conditions for the final acquisition of the reference star HD 39925 on February 13.

For photometric calibration, each of the saturated sequences was immediately followed by unsaturated observations of AB Aur using a neutral density filter with a transmission of 0.9%.

K_s-band (λ₀ = 2.16 μm, Δλ = 0.32 μm) observations of AB Aur are described and analyzed in detail in Paper II (Betti et al. 2022) and used here for completeness. The observations were performed on the night of 2015 January 4, using only the left (SX) side telescope. Individual exposure times were 2 s, and observations were taken at two nod positions with nodding every 5 minutes. Two PSF reference stars, HIP 22138 (G8III, K = 4.627, L = 4.582) and HIP 24447 (K0, K = 4.219, L = 4.136), were observed before and after the observations of AB Aur, respectively, with identical instrument settings.

We employed a custom-built IDL pipeline (Bonnefoy et al. 2014) for applying bad-pixel correction, nod-subtraction, centering of the star by fitting a Moffat profile, aligning the frames, and cropping them to a 300 × 300 pixels field of view. With a plate scale of 10.7 mas px⁻¹ (Maire et al. 2015), this corresponds to 321×321. For AB Aur, frames where the loop was open or where the AO correction was very bad were removed based on principal component analysis (PCA), with some additional bad frames not detected by this method removed manually. The PCA analysis also revealed that the PSF shape differed significantly between the two nod positions. This was attributed to the dichroic beamsplitter that sends light to both the LMIRCam and the 8–13 μm camera Nulling Optimized Mid-Infrared Camera (NOMIC) at LBTI (Hinz et al.2016) and can cause different internal reflection and diffraction effects at different positions. We therefore treated the data from the two positions separately for the creation and subtraction of the PSF and recombined them again afterwards. The final number of usable frames in each data set is listed in Table 1.

All the observations were post-processed using the IPAG-ADI pipeline (Chauvin et al. 2012). We subtracted the PSF for the data that had sufficient field rotation with closed loop through meridian transit (data sets 1 and 2 for AB Aur) using multiple ADI techniques to search for point sources and to reveal asymmetric disk structures. For those data sets where the reference star was observed immediately before or afterwards (data sets 2–7 for AB Aur), RDI processing was applied to reveal the rotationally symmetric disk structures without self-subtraction.

3.2.1. Angular Differential Imaging

3.2.2. Reference Differential Imaging

To explore the point-source sensitivity of the LBT observations, pixel-to-pixel noise maps of each ADI reduction were estimated within a sliding box of 1.5 × 1.5 FWHM in the processed LBT field of view. We injected regularly spaced fake planets (every 10 pixels at 3 different position angles, with a flux corresponding to 100 ADU) in the original data cubes to evaluate the flux losses caused by the ADI process (e.g., Chauvin et al. 2010; Rameau et al. 2013). The final 5σ contrast maps were obtained using the pixel-to-pixel noise maps divided by the flux loss and normalized by the relative calibration with the primary star, and were also corrected from small number statistics following the prescription of Mawet et al. (2014) to adapt our 5σ confidence level at small angles.

The detection limits for data sets 1 and 2, obtained from the cADI processing, are shown in Figure 2. On the second night, the achieved 5σ mass limits were up to 1 mag better than on the first night, depending on the separation. We reached a contrast of 10 mag at 025, 12.5 mag at 05, and 14 mag at 1''. The 5σ contrast curves were then translated to mass limits based on the mass–luminosity relation for giant planets predicted by the ATMO 2020 (Phillips et al. 2020) and BEX-DUSTY-cold-start (Marleau et al. 2017; Asensio-Torres et al. 2021) evolutionary models. Mass limits were derived considering either a 2 and 4 Myr system. For the more optimistic scenario of 2 Myr of age, the mass limits for companions around AB Aur from the ATMO 2020 models translate to: 3–4 M_Jup beyond 06 (∼100 au), 6 M_Jup at 05 (81 au), and 14 M_Jup at 03 (50 au). The inner working angle of the observations is approximately 02 (28 au), where we would have been sensitive to brown dwarf companions with masses above ≈35 M_Jup . In the case of the BEX-DUSTY-cold-start, models we recover a mass limit of 25–20 M_Jup between 02 and 04, 20–15 M_Jup between 04 and 06 and 15–11 M_Jup beyond 06. Although this results clearly indicate that we are only sensitive for high-mass objects arising from this cold-start or low entropy scenario, it is important to point out that recent works (e.g., Berardo et al. 2017; Marleau et al. 2017) have showed that cold start is actually unlikely to occur for giant planet formation and that these objects are more likely associated with a high initial entropy (Mordasini et al. 2017; Marleau et al. 2019). Because of this, further analysis regarding the possible presence of planetary companions on the system is based on our derived ATMO 2020 detection limits exclusively.

Zoom In Zoom Out Reset image size

Based on our obtained mass limits, we derive a detection probability map for the second night, right-eye observations, as they provide a higher sensitivity to point sources, and assuming an age of 2 Myr. This map was obtained by using the Multi-purpose Exoplanet Simulation System code, a Monte Carlo tool for the predictions of exoplanet search results (Bonavita et al. 2012). We generated a uniform grid of masses and semimajor axis in the interval [0.5, 80] M_Jup and [10, 1000] au with a sampling of 0.5 M_Jup and 1 au, respectively. For each point in the grid, 10⁴ orbits were generated with a fixed inclination and position angle (based on the inclination and position angle of the disk from Tang et al. 2017) but randomly oriented in space from uniform distributions in ω, e < 0.1, and M, which correspond to the argument of periastron with respect to the line of nodes, eccentricity, and mean anomaly, respectively. The detection probability map is then built by counting the number of detected planets over the number of generated ones, by comparing the on-sky projected position (separation and position angle) of each synthetic planet with the 2D mass maps at 5σ.

Tang et al. (2017) proposed that the observed inner CO spirals might be triggered by the presence of a planetary companion at either 60–80 au or 30 au from the central star, while Boccaletti et al. (2020) recovered the signal of two apparent point sources located at a separation of 016 (∼26 au; f1) and 0681 (∼111 au; f2), with f1 seemingly associated with one of the observed spirals in the H band, consistent with the predictions from the ALMA CO observations. The location and estimated masses of f1 and f2 are reported in Figures 1 and 2. None of these point sources are recovered in any of our ADI and RDI reductions at the L' and K_s bands. As we can only resolve objects down to a separation of approximately 02, detection of f1 is not possible on our L' and K_s observations as is the case to the possible binary companion proposed by Poblete et al (2020). In the case of f2, the contrast derived by Boccaletti et al. (2020) is consistent with a mass ranging from 2 to 4 M_Jup based on the AMES-COND models. Boccaletti et al. (2020) also obtained a preliminary mass estimation of ∼3 M_Jup based on dynamical considerations derived by Wisdom (1980), which provides a relation between location of the inner edge of the observed cavity, the star mass, and the planet's mass and distance.

This predicted mass lies in the 5σ mass limit range derived from the second night of LBTI L' observations and has a detection probability of 50% at the deprojected position of f2, for an age of 2 Myr. At first glance, this detection probability is not enough to properly confirm or discard the existence of f2. However, it is important to take into account that the derived detection probability for a given semimajor axis (orbit) does not imply that the detection probability is the same at every point of that orbit. This is because this probability is only based on the count of detected planets along the full extension of the orbit, not considering where in the orbit they were specifically recovered. We can solve this problem by looking at the bidimensional mass detection limit at the specific location of f2 (Figure 5). At this location, we obtained a 5σ mass detection limit of ∼3.16 M_Jup , very close to the mass limit from Boccaletti et al. (2020) obtained at shorter wavelengths (hence, likely more impacted by disk foreground extinction) and implying that the detection probability at this location should be higher than 50%. Based on these results, the lack of detection of f2 suggests it is not a real companion.

We detect disk emission at the L' band up to ∼1'' separation. The innermost extended emission (<05) at the east and west sides of the disk are in good agreement with the location of the two main spirals detected by Boccaletti et al. (2020), with the western spiral (S2) being more clearly detected, while only the starting part of the eastern spiral (S1) appears to be recovered. The consistency of the appearance of these structures for both reduction methods (ADI and RDI), as well as for each of the independent ADI reductions (see Figure 5, Appendix B) suggests that the recovered signal effectively corresponds to the inner spirals, while the fragmented appearance of S2 is most likely caused by nonoptimal halo subtraction.

A brighter ring-like emission centered at a radius of ≈056 is evidenced in the L' and K_s LBTI images. It corresponds to a similar structure evidenced in the Subaru/HiCIAO (Hashimoto et al. 2011) and VLT/SPHERE polarized intensity H-band observations of Boccaletti et al. (2020) and shows a strong brightness asymmetry. This asymmetry is seen in the HiCIAO images but not recovered in the SPHERE observations. This suggests it to be caused by finer substructures left unresolved in our observations and Hashimoto et al. (2011) ones.

The ADI and RDI post-processed K_s-band images correspond to the (d) and (e) panels of Figure 1, respectively. The inner regions (<05) in the ADI images appears to be polluted by speckles. Instead, this reduction appears to be more sensitive at the outer parts of the disk, as faint emission from different structures, and particularly spirals, appears to be recovered, which is in good agreement with previously observed spirals (Hashimoto et al. 2011). Finally, the RDI reduction reveals the ring-like emission evidenced at the H and L' bands. The structure also shows the asymmetry reported at the L' band; hence it was possibly caused by the coarser angular resolution of these observations.

The ring-like structure lies at the H, K_s, and L' bands within the dust ring detected in the ALMA continuum observations (1.3 mm; Tang et al. 2017) as shown in Figure 1. This difference between observed structures at different wavelengths can be explained by considering the effect of dust trapping in a disk, which can be triggered by planet-disk interactions of one or more planetary companions. The presence of a massive planet, such as the case of f1, induces a pressure enhancement on the gas distribution of the disk, which traps large particles in this pressure maxima, while smaller particles are allowed to drift into the inner cavity that is depleted of large grains (de Juan Ovelar et al. 2013; Pinilla et al. 2015; Keppler et al. 2018). This could explain why a sharp inner edge can be observed for the dust ring at 1.3 mm, which traces large particles, while dust grains of smaller size can still be detected in the cavity, producing the features observed in the NIR.

Another possible explanation for the differences in wavelength of the observed structures, specifically the asymmetric ring-like structure, arises when considering the time variability of Herbig Ae/Be stars. It has been showed that these types of systems present irregular photometric (Eiroa et al. 2002) and spectroscopic (Mendigutía et al. 2011) variations, which have been attributed to shadowing from the inner ring structure (Vioque et al. 2018; Chen et al. 2019). If that is the case for AB Aur, then it is possible that the observed differences are only product of this variability between the different epochs of observation.

Although out of the scope of this study, further modeling of the disk would allow to determine the degree asymmetry of the phase function for the dust scattering (Henyey & Greenstein 1941) and subsequently the dust grain size of the observed region, and to put strong constraints on the properties of both possible planetary companions and the protoplanetary disk, as dust trapping is heavily dependent on those parameters. At the same time, further observations at different NIR bands at similar epochs would allow to properly determine the possible time variability of AB Aur and further identify if the differences between observed structures are also influenced by these variations.